Print Page | Close Window

Threadripper 1950x low benchmark score

Printed From: ASRock.com
Category: Technical Support
Forum Name: AMD Motherboards
Forum Description: Question about ASRock AMD motherboards
URL: https://forum.asrock.com/forum_posts.asp?TID=7155
Printed Date: 03 Dec 2024 at 1:31pm
Software Version: Web Wiz Forums 12.04 - http://www.webwizforums.com


Topic: Threadripper 1950x low benchmark score
Posted By: ventralstreams
Subject: Threadripper 1950x low benchmark score
Date Posted: 03 Jan 2018 at 10:23pm
" rel="nofollow - Hello everybody. I have setup a workstation based on a Taichi x399 and a TR 1950x and have been working on it without any obvious problems. Until I decided to benchmark the system that is. The score I get (on performance test 9.0) is 15742 which is too low in comparison to the average (about 21000). Any suggestions on what could be wrong? Thank you in advance.

-------------
Asrock Taichi x399, Threadripper 1950X, 32Gb G.Skill flare X 3200, MSI GTX1080, CoolerMaster MasterLiquid 240,SSD Samsung 850 EVO 250 GB, SSD Crucial CT256MX100, SSD Sandisk 480GB, HDD SG Baracuda 1TB



Replies:
Posted By: zlobster
Date Posted: 03 Jan 2018 at 11:17pm
" rel="nofollow - Never been a fan of PassMark. IDK if this benchmark will let you see which sub-system is the contributor for the low score. Theoretically, it could be everything, incl. bad optimization of PassMark for TR.

Try running separate benchmarks for the separate sub-systems, i.e. CrystalDisk for your SSD(?), 3DMark for the GPU, Blender & Cinebench for the CPU, etc. Then compare online with similar systems and see which one(s) deviate.

Do let us know your results.





-------------
1700X ZP-B1 (stock); X370 Taichi (UEFI 3.10); 16GB F4-3200C14-8GFX XMP; 256GB 960 EVO; RX 580 NITRO+ 8GB


Posted By: ventralstreams
Date Posted: 03 Jan 2018 at 11:32pm
I tested the CPU with Cinebench and the result seems to be normal in comparison (3019cb  @ 3.4) so I assume that it is an issue with PerformanceTest's optimization. (phew. I almost re-installed Windows).

-------------
Asrock Taichi x399, Threadripper 1950X, 32Gb G.Skill flare X 3200, MSI GTX1080, CoolerMaster MasterLiquid 240,SSD Samsung 850 EVO 250 GB, SSD Crucial CT256MX100, SSD Sandisk 480GB, HDD SG Baracuda 1TB


Posted By: PetrolHead
Date Posted: 03 Jan 2018 at 11:45pm
PerformanceTest 9.0 gives a pretty good breakdown of the total results into different sub-categories (CPU, memory etc.) and their sub-categories (floating point operations, physics etc.). I suggest you download a baseline close to your own system specs (PerformanceTest 9.0 has a search function that you can use for this). That way you can compare the results of the different CPU tests to that baseline and see which test(s) are causing your system to lag behind in the total CPU result.


-------------
Ryzen 5 1500X, ASRock AB350M Pro4, 2x8 GB G.Skill Trident Z 3466CL16, Sapphire Pulse RX Vega56 8G HBM2, Corsair RM550x, Samsung 960 EVO SSD (NVMe) 250GB, Samsung 850 EVO SSD 500 GB, Windows 10 64-bit


Posted By: ventralstreams
Date Posted: 04 Jan 2018 at 12:03am

PT9 gave very low floating point math result and low results allover. 








-------------
Asrock Taichi x399, Threadripper 1950X, 32Gb G.Skill flare X 3200, MSI GTX1080, CoolerMaster MasterLiquid 240,SSD Samsung 850 EVO 250 GB, SSD Crucial CT256MX100, SSD Sandisk 480GB, HDD SG Baracuda 1TB


Posted By: zlobster
Date Posted: 04 Jan 2018 at 12:18am
From their web page:

PassMark Software is a Microsoft Registered Partner and an Intel Software Partner.

