GPUPI - SuperPI on the GPU

[_mat_]

The submission node is the root node of the HWBOT validation file. But the errors before that already make it very clear, that the data could not be decrypted correctly. Sry, that's bad news. Seems like the file was damaged at some point.

[niobium615]

It's happening to every file I create, I guess I'll try reinstalling GPUPi and see if that helps.

 

Oddly, when I just tried, HWBot seems to have no trouble decrypting them.

 

Edit: Same issue with a fresh install of GPUPi, but HWBot is still happy with the scores. Strange.

Edited July 2, 2017 by niobium615

[_mat_]

That's strange indeed. Please send one of these result files to matthias at hwbot org, I will have a look at it. Thank you.

[_mat_]

What? Skylake X is not supported by Intel's official OpenCL drivers? Now I see why the current scores where made with the AMD OpenCL driver.

 

That's really sad.

[_mat_]

If anyone is up for testing a hot of the press GPUPI version, you can download it here: GPUPI 3.0

 

• Supports calculations with mixed APIs like AMD Radeon on OpenCL and NVIDIA on CUDA.
  
• Detection of clock frequencies with integrated HWiNFO library.
  
• Full support for CUDA 9.1 including various performance improvements!
  
• Early support for OpenCL 2.0, currently no changes in performance.
  
• CUDA + Titan V is now a full win. Try it and get your gold cup! BIG thx to Joe "the slacker" Steponz for letting me into his rig.
  
• Submission to hwbot.org is now encrypted via https.
  
• Screenshots are now mandatory to help with the moderation of scores.
  
• Improved device detection for AMD Radeon cards by a lot
  
• Bugfix for result file validation
  
• No HPET timer necessary for approved platforms. That is currently: Skylake, Kaby Lake, Kaby Lake X and Skylake X
  
• Memory Reduction greatly optimized for CUDA (uses Warp Shuffling, if available)
  
• Added a new Reduction Size: 1024
  
• New settings checkbox: "Run Confirmation" - you can now deliberately enable or disable the message box before each run.
  
• Brand new command line mode via GPUPI-CLI.exe
   
  Examples:
   
  => GPUPI-CLI.exe -h ... Help
  => GPUPI-CLI.exe -l ... List all available devices
  => GPUPI-CLI.exe -d 1B -g -a CUDA -b 20M -r 512 ... 1B on all CUDA graphics cards
  => GPUPI-CLI.exe -d 1B -g -a OpenCL -b 20M -r 512 ... 1B on all OpenCL graphics cards
  => GPUPI-CLI.exe -d 100M -c -b 10M -r 128 ... 100M on all CPUs and APIs available in the system (can take a while)

 

 

After some testing it will be officially announced. Early feedback is appreciated though.

Edited December 19, 2017 by _mat_

[websmile]

We might have a problem, maybe you can give me an explanation for results like these

http://hwbot.org/submission/3672434_sergeyr_gpupi_for_cpu___1b_2x_xeon_x5690_46sec_718ms/

in comparison with this

http://hwbot.org/submission/3410802_sparkey247_gpupi_for_cpu___1b_core_i7_extreme_980x_4min_29sec_662ms

and this

http://hwbot.org/submission/3668752_splave_gpupi_for_cpu___1b_core_i9_7920x_1min_26sec_607ms/

and especially these

http://hwbot.org/submission/2860344_skulstation_gpupi_for_cpu___1b_xeon_x5690_4min_47sec_284ms

http://hwbot.org/submission/3626243_skulstation_gpupi_for_cpu___1b_2x_xeon_e5645_59sec_189ms

I think there definitely is a more than weird scaling and results of multi cpu xeons are real dodgy

[skulstation]

about the scores whit the sr-2.

i promisd mate some time a go thad he can use my sr-2 setup for testing.

as i show him some real bugged runs thad i have made whit the sr-2 sub 10sec for 1B

but family stuf go first

in my new home i gone make time for it

Edited December 29, 2017 by skulstation

[_mat_]

Whatever happens here is impossible to see by just examing the screenshots. The performance is not the issue here, but the timer might be. Those scores were done with the RTC timer, which is currently approved by HWBOT for older Windows versions. It might get skewed due to improper handling on the SR-2 or on every dual-socket board for 1366.

