Description

The goals for this build were:

  • It absolutely must be silent!
  • It must be compatible with Linux kernel 4.4.
  • It must have good interactive desktop performance

My applications are mainly:

  • web browsing with Firefox (often using several profiles and hundreds of tabs in each of several browser windows)
  • spreadsheet editing
  • KDE Plasma 5
  • running custom in-house applications

My constraints:

  • I do not plan to overclock.
  • No gaming. None.
  • I do not want a watercooled rig. I want my rig (hardware) to be relatively maintenance-free.

Results: very disappointing performance :-(

UPDATED RESULTS; I finally have everything sorted out! I changed out several components. I'm using a GTX 1070 GPU and I upgraded the CPU's to Xeon E5 2630-v4. I also upgraded to 256 GB of RAM.

I also decided to used different monitors. I'm using two Dell UltraSharp U3415W 34-Inch Curved monitors and one Dell 24" for a three-monitor setup.

I had to do a lot of troubleshooting to get this build sorted out. However, my main problem, it seems was that I bought the original Xeon CPU's on ebay and they were engineering samples. Sometimes ES CPU's are good, but in my case, the person sold me some that had been overheated or something. Eventually one CPU failed completely and the computer would not boot up. I replaced both CPU's with new v4 units and all is good!

See original discussion thread here:

$5500 budget dual Xeon ECC RAM with dual GPUs and dual 4K monitors for Linux and must be silent - Forums - PCPartPicker https://pcpartpicker.com/forums/topic/129898-5500-budget-dual-xeon-ecc-ram-with-dual-gpus-and-dual-4k-monitors-for-linux-and-must-be-silent

Of course, I blew through that budget. It was a lot of work, but in the end I"m happy with the result. Of course, it is NOT slow once I got the CPU issues sorted out.

Comments

  • 42 months ago
  • 10 points

Why buy a 970 if you don't want to game?

  • 42 months ago
  • 8 points

"web browsing with Firefox ...spreadsheet editing" with 20 cores and here i am with my little two core crying in my arms. D:

  • 42 months ago
  • 0 points

I'm sure your system is faster than mine. I should be the one crying. ;-)

  • 42 months ago
  • 3 points

"web browsing with Firefox ...spreadsheet editing"

Well dont those things favor stronger single core performance?

  • 11 months ago
  • 2 points

yes

  • 42 months ago
  • 0 points

Yes. You're right.

  • 42 months ago
  • 6 points

It's so good it's bad.

jk

  • 42 months ago
  • 1 point

Have you tried disconecting one monitor or both and replace the with a 1080p one ,i dont know anything about arclinux but can it handle so much 4k ,my way of troubleshooting is take bit off and seeif it makes a difference..honestly you should try with just 1 cpu see if that makes a difference

  • 42 months ago
  • 2 points

I did test with 1 CPU. No difference in GUI performance.

I did most testing with just one 1080p monitor.

  • 42 months ago
  • 1 point

So ...even if you dont want to you could try windows runs same benchmarks..if the results are the same then you have hardware software incoparibilty so you need to change one of the ,ask a friend with a win 10 hdd for a bit and second ...you ram seems a bit funny

  • 42 months ago
  • 1 point

I just downloaded Windows 10. I'll test it. But first I have to buy another storage drive. My drives are 100% filled with btrfs right now. Now room for Windows on any of them and I don't want to reformat these drives. So I'll get another drive for the Windows test. But that will also mean that the Windows test won't be running on exactly the same hardware...

What's funny about the RAM? It is high quality ECC server RAM. Not desktop RAM.

  • 39 months ago
  • 1 point

Well.. that might be the thing, unless you're running a server for your spreadsheets, then you should just get desktop components.

  • 42 months ago
  • 1 point

I have windows 10 installed. I just ran a Fire Strike benchmark.

My Fire Strike (v1.1) score was 10712 with a GTX 970 http://www.3dmark.com/3dm/12560080

  • 42 months ago
  • 1 point

oh a 7600$ should do much better but as you said it wasn't built for gaming so yeah

  • 42 months ago
  • 0 points

lol

  • 42 months ago
  • 3 points

What are you using this for that called for this kind of horse power? If it is just web browsing and spreadsheet editing, I would hate to see the spread sheets you are editing. This is a beast of a PC, I would expect to see something like this in a sever room for a small business. Good lord 40 threads, 256GB of RAM, 30TB of HDD storage, I am just going to sit here and admire this parts list for a bit. Very nice touch adding the thermal pics.

Do you have any of the HDDs in a RAID setup?

I would ask why you have a $300 GPU in a build that will not be gaming, but when you total price is 8k what is an extra $300. lol

– No gaming. None.

Have you tried disabling one of your CPU's to see if that changes anything. I could consider disabling a CPU, and removing all unneeded RAM and test one piece at a time. What are you having the performance lags on? General computing or something specific? I would suspect you may have a faulty part, and you have a lot of parts to test. :/

  • 42 months ago
  • 1 point

Thanks for your comment. It is indeed a nice parts list. And the finished build looks very nice and clean. I'm happy with the way it went together.

I am not using RAID.

I have tested with one CPU disabled in software. I also tested further and disabled all cores except one (in a separate test). The desktop GUI performance (which is painfully slow) did not change (subjectively).

Applications are slow to load. Applications are also slow to respond. For example, there is a noticeable delay when clicking menu items. The GUI is generally slow.

I notice the slowness in all interactive desktop GUI applications. It is both graphics (resizing windows) and I/O (opening or saving files).

I also suspect a faulty part, but nothing has shown up in any benchmarks yet.

Here are some benchmarks:

Look for Supermicro X10DAL 2x 2660-v3 Radeon 390X

Jmeter1 - 16 X Intel Xeon E5-2690 V3 @ 2.59GHz Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601276-HA-1601260HA52

Fiteyes Benchmarks, Linux Performance - OpenBenchmarking.org http://openbenchmarking.org/s/fiteyes

FitEyes Testing Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601278-HA-1601275HA22

FitEyes Testing Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601277-HA-1601275HA66

Jmeter1 - 16 X Intel Xeon E5-2690 V3 @ 2.59GHz Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601276-HA-1601260HA52

Jmeter1 - 16 X Intel Xeon E5-2690 V3 @ 2.59GHz Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601276-HA-1601276HA91

  • 42 months ago
  • 1 point

Ya, the bench marks don't really point at the location of the issue. :/

Have you watched CPU/RAM/GPU/SSD/HDD utilization when testing the PC? I'm wondering what percentage of the hardware is being utilized. With general UI jumpiness I would figure you are not hitting a CPU or RAM cap (or at least I hope not with this kind of hardware), but a disk usage limit. I know this is a dumb question but you did install on the SSD right? You know our fingers like to do dumb things when typing. I'm assuming you are using the M.2 for the boot drive, I wonder if there may be a driver issue with PCIe storage with that version of Linux. Have you tried the OS on the 2.5 SSD or one of the HDDs?

General slowness like that could very well come from poor preforming storage devices, or even a fault motherboard. Just a thought, and if it was either of those it would not really show up on your benchmarks.

  • 42 months ago
  • 1 point

I'm wondering what percentage of the hardware is being utilized.

Very low utilization for the most part.

The NVMe drive gives read performance of 1.75 GB/s so I think it must be installed right. It is the root ("/") drive.

The NVMe drive is the fastest drive in the system.

It could be a faulty motherboard, but I am not getting an indication of that on any benchmarks. However, as you said, "the bench marks don't really point at the location of the issue"

  • 42 months ago
  • 3 points

Xeons are excellent for highly parallelised loads where the lower clock speed doesn't matter when you have 20 cores. For primarily single-threaded workloads like browsing and desktops a Xeon is simply worse than a cheap i5 because it has half the clock speed.

The ram certainly isn't helping - 8 single sticks from different packs rather than larger 2x32 or 4x32GiB modules will limit the memory bandwidth - running possibly in single channel rather than double or quadruple. Don't know about your motherboard though.

Finally - what are the thermals on the CPUs? How good is the cooling and airflow with two giant air coolers almost in contact?

Linux is great, and Arch should be particularly light weight but for what you want this hardware seems worse than a £300 netbook. If there weren't picture I might think it was trolling.

  • 42 months ago
  • 2 points

Yes, the GUI performance is slightly worse than my $100 Gigabyte BRIX GB-BXBT-2807 Celeron (which is also running KDE).

The CPUs are running very cool.

This is good quality RAM. I don't think you can get ECC RAM in the larger bundles like you see for desktop systems... this RAM was specifically recommended to me by a respected guy on PCPP.

Here's part of the RAM discussion we had:

The "load" on the memory controller with registered DDR4 is going to be well below maximum supported either way. The difference in performance from interleaving characteristics will be effectively nill. It just doesn't matter.

If we were talking about a build on a motherboard with 2 or more DIMM slots per channel, then I would absolutely advise the dual rank DIMM's over the quad rank to keep the load on the memory controller down, and keep more upgrade options available.

  • 42 months ago
  • 1 point

Finally - what are the thermals on the CPUs? How good is the cooling and airflow with two giant air coolers almost in contact?

Notice that the airflow direction is upward. There are two more fans above.

And there are two fans in the front of the case. Air comes in the front, flows through the CPUs and out the top. CPUs run very cool.

  • 42 months ago
  • 1 point

Xeons are excellent for highly parallelised loads where the lower clock speed doesn't matter when you have 20 cores. For primarily single-threaded workloads like browsing and desktops a Xeon is simply worse than a cheap i5 because it has half the clock speed.

The benchmarks seem to be proving this out. The Fire Strike score (10712) is on par with an i5 with the same GTX 970 card.

  • 42 months ago
  • 2 points

Motherboard bios up to date? Bios settings correct?

Tried using Windows 10? .... Maybe Linux is unable to recognise and fully utilise those CPUs??

  • 42 months ago
  • 1 point

Yes, BIOS settings appear correct.

​Linux still rules supercomputing | ZDNet http://www.zdnet.com/article/linux-still-rules-supercomputing/

All 20 cores / 40 CPU threads are operating correctly according to the tests I have run.