I'll stop here.


-------------
1700X ZP-B1 (stock); X370 Taichi (UEFI 3.10); 16GB F4-3200C14-8GFX XMP; 256GB 960 EVO; RX 580 NITRO+ 8GB


Posted By: zlobster
Date Posted: 04 Jan 2018 at 12:22am
If ANY test involves heavy CCX thread switching TR is dead in the water. CCX context switching is penalized heavily.

Then again, you never told us even what version of Windows you're using. Yes, we can see your screenshot and assume it's Win 10, however many folks claim TR needs 1703 and above for full potential to be utilized.

Worth checking if you have NUMA enabled or not, and flash the latest UEFI.


-------------
1700X ZP-B1 (stock); X370 Taichi (UEFI 3.10); 16GB F4-3200C14-8GFX XMP; 256GB 960 EVO; RX 580 NITRO+ 8GB


Posted By: ventralstreams
Date Posted: 04 Jan 2018 at 12:28am
Oh yeah. Sorry. I use Windows 10 pro (latest update and Latest UEFI). 

EDIT: I can't find a NUMA setting in the BIOS menu but almost all of the memory related options are set to AUTO.


-------------
Asrock Taichi x399, Threadripper 1950X, 32Gb G.Skill flare X 3200, MSI GTX1080, CoolerMaster MasterLiquid 240,SSD Samsung 850 EVO 250 GB, SSD Crucial CT256MX100, SSD Sandisk 480GB, HDD SG Baracuda 1TB


Posted By: PetrolHead
Date Posted: 04 Jan 2018 at 12:51am
" rel="nofollow - ventralstreams, that floating point result is really bad. Even my old Phenom II scored a lot better (7508 @ 3.7GHz, five run average), and so does my Ryzen 5 1500X (7992 @ 3.7GHz, five run average). In integer math you're doing less than 30% better than my Ryzen 5 1500X. Then again in SSE, encryption, sorting and compression your results are roughly 4 times better, which means these results have scaled with the amount of cores pretty well. Prime number and physics results are roughly twice what my 1500X achieves.

You can actually compare those sub results to other results with the same CPU directly from that screen by choosing the icon below the globe in the upper right hand corner, but I think it's pretty clear something's not right. Those CPU results should be pretty consistent between benchmark runs. Are you running the CPU in game mode or creator mode?

What about other benchmarks (in addition to Cinebench)? CPU-Z? Timespy and Firestrike CPU benchmarks? Geekbench 4.2?

Originally posted by zlobster zlobster wrote:

From their web page:

PassMark Software is a Microsoft Registered Partner and an Intel Software Partner.

I'll stop here.


That doesn't explain why OP's Threadripper seems to be performing a lot worse than other Threadrippes CPUs.


-------------
Ryzen 5 1500X, ASRock AB350M Pro4, 2x8 GB G.Skill Trident Z 3466CL16, Sapphire Pulse RX Vega56 8G HBM2, Corsair RM550x, Samsung 960 EVO SSD (NVMe) 250GB, Samsung 850 EVO SSD 500 GB, Windows 10 64-bit


Posted By: ventralstreams
Date Posted: 04 Jan 2018 at 1:26am
Yeah. Last test was even worse. 14914 and as expected the result is in the low end of the chart. In all categories except integer and floating point math I about average. On those two categories I am dead last. On top of that I don't see any suggestions except reinstalling windows anywhere.




