Folding Community Board

[Gorgon]

On 8/9/2019 at 1:43 PM, palespartan said:

Best way of announcing this. lol.

Cool. I had to reinstall Kali anyways (pretty sure the nvidia driver broke the settings app) so Ill check about installing the python 2 support packages first this time.

Did a bunch of testing today for the Linux Install Guide towards getting FAHControl working in a fresh install of Ubuntu 18.04.3. Discovered that FAHControl requires EITHER the python-support package from the external repository OR the dh-python package which is available as part of Ubuntu (Debian also?)

This should do the trick:

sudo apt install python dh-python python-gnome2
wget https://download.foldingathome.org/releases/public/release/fahcontrol/debian-stable-64bit/v7.5/fahcontrol_7.5.1-1_all.deb
sudo dpkg -i fahcontrol_7.5.1-1_all.deb

[RollinLower]

i'm getting about 270K PPD on my Titan X/1070 system and only 60K PPD on my 2x 290 system?

what the hell? is anyone else seeing bizarre numbers like these?

[Gorgon]

1 hour ago, RollinLower said:

i'm getting about 270K PPD on my Titan X/1070 system and only 60K PPD on my 2x 290 system?

what the hell? is anyone else seeing bizarre numbers like these?

 

[Favebook]

On 8/2/2019 at 9:30 AM, Gorgon said:

Go to your favourite CPU performance database and compare the performance in TFLOPS. That should, in general, answer it.  The other considerations are a higher TDP causing your power bill to go up and, different memory and cooler requirements.

 

Simple math says 40*1.7=68 and 24*3=72 so assuming they’re both Ivy Bridge with similar IPCs, AVX instruction sets probably not worth it.

 

Im rather fond of “Big Iron” but Ryzen Gen 3 I suspect will really disrupt the HEDT market. With the 3900X out the prices here on the 2920WX have been slashed. It’s causing me to rethink my new workstation build. I’ll have to wait for the 16 core model to come and after looking at the specs figure out if the better IOMMU groups, ECC Memory with Encryption and extra PCIe lanes are worth the extra expense for threadripper.

 

I built a new KVM machine recently using a 2700 and 32GB DDR4-3200 CL16 in a B450 board and the cost was quite reasonable and I just use the excess threads for BOINC.

I see that you are knowledgeable so I hope you won't mind me asking...

If for BOINC CPU's should be compared by their performance in TFLOPS, how do we compare GPU's for F@H? Also, is there a spreadsheet somewhere with approximate speeds of GPU's?

[GOTSpectrum]

6 hours ago, Favebook said:

I see that you are knowledgeable so I hope you won't mind me asking...

If for BOINC CPU's should be compared by their performance in TFLOPS, how do we compare GPU's for F@H? Also, is there a spreadsheet somewhere with approximate speeds of GPU's?

https://www.overclock.net/forum/55-overclock-net-folding-home-team/475163-gpu-projects-ppd-database.html

[Gorgon]

2 hours ago, GOTSpectrum said:

https://www.overclock.net/forum/55-overclock-net-folding-home-team/475163-gpu-projects-ppd-database.html

This spreadsheet is a whole lot less useful than it was as it no longer shows ndividual data points but just the aggregate values so I’ve stopped submitting results to it as the data tends to be skewed.

 

In general with FAH core 21 Work Units (WUs) Linux tends to out-perform Windows by 5-10% so that can’t be seen on the overclock sheet anymore nor can you gauge the overclock anymore or which WU was reported.

 

AnandTech used to include FAHBench in their GPU test suites but of late has stopped doing so but hopefully will resume testing with it again. FAHBench, however, reports in ns/day, but the results appear to translate somewhat linearly as a measure or performance. If you go to their “Bench” results and look at results for Bench 2017 etc. it should give you an idea of relative performance.

 

Note that the new AMD 5700 and XT currently are unsupported due to AMD driver issues.

 

If you have questions about specific cards generally you will find a few people here running them happy to share their results. 