Here's some background:

https://pcpartpicker.com/forums/topic/129898-5500-budget-dual-xeon-ecc-ram-with-dual-gpus-and-dual-4k-monitors-for-linux-and-must-be-silent#cx1425655

  • 42 months ago
  • 2 points

Linux ruling super computer has nothing to do with troubleshooting. It is still worth tossing a win10 or server 2016 (use an evaluation copy, don't actually buy windows) on the box to see if it runs any better. If your distro is not liking this hardware it does not matter how good the software it at super computing you are shooting yourself in the foot. I could see single core performance being pretty meh in this build but it should crush anything in multi core. There is a chance your distro is not optimized for dual CPU 20 core 40 thread hardware.

  • 42 months ago
  • 1 point

I have tested 3 different distros (4 versions total). I tested Kubuntu 14.04 and 16.04. I tested opensuse. I am currently running Arch Linux. The Arch install is super clean and very optimized for everything I know to tweak.

This is not an OS problem. When I initially built this system I also built an i7 system. I installed Kubuntu 14.04 on both. The i7 was as fast as expected. This system was a turtle.

There is a chance your distro is not optimized for dual CPU 20 core 40 thread hardware.

All 40 CPU threads are active and responding. NUMA is enabled properly.

However, I have also tested with all but one core disabled and it doesn't help.

  • 42 months ago
  • 2 points

Which distro do you use? Did you ask in a Linux forum as well? Maybe this is a kernel or distribution problem.

  • 42 months ago
  • 1 point

It's running Arch Linux. I had the same issues with Kubuntu and OpenSuse, so it is not likely a distro problem. I have tried kernels ranging from 4.2 to 4.6. No difference, so it is not a specific kernel problem.

  • 42 months ago
  • 2 points

idk if the linux kernels support dual CPUs. windows is likely to show massive performance increases.

  • 42 months ago
  • 1 point

Nope. No performance increase in Windows at all.

  • 42 months ago
  • 1 point

windows server?

  • 42 months ago
  • 1 point

Not the server edition. I installed Windows 10 64-bit on a trial. I'm going to remove it and go back to Arch Linux when I finish running these benchmarks.

  • 42 months ago
  • 1 point

Looks like it requires Windows Server edition to use both my CPUs... is that true?

I guess it doesn't matter much for this benchmarking. I'm going back to Linux when I finish and Linux definitely uses both CPUs.

  • 42 months ago
  • 2 points

i don't think xeon was ever made for browsing , try i7 extreme

  • 42 months ago
  • 2 points

the xeon is probably best for running servers

  • 42 months ago
  • 1 point

yeah true

  • 42 months ago
  • 1 point

the pc seemed to be total overkill for what he wanted it for anyway. rip money

  • 42 months ago
  • 1 point

yeah , he would've made 2x i7 pcs using them i don't know if he can rma these things

  • 42 months ago
  • 1 point

oh right, fair enough then.

  • 42 months ago
  • 0 points

I built an i7 system and it is much faster for everything I use it for.

  • 42 months ago
  • 1 point

good decision

  • 42 months ago
  • 1 point

a one i7 extreme is way better than those xeons

[comment deleted]
  • 42 months ago
  • 2 points

This may be a shot in the dark to try since I'm not overly familiar with Xeons and Supermicro boards, but check inside your BIOS to see if your CPU L1, L2, or L3 cache is disabled.

I remember years ago when I had my Pentium 60 having similar poor performance after fiddling with those settings in my BIOS.

  • 42 months ago
  • 2 points

Thanks. I'll check that...

  • 42 months ago
  • 2 points

its probably the low clock speed processors.

  • 42 months ago
  • 1 point

the clock speed isn't that low for a Xeon.

Also, the system is slower than my Gigabyte BRIX GB-BXBT-2807 NUC.

CPU: Intel Celeron N2807 / 1.58 GHz Max Turbo Speed: 2.16 GHz

These Xeons clock up to 3.0 GHz.

  • 42 months ago
  • 1 point

Im saying clock per core. So the relative clock speed in conjunction and correlation with the amount of cores.

  • 42 months ago
  • 1 point

Each core can clock up to 3.0 GHz. So I don't understand how this system can have Firefox performance that is similar to my Celeron-based NUC. In fact, Firefox opens faster on that NUC. Both are running Linux, so it's not a difference due to OS.

Anyway, thanks for your comments.

  • 42 months ago
  • 1 point

geez that sucks man, i hope you can get things sorted out.

  • 42 months ago
  • 1 point

Thanks! I'm trying some of these suggestions right now.

  • 42 months ago
  • 2 points

You're using a server optimized board for end-user GUI workloads, should have tried the Asus workstation ones. Also, check if Intel's turbo boost is on, most apps still use only one core.

  • 42 months ago
  • 2 points

You are right. My i7 system is much faster. Still, the dual Xeon system shouldn't be slower than my Gigabyte BRIX GB-BXBT-2807 Celeron should it? (Both run Linux with KDE.)

The CPU's do increase clock speed under load. What do you have in mind that I should check to see if Intel's turbo boost is on? The BIOS? I checked all BIOS settings and everything looks correct.

  • 42 months ago
  • 2 points

So I have to ask, are you attempting to power 2x 4k monitors (effectively creating 8k of real estate on your desktop?), because there are 2 problems with this. The GTX 970 is not designed to handled 8k resolutions. Nvidia states the maximum digital resolution of the gpu would be 5120x3200 (or 16.3MP), where as 2x 4k resolutions (assuming your using 3840x2160), would total over 16.5MP (7680x2160) at the smallest configuration, if you intend to use DCI 4k (4096x2160) or even Ultra-wide TV 4k (5120x2160), Your looking at 17.7MP and 22.1MP respectively, which is way beyond what that GPU can do even just on the desktop.

Quick way to test this would be to just try one monitor, if performance improves, then you are pushing your single 970 outside its bounds of capabilities.

Also a side note, I have 2x 980 Ti's (6GB) Gpu's in SLI, and just sitting on my desktop at 7680x2160 uses almost 3GB of VRAM. Boot up a simple 4k youtube video, and my VRAM usage goes to 4.5GB (Full screened), try and load into any 3D accelerated application, and I sit pegged at 6GB, with horrible performance. Even at 4k the 6GB 980Ti's are struggling to get away from that 6GB VRAM usage mark.

All of what I mentioned above might just be mute if I misunderstood the 4k mention you made at the end of your description. But if you are going to be putting this kind of money into a linux 8k rig, you may want to look into a GPU with more VRAM, especially if you intend to do any heavy 2D/3D workstation loads, you may want to look into the quaddro cards, as many of them are already into the 20+GB VRAM range, and are designed for 4,8,12,16,20k workstation setups. Something like the Quadro M5000 supports up to 16k of desktop real estate with a max supported resolution of 4096x2160 from each DP1.2 port. For a maximum display real estate of something like 8192x4320 (or 35.4MP).

Good luck on getting this figured out, hope I helped.

  • 42 months ago
  • 1 point

That is all very helpful information! It will come in handy after I solve the performance problems. Thank you.

I'm not running any 4k monitors. I'm doing my testing with one 1080p monitor.

  • 42 months ago
  • 2 points

Ya know what...I just realized your using a Muskin reactor. I had one of those SSD's a while back, and it did nothing but bog my system down, and I ultimately ended up replacing it, because I would have read and write speeds that were slower than my laptop 5200 RPM drives... I'm not sure if you've tried it yet, but you may want to see about either just trying a different drive, or maybe just installing linux onto a thumb drive, and run from that, and see if performance is better.

  • 42 months ago
  • 1 point

That's interesting. Thanks for the idea. I'll try it today.

  • 42 months ago
  • 2 points

LOL. You have built a server and setup it up like a workstation. You need to install something like OpenVZ and install your OS as a virtual machine.

  • 42 months ago
  • 1 point

Yeah, it is supposed to be a workstation...

  • 28 months ago
  • 2 points

The OS can play a big role in how a PC preforms. I ran a dual CPU Xeon in Windows 10 and had horrible performance. I tried a copy of Windows Server 2012 R2 and the performance was amazing. You should also make sure all of your drivers are updated and configured properly.

  • 42 months ago
  • 1 point

Over the top build! +1

So, how it it slower than your other PC?

  • 42 months ago
  • 1 point

It was fun to build. I learned a lot. The build turned out very clean as you see.

I had to do some minor case mods (drilling and tapping two holes for the motherboard mount screws).

Why is it slower? That's the $7,000.00 question!!!

BTW, it's not just a little slower. It's a LOT slower.

  • 42 months ago
  • 2 points

try running windows 10 as your primary OS.

  • 42 months ago
  • 1 point

Just tried it. Windows sucks worse than Linux.

  • 42 months ago
  • 1 point

When you say "slower" can you give a little more detail? What's slower?

  • Does it boot slower?
  • Render slower?
  • Are you benchmark numbers way off?
  • Applications take an eternity to load?
  • 42 months ago
  • 1 point

Every benchmark I have run seems normal. However, I believe I need some advice regarding benchmarks. I would like to re-run all benchmarks carefully.

Applications are slow to load. Applications are also slow to respond. For example, there is a noticeable delay when clicking menu items. The GUI is generally slow.

I notice the slowness in all interactive desktop GUI applications. It is both graphics (resizing windows) and I/O (opening or saving files).

The OS is installed on the NVMe drive.

It boots very slow but that is because it is a server motherboard. That's a different issue. Beyond the server BIOS, booting up seems relatively normal.

  • 42 months ago
  • 2 points

I see, good point on the boot thing. But the NVMe drive has got me curious. Is it really running at spec on that board? Perhaps you can do an I/O test and see.

As for the GUI, have you installed the latest drivers for the GPU or maybe tried installing a different one to see if anything changes?

  • 42 months ago
  • 1 point

Thanks for your reply.

I'm running the proprietary nvidia drivers from Arch Linux. They are installed correctly. (I have reinstalled several times. I have tested different settings. There is no big difference even when I use different nvidia settings.)

I also tested with a Radeon 390X and the system was still slow.

I ran some I/O tests on the NVMe drive. I got read speeds of 1.75 GB/s which is about what is expected I think. I was hoping for slightly higher speeds, but those speeds are still far faster than most SSD's so they could not be the cause of my very slow performance.