-------------
Asrock Taichi x399, Threadripper 1950X, 32Gb G.Skill flare X 3200, MSI GTX1080, CoolerMaster MasterLiquid 240,SSD Samsung 850 EVO 250 GB, SSD Crucial CT256MX100, SSD Sandisk 480GB, HDD SG Baracuda 1TB


Posted By: PetrolHead
Date Posted: 04 Jan 2018 at 1:38am
You don't seem to be alone, however. Hmm. I wonder what the "sub 17000 club" has in common. The memory setting (gamer or creator) is the first that comes to mind and the Taichi should for sure have the option to change it somewhere in the BIOS. If I were you, I'd first locate that setting and see if changing it has any effect.


-------------
Ryzen 5 1500X, ASRock AB350M Pro4, 2x8 GB G.Skill Trident Z 3466CL16, Sapphire Pulse RX Vega56 8G HBM2, Corsair RM550x, Samsung 960 EVO SSD (NVMe) 250GB, Samsung 850 EVO SSD 500 GB, Windows 10 64-bit


Posted By: zlobster
Date Posted: 04 Jan 2018 at 1:38am
Maybe it's the earthquakes? Big smile Sorry, saw your location and I couldn't resist. Handshake ?

On the point - try AIDA64 benches and see how they compare.


-------------
1700X ZP-B1 (stock); X370 Taichi (UEFI 3.10); 16GB F4-3200C14-8GFX XMP; 256GB 960 EVO; RX 580 NITRO+ 8GB


Posted By: ventralstreams
Date Posted: 04 Jan 2018 at 2:02am
lol. We ve been shaking a bit the last few days but we're still standing. Anyway. Will try this AIDA thing and get back to you. @PetrolHead i can't see any "mode" setting in the BIOS. Most settings are voltage based (and everything is set to auto). (BTW, thanks for all the help guys.)

-------------
Asrock Taichi x399, Threadripper 1950X, 32Gb G.Skill flare X 3200, MSI GTX1080, CoolerMaster MasterLiquid 240,SSD Samsung 850 EVO 250 GB, SSD Crucial CT256MX100, SSD Sandisk 480GB, HDD SG Baracuda 1TB


Posted By: ventralstreams
Date Posted: 04 Jan 2018 at 2:25am
So i tested it with this AIDA software and got a bunch of numbers but nothing to compare them with. 

-------------
Asrock Taichi x399, Threadripper 1950X, 32Gb G.Skill flare X 3200, MSI GTX1080, CoolerMaster MasterLiquid 240,SSD Samsung 850 EVO 250 GB, SSD Crucial CT256MX100, SSD Sandisk 480GB, HDD SG Baracuda 1TB


Posted By: ventralstreams
Date Posted: 04 Jan 2018 at 2:40am
*I also did the cpuz benchmark and I get score close to the average, so it might not have to do with the processor after all.

-------------
Asrock Taichi x399, Threadripper 1950X, 32Gb G.Skill flare X 3200, MSI GTX1080, CoolerMaster MasterLiquid 240,SSD Samsung 850 EVO 250 GB, SSD Crucial CT256MX100, SSD Sandisk 480GB, HDD SG Baracuda 1TB


Posted By: PetrolHead
Date Posted: 04 Jan 2018 at 2:42am
I have the full version of Aida64 and my x64 CPU results for that same test (just one run) are as follows:

Memory Read: 46948 MB/s
Memory Write: 45115 MB/s
Memory Copy: 41229 MB/s
Single-Precision FLOPS: 236.1 GFLOPS
Double-Precision FLOPS: 118.1 GFLOPS
24-bit Integer IOPS: 59.03 GIOPS
32-bit Integer IOPS: 59.03 GIOPS
64-bit Integer IOPS: 29.50 GIOPS
AES-256: 32928 MB/s
SHA-1 Hash: 11133 MB/s
Single-Precision Julia: 96.21 FPS
Double-Precision Mandel: 50.15 FPS

These with a Ryzen 5 1500X @ 3.7 GHz and 3200MHz (14-18-18-18-38 CR1) RAM. It looks like your getting roughly 4 times bigger scores in many of the tests, which is what you should be getting if these tests scale well with cores.