[GOTSpectrum]

1 minute ago, Gorgon said:

This spreadsheet is a whole lot less useful than it was as it no longer shows ndividual data points but just the aggregate values so I’ve stopped submitting results to it as the data tends to be skewed.

 

In general with FAH core 21 Work Units (WUs) Linux tends to out-perform Windows by 5-10% so that can’t be seen on the overclock sheet anymore nor can you gauge the overclock anymore or which WU was reported.

 

AnandTech used to include FAHBench in their GPU test suites but of late has stopped doing so but hopefully will resume testing with it again. FAHBench, however, reports in ns/day, but the results appear to translate somewhat linearly as a measure or performance. If you go to their “Bench” results and look at results for Bench 2017 etc. it should give you an idea of relative performance.

 

Note that the new AMD 5700 and XT currently are unsupported due to AMD driver issues.

 

If you have questions about specific cards generally you will find a few people here running them happy to share their results. 

Very true, I use it as rough guide only, I run the following cards if anyone wants into, 1080 ti hybrid, 1080 hybrid, 1080 FE, 980 FE and 980ti super jetstream. 

[justpoet]

How long does it normally take for a new arch to be supported, like Navi or the rtx cores?  Been thinking of a Navi card for when I build up a gaming rig.

[GOTSpectrum]

19 minutes ago, justpoet said:

How long does it normally take for a new arch to be supported, like Navi or the rtx cores?  Been thinking of a Navi card for when I build up a gaming rig.

Currently the RTX cores are not utilised for folding,I don't think they are any plans to do so either

[palespartan]

On 8/10/2019 at 5:16 PM, Gorgon said:

Did a bunch of testing today for the Linux Install Guide towards getting FAHControl working in a fresh install of Ubuntu 18.04.3. Discovered that FAHControl requires EITHER the python-support package from the external repository OR the dh-python package which is available as part of Ubuntu (Debian also?)

This should do the trick:

sudo apt install python dh-python python-gnome2
wget https://download.foldingathome.org/releases/public/release/fahcontrol/debian-stable-64bit/v7.5/fahcontrol_7.5.1-1_all.deb
sudo dpkg -i fahcontrol_7.5.1-1_all.deb

Awesome, actually on a new fresh install of kali again so I will test this after I get all the remote management setup again.

 

11 hours ago, Favebook said:

I see that you are knowledgeable so I hope you won't mind me asking...

If for BOINC CPU's should be compared by their performance in TFLOPS, how do we compare GPU's for F@H? Also, is there a spreadsheet somewhere with approximate speeds of GPU's?

For folding it is actually almost the same all gpus are also measured in tflops for compute performance. These stats are usually hard to find though so if you have a specific gpu you can ask us and we can let you know if any of us know the ppd but honestly Ive gotten away with googling folding performance of card X to get rough ppd.

[Gorgon]

8 hours ago, justpoet said:

How long does it normally take for a new arch to be supported, like Navi or the rtx cores?  Been thinking of a Navi card for when I build up a gaming rig.

The issue with Navi is AMD hasn’t gotten the OpenCL working in the drivers yet. Something to do with not having the right hooks for the new architecture. Once that’s done it should just work once Folding at Home re enables the GPU IDs in the whitelist.

 

Using RTX cores would require the team whom manage the OpenMM libraries to investigate if they can be used at all and to integrate it into a new library and then have the Folding at Home team compile their cores against the new library.

 

Current cores use the OpenMM version 21 library and they are working on the version 22 integration which was in Beta testing but the developers have gone back to the drawing board so they’ve dried up for now but should be back at some point.

 

So with respect to using RTX cores it could be never if they don’t offer any useful benefit or if they do it would likely be a couple of years.

[Gorgon]

FYI

[palespartan]

13 hours ago, Gorgon said:

The issue with Navi is AMD hasn’t gotten the OpenCL working in the drivers yet. Something to do with not having the right hooks for the new architecture. Once that’s done it should just work once Folding at Home re enables the GPU IDs in the whitelist.