Some of my benchmarking threads are here: https://pcpartpicker.com/forums/topic/129898-5500-budget-dual-xeon-ecc-ram-with-dual-gpus-and-dual-4k-monitors-for-linux-and-must-be-silent#cx1339942

https://pcpartpicker.com/forums/topic/129898-5500-budget-dual-xeon-ecc-ram-with-dual-gpus-and-dual-4k-monitors-for-linux-and-must-be-silent#cx1341219

You can see some of my other benchmarks here. Look for Supermicro X10DAL Radeon 390X.

Jmeter1 - 16 X Intel Xeon E5-2690 V3 @ 2.59GHz Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601276-HA-1601260HA52

Testing Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601278-HA-1601275HA22

  • 42 months ago
  • 1 point

4 months ago?! Yikes... I'm sorry you've been suffering for so long.

Have you been able to confirm that the kernel/OS are fully utilizing the CPU's? I'd start some workloads and run htop to see where each core/thread stands.

  • 42 months ago
  • 1 point

Yeah, been struggling with this performance issue for months!

Do you have any specific suggestions for creating workloads to measure with htop?

I just ran interbench for about the lasts 5 hours and every test passed with 100% of deadlines met. It's weird that the system runs so slow but the benchmarks don't seem to reveal any major problems (yet).

  • 42 months ago
  • 1 point

So... High access times? Am I reading this right?

  • 42 months ago
  • 1 point

Some of my benchmark results. Look for system named "Supermicro X10DAL 2x 2660-v3 Radeon 390X". I ran those benchmarks with the Radeon card. (I get similar results with the nvidia GTX 970.)

Jmeter1 - 16 X Intel Xeon E5-2690 V3 @ 2.59GHz Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601276-HA-1601260HA52

Fiteyes Benchmarks, Linux Performance - OpenBenchmarking.org http://openbenchmarking.org/s/fiteyes

FitEyes Testing Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601278-HA-1601275HA22

FitEyes Testing Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601277-HA-1601275HA66

Jmeter1 - 16 X Intel Xeon E5-2690 V3 @ 2.59GHz Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601276-HA-1601260HA52

Jmeter1 - 16 X Intel Xeon E5-2690 V3 @ 2.59GHz Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601276-HA-1601276HA91

  • 42 months ago
  • 1 point

The image showing the desktop shows how slow the GUI responds to window resizing. The mouse is down in the lower right corner but the window resize is lagging far behind.

  • 42 months ago
  • 1 point

Whadda ya mean by "slower"? Game benchmarks or rendering or what?

  • 42 months ago
  • 1 point

Applications are slow to load. Applications are also slow to respond. For example, there is a noticeable delay when clicking menu items. The GUI is generally slow.

Windows are slow to resize. See image of me resizing a window here: https://pcpartpicker.com/b/r7zMnQ

I notice the slowness in all interactive desktop GUI applications. It is both graphics (resizing windows) and I/O (opening or saving files).

Graphics frame rates are low. I don't have any benchmark numbers to quote on those. But I do have other benchmarks. All the non-graphics benchmarks are fine. See links in my other comment.

  • 42 months ago
  • 1 point

I suggest troubleshooting ur PC by removing some parts and starting with the basic system, then gradually add more parts onto it or run stress tests and check system utilization, and see if the problem is Linux not able to fully utilize your system.

  • 42 months ago
  • 1 point

Yes, that is a good idea. I did that, but maybe I need to start over and do it again.

I initiallly started with just one storage drive (the NVMe drive) and one monitor connected and just 4 sticks of memory. I swapped different GPU's in and out. The system was pretty bare at that point. Performance was terrible still.

I also tested it with only one CPU during that time.

I've been trying to resolve this for six months! :-( I gave up and took a break for a couple months out of those six.

see if the problem is Linux not able to fully utilize your system.

Linux is definitely not the problem.

  • 42 months ago
  • 1 point

Have you tried running windows on it?

  • 42 months ago
  • 1 point

Not yet, but many people have suggested that as a test. I plan to try it this weekend.

  • 42 months ago
  • 1 point

Are your other drives not bottlenecking you? Might also be the RAM and mobs, but I doubt it

  • 42 months ago
  • 1 point

the only pseudo-similar problem I had remotely related to your is this one: brand new Lenovo PC (M93p sff I think) running Ubuntu 14.xx LTS and when plugging in a 3rd monitor... it would crap the gui everywhere. We discovered ubuntu+M93p disliked 3 monitors and also disliked having to feed 2 VGA display. So we ditched the old vga monitors and used only 1 VGA cable and all the other output go through displayport or DVI. So 1 analog only and all the rest digital.

like somebody said, try to reduce the number of graphic output and reduce resolution. Test at lower graphic specs and see if it improves. http://ca.pcpartpicker.com/b/r7zMnQ#cx1760465

  • 42 months ago
  • 1 point

I have done almost all testing with just a single 1080p monitor attached. And it is super slow.

  • 42 months ago
  • 1 point

KDE Plasma 5

Well there's your problem. Not running i3? That's a paddlin'. /s

Theoretically this bad boy should be destroying your workload. As much as I hate Windows it might be worthwhile to install and see if you see the same issues. Then blow that crap away and put a modern OS back on your rig.

Oh try KDE Neon too.

  • 42 months ago
  • 1 point

Thanks for the ideas.

I plan to test with Windows, but I will have to buy another storage drive to do the testing. I'm not going to screw with one of my btrfs drives for a temporary test of Windows. (My current storage configuration is set up really nicely. I put a lot of work into it. )

KDE Neon looks very nice. But I like Arch Linux too much to switch. :-)

  • 42 months ago
  • 1 point

You and I both brother. My current OS of choice since the AUR has such wonderful packages with great maintainers. Arch is only marginally behind Neon usually by less than a couple of hours.

Yuck on btrfs. I don't have any storage use case beyond good old, reliable ext4.

  • 42 months ago
  • 1 point

I love btrfs now. I use snapper and btrbk. It's really quite awesome! :-)

  • 42 months ago
  • 1 point

I installed Windows 10. Here are some benchmark results.

My Fire Strike (v1.1) score was 10712 with a GTX 970 http://www.3dmark.com/3dm/12560080

That score is very comparable (even slightly better) than Xeon systems with similar graphics cards. Here's one: http://www.3dmark.com/fs/8833234

  • 42 months ago
  • 1 point

Hmm sadly that doesn't tell us a whole lot as to what would be going on...

I can't think of an IRC channel that might be able to help you with this, maybe check FreeNode?

  • 42 months ago
  • 1 point

I'm about to conclude that Xeon systems really suck for desktop application GUI responsiveness. The system seems to be performing properly in other areas. It's just super laggy when it comes to the GUI.

  • 42 months ago
  • 1 point

Just a ball park guess, but have you tried moving the GPU up to the top slot? Server boards can be very finicky with GPU placement in their PCI-E Slots. Also, are both CPU power cables connected together/on the same rail? It's possible to saturate the power link, forcing your CPUs to run less responsively, rather than not at all. It could very well be a power issue. If you have a spare PSU it doesn't hurt to give it a try with that.

  • 42 months ago
  • 1 point

Good suggestion. I did try moving the GPU to different slots. I also tested two other GPUs (a GTX 960 and a Radeon 390X). It was slow with all of those.

Yes, both power cables are connected to the GPU and both connectors come from the same cable.

I think it might be a faulty part. I don't know which one, but I am willing to try another PSU. I'll have to buy one.

It's possible to saturate the power link, forcing your CPUs to run less responsively, rather than not at all.

The benchmarks don't seem to indicate this, but maybe that's because I don't know how to read the benchmarks well enough or I'm not running the right ones. But CPU performance on various pts tests was normal.

  • 42 months ago
  • 1 point

Benchmarks will not be an indicator of power error in this situation, they technically supply enough power to run the CPUs at full bore. The problem is they can not do it fast enough when they suddenly spike in performance requirements, which may cause the slow down you are witnessing. Xeons are power efficient, but there are two of them and they do take power. The PSU you have is made for gamers who overclock on a single CPU system(as it's advertising says). Looking at the I/O of the PSU, it doesn't have a second CPU power cable output. Which means you are probably overloading a power rail. I'd recommend giving this a shot. It has 2 dedicated CPU outs, and EVGA in general just makes amazing PSUs in every category. (10/10 in Jonnyguru for that one). Also, all black cables.

  • 42 months ago
  • 1 point

I will try that PSU as a test. That's a good suggestion and I appreciate it.

However, the PSU I have now DOES have a second CPU power cable. I have all required power cables connected for the CPUs.

  • 42 months ago
  • 1 point

It does? Well, that's the first thing you should try. Get each CPU on it's own power cable and see what happens. If that doesn't do anything then I don't think there is a point in buying the EVGA PSU.

  • 42 months ago
  • 1 point

Yes. I just took some pictures to show that there are two separate CPU power cables connected from the PSU to the MB. I'll upload the pics later tonight.

  • 42 months ago
  • 1 point

Has to grab a calculator in order to count the amount of RAM. +1

  • 42 months ago
  • 1 point

lol

I initially bought half that much RAM. But then I learned that for best performance you need to balance the RAM. That means all channels of each CPU should be populated evenly. With this motherboard, that means filling all RAM slots. So I had to buy 4 more sticks of RAM.

I did this because of the performance problems I am having. It was one thing to check off the list.

See here for a problem that sounded similar to mine:

single i7 MUCH faster than dual xeon E5-2650 v3 !!! -- CFD Online Discussion Forums http://www.cfd-online.com/Forums/hardware/144936-single-i7-much-faster-than-dual-xeon-e5-2650-v3.html

  • 42 months ago
  • 1 point

I understand your logic but I don't think it's the cause of your problem. I'm not familiar with Linux, and I know you've tried with different versions, but I feel that the bottleneck might come from Linux incompatibility. From what I see, you're using a PCIe adapter for your M.2 drive. I believe your OS and your applications are installed on that drive. Maybe the adapter is faulty, maybe the drive itself is faulty. I would try plugging the adapter in different PCIe slots (if you haven't already tryed that).

  • 42 months ago
  • 1 point

I understand your logic but I don't think it's the cause of your problem.

Right. I proved that isn't the cause of my problems. But it was the best theory at the time so I had to test it.

I feel that the bottleneck might come from Linux incompatibility.

Incompatibility with what? Everything I'm using is known to be compatible with Linux.

Maybe the adapter is faulty, maybe the drive itself is faulty.