-------------
Ryzen 5 1500X, ASRock AB350M Pro4, 2x8 GB G.Skill Trident Z 3466CL16, Sapphire Pulse RX Vega56 8G HBM2, Corsair RM550x, Samsung 960 EVO SSD (NVMe) 250GB, Samsung 850 EVO SSD 500 GB, Windows 10 64-bit


Posted By: ventralstreams
Date Posted: 04 Jan 2018 at 2:56am
Yeah. I'm really not sure how to proceed with this. I actually don't know if I have a real problem or not.

-------------
Asrock Taichi x399, Threadripper 1950X, 32Gb G.Skill flare X 3200, MSI GTX1080, CoolerMaster MasterLiquid 240,SSD Samsung 850 EVO 250 GB, SSD Crucial CT256MX100, SSD Sandisk 480GB, HDD SG Baracuda 1TB


Posted By: zlobster
Date Posted: 04 Jan 2018 at 3:01am
" rel="nofollow -
Originally posted by ventralstreams ventralstreams wrote:

Yeah. I'm really not sure how to proceed with this. I actually don't know if I have a real problem or not.


Since your problem is with FPU in PassMark (a synthetic benchmark), try running AIDA64 dedicated FPU tests and see how they compare to others'. Try YouTube for comparison runs, there are quite a few videos doing AIDA64 benches on TR 1950X. Example: https://www.youtube.com/watch?v=8c8Jwva5n_Q


-------------
1700X ZP-B1 (stock); X370 Taichi (UEFI 3.10); 16GB F4-3200C14-8GFX XMP; 256GB 960 EVO; RX 580 NITRO+ 8GB


Posted By: ventralstreams
Date Posted: 04 Jan 2018 at 3:22am
Ok. Will do that. Thank you.

-------------
Asrock Taichi x399, Threadripper 1950X, 32Gb G.Skill flare X 3200, MSI GTX1080, CoolerMaster MasterLiquid 240,SSD Samsung 850 EVO 250 GB, SSD Crucial CT256MX100, SSD Sandisk 480GB, HDD SG Baracuda 1TB


Posted By: MisterJ
Date Posted: 04 Jan 2018 at 3:55am
ventralstreams, here are my AIDA64 results.  One thing for sure is you need to post all your system specifications as I have.  I recommend you do so in your signature so they appear on each post.  I noticed that my AIDA64 is not up to date.  I have updated it and will rerun (when time permits) and post if things change much.  At least you can see all test results.  Enjoy, John.

https://postimg.org/image/3ks2ij6k5/" rel="nofollow">


-------------
Fat1 X399 Pro Gaming, TR 1950X, RAID0 3xSamsung SSD 960 EVO, G.SKILL FlareX F4-3200C14Q-32GFX, Win 10 x64 Pro, Enermx Platimax 850, Enermx Liqtech TR4 CPU Cooler, Radeon RX580, BIOS 2.00, 2xHDDs WD


Posted By: kerberos_20
Date Posted: 04 Jan 2018 at 5:53am
hi, even tho i dont have threadripper, but a plain 1600X
anway...
passmark's performance test - cpu floating point test
that one gets crappy low score with HPET disabled



Posted By: TANWare
Date Posted: 04 Jan 2018 at 6:16am
I had the same issue. I found also using an older single thread app SuperPI MOD hat on 16K it went from 0.140 Sec to 0.350 Sec. It was Windows10 that did it, I went and did the system restore and got all of the performance back.


Posted By: ventralstreams
Date Posted: 04 Jan 2018 at 7:14am
I have no idea when this started happening so I wouldn't know where to restore the system to. reinstalling windows is a total nightmare right now. I'm gonna leave this drastic solution as a last resort and keep looking for a way out. Unhappy

-------------
Asrock Taichi x399, Threadripper 1950X, 32Gb G.Skill flare X 3200, MSI GTX1080, CoolerMaster MasterLiquid 240,SSD Samsung 850 EVO 250 GB, SSD Crucial CT256MX100, SSD Sandisk 480GB, HDD SG Baracuda 1TB


Posted By: TANWare
Date Posted: 04 Jan 2018 at 2:30pm
I used the restore/reset and keep data files. I had to then reinstall all my apps. Yes I know it was a pain.......


