ehw foldathon ExtremeHW Folding@Home Competition, January 20th - 22nd

[axipher]

Hello all,

 

Welcome to the ExtremeHW Folding Competition, January 20th - 22nd, Midnight UTC.

 

Countdown to Event Start

Countdown to Event End

 

 

 

In the spirit of folding for the cause and good old friendly competition between members, we are hosting another folding competition. For more details please see below.

 

How to participate:

• During the competition you must be folding for Team 239902
• You must post in this thread with "IN" followed by your folding username.

 

Prizes:

• USD $25 PayPal (@axipher) Won by @Minotaurtoo
  • Total amount = $5 per 100 Total Work Units up to $25 Maximum
    • 208 Total Work Units = $10
    • 513 Total Work Units = $25
• 1 year Gold Membership! (@Bastiaan_NL) Won by @Sgt_Swanny

 

Folding Time Prizes:

• 1 Week of 2080 Super (@Bastiaan_NL) (If we get 20 participants)
• 1 Week of 1080 Ti (@Cerberus) Won by @ENTERPRISE
• 1 Week of 2080 Ti (@Supercrumpet) (If we get 20 participants)

 

 

 

All "Folding Time" prizes will require your username/passkey to be given to the donator, this will not give them any of your personal information, it just let's them be able to contribute full points to your username.  Passkeys need at least 10 full Work Units to start getting the full Quick Return Bonus.  Alternatively, you could setup another passkey for your Folding@Home username with a different email address and fold your first 10 units on that.  This would still put the points towards your username, but they would also be tracked by a different passkey.

 

Please note all winners must claim within 30 days from announcement of the winners.

 

 

Current Event Stats:

https://docs.google.com/spreadsheets/d/e/2PACX-1vSkztaGPFX12T6GazTilrFS1iy8OsssANF9sv3yMT6qtVFdBtoRNfFyQghtZiFNz2L_uPfr5Rxfr1TQ/pubhtml

 

 

Team overall stats:

https://docs.google.com/spreadsheets/d/e/2PACX-1vRlBI5oz7kCqBzp8hktCt31IWFPMXxDsOwNT7ipvAO6vvlD98sO-5yljfr6kAwSbimo8jpwir-Kt5WY/pubhtml

 

 

Note: Stats are updated every hour at XX:25 for previous hour of submitted Work Units.

 

 

 

Good luck to all!

axipher

Edited January 28, 2021 by axipher

[Bastiaan_NL]

In Bastiaan_NL

 

I will donate a 1 year Gold Membership!

And if we have more than 20 participants for the Competition I'll add a week of folding time on the 2080 super, worth anywhere between 3 and 4 mil PPD

[axipher]

20 minutes ago, Bastiaan_NL said:

In Bastiaan_NL

 

I will donate a 1 year Gold Membership!

And if we have more than 20 participants for the Competition I'll add a week of folding time on the 2080 super, worth anywhere between 3 and 4 mil PPD

 

Thanks mate, added to the first post.

[Cerberus]

In as Cerberus, will donate another week of 1080ti, too

[axipher]

5 minutes ago, Cerberus said:

In as Cerberus, will donate another week of 1080ti, too

 

Thanks mate, added that to the OP as well.

[Minotaurtoo]

in as minotaurtoo

[Supercrumpet]

I like that idea @Bastiaan_NL, I'll do the same with a 2080 Ti if we get over 20 folders.

 

In as CptAsian, of course.

[axipher]

45 minutes ago, Supercrumpet said:

I like that idea @Bastiaan_NL, I'll do the same with a 2080 Ti if we get over 20 folders.

 

In as CptAsian, of course.

 

Thanks as well, added to OP.

 

I also removed the estimated PPD numbers since those can change anyway with units and it's free PPD for the winner anyway.

[Bastiaan_NL]

@axipher Those PPD numbers are all over the place, I see over 5 mil difference between my lowest and highest average PPD

[bonami2]

IN mrbonami2

 

 

[Sgt_Swanny]

In Sgt-Swanny

[ENTERPRISE]

Nice to see see the usual faces and a new one . Welcome to the fold @Sgt_Swanny

[tictoc]

Have a few GPUs folding for EHW while testing the latest ROCm.  I'll keeping them going through the comp.

[ENTERPRISE]

Nice....What is ROCm ?

[Sgt_Swanny]

On 1/12/2021 at 3:52 PM, ENTERPRISE said:

Nice to see see the usual faces and a new one . Welcome to the fold @Sgt_Swanny

I hope my 1050ti can keep up haha. 

[Bastiaan_NL]

I'll be switching all the hardware tomorrow evening after the FaT, should be fun!

[Diffident]

15 hours ago, ENTERPRISE said:

Nice....What is ROCm ?

 

ROCm is AMD's open source OpenCL driver for Linux that only works with a hand full of applications.

[tictoc]

17 hours ago, ENTERPRISE said:

Nice....What is ROCm ?

ROCm is AMD's open source compute stack of software. https://www.amd.com/en/graphics/servers-solutions-rocm

 

1 hour ago, Diffident said:

 

ROCm is AMD's open source OpenCL driver for Linux that only works with a hand full of applications.

