GPU, didn't expect that
Moderators: Site Moderators, FAHC Science Team
GPU, didn't expect that
Hi Folding colleagues,
I've been running the folding client, mainly on dedicated hardware, since around 2010. Originally on three PCs running XP. Various progressions from there and now running three HP Proliant rack servers and an HP Z600 workstation. Everything runs Ubuntu 18.04 (Desktop and server). I remote manage the servers from the desktop control panel which works really well. I generally run 10 slots across this hardware with no less than 8 logical cores per slot. Generally a job can run from between under an hour, right out to a day.
While I have some 80 cores available, I reasoned that this isn't a patch on the GPU approach. Not being a gaming person the need for a high performance graphics module isn't there. However, recently I realised that the Nvidia Quadro K600 card in my workstation, while not the most powerful by any means, was taskable. Yesterday I fiddled around to install the latest Nvidia driver and OpenCL, and without too much difficulty I had the GPU available. Additional fiddling to get manual control of the cooling and we were running. Never done this before so I was pleased with this result.
So, the reason for this post is that I was expecting the GPU to be picking up jobs and the resulting progress to be burning past the CPUs. But, on the contrary, the GPU task is forecast to run for two days and it is crawling along. I suspect that the GPU work package is a lot larger, which is what I wanted to confirm. The Nvidia control panel is showing 100% commitment by the 192 CUDA cores, so it seems to be doing what is required of it. The CPU clusters are tearing through jobs in the meantime.
Does this sound right?
Many thanks, Dave
I've been running the folding client, mainly on dedicated hardware, since around 2010. Originally on three PCs running XP. Various progressions from there and now running three HP Proliant rack servers and an HP Z600 workstation. Everything runs Ubuntu 18.04 (Desktop and server). I remote manage the servers from the desktop control panel which works really well. I generally run 10 slots across this hardware with no less than 8 logical cores per slot. Generally a job can run from between under an hour, right out to a day.
While I have some 80 cores available, I reasoned that this isn't a patch on the GPU approach. Not being a gaming person the need for a high performance graphics module isn't there. However, recently I realised that the Nvidia Quadro K600 card in my workstation, while not the most powerful by any means, was taskable. Yesterday I fiddled around to install the latest Nvidia driver and OpenCL, and without too much difficulty I had the GPU available. Additional fiddling to get manual control of the cooling and we were running. Never done this before so I was pleased with this result.
So, the reason for this post is that I was expecting the GPU to be picking up jobs and the resulting progress to be burning past the CPUs. But, on the contrary, the GPU task is forecast to run for two days and it is crawling along. I suspect that the GPU work package is a lot larger, which is what I wanted to confirm. The Nvidia control panel is showing 100% commitment by the 192 CUDA cores, so it seems to be doing what is required of it. The CPU clusters are tearing through jobs in the meantime.
Does this sound right?
Many thanks, Dave
-
- Site Moderator
- Posts: 6986
- Joined: Wed Dec 23, 2009 9:33 am
- Hardware configuration: V7.6.21 -> Multi-purpose 24/7
Windows 10 64-bit
CPU:2/3/4/6 -> Intel i7-6700K
GPU:1 -> Nvidia GTX 1080 Ti
§
Retired:
2x Nvidia GTX 1070
Nvidia GTX 675M
Nvidia GTX 660 Ti
Nvidia GTX 650 SC
Nvidia GTX 260 896 MB SOC
Nvidia 9600GT 1 GB OC
Nvidia 9500M GS
Nvidia 8800GTS 320 MB
Intel Core i7-860
Intel Core i7-3840QM
Intel i3-3240
Intel Core 2 Duo E8200
Intel Core 2 Duo E6550
Intel Core 2 Duo T8300
Intel Pentium E5500
Intel Pentium E5400 - Location: Land Of The Long White Cloud
- Contact:
Re: GPU, didn't expect that
Welcome to the F@H Forum Darmain,
It would be great if you can post the log file which has the system configuration so we can see how the client is setup.
In the meantime, I can make some general recommendations:
1) Are you using a passkey? If not, have a read of this and think about it: https://foldingathome.org/support/faq/points/passkey/
2) I would suggest that instead of running a CPU slot with 8 CPUs, up that number to 12, i.e. 12 CPUs per CPU Slot. This is the combination that has the most number of WUs available. Other numbers are 16, 24, 32 & 64. However, there might be some issues getting WUs for 32/64 CPUs.