I do suspect some part is faulty. I am still not sure which part. The NVMe drive shows read speeds of 1.75 GB/s via hdparm, so the drive and adapter don't seem to be faulty. But maybe that test just doesn't expose the problem.

I would try plugging the adapter in different PCIe slots (if you haven't already tryed that).

Yes, already tried that.

  • 42 months ago
  • 1 point

The reason. NVME drives are meant for things like large,quick storage not for apps. Have you tried doing some updates on arch?

  • 42 months ago
  • 1 point

I keep Arch updated. My Arch install is very clean too. I didn't even install all of KDE -- just the exact stuff I wanted.

So you think the NVMe drive is the problem? I guess that's possible. I had not considered that because on hdparm the NVMe drive is fast.

  • 42 months ago
  • 1 point

NVME needs time to initialise and apps loading from it will be slow as NVME drives take 5-10 minutes to reach its maximum potential

  • 42 months ago
  • 1 point

I never heard of this. However, the system is still slow after having run for more than 10 minutes. It is the same right after booting up, after an hour or after running for a couple days.

  • 42 months ago
  • 1 point

Maybe try a BIOS update.

  • 41 months ago
  • 1 point

Maybe try a kernel upgrade as well.

  • 41 months ago
  • 1 point

Already did.

  • 42 months ago
  • 1 point

Dumb question but is that m.2 ssd set as C: boot drive, and if so are its read/write speeds checking out? It's possible you have a bum ssd or rotten m.2 slot on that board.

  • 42 months ago
  • 1 point

I have the m.2 drive set as "/" (root). /boot is on a separate drive.

The m.2 drive has read speeds of 1.75 GB/s. I tried it in different slots and there was a difference in speed, so I left it in the fastest slot.

  • 42 months ago
  • 1 point

So you do have Linux installed to the m.2 drive then.

  • 42 months ago
  • 1 point

Yes, Linux is installed on the m.2 drive.

  • 42 months ago
  • 1 point

did you put thermal paste on your cpus? That's my guess.. Also, Why in the HELL did you spend SO MUCH on a computer that would only be used for a FRACTION of it's potential?? You couldve EASILY spent a cool 500 bucks for a computer that could do what you needed and still have the chance to upgrade, if needed.

  • 42 months ago
  • 1 point

did you put thermal paste on your cpus?

Yes, of course. The CPUs run cool. See my thermal images of the running system (under heavy load from interbench).

You couldve EASILY spent a cool 500 bucks for a computer that could do what you needed

I built a $1500 i7 at the same time. It performs much better!

However, the fact is that I need at least that i7 level performance. So a $500 build would not come close. In fact, the i7 doesn't have enough storage. But it does have much better desktop application performance.

  • 42 months ago
  • 1 point

T_T dude, you built this thing for WEB BROWSING.

  • 42 months ago
  • 1 point

ya don't need 20 cores for youtube.

  • 42 months ago
  • 1 point

We tend to think web browsing is not demanding on a system at all. But web browsers (with lots of tabs open) can put a pretty heavy demand on a system.

In all seriousness, I use my system very heavily for a lot of different stuff. including audio and video editing, photo editing, large spreadsheets, etc. But Firefox (multiple browser profiles, each with dozens or even hundreds of open tabs) taxes it as hard as anything else I do.

  • 42 months ago
  • 1 point

You do video editing? Then why on earth didn't you say so in the first place? And I suppose if you want hundreds of tabs open I can understand the need... But why? Why so many tabs? Nevermind, you don't have to answer that. Thank you for clearing that up for me. XD

  • 42 months ago
  • 1 point

My prior system was a single CPU Xeon workstation. I knew I wanted/needed something more powerful than what I had. So I made sure this one was an upgrade in terms of hardware. Unfortunately, it is significantly slower than my prior system. Here are the specs for the prior one:

  • Xeon Qc E3-1230 Processor
  • 4 x Kingston 8GB (2 x 4GB) 240-Pin DDR3 SDRAM DDR3 1333
  • Asus P8B WS LGA1155/ Intel C206/ Quad CrossFireX/ SATA3&USB3.0/ A&2GbE/ ATX Motherboard
  • PowerColor Go! Green ATI Radeon HD6750 1 GB DDR5 2DVI/HDMI/DisplayPort PCI-Express Video - - Card AX6750 1GBD5-NS3DHG
  • Crucial 512 GB m4 2.5-Inch Solid State Drive SATA 6Gb/s CT512M4SSD2
  • Samsung Spinpoint F4EG 2 TB SATA2 5400rpm 32 MB Hard Drive HD204UI/Z4 (internal)
  • Samsung Spinpoint F4EG 2 TB SATA2 5400rpm 32 MB Hard Drive HD204UI/Z4 (internal)
  • Samsung Spinpoint F4EG 2 TB SATA2 5400rpm 32 MB Hard Drive HD204UI/Z4 (internal) etc.

Obviously, I'm asking myself how the heck can this system be slower than that one??? (They both run Linux)

  • 42 months ago
  • 1 point

Can you somehow check if TRIM is enabled/active?

I have a feeling that the slowness is related to your storage drive(s).

You may have to play around with mirroring your install onto each of your other drives to see if the UI feels more responsive. You may be able to identify a problematic drive that's causing the UI slowness.

  • 42 months ago
  • 1 point

NAMe drives should not have trim enabled.

  • 42 months ago
  • 1 point

https://wiki.archlinux.org/index.php/Solid_State_Drives/NVMe

Have you attempted to apply fstrim manually?

Are you using the appropriate sector size for the formatted partition for your OS?

  • 42 months ago
  • 1 point

Thanks for the link. From that link:

Note: Contrary to recommendations for SSDs, NVMe devices should not be issued discards.

And that's something I took into account when setting the drive up, so I did it properly.

Regarding alignment and sector size, I'm not 100% sure I have everything perfect. I created the first partition at sector 2048 on all the different disks. I read that this was sufficient. But, as I said, I'm not 100% convinced.

Here's the partition info for the nvme drive:

Found valid GPT with protective MBR; using GPT.
Disk /dev/nvme0n1: 1000215216 sectors, 476.9 GiB
Logical sector size: 512 bytes
Partition table holds up to 128 entries
First usable sector is 34, last usable sector is 1000215182
Partitions will be aligned on 2048-sector boundaries
Total free space is 2014 sectors (1007.0 KiB)
Number  Start (sector)    End (sector)  Size       Code  Name
1            2048         2099199   1024.0 MiB  EF00  EFI System
2         2099200       987760639   470.0 GiB   8300  Linux filesystem
3       987760640      1000215182   5.9 GiB     8200  Linux swap
  • 42 months ago
  • 1 point

WHAT THE **** HOW MUCH RAM? but srsly how is this not powerful enough?! You've got 20 2.6Ghz cores.

  • 42 months ago
  • 1 point

128 GB RAM

I know... Seriously, how can this machine be slower than my laptop??? It doesn't make sense. But I have several new ideas from all the comments here. I'll be testing more ... and hopefully I'll get it resolved soon

  • 42 months ago
  • 1 point

What do you define as 'slow'?

  • 42 months ago
  • 1 point

see screen shot of window resizing. I also gave more details in other comments.

  • 41 months ago
  • 1 point

It is actually 256 GB of RAM. (I started with only 128, but as part of my troubleshooting I populated all banks for all CPUs.)

  • 42 months ago
  • 1 point
  • 42 months ago
  • 2 points

OK! He's got a good channel. :-)

  • 42 months ago
  • 2 points

Here is the forum post:

My highest end build ever turned out to be my slowest PC :-( - Troubleshooting - Linus Tech Tips https://linustechtips.com/main/topic/612805-my-highest-end-build-ever-turned-out-to-be-my-slowest-pc/

[comment deleted]
  • 42 months ago
  • 1 point

I would have contacted the motherboard manufacturer and ask for help. On a side note, I would have gone with an ASUS WS motherboard like the Z10PE. But as other people have said, try Windows first to rule out OS compatibility problem. After all, you have tried to switch pretty much everything except your motherboard and CPUs. Good luck, looks like a tough one!

  • 42 months ago
  • 1 point

Doctor Frankenstein, I'm afraid you've created a monster. :V That is an absolutely insane amount to spend on disappointment >.>

  • 42 months ago
  • 1 point

You are really better off making this xeon build a compute system for dedicated tasks, like video transcoding or batch editing, as well as a server for the business or house.

Also, why is the GPU on a lower pci-e lane. Can you move it up to the top one? You may be running at less than optimal on the 2nd slot.

  • 42 months ago
  • 1 point

You know you cannot put Petrol in Diesel operated Engines. Same goes here. Just saw your list and incompatibilities and saw these 2 issues:

Potential Issues / Incompatibilities

1 additional M.2 2280-M compatible slot is needed. Samsung 950 PRO 256GB M.2-2280 Solid State Drive and Supermicro MBD-X10DAL-I-O ATX Dual-CPU LGA2011-3 Motherboard are not compatible.

May be look for the solution. You know I made this account just to answer that :) Anyways one time I purchased Alienware high end laptop for gaming and it was running lower than my old laptop. I changed it and was happy with my current 5 years old MSI.

Motive is sometime you need to let it go and get new/old ones.

  • 42 months ago
  • 1 point

Those incompatibilities are wrong.

The PCI card resolves the issue.

Right now this is as much a learning project as anything. So I'm going to keep going as long as I'm learning.

The stress-ng, sar, sysstat and other tools are new to me. I'll enjoy getting experience with them.

  • 42 months ago
  • 1 point

Have you installed all the drivers for your parts?

  • 42 months ago
  • 1 point

I don't know what is supported by this motherboard, but have you tried non ECC RAM?

  • 42 months ago
  • 1 point

It only works with ECC RAM

  • 42 months ago
  • 1 point

I'm so confused why you did this

  • 42 months ago
  • 1 point

Had the same problem, reinstalled pulseaudio, worked like a charm. Good luck, OP.

  • 42 months ago
  • 1 point

thanks. Arch Linux doesn't tend to need stuff reinstalled like Ubuntu. I feel pretty comfortable that this Arch install is very clean. I actually did install Arch on this system 8 different times! I also tested Kubuntu 14.04 and 16.04 as well as opensuse. All showed the exact same kind of slow GUI / application performance. But I run Linux on all my other machines and it is fine on those. So this isn't something I can blame on Linux. I haven't run Windows in over ten years and Linux has generally been great for me. It's still great on my other computers. (It's on all my laptops, 3 NUCs, and a couple other desktop systems.)