 

As always: If a run takes 4 minutes in real time and is submitted as a sub 1 minute result, it should be reported by the overclocker instead of being submitted.

 

Btw, if this is really a timer issue then I would recommend to remove all dual xeons scores from the bot. Other benchmarks could share the same problems, I can help to point out which ones. But it's too early to say yet, so let's not jump to any conclusions.

 

If anyone could give me remote access to a setup with sr-2, I would very much appreciate it. Any good offers on a cheap sr-2 are welcome as well.

Edited December 29, 2017 by _mat_

[yosarianilives]

Whatever happens here is impossible to see by just examing the screenshots. The performance is not the issue here, but the timer might be. Those scores were done with the RTC timer, which is currently approved by HWBOT for older Windows versions. It might get skewed due to improper handling on the SR-2 or on every dual-socket board for 1366.

 

As always: If a run takes 4 minutes in real time and is submitted as a sub 1 minute result, it should be reported by the overclocker instead of being submitted.

 

Btw, if this is really a timer issue then I would recommend to remove all dual xeons scores from the bot. Other benchmarks could share the same problems, I can help to point out which ones. But it's too early to say yet, so let's not jump to any conclusions.

 

If anyone could give me remote access to a setup with sr-2, I would very much appreciate it. Any good offers on a cheap sr-2 are welcome as well.

 

I actually have an sr-2. I'd have to set it up and there's no time this weekend but I could definitely work with you on this. My best guess is that it has to do with changing bclk in OS but I could be 100% off on that. If it is legit then I'd love to see how far I can push this gal on dice, 5 ghz 12c westmere beating 6 ghz 7980xe is big if true

[_mat_]

New version with a few bugfixes:

 

GPUPI 3.0.1

 

• Bugfix for the initialization of older opencl devices like HD 7xxx series or Intel Core 2 Duo
  
• Update for the integrated HWiNFO library
  
• Improved formatting of errors

 

Download

[_mat_]

I actually have an sr-2. I'd have to set it up and there's no time this weekend but I could definitely work with you on this.

That would be very much appreciated! Thank you in advance and please contact me any time on matthias at hwbot.org.

[websmile]

Any update on the issue?

[bolc]

New version with a few bugfixes:

 

GPUPI 3.0.1

 

• Bugfix for the initialization of older opencl devices like HD 7xxx series or Intel Core 2 Duo
  
• Update for the integrated HWiNFO library
  
• Improved formatting of errors

 

Download

 

Hello

I was about to ask if using older gpupi version like 2.2 may help benching old ati card like hd4850 (which 2.3.4 does not recognize...) but 2.2 is not allowed for some time already

...

may this new 3.0.x version do the job for hd4850 and alike as well as for c2d cpus ...?

will test it soon.

is 3.0.1 presently valid/accepted for submitting results ?

Thanks.

Cheers

Edited January 10, 2018 by bolc

[bolc]

not exactly on the dual xeon bug issue, but what about submission with dual xeon stations, competing with single xeon submissions...? (submitted by mistake on both I think)

 

same score submitted as both 4x and 8x

http://hwbot.org/submission/3607721_stingeryar_gpupi_for_cpu___100m_xeon_e5440_45sec_130ms

http://hwbot.org/submission/3608472_stingeryar_gpupi_for_cpu___100m_2x_xeon_e5440_45sec_130ms

 

8x submitted as 4x

http://hwbot.org/submission/3607549_stingeryar_gpupi_for_cpu___1b_xeon_e5410_12min_19sec_2ms

[_mat_]

Your HD 4850 should work fine with GPUPI 3.0.1. It supports OpenCL 1.0 and double precision, so it's good to go. If you encounter any problems, please let me know here.