OpenCL is still very hit and miss on the open source stack.  I am actually testing some HIP code that was ported from CUDA and it is working surprisingly well.

 

For F@H the ROCm opencl-runtime is definitely the way to go, at least on anything Vega.  Here's a quick little comparison of some WUs running on the AMD closed source OpenCL driver and on the ROCm 4.0 OpenCL driver.

 

Radeon VII

amdgpu-pro-20.30_1109583
    Slot 01 p14914 - tpf=3:17
    Slot 00 p14911 - tpf=3:10
    Slot 02 p14912 - tpf=5:31
    Slot 03 p13442 - tpf=11:46

ROCm 4.0
    Slot 01 p14914 - tpf=1:14
    Slot 00 p14911 - tpf=1:17
    Slot 02 p14912 - tpf=1:16
    Slot 03 p13442 - tpf=1:42

 

The p1491x WUs are pretty brutal on AMD GPUs, only about 800-900k ppd on a Radeon VII.  That is at least tolerable compared to the 250k ppd on the amdgpu-pro driver.  Some other WUs are actually pretty decent, like the p17321, which is around 2.9M ppd.

Edited January 20, 2021 by tictoc
typos

[Diffident]

I can't get FAH to work with ROCm 4.0 on my Radeon VII in Gentoo.  I guess I'll have to boot my Arch drive and try it.

 

ROCm is killer with Einstein@home.

[tictoc]

Not sure what the special sauce is for F@H on ROCm.  Previously I had no issues running the closed source bits of the pro driver, but now the performance for F@H has really tanked.

 

Just took a quick look at the rocm-opencl-runtime Gentoo package, and as far as I can tell, all of the necessary libs, compilers, and applications should be getting built.  I build a bit more of the ROCm stack, but that is all for HIP, profiling, and debugging.

 

FWIW I've only tested and ran F@H+ROCm_4.0 on kernels 5.10.7 and 5.10.8.

[tictoc]

Sorry for the somewhat OT double post.

 

@Diffident I'm not a fan of the way that the latest F@H client is packaged in the AUR, and I think it might actually cause some issues when running F@H on ROCm.  I didn't have time to fully test it, so I've just continued to use the older client.  The new client in the AUR broke some scripts that I have for F@H along with some other things that I don't really like about how it's installed.  I'm currently running the 7.5.1 client. 

 

If you want to give that a go, here's my PKGBUILD along with the post-install and SystemD service. foldingathome.zip

 

[ENTERPRISE]

9 hours ago, Diffident said:

 

ROCm is AMD's open source OpenCL driver for Linux that only works with a hand full of applications.

 

The veil has been lifted and I am now educated  

 

7 hours ago, tictoc said:

ROCm is AMD's open source compute stack of software. https://www.amd.com/en/graphics/servers-solutions-rocm

 

OpenCL is still very hit and miss on the open source stack.  I am actually testing some HIP code that was ported from CUDA and it is working surprisingly well.

 

For F@H the ROCm opencl-runtime is definitely the way to go, at least on anything Vega.  Here's a quick little comparison of some WUs running on the AMD closed source OpenCL driver and on the ROCm 4.0 OpenCL driver.

 

Radeon VII

amdgpu-pro-20.30_1109583
    Slot 01 p14914 - tpf=3:17
    Slot 00 p14911 - tpf=3:10
    Slot 02 p14912 - tpf=5:31
    Slot 03 p13442 - tpf=11:46

ROCm 4.0
    Slot 01 p14914 - tpf=1:14
    Slot 00 p14911 - tpf=1:17
    Slot 02 p14912 - tpf=1:16
    Slot 03 p13442 - tpf=1:42

 

The p1491x WUs are pretty brutal on AMD GPUs, only about 800-900k ppd on a Radeon VII.  That is at least tolerable compared to the 250k ppd on the amdgpu-pro driver.  Some other WUs are actually pretty decent, like the p17321, which is around 2.9M ppd.

Wow...That is a huge difference in TPF (Time Per Frame, lower is better for those who are unaware).

[Minotaurtoo]

last night before going to bed I loaded my power profiles for long term folding.   CPU power limited to 65w GPU limited to 145w  Current CPU tpf: 3m 42s  Current GPU tpf: 1m 37s   Wondering now if that could be improved with some other tuning to increase power efficiency further.

[Diffident]

7 hours ago, tictoc said:

Not sure what the special sauce is for F@H on ROCm.  Previously I had no issues running the closed source bits of the pro driver, but now the performance for F@H has really tanked.

 

Just took a quick look at the rocm-opencl-runtime Gentoo package, and as far as I can tell, all of the necessary libs, compilers, and applications should be getting built.  I build a bit more of the ROCm stack, but that is all for HIP, profiling, and debugging.

 

FWIW I've only tested and ran F@H+ROCm_4.0 on kernels 5.10.7 and 5.10.8.

For some reason it doesn't work.  FAH detects the GPU but says OpenCL not detected.  I know ROCm is working since I can run Einstein tasks.

[tictoc]

It might be the client.  I had issues with the latest one, and didn't take the time to sort it out.  Maybe try rolling back to an older version of the F@H client.