  • 42 months ago
  • 1 point

Very very weird.

Have you looked into Google's stressapptest? You can have it beat up and benchmark RAM, SATA I/O, and a few other things all at once.

You said you have the NVIDIA proprietary driver installed. Are you sure? The nouveau driver can sometimes hang on a bit more tenaciously than expected, causing the NVIDIA drivers to fail to load.

  • 42 months ago
  • 1 point

Thanks for you suggestions. See my prior comment for the exact details of the nvidia driver install.

I just learned about Google's stressapptest yesterday. I will look into it. The last 24 hours I have been running tests with stress-ng.

  • 42 months ago
  • 1 point

what made you make this build

  • 42 months ago
  • 1 point

My last build is almost 5 years old. I wanted something faster. See my other comment for the details of my last build. It was a single CPU Xeon workstation.

  • 42 months ago
  • 1 point

Have you installed the latest proprietary Nvidia GPU driver? If you haven't this is probably the source of your problem. Ubuntu based GNU/Linux distributions default to the open source nouveau display driver for Nvidia graphics cards which is really poorly implemented and results in really crappy interface performance, let alone gaming performance. It even lacks hardware acceleration. Here's how to install the latest proprietary driver on Ubuntu/Kubuntu 16.04: (in terminal)

sudo add-apt-repository ppa:graphics-drivers/ppa

sudo apt update

sudo apt install nvidia-364

sudo reboot

  • 42 months ago
  • 1 point

This build is a wreck...

  • 42 months ago
  • 1 point

yep

  • 42 months ago
  • 1 point

I hope you get this monster figured out. I built a server last fall for my office that is very similar to your build. I'm running Windows Server 2012 on it. It is not the fastest horse in the stable either, but it chews relentlessly through anything we throw at it. Like others have said, I think I would try out Windows to eliminate any compatibility issues. I am going to follow this and see where it goes. Good Luck!!!

  • 42 months ago
  • 1 point

I appreciate the feedback!

  • 42 months ago
  • 1 point

I installed WIndows 10 and ran a couple benchmarks. Resizing applications windows has lag just like in Linux. Also, just like under Linux, the benchmark scores don't indicate any serious problems. I posted links to my Fire Strike score in other comments. The score seems appropriate for this hardware. It performs about the same as an i5 system... like you said, not the fastest horse in the stable.

I'm about to concllude that Xeon systems suck for desktop applications. Several people have told me that, but just as many disagree. My testing seems to support those who say this laggy performance is the best I'll get from a Xeon system.

  • 42 months ago
  • 1 point

Hey, there is one thing you could do for me -- could you share a screen cap showing your CPU info in Task Manager or some other program such as PassMark Performance Test?

My CPU shows up as "Genuine Intel (R) @ 2.40 GHz". It's the same in both Task Manager Performance tab and PassMark System tab. It does not say Xeon.

How does your CPU info show up?

  • 42 months ago
  • 1 point

I am out of town till Monday, but yes, I can absolutely do that when I get back. I wish I knew off the top of my head head how it shows up.

  • 42 months ago
  • 1 point

Thanks. I attached a screen shot of CPU-Z. My CPU is listed as:
Genuine Intel (R) CPU @ 2.40GHz (ES)

That's not what it should show. Furthermore, the (ES) probably means engineering sample.

Given my unusual problems, I think I should focus on the CPUs as the culprit...

  • 42 months ago
  • 1 point

WHY HAVE YOU DONE THIS TO ME, i spent 10 hours on learning basic Linux, installing Arch Linux onto one of my Vms. no, but seriously ns build i hope everything works out. +1 for the inspiration.

  • 42 months ago
  • 1 point

Let me know how it goes. The best thing about this build was learning about Arch Linux in more depth. (Still learning...)

  • 42 months ago
  • 1 point

I'm not very experienced with GPU stuff. But, could your vide card be bottlenecking the processors? Again, I have no idea, I'm just makin a guess.

  • 42 months ago
  • 1 point

If you haven't properly done troubleshooting I would do that. Take EVERYTHING out and boot the system with as little as you can. See if you can get the performance your looking for with one cpu, one stick of ram, one HDD installed (in your case the nvme ssd) and slowly add parts until you find out what's causing the performance issues.

I don't have any experience with server grade hardware so take my suggestion with a grain of salt. I hate to see someone spend this much money and not get the performance you expect for several months. My wallet cries for you, my friend.

  • 42 months ago
  • 1 point

Thanks! I have done some of that troubleshooting. For example, I disabled one CPU and just populated the memory on the other CPU. Then I even disabled all but one core of the one CPU. There was almost no difference in GUI and application performance from that test.

I also swapped video cards and tried 3 different ones. I also moved the cards to different PCI slots.

One thing I have not done is install the OS on a different drive. I've been using the NVMe drive. It benchmarks really well, so I don't think it has a problem, but I plan to try a test with just the Mushkin SSD installed and the OS on it (because that's one thing I haven't tried).

For the last six months I had this system running on a test bench so I could more easily troubleshoot this problem. When I could not resolve it after that time I finally put everything in the case so at least I could say I completed the build. But while I had it on the test bench I was swapping a lot of parts trying to figure out this issue.

The motherboard is another thing that has been constant. I have not tried a different motherboard (and that will not be an easy thing to try!)

  • 42 months ago
  • 1 point

Yeah, just swapping won't be that easy since you only have ECC memory and idk what consumer boards support those cpu's you have. Worst case scenario you could grab the cheapest board that supports the rest of the hardware that you have there and see if that helps.

  • 42 months ago
  • 1 point

I think Puget Systems has a list of workstation motherboards that work with Xeon CPUs and ECC RAM.

This is one MB I will consider:

Configure PC w/ Asus Z10PE-D8 WS (Intel C612 EATX) DDR4 REG ECC https://www.pugetsystems.com/parts/Motherboard/Asus-Z10PE-D8-WS-10563

BTW, here's one link I read when I decided to go with the Xeon system.

Why to Choose a Xeon - Puget Custom Computers https://www.pugetsystems.com/blog/2015/04/07/Why-to-Choose-a-Xeon-639/

  • 41 months ago
  • 1 point

Asus Z10PA-D8 WS is just as bad...

  • 42 months ago
  • 1 point

LOL ur slowest pc are you comparing this from the one you saw at ur nasa field trip?

  • 42 months ago
  • 1 point

Actually I wasn't joking. It's slower than my old laptop. I'm trying to figure out why.

  • 42 months ago
  • 1 point

wait so 2 xeon's and a desktop 970 and 256gb of ram didnt match the speed of ur old laptop...mind giving me some info and specs of the laptop.

  • 42 months ago
  • 2 points

I built it to replace the following system and I wanted something significantly faster. Right now it is significantly slower than this:

Xeon Qc E3-1230 Processor
4 x Kingston 8GB (2 x 4GB) 240-Pin DDR3 SDRAM DDR3 1333
Asus P8B WS LGA1155/ Intel C206/ Quad CrossFireX/ SATA3&USB3.0/ A&2GbE/ ATX Motherboard
PowerColor Go! Green ATI Radeon HD6750 1 GB DDR5 2DVI/HDMI/DisplayPort PCI-Express Video - - Card AX6750 1GBD5-NS3DHG
Crucial 512 GB m4 2.5-Inch Solid State Drive SATA 6Gb/s CT512M4SSD2
Samsung Spinpoint F4EG 2 TB SATA2 5400rpm 32 MB Hard Drive HD204UI/Z4 (internal)
Samsung Spinpoint F4EG 2 TB SATA2 5400rpm 32 MB Hard Drive HD204UI/Z4 (internal)
Samsung Spinpoint F4EG 2 TB SATA2 5400rpm 32 MB Hard Drive HD204UI/Z4 (internal) etc.

At the same time I was building this new system, I also built this i7 for a fraction of the cost. The i7, running the same operating system and the same application software is significantly faster for general desktop use.

Intel Core i7-6700K 4.0GHz Quad-Core Processor
Asus Z170-A ATX LGA1151 Motherboard
G.Skill Ripjaws V Series 32GB (2 x 16GB) DDR4-2666 Memory
Samsung 850 Pro Series 512GB 2.5" Solid State Drive

As mentioned, my laptops are faster. I have several laptops, all running the same Linux operating system I installed on this build. One is a first generation Lenovo X1 Carbon with a mobile i5

In fact, my NUC does about the same, or maybe better, for web browsing than this new system. My NUC is a Gigabyte BRIX GB-BXBT-2807 Celeron.

  • 42 months ago
  • 1 point

Have you tried switching DE's? Try something non-qt like GNOME, XFCE or LXDE and see if anything changes. Also, have you checked any logs for errors?

  • 42 months ago
  • 1 point

Yes, I have checked the logs and resolved any errors found there. I have only tested KDE. However, right now I am testing Windows 10. I'm looking for some Windows benchmarks to run... (but first I have to get my trackpoint working in Windows)

  • 42 months ago
  • 1 point

My current (18-June-2016 3pm) benchmark scores:

Under Linux - Look for Supermicro X10DAL Radeon 390X.

Jmeter1 - 16 X Intel Xeon E5-2690 V3 @ 2.59GHz Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601276-HA-1601260HA52

Benchmarks, Linux Performance - OpenBenchmarking.org http://openbenchmarking.org/s/fiteyes

Testing Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601278-HA-1601275HA22

Testing Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601277-HA-1601275HA66

Jmeter1 - 16 X Intel Xeon E5-2690 V3 @ 2.59GHz Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601276-HA-1601260HA52

Jmeter1 - 16 X Intel Xeon E5-2690 V3 @ 2.59GHz Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601276-HA-1601276HA91

"hdparm -tT --direct" gives read speeds of up to 1.75 GB/s which is about what is expected for this NVMe drive.