 

Just be aware that GPUPI 3.0 currently works only on 64 bit operating systems and Windows 7 upwards. It should work fine on Vista though, but it's untested yet.

 

For Windows XP support use the Legacy version of GPUPI 2.3.4 (Download it here: https://www.overclockers.at/news/gpupi-international-support-thread). If you need 32 bit support, let me know as well. I can easily build one, although I prefer to phase out 32 bit support if possible.

 

GPUPI 3.0 and all its minor versions are currently allowed for submission.

 

Regarding the results you have posted: Please report them so the moderators can have a look.

[bolc]

Your HD 4850 should work fine with GPUPI 3.0.1. It supports OpenCL 1.0 and double precision, so it's good to go. If you encounter any problems, please let me know here.

 

Just be aware that GPUPI 3.0 currently works only on 64 bit operating systems and Windows 7 upwards. It should work fine on Vista though, but it's untested yet.

 

For Windows XP support use the Legacy version of GPUPI 2.3.4 (Download it here: https://www.overclockers.at/news/gpupi-international-support-thread). If you need 32 bit support, let me know as well. I can easily build one, although I prefer to phase out 32 bit support if possible.

 

GPUPI 3.0 and all its minor versions are currently allowed for submission.

 

Regarding the results you have posted: Please report them so the moderators can have a look.

 

 

Great to hear the 4850 will work with it, I use win7 64 so it is fine

Thank you

[websmile]

not exactly on the dual xeon bug issue, but what about submission with dual xeon stations, competing with single xeon submissions...? (submitted by mistake on both I think)

 

same score submitted as both 4x and 8x

http://hwbot.org/submission/3607721_stingeryar_gpupi_for_cpu___100m_xeon_e5440_45sec_130ms

http://hwbot.org/submission/3608472_stingeryar_gpupi_for_cpu___100m_2x_xeon_e5440_45sec_130ms

 

8x submitted as 4x

http://hwbot.org/submission/3607549_stingeryar_gpupi_for_cpu___1b_xeon_e5410_12min_19sec_2ms

 

We see this a lot lately, people seem to think this is acceptable behaviour to submit exact same results to several categories to steal points. We consider this cheating, I already banned some members because of similar actions lately. Maybe it is time to enforce this even more..

[bolc]

We see this a lot lately, people seem to think this is acceptable behaviour to submit exact same results to several categories to steal points. We consider this cheating, I already banned some members because of similar actions lately. Maybe it is time to enforce this even more..

 

 

hopefully it was temporary bans...

 

i d agree with you if this would be consistently done. in his case, he has lots (really a lot!) of xeon x1 and x2 subs, he may have missed those 2 out of his multiple dozens...

 

a bit like when submitting amd unlocked cpu scores, sometimes, the system bugs and does not let you adjust the number of cores although you edit. it can get frustrating and hard to find the faulty sub later on, or manage to get it right.

not sure if that applies to multi cpu benchmarks in term of "submission difficulty"

 

if the results "steals" lots of points, and the owner doe not do anything, yeah, that sounds like cheating. but for those little guys lost in the darkness of 771 benching, I do not think that strongly.

 

but yeay, you are the rule enforcer

 

Cheers

[websmile]

Everyone can make a mistake, which is why we watch this closely - I see no pattern for example at this user, but we have/had others that did stuff like this systematically. On these we decided to not tolerate this anymore, and we also hand out warnings now for stuff like posted here to show that this is no joke.

[_mat_]

Thanks to skulstation I could work on a system with the SR-2 today to have a look at the timing issues. I could not reproduce problems with the RTC timer via remote access myself and I am pretty sure that this has nothing to do with already known bclock skewing that impacts the TSC timers.