Posted By: ventralstreams
Date Posted: 04 Jan 2018 at 6:35pm
Ok. With heavy heart I reset windows but was really happy to get performance back. I will have to be careful with what I install from now on I guess. Thank you all.



-------------
Asrock Taichi x399, Threadripper 1950X, 32Gb G.Skill flare X 3200, MSI GTX1080, CoolerMaster MasterLiquid 240,SSD Samsung 850 EVO 250 GB, SSD Crucial CT256MX100, SSD Sandisk 480GB, HDD SG Baracuda 1TB


Posted By: PetrolHead
Date Posted: 04 Jan 2018 at 7:57pm
Great to see that it worked, though! It might be worthwhile to take system image backups now and then, so you don't need to reset Windows totally if this occurs again. It might also help pinpoint which change in the system caused it to happen.


-------------
Ryzen 5 1500X, ASRock AB350M Pro4, 2x8 GB G.Skill Trident Z 3466CL16, Sapphire Pulse RX Vega56 8G HBM2, Corsair RM550x, Samsung 960 EVO SSD (NVMe) 250GB, Samsung 850 EVO SSD 500 GB, Windows 10 64-bit


Posted By: zlobster
Date Posted: 04 Jan 2018 at 10:31pm
Windows restore is a waste of space and time for me at least.

Re-deploying from the latest .iso is a matter of 15 minutes. 15 more to plop the other software. Plus, it's always the better option. Images, videos, music, work - all on the home NAS (RAID 10 for best of both worlds). This way I can keep bare minimal working environment on all my PCs.


-------------
1700X ZP-B1 (stock); X370 Taichi (UEFI 3.10); 16GB F4-3200C14-8GFX XMP; 256GB 960 EVO; RX 580 NITRO+ 8GB


Posted By: TANWare
Date Posted: 05 Jan 2018 at 6:43am
Mine again went south. I tried restore to a point I knew was good but this did not restore the performance. I had to reset again.
This time I am installing Macrium Reflect, and hopefully I can track the performance better.
 
Edit; I should note even though task manager shows 100% usage the integer and then floating point scores are the ones reduced that give it the overall low score.


Posted By: JayB33
Date Posted: 05 Jan 2018 at 12:34pm
Here's mine:





Mine's been overclocked to 4.0GHz w/ 3200MHz memory.

I'm not sure what it could be that saps your performance. If it's not running as a process or eating cycles, which would be visible in Task Manager, then what other effect could software be having on your CPU?

Could it be that you somehow enabled Game Mode? That would halve the number of cores of your CPU and improve latency. I'd have thought you would know if you had enabled it, though. I think it has to be done via Ryzen Master.

https://www.anandtech.com/show/11726/retesting-amd-ryzen-threadrippers-game-mode-halving-cores-for-more-performance" rel="nofollow - https://www.anandtech.com/show/11726/retesting-amd-ryzen-threadrippers-game-mode-halving-cores-for-more-performance


Posted By: TANWare
Date Posted: 05 Jan 2018 at 2:20pm
and mine now;
 


Posted By: JayB33
Date Posted: 05 Jan 2018 at 9:40pm
So I suppose you have no issues now. :P


Posted By: TANWare
Date Posted: 06 Jan 2018 at 2:53am
Me, no. I have installed mostly everything with Macrium run at each step to create images, of course I tested them first. I also just went to the 1709 Accumulative Update as well. Still have a few things to install, like a PDF reader etc..
 
I put in Macrium free edition but I think the home paid one is better in the long run.