Under Windows 10:

Cinebench R15:

  • OpenGL 126.90 fps (rank #1); Ref Match 99.6%
  • CPU 1306 cb (rank #1)
  • CPU (Single Core) 114 cb; MP Ratio: 11.47x

Fire Strike v1.1 score 10712
http://www.3dmark.com/3dm/12560080

PassMark Rating (Performance test 8.0): 3078.2 -- 3.5 stars overall (all tests)
http://www.passmark.com/baselines/V8/display.php?id=63558747340

Individual PassMark Benchmark Scores

CPU Mark - 5 stars
15,730

2D Graphics Mark - 3.5 stars
660

3D Graphics Mark - 4.5 stars
8,443

Memory Mark - 3.5 stars
2,113

Disk Mark - not relevant because I'm not running Windows in my NVMe drive. 893

I'm running Windows from a spinning disk, so disk benchmarks under Windows are not of interest.


None of the benchmarks show a clear problem.

  • 42 months ago
  • 1 point

In thePassMark Performance test benchmark reported above, the other comparable systems with this CPU report the CPU as:
Ïntel Xeon E5-2660 @ 2.60GHz.

That app reports my CPU as:
Intel @ 2.40GHz.
Also the codename (Haswell E), socket and lithography are missing from this report.

See the screen shot.

https://cdn.pcpartpicker.com/static/forever/images/userbuild/172654.a6112467b8e5645fc64e93a44d608f82.a6ce28277d9419521ad187476167d805.1600.jpg

On the other hand, CPU-Z reports Intel Xeon E5 v3, so maybe they are OK.

screen shot: https://cdn.pcpartpicker.com/static/forever/images/userbuild/172654.0b8cb65e8bd3edfc25e3abe83e0ef8da.f0c87b06c38b352fb593c9d668725805.1600.jpg

  • 42 months ago
  • 1 point

I've never had KDE run well, but it seems that it's just me. It's likely a software/driver issue.

  • 42 months ago
  • 1 point

KDE runs great for me. I've been using it long enough to be familiar with tweaking performance. No issues with KDE on my other computers. I built an i7 at the same time as this system and install the same OS on both. The i7 runs fine. So I think it's a hardware issue. Becoming more convinced of that over time with continued testing.

  • 42 months ago
  • 1 point

:/ Idk then. Not experienced enough in hardware to advise, and I haven't run KDE long enough to know it very well.

  • 42 months ago
  • 1 point

try doing a cinebench test

  • 42 months ago
  • 1 point

already did. see other comments

  • 42 months ago
  • 1 point

also are you A millionaire by any chance?

  • 42 months ago
  • 1 point

It's a work PC, paid for by my company

  • 42 months ago
  • 1 point

oh right haha i thought you had 7k laying around

  • 42 months ago
  • 1 point

No expert here. But isn't it recommended to run your OS as a virtual machine with this kind of hardware.

  • 42 months ago
  • 1 point

No, not necessarily. It the host operating system isn't performing well, the guess will likely perform even worse.

  • 42 months ago
  • 1 point

I have read in the past that too much RAM could be a bad thing and cause systems to slow down in certain circumstances. Have you tried removing some of the sticks? Maybe try just two sticks and see if that helps. Just a shot in the dark, not sure if it will help. You can also try removing any non-necessary components (extra HDDs, etc) and see if any of that helps.

  • 42 months ago
  • 1 point

That's interesting. I initially did this build with half this much RAM. I also tested it with just two sticks while looking for potential memory errors.

As part of my troubleshooting, I decided to populate all the memory slots. I learned that having unbalanced memory configurations can lead to performance problems (see links I posted in another comment for interesting details). Anyway, balancing the memory did not solve my problems.

I am still trying to figure it out.

  • 42 months ago
  • 1 point

I am in the middle of a very similar build (same MOBO and almost the same case), did you have trouble mounting the MOBO in ? In particular, this MOBO doesn't seem to have two of the center holes required for ATX, and the case has a solid center pin that is just pressing into the MOBO. If you noticed this, please let me know what you did. If not, maybe I'm doing something very wrong.

  • 42 months ago
  • 1 point

Yes, I faced these issues. I removed the center pin.

Then I aligned the motherboard in the case where it should go. I used a long thing pencil lead through the MB holes that were missing standoffs to mark the case. Then I center-punched the case at those marks. (I have a spring-loaded center-punch that doesn't require a hammer.)

I drilled the proper size hole and then I threaded the case using a tap.

The tap I used is a 6-32 NC (National Course Thread). This matches my MB standoffs and I assume it will match yours. It's the same size as the other holes in the case.

The drill bit I used measures 0.95" dia. (I don't know the actual fractional size because I picked the drill bit out of a pile I had.)

The drill bit must be smaller than the thread size (but not too small). To pick a drill bit size, simply stick the smooth end of the drill bit into a threaded MB standoff and make sure it has some clearance. A smaller drill bit is better than one that is too big. But if you go too small, you can break the tap.

When tapping holes, turn the tap 1/4 turn in, then back out. Then progress a further 1/4 in, back out, repeat. I have tap fluid, but it isn't really necessary. You could use some oil on the tap, but you can also get by without any.

After tapping the holes, clean off any burrs, then screw in the standard MB standoffs and mount motherboard as normal. (I have a deburring tool, but you can just use a knife if needed.)

My case looks like it was factory made for this MB now.

Let me know how your build goes. I still can't figure out my GUI lag issue, but it isn't the MB or the Xeon CPUs or the GPU or Linux.

Tools required:

  • good lighting
  • center punch
  • 6-32 NC tap (and tap handle or wrench to fit)
  • 0.95" dia drill bit or similar
  • drill
  • tap fluid or drilling oil or similar
  • deburring tool or knife (optional)
  • long pencil lead (or anything to reach through MB holes and mark case)
  • extra motherboard standoffs to use for the new holes
  • 41 months ago
  • 1 point

Thanks for your reply ! Sorry I didn't check the site for a while and didn't see it. I just removed the center pin, I didn't go and properly add other standoffs for the unused holes. Besides that, my build went well. My use case is quite different: I want this for compiling/running large science codes. Still, I did find that regular Ubuntu would use too much CPU for its "Compiz" GUI environment, so I switch to Lubuntu, which feels much better.

  • 41 months ago
  • 1 point

congrats on your build. As you probably saw, I converted my build to a headless server. It had been intended as a workstation, but the GUI performance wasn't up to my expectations. It does fine as a headless server. I built an overclocked i7 6700K Z170 workstation, which I like much better.

I may post the above comment over on your build page in case anyone else attempts a similar build. I was able to unscrew the centerpin. I'm not sure why you needed a dremel to remove it. Also, as far as I know, ALL SuperMicro server boards use their own proprietary non-ATX mounting hole positions. The boards are designed for SuperMicro cases.

  • 41 months ago
  • 1 point

Update: 2016.07.03

Need help troubleshooting poor GUI application performance (Linux / KDE) on specific hardware.

The problem shows up only on certain classes of hardware, which I'll describe. (On other hardware, the exact same operating system has no problems.) The problems I'm encountering are:

Application windows are slow to resize. Window resizing is choppy. Content redrawing is slow / choppy. See image of me resizing a window here:

gui-performance-slow-window-resize6.jpg
http://kuuv.io/i/lN2UzDw.jpg

Notice how far the mouse cursor is from the edge of the window (indicating GUI lag).

Applications are also slow to respond to the mouse. For example, there is a noticeable delay when clicking menu items. The GUI is generally slow.

GUI applications are slow to load.

I notice the issues most in Libre Office Calc and Firefox but that's probably because those are the apps I use the most.

I notice the slowness in all interactive desktop GUI applications. It is both graphics (resizing windows) and, to a lesser extent, I/O (opening files).

It is slow to the point of being annoying and decreasing productivity.

The issue does not show up in any benchmarks I have run so far. See some benchmark scores below.

Software

My testing has been done with KDE 4.X and KDE Plasma 5.X. Specifically, I have tested with Kubuntu 14.04, Kubuntu 16.04, opensuse 13.2, and Arch Linux (fully updated throughout testing from December 2015 to July 2016).

I have tested a number fresh installs. (I installed Kubuntu about a dozen times and Arch Linux ten times.) I also have a clean Arch Linux KDE installation on an SSD that I transfer between hardware so that I can rule out differences in OS, DE or software application settings.

Hardware

All testing has been done with one or two monitors attached. I have tested different monitors configurations. The most common monitor configuration I am testing with is a single 1080p display. I also test with dual monitors where I use one 2560 x 1440 display and one 1920 x 1280 display.

Hardware WITHOUT any problems:

These are just some examples of hardware that runs KDE fine for me. Generally, most systems I have used run KDE, Firefox and Libre Office Calc fine.

NUC: System #5 Gigabyte BRIX GB-BXBT-2807 Celeron This system is a bit slow when lots of applications are open or when multiple Firefox windows, each with many dozens of tabs, are simulatenously open. However, the GUI is free of the problems described above even when pushing this small system past its comfort zone.

Laptop: System #6 First generation Lenovo X1 Carbon with a mobile i5 http://www.lenovo.com/psref/pdf/withdraw/ThinkPad_X1_Carbon_1st_Gen_WE.pdf

Laptop: System #7 ThinkPad W550s | Ultrabook & Mobile Workstation | | Lenovo US http://shop.lenovo.com/us/en/laptops/thinkpad/w-series/w550s/#tab-tech_specs

Skylake Desktop: System #8 Intel Core i7-6700K 4.0GHz Quad-Core Processor Asus Z170-A ATX LGA1151 Motherboard G.Skill Ripjaws V Series 32GB (2 x 16GB) DDR4-2666 Memory Mushkin 1TB 2.5" Solid State Drive This system has fantastic performance and is completely free of the problems described above.