 

Using RTX cores would require the team whom manage the OpenMM libraries to investigate if they can be used at all and to integrate it into a new library and then have the Folding at Home team compile their cores against the new library.

 

Current cores use the OpenMM version 21 library and they are working on the version 22 integration which was in Beta testing but the developers have gone back to the drawing board so they’ve dried up for now but should be back at some point.

 

So with respect to using RTX cores it could be never if they don’t offer any useful benefit or if they do it would likely be a couple of years.

I think unless NVIDIA releases an open compute API that can use the RTX cores then it just wont happen. Right now they seem pretty locked on the fact that they just want them used for ray tracing and not interested in putting in the time for anything else. Unlike the tensor cores on the other cards that were intended for deep learning compute.

On 8/11/2019 at 1:02 PM, Gorgon said:

 

Does anyone know an ETA on when this will be fixed. I just had the problem yesterday when I fired up one of my rigs where it was assigning bonus to one of my cards folding but not the other.

[Gorgon]

On 8/13/2019 at 12:40 PM, palespartan said:

... Does anyone know an ETA on when this will be fixed. I just had the problem yesterday when I fired up one of my rigs where it was assigning bonus to one of my cards folding but not the other.

 

[Gorgon]

And speaking of vacation I’m off to a cottage for a week. Hopefully my folding rigs will stay up as I’m not driving the two hours back just to press the reset button.

[NoneScope92]

if you have any problem with P-state of GTX 1080/1070... it works when you open a 3D model in 3D Paint .... then its always on full power  (that nvidia driver inspector didnt work for me)

[GOTSpectrum]

On 8/9/2019 at 6:43 PM, palespartan said:

Best way of announcing this. lol.

I'm glad that you liked the announcement method haha 

[GOTSpectrum]

On 8/18/2019 at 8:55 PM, yaboistar said:

folks, a serious question here. should we write a new folding guide to match modern hardware? times have moved on since the old one was written. NACL is dead, a new generation of GPU's is out and there are some common slot settings that i can't see covered in the old guide.

I've been wanting too, but it seems like the job for the folding team leader and alas, that is not I

[Kitsep]

Out of curiosity, how does someone donate to the folding month prize pool? Are physical donations (for example, a graphics card or CPU) allowed? On lunch at work, didn't have too much time to go searching through the forums so I apologize if I missed a post with those details! 

[GOTSpectrum]

2 minutes ago, Kitsep said:

Out of curiosity, how does someone donate to the folding month prize pool? Are physical donations (for example, a graphics card or CPU) allowed? On lunch at work, didn't have too much time to go searching through the forums so I apologize if I missed a post with those details! 

If it's digital send me the code, if not then you will have to be responsible for covering all postage, packing and any customs fees induced. 

[Kitsep]

Awesome, thanks! I'll check with the rest of my mini-team (lowspecactionsquad) and see if we can donate something, be it steam keys, gift cards or otherwise

[RollinLower]

500M boys!

[Spotty]

5 minutes ago, RollinLower said:

500M boys!

Congratulations!

[RollinLower]

now that i maxed out the F@H badge i won;t be doing 24/7 folding on my Titan anymore. 

i just recieved a note from the power company about how my bills are gonna be much, much higher. so power is definitely not something i have plenty of anymore.

that, and i recently bought some pretty power hungry hardware for BOINC aswell. so yeah, just the 1070 and my vouwfabriek server for now.

untill i can get my hands on some more power efficient hardware that is!

[justpoet]

22 minutes ago, RollinLower said:

now that i maxed out the F@H badge i won;t be doing 24/7 folding on my Titan anymore. 

i just recieved a note from the power company about how my bills are gonna be much, much higher. so power is definitely not something i have plenty of anymore.

that, and i recently bought some pretty power hungry hardware for BOINC aswell. so yeah, just the 1070 and my vouwfabriek server for now.

untill i can get my hands on some more power efficient hardware that is!

Somebody make the Lower one another tier of badge to shoot for!