 

This is something completely different and it skews the RTC on Windows 7 and possibly every other Windows version as well. That means that every benchmark that uses RTC - Aquamark for example - will be impacted by this problem as well, not just GPUPI.

 

As a countermeasure I have implemented a safeguard for GPUPI 3.1 that prohibits the usage of the RTC timer on the SR-2. So you will need to enable the High Precision Event Timer on those systems otherwise GPUPI 3.1 will not let you run the benchmark. skulstation is currently testing the new version as we speak and will report any problems or findings. Let's see if that fixes the issue. If no further problems arise I will release GPUPI 3.1 officially and disable benchmark submission for version 2.3.x and 3.0.

 

As for the other benchmarks like Aquamark, that should be looked at as well. I can compile a list of benchmarks that use the RTC timer, so we can work on that. But let's fix GPUPI first to be sure that the RTC timer is the issue. Afterwards it should be discussed what is going to happen with the current SR-2 runs. If nobody can reproduce a 46 seconds run for GPUPI for CPU 1B with two hexacore Xeons with GPUPI 3.1, I am recommending to delete these scores.

[_mat_]

If you want to test as well, feel free to download GPUPI 3.1 right here: https://www.overclockers.at/downloads/GPUPI%203.1.zip

[bolc]

If you want to test as well, feel free to download GPUPI 3.1 right here: https://www.overclockers.at/downloads/GPUPI%203.1.zip

 

 

Hi

I have tried running Win7 64 with HD4850, catalyst driver 13.9, latest on amd site, with gpupi 3.0.1, and the device is not seen un gpupi. It seels intel opencl and amd opencl for the cpu only.

same thing with 2.2, 2.3.4 legacy and 3.1

any hints where to look after?

Thank you

Cheers

Edited January 14, 2018 by bolc

[_mat_]

If the device is not shown, it is a driver issue. The enumeration of devices is a pretty basic thing an relies on what the OpenCL driver returns.

 

Just to be sure: Have you tried Luxmark or other OpenCL applications? Do you have the same problems there?

[bolc]

If the device is not shown, it is a driver issue. The enumeration of devices is a pretty basic thing an relies on what the OpenCL driver returns.

 

Just to be sure: Have you tried Luxmark or other OpenCL applications? Do you have the same problems there?

 

Hi

Thanks.

the strange thing is that some suceeded to detect the card

http://hwbot.org/submission/3319863_nicegab_gpupi___1b_radeon_hd_4850_gddr3_14min_6sec_376ms

with catalyst 13.9, opencl 1.2

 

on my hd7950, installing catalyst 15.7.1 installs opencl driver.

 

 

it seems this carsd is only compliant with opencl 1.0 (these sites say 1.0 but gpupi worked in 1.2 so I guess 1.x is ok....?)

https://www.techpowerup.com/gpudb/493/radeon-hd-4850

https://community.amd.com/thread/152535

here they dismiss 48xx from 1.1, suggesting that 1.0 is the only driver good for 4850.

https://www.mql5.com/en/articles/690

but since it worked on gpupi for several hwbot subs, 1.2 should be ok.

 

 

the only way I know so far to install opencl is via catalyst center. it installs OpenCL Driver 10.0.937.2.

 

perhaps there is another way to force 1.2 installation instead of 2.0 ?

 

 

THERE MAY BE HOPE HERE:

http://www.geeks3d.com/20091221/how-to-enable-opencl-support-on-nvidia-and-amd-platforms/

 

I will try luxmark too.

Cheers

Edited January 15, 2018 by bolc

[havli]

IIRC HD 4800 stopped working with GPUPI long time ago. I think when version 2.3.x was released.

 

@_mat_

could you please also make 32-bit version of GPUPI 3.1 if it becomes mandatory? There is quite a lot of 32-bit only NetBurst CPUs otherwise capable of running GPUPI. And it is always nice to get free points and gold cups