Hardware WITH problems:

Dual Xeon System #1
Dual Intel Xeon E5-2660 V3 2.6GHz 10-Core Processor
2x CPU Cooler: Noctua NH-D9DX i4 3U 46.4 CFM CPU Cooler
SUPERMICRO MBD-X10DAL-I-O ATX Server Motherboard Dual LGA 2011 Intel C612
256GB RAM 8x Samsung DDR4-2133 32GB/4Gx72 ECC/REG CL15 Server Memory
EVGA GeForce GTX 970 4GB SSC Gaming ACX 2.0+ Cooling Graphics Card (04G-P4-3975-KR)
1x SAMSUNG 950 PRO M.2 512GB PCI-Express 3.0 x4 Internal Solid State Drive (SSD) MZ-V5P512BW
Storage: Mushkin Reactor 1TB 2.5" Solid State Drive
WD Red 6TB NAS Desktop Hard Disk Drive - Intellipower SATA 6 Gb/s 64MB Cache 3.5 Inch - WD60EFRX
2x Seagate 8TB Archive HDD SATA 6GBps 128MB Cache SATA Internal Bare Drive (ST8000AS0002)
Case: Fractal Design Define R5 Blackout Edition w/ Window ATX Mid Tower Case
Power Supply: FSP Group 1200W 80+ Platinum Certified Fully-Modular ATX Power Supply
Optical Drive: Asus DRW-24B1ST/BLK/B/AS DVD/CD Writer
Other: Addonics ADM2PX4 M2 Pcie Ssd Pcie 3.0 4-lane Accs Adapter

Same system as #1 but with different GPU:
Radeon 390X

Dual Xeon System #2
Dual Intel Xeon E5-2630 V4 CPU
2x CPU Cooler: Noctua NH-D9DX i4 3U 46.4 CFM CPU Cooler
Asus Z10PE-D8 WS: https://www.asus.com/us/Motherboards/Z10PED8_WS/specifications/
128GB RAM 8x Samsung DDR4-2133 32GB/4Gx72 ECC/REG CL15 Server Memory
EVGA GeForce GTX 970 4GB SSC Gaming ACX 2.0+ Cooling Graphics Card (04G-P4-3975-KR)
Storage: Mushkin Reactor 1TB 2.5" Solid State Drive
Power Supply: EVGA Supernova 1000 T2 1000W

Same system as #2 but with different GPU:
EVGA nvidia GeForce GT 720 2GB DDR3 PCIe 2.0

Broadwell-E X99 i7 6850K System #3
Intel Core i7-6850K 3.6GHz 6-Core Processor
Noctua NH-D15 82.5 CFM CPU Cooler
Asus X99-DELUXE II ATX LGA2011-3 Motherboard
Patriot Memory VIPER 4 Series 3000MHz (PC4 24000) 16GB Dual Channel DDR4 Kit PV416G300C6K
EVGA GeForce GTX 970 4GB SSC Gaming ACX 2.0+ Cooling Graphics Card (04G-P4-3975-KR)
Samsung 950 PRO 512GB M.2-2280 Solid State Drive
EVGA 1000W 80+ Platinum Certified Fully-Modular ATX Power Supply

Same system as #3 but with different storage:
Mushkin Reactor 1TB 2.5" Solid State Drive

I first noticed the problem when I built system #1 above to replace the following aging system:

System #4 Xeon Qc E3-1230 Processor
4 x Kingston 8GB (2 x 4GB) 240-Pin DDR3 SDRAM DDR3 1333
Asus P8B WS LGA1155/ Intel C206/ Quad CrossFireX/ SATA3&USB3.0/ A&2GbE/ ATX Motherboard
PowerColor Go! Green ATI Radeon HD6750 1 GB DDR5 2DVI/HDMI/DisplayPort PCI-Express Video - - Card AX6750 1GBD5-NS3DHG
Crucial 512 GB m4 2.5-Inch Solid State Drive SATA 6Gb/s CT512M4SSD2
Samsung Spinpoint F4EG 2 TB SATA2 5400rpm 32 MB Hard Drive HD204UI/Z4 (internal)
Samsung Spinpoint F4EG 2 TB SATA2 5400rpm 32 MB Hard Drive HD204UI/Z4 (internal)
Samsung Spinpoint F4EG 2 TB SATA2 5400rpm 32 MB Hard Drive HD204UI/Z4 (internal) etc.

System #4 has always had fairly slow GUI performance, but I thought it was due to the Radeon HD6750. I built System #1 to resolve this specific issue. However, I was shocked to find that the GUI performance was significantly worse in System #1 comparaed to the againg System #4. This observation made me realize that System #4 has slower GUI performance than several lower-powered systems.

In summary all of these workstation-class systems have very poor GUI performance under KDE. In contrast, the Skylake i7 system has fantastic performance. And even the NUC has acceptable performance.

I have been troubleshooting this off and on for seven months. I swapped hardware extensively. In recent tests, I have taken the same SSD boot and OS drive and tested it in the following different hardware systems:

  • Skylake Desktop: System #8
  • Haswell-EP Dual Xeon System #1
  • Broadwell-EP Dual Xeon System #2
  • Broadwell-E X99 i7 6850K System #3

By keeping the software and software configuration constant and changing the hardware (including more extensive changes than listed here), it is clear that KDE Linux performs very well on Skylake systems (as well as many other lower-end systems). However, it performs poorly on the single and dual CPU Xeon systems tested as well as on the Broadwell-E consumer enthusiast platform.

What I want to know is why.

Benchmark notes:

The following benchmarks are for System #1 with the Radeon 390X and the open source driver. I tested both proprietary and open source AMD drivers. Look for the system named "Supermicro X10DAL Radeon 390X".

Jmeter1 - 16 X Intel Xeon E5-2690 V3 @ 2.59GHz Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601276-HA-1601260HA52

Benchmarks, Linux Performance - OpenBenchmarking.org http://openbenchmarking.org/s/fiteyes

Testing Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601278-HA-1601275HA22

Testing Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601277-HA-1601275HA66

Jmeter1 - 16 X Intel Xeon E5-2690 V3 @ 2.59GHz Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601276-HA-1601260HA52

Jmeter1 - 16 X Intel Xeon E5-2690 V3 @ 2.59GHz Benchmarks - OpenBenchmarking.org http://openbenchmarking.org/result/1601276-HA-1601276HA91

"hdparm -tT --direct" gives read speeds of up to 1.75 GB/s which is about what is expected for this NVMe drive.

System #1 performed roughly the same with three different nvidia cards (using proprietary drivers): MSI GTX 960, Asus GTX 970 and EVGA GTX 970.

Later I tested systems #2 and #3. Given that the benchmarks are not reflecting the problems I'm experiencing in real life usage, my testing now consists of repeating certain tasks manually such as resizing a Firefox browser window. For both System #2 and #3 the problem was the same as in System #1 with Linux.

For comparison (since my testing is now subjective / no formal benchmarks), I tested Skylake Desktop System #8 again. Repeated tests confirm it is free of the performance issues being discussed. The difference is striking; no benchmarks are needed to see the difference.

I installed WIndows 10 on System #1 (with nvidia GTX 970) and ran a couple benchmarks. Resizing applications windows has some lag similar to, but not as bad as, Linux. Also, like under Linux, the Windows 10 benchmark scores don't indicate any serious problems with System #1 hardware.

Benchmarks Under Windows 10:

Cinebench R15:

OpenGL 126.90 fps (rank #1); Ref Match 99.6%
CPU 1306 cb (rank #1)
CPU (Single Core) 114 cb; MP Ratio: 11.47x

Fire Strike v1.1 score 10712 http://www.3dmark.com/3dm/12560080

PassMark Rating (Performance test 8.0): 3078.2 -- 3.5 stars overall (all tests) http://www.passmark.com/baselines/V8/display.php?id=63558747340

Individual PassMark Benchmark Scores

CPU Mark - 5 stars 15,730

2D Graphics Mark - 3.5 stars 660

3D Graphics Mark - 4.5 stars 8,443

Memory Mark - 3.5 stars 2,113

Disk Mark - not relevant because I'm not running Windows in my NVMe drive. 893

I'm running Windows from a spinning disk, so disk benchmarks under Windows are not of interest.

  • 41 months ago
  • 1 point

Your firestrike score says the driver was not approved, maybe look into that. Could also be that your hard drives are slowing up the system as they're practically full. i dunno +1 anyway

  • 41 months ago
  • 1 point

Already updated the driver a while back and it didn't make any difference. (The message wasn't right anyway. The driver was fine.)

No, my drives are not full. I was testing with different storage at different times. There are notes about this in my posts regarding the benchmarks.

My main OS drive is a 950 Pro 512 GB and it is mostly empty.

  • 41 months ago
  • 1 point

Ah ok, sorry about the build. Good luck with trying to fix it, I'm sure it'll pay out in the end.

  • 41 months ago
  • 1 point

Thanks. I just told @haydood we now repurposed this computer it as a pure headless server. And I built myself an i7 workstation. I"m happier with the i7. I'll put pictures of it up soon.

  • 41 months ago
  • 1 point

Nice, have fun with it.

  • 41 months ago
  • 1 point

THAT'S SO MUCH RAM, WHY?

  • 41 months ago
  • 1 point

I started with half that ram, but when the performance wasn't as good as I expected, one of the things I learned in my research is that ram needs to be balanced. I increased the ram to see if having all channels populated made a difference. It didn't.

  • 41 months ago
  • 1 point

Ah ok. Well I guess it doesn't really matter because it's technically your work's computer. Also, good luck with getting your computer working to it's best ability.

  • 41 months ago
  • 1 point

Thanks!

BTW, now we repurposed this computer it as a pure headless server. And I built myself an i7 workstation. This is a better solution :-)

  • 41 months ago
  • 1 point

Greetings!

I know this thread is a touch old, and you decided a different use for the system, but I wanted to dig a little farther into one detail for I am planning on building a similar system. Your CPUs. Where did you get them? Were they new and retail or did you purchase them at a savings off of ebay?

The reason I am asking is because your screenshot of CPUZ-ID shows your CPU as an 'ES' which generally refers to an engineering sample. You can generally get pretty lucky with these, but often they are not retail ready versions and have been pounded by testers/reviewers to a point of problematic behavior due to improper/abusive stressing. Basically, the silicone lottery is tricky enough without adding this turkey shoot into the mix. This could easily be the root of your problem if you purchased 'ES' processors third party...also, it is actually illegal to sell them, so the sources for these processors aren't always on the up-and-up to begin with.

I realize this is the most expensive fix for the problem (if the ES cpus are the problem), but I feel this is an important place to look, and a point that was glazed over in this discussion too quickly. I know there are many that will say ES cpus are fine, but there are an equal share to say the opposite, and being the single most important component in a computer, it is not something that should be overlooked as a possible source of the issue...in my opinion, cpus should be the last thing (along with psus) that one should take a chance on.

Is it possible that an ES cpu may have an effect on pcie lane capability, memory or driver interface? You betcha! Worth looking into? For the cost of your system, yes! An expensive fix? Yes and no. You can get two E5-2630 v4 10c/20t processors for around $1300.00...expensive, but still less than only one of your E5-2660 v3s at full retail.