Posted By: JayB33
Date Posted: 06 Jan 2018 at 2:54am
" rel="nofollow - So you don't think you accidentally enabled Game Mode in Ryzen Master which may have caused your initial performance problem?


Posted By: ventralstreams
Date Posted: 06 Jan 2018 at 7:10am
Personally I have no idea what had caused the issue. I am now monitoring performance after every action i take just to be sure.

-------------
Asrock Taichi x399, Threadripper 1950X, 32Gb G.Skill flare X 3200, MSI GTX1080, CoolerMaster MasterLiquid 240,SSD Samsung 850 EVO 250 GB, SSD Crucial CT256MX100, SSD Sandisk 480GB, HDD SG Baracuda 1TB


Posted By: zlobster
Date Posted: 06 Jan 2018 at 9:49pm
Funny enough, but after M$ installed the Spectre/Meltdown patch my CPU scores are down 5% in CPU-Z, Cinebench R15 & 3DMark Firestrike. PT9 CPU scores tanked pretty low!


-------------
1700X ZP-B1 (stock); X370 Taichi (UEFI 3.10); 16GB F4-3200C14-8GFX XMP; 256GB 960 EVO; RX 580 NITRO+ 8GB


Posted By: TANWare
Date Posted: 07 Jan 2018 at 12:50am
I only have Passmark and CB R15 and neither went down.


Posted By: zlobster
Date Posted: 07 Jan 2018 at 1:52am
Originally posted by TANWare TANWare wrote:

I only have Passmark and CB R15 and neither went down.


Fok! Double fok!

CrystalDisk 4kT1Q4 tanked by 50%!!! Both R&W... Fok that!


-------------
1700X ZP-B1 (stock); X370 Taichi (UEFI 3.10); 16GB F4-3200C14-8GFX XMP; 256GB 960 EVO; RX 580 NITRO+ 8GB


Posted By: TANWare
Date Posted: 08 Jan 2018 at 10:37am
It seems I have lost about 2%-5% of my CDM scores. This is with an SATA 1TB drive. I had heard that M.2 drives are more affected though. Since AMD supposedly is only in jeopardy if the current user is there to place it in troubles way we should be given the opportunity not to use the patch, at least that is my opinion. 


Posted By: zlobster
Date Posted: 08 Jan 2018 at 5:48pm
Time to dust off my IOMMU skills and get that Windoze cr@p in the VM, w/ host OS being not-so-retarded Linux.


-------------
1700X ZP-B1 (stock); X370 Taichi (UEFI 3.10); 16GB F4-3200C14-8GFX XMP; 256GB 960 EVO; RX 580 NITRO+ 8GB


Posted By: JayB33
Date Posted: 09 Jan 2018 at 8:47am
Pre-Spectre / Meltdown Patch, Windows is installed on this Samsung 960 Pro:



Post-Spectre / Meltdown Patch, Windows is installed on this Samsung 960 Pro:



Pre-Spectre / Meltdown Patch, Windows is not installed on this Samsung 960 Pro:



Post-Spectre / Meltdown Patch, Windows is not installed on this Samsung 960 Pro:



Conclusion on the Windows drive: Q32T1 Read results appear to have tanked by 450 MB/s with Windows installed on the same drive after the Spectre / Meltdown patch, although results always seem to be inconsistent when Windows is installed on the drive to be benchmarked. 4 KiB Q8T8 Write INCREASED by 250MB/s after after the Spectre / Meltdown patch on the Windows drive. Margin of error on an inconsistent benchmark? I dunno.

Conclusion on the non-Windows drive: Within margin of error on all tests. The Spectre / Meltdown patch does not appear to have affected drives that do not have Windows installed, at least with ASRock X399 & MS Drivers.



Print Page | Close Window

Forum Software by Web Wiz Forums® version 12.04 - http://www.webwizforums.com
Copyright ©2001-2021 Web Wiz Ltd. - https://www.webwiz.net