If your cpus were new/retail then my question is totally moot, but I felt this was an important enough point to bring into the discussion:-)

  • 41 months ago
  • 1 point

Hi. Thanks for all the information. I learned a lot from this build, and I'm still learning from it. Your reply taught me a few more things.

I bought my CPUs used on ebay. I have a thread here asking about them as the potential source of my problems.

Did I get scammed on my CPUs? - Forums - PCPartPicker https://pcpartpicker.com/forums/topic/165245-did-i-get-scammed-on-my-cpus

However, I ruled out the CPUs being the main cause of my problems by the following test. I replaced the motherboard, CPUs and RAM of this build with an Asus Z10PAD8, two brand new Xeon E5 2630 V4 CPUs and 8x Samsung DDR4 2133MHzCL15 8GB RegECC 1RX4 (PC4 2133) M393A1G40DB0-CPB.

WIth these new parts this was essentially a new build. However, I carried over the Samsung 950 Pro NVMe drive (with the existing Arch Linux KDE install). This "new" system had the same poor GUI performance! It was absolutely no better. I used it for one day, and it was so disappointing that I returned the new parts and kept the used ES CPUs. (As you know, I also converted this build to a GPU-less / headless server, for which it is tremendous overkill and a huge waste of money.)

From all this testing I was able to decide that the ES CPUs were not the root of my problems. However, having learned what I learned, and to your point, I would not buy used ES CPUs from ebay again for another build. At the time I bought them I did not know what ES meant. (I actually thought it said E5 at first glance when I bought them.)

Of interest, putting the same 950 Pro into a a new i7 6700K build resulted in good performance. This i7 is now my main workstation and I am very happy with it.

In a relatively extreme attempt to get to the root of this issue, I tested a brand new overclocked i7 6850K X99 system with this same 950 Pro (with the same exact OS installation used in prior tests). It also had relatively poor GUI performance compared to the 6700K. (It was not as bad as the dual Xeon systems, but still less performant than the Skylake system.)

I wrote up a detailed report of all these comparisons and posted it on a forum, but the forum moderators rejected it as inappropriate. However, I do feel there is something to learn from all this about KDE GUI performance. My tentative conclusion is that KDE runs better on the Skylake platform. Higher CPU frequency obviously helps, but the overclocked 6850K didn't perform as well on my tests even with an overclocked freq matching the 6700K.

After spending six months on this and building SIX different new systems (two dual xeon, two i7 6700K, one i7 6850K and one i5 6600) and installing linux on all of them plus reinstalling linux on several laptops, I can say that I took my knowledge to a new level. But unfortunately, I don't have a rock solid definitive conclusion about why the KDE GUI performance is unacceptably poor on certain systems but totally acceptable on other systems (especially Skylake).

If I had time (which I don't) I would make an even more detailed writeup with all my different tests and comparisons. For example, I tested the same OS (Kubuntu 14.04) with the same settings on different builds and saw dramatically different performance -- slower performance on the more powerful system. I saw this time and again and if I were a developer (or working in this field in any way), I would certainly want to study this in greater detail.

  • 40 months ago
  • 1 point

FWIW, I just finished a build using the same motherboard and dual xeon 2658 v3. running 16gb ram, R9 295x2. It is not laggy or doing any of the issues you are describing. Im still working out the kinks and optimizing it all. Ill be runing windows 10 and got a cinebench CPU score over 2500.

  • 40 months ago
  • 1 point

That's good to hear. Thanks for letting me know.

  • 39 months ago
  • 1 point

Is this a server build?

  • 39 months ago
  • 1 point

Tu tarjeta de video es el problema... te da cuello de botella, nesecitas una tarjeta de video Nvidia Cuadro M4000 de 8gb.

  • 37 months ago
  • 1 point

Why would you need 256GB of ram?!

  • 36 months ago
  • 1 point

No hate, but why would you buy 7k worth of computer parts, including freaking TWO 10 core processors and 970, only for web browsing...

Now it's not that bad of a PC, and I still 1+ this, but I think this was made for just the wrong purpose (serious with no gaming?)

  • 36 months ago
  • 1 point

Many people seem to misunderstand the demands of heavy web research. And that is not the only thing the build was intended for. The computer hardware choices had a purpose and it was explained somewhere in the posts and various comments already (more than once, actually).

The main expert who advised on the build agreed with the component selection. He recommended most of the components based on a deeper understanding of the use-case than those who are commenting that it is a waste of money.

  • 36 months ago
  • 1 point

Ok, I understand this may be a productivity setup, but what do you mean by "heavy web research." Like do you find information and need this to store more cache or...

I'm trying not to be negative here and I really think it is a good build.

  • 33 months ago
  • 1 point

looks like one of your CPUs died or wasn't installed correctly. The data you provided shows that you have one cpu instead of two

  • 33 months ago
  • 1 point

This is an old post, not sure if you resolved speed issue. I saw Linus over at LinusTechTips on YouTube talking about curious speed issues related to high core count CPUs, not sure which video but believe the title includes a reference to "Xeon" or "Cores".

  • 32 months ago
  • 1 point

You: blows 8 grand on a PC , PC is slower than 5 year old PC. me: claps slowly and walks away

  • 29 months ago
  • 1 point

the 970 bottlenecks.

Crap build

  • 24 months ago
  • 1 point

u must be rily rich to do this :D

  • 23 months ago
  • 1 point

Y so much ram for a 4 port mobo? just wastin yo money

  • 23 months ago
  • 1 point

I already answered this question. The comments on this thread are just going downhill. Please don't comment unless you have something new and constructive to say.

  • 22 months ago
  • 1 point

this makes no sense

  • 22 months ago
  • 1 point

that's probably because you didn't try to understand it (and you probably didn't read the most important comments either)

  • 38 months ago
  • 0 points

This is so bad What a waste Also a bit over over the top You don't even need more than 32 gb what did u even need it for comletely ridiculous

  • 11 months ago
  • 0 points

this is a stupid build. and also why do you have so many tabs open?

  • 10 months ago
  • 2 points

this is a stupid comment. Why did you waste my time?

  • 42 months ago
  • -1 points

The problem with this build is that the ram modules you chose are highly inefficient, the best ram have extremely low CAS latency, you chose some with 15. Also you chose a Mushkin hard drive which aren't known for speed anyway, also there are way too many parts on this motherboard. You're better off going with i7 processor, Seagate HDD, Kingston RAM, along with some other MSI parts, if you want a fast computer.

  • 42 months ago
  • 2 points

Ddr3 has Cas latencies around 10. DDR4 is around 15. That doesn't make it inefficient. It's different architecture. He also has a Samsung m.2 ssd running the OS. Even if it was on the mushkin its still an SSD, so it shouldn't be as slow as he's reporting unless it's a bad drive. Nor gonna argue with the rest though. It seems like he didn't NEED this server hardware and just wanted to experiment and there's nothing wrong with that if he's got the money.

  • 42 months ago
  • 1 point

Thanks for the comments. My prior system was a single CPU Xeon E3 1230 with SSD drives. It was slow enough to hinder my productivity. So I wanted this system to be an upgrade. Initially I was leaning toward an i7, but in the thread on PCPP where I asked for build advice I got talked into a dual Xeon system. In hindsight, that appears to have been a bad decision...

  • 42 months ago
  • 1 point

Just out of curiosity can you link me to the post?

  • 42 months ago
  • 1 point

which post? I linked this one above. Is there another post I need to link?

  • 42 months ago
  • 1 point

That's the one. There's a lot of things you mentioned in that original post that you left out of the pc build page. Any further luck in trouble shooting?

  • 42 months ago
  • 1 point

I built a system with:

  • Intel Core i7-6700K 4.0GHz Quad-Core Processor
  • Asus Z170-A ATX LGA1151 Motherboard
  • G.Skill Ripjaws V Series 32GB (2 x 16GB) DDR4-2666 Memory
  • Samsung 850 Pro Series 512GB 2.5" Solid State Drive

It runs Kubuntu 14.04. It is far faster than this dual Xeon machine. However, even my Gigabyte BRIX GB-BXBT-2807 with an Intel Celeron N2807 / 1.58 GHz is faster than this system.

Therefore, this problem isn't just that my Mushkin drive isn't the fastest or my RAM isn't the fastest. There is something else wrong. I don't know what.

But the system should still be fast with these parts.

BTW, the RAM is ECC. That's why I picked that RAM.

[comment deleted]
[comment deleted by staff]
[comment deleted by staff]
[comment deleted by staff]
[comment deleted by staff]
[comment deleted by staff]
[comment deleted by staff]
[comment deleted]
  • 42 months ago
  • 1 point

That's interesting... I am tempted to try another motherboard (because I have tried almost everything else I can think to try).

[comment deleted]
  • 42 months ago
  • 1 point

Yeah, I use my system for a lot of stuff, some of it is pretty demanding. My prior system was a single Xeon workstation and I needed something more performant, so I went with this.

[comment deleted]
  • 42 months ago
  • 1 point

My company paid for it. It's for work.

  • 42 months ago
  • 1 point

I would really like to stick with an ATX form-factor board...

[comment deleted]
[comment deleted]
  • 42 months ago
  • 1 point

What do you think about the Asus Z10PE-D8 WS Motherboard?

[comment deleted]
  • 42 months ago
  • 1 point

I just noticed that that's also an EEB size board... :-(

[comment deleted]
  • 42 months ago
  • 1 point

The ASUS Z9PA-D8 ATX Server Motherboard is ATX, but it has inferior specs to my current Super Micro board.

  • 42 months ago
  • 1 point

I guess I could try a single CPU board... at this point I don't have much to lose. It can't be any worse. Maybe I could turn this build into two different systems for relatively little more money.

[comment deleted]
  • 42 months ago
  • 3 points

So want to explain why any self respectful datacenter offers Xeons? Eg: SingleHOP RackSpace

Ask them for an i5 or an i7 and you'll be laughed at by every tech in the datacenter.

  • 42 months ago
  • 2 points

BECAUSE... you used a Xeon! They are B-A-D.

This may be right...

Also, why Linux?

I love KDE and Arch Linux. I have no interest in running Windows. Arch Linux and KDE run fine on my other computers. I have no problems with Linux. My only problem is this build.

[comment deleted by staff]
[comment deleted by staff]
[comment deleted by staff]