3) Your GPU is 7 years old and is technically supported by F@H. However, given it's age, it will take a wee while for it to fold and it is useful as long as it is returned before the Timeout date.
It would be great if you can post the log file which has the system configuration so we can see how the client is setup.
In the meantime, I can make some general recommendations:
1) Are you using a passkey? If not, have a read of this and think about it: https://foldingathome.org/support/faq/points/passkey/
2) I would suggest that instead of running a CPU slot with 8 CPUs, up that number to 12, i.e. 12 CPUs per CPU Slot. This is the combination that has the most number of WUs available. Other numbers are 16, 24, 32 & 64. However, there might be some issues getting WUs for 32/64 CPUs.
3) Your GPU is 7 years old and is technically supported by F@H. However, given it's age, it will take a wee while for it to fold and it is useful as long as it is returned before the Timeout date.
ETA:
Now ↞ Very Soon ↔ Soon ↔ Soon-ish ↔ Not Soon ↠ End Of Time
Welcome To The F@H Support Forum Ӂ Troubleshooting Bad WUs Ӂ Troubleshooting Server Connectivity Issues
Now ↞ Very Soon ↔ Soon ↔ Soon-ish ↔ Not Soon ↠ End Of Time
Welcome To The F@H Support Forum Ӂ Troubleshooting Bad WUs Ӂ Troubleshooting Server Connectivity Issues
Re: GPU, didn't expect that
Hi PantherX,PantherX wrote:Welcome to the F@H Forum Darmain,
It would be great if you can post the log file which has the system configuration so we can see how the client is setup.
In the meantime, I can make some general recommendations:
1) Are you using a passkey? If not, have a read of this and think about it: https://foldingathome.org/support/faq/points/passkey/
2) I would suggest that instead of running a CPU slot with 8 CPUs, up that number to 12, i.e. 12 CPUs per CPU Slot. This is the combination that has the most number of WUs available. Other numbers are 16, 24, 32 & 64. However, there might be some issues getting WUs for 32/64 CPUs.
3) Your GPU is 7 years old and is technically supported by F@H. However, given it's age, it will take a wee while for it to fold and it is useful as long as it is returned before the Timeout date.
Many thanks for your response.
1). Yes, I am using a passkey.
2). The 8 core clusters are due to the architecture of the machines. Some have 16 cares, other have 24 cores. I can obviously run the machines on a single slot each. I gather that would be a more profitable approach, both from the folding throughput and the points perspective (?).
3). Yes, the GPU is old I agree. I thought it worth a try to see if I could make it work for us. If it is going to fail to return in time then that is a disaster and I'll pull the plug on it.
Many thanks, Dave
Re: GPU, didn't expect that
So, to explore the option, what would be a good graphics card to choose for Folding, without going cost stupid?
-
- Site Moderator
- Posts: 6986
- Joined: Wed Dec 23, 2009 9:33 am
- Hardware configuration: V7.6.21 -> Multi-purpose 24/7
Windows 10 64-bit
CPU:2/3/4/6 -> Intel i7-6700K
GPU:1 -> Nvidia GTX 1080 Ti
§
Retired:
2x Nvidia GTX 1070
Nvidia GTX 675M
Nvidia GTX 660 Ti
Nvidia GTX 650 SC
Nvidia GTX 260 896 MB SOC
Nvidia 9600GT 1 GB OC
Nvidia 9500M GS
Nvidia 8800GTS 320 MB
Intel Core i7-860
Intel Core i7-3840QM
Intel i3-3240
Intel Core 2 Duo E8200
Intel Core 2 Duo E6550
Intel Core 2 Duo T8300
Intel Pentium E5500
Intel Pentium E5400 - Location: Land Of The Long White Cloud
- Contact:
Re: GPU, didn't expect that
Yep, that's correct. You will be able to do more science hence the increase in PPD. I do know that up to 12 CPUs will have the maximum WUs available. However, there should be a decent amount of WUs for 16 and 24 CPUs. You can experiment and see what works best for your setup and fold with it.Darmain wrote:...Some have 16 cares, other have 24 cores. I can obviously run the machines on a single slot each. I gather that would be a more profitable approach, both from the folding throughput and the points perspective (?)...
Since Linux has a bit of speed advantage over Windows, I do think that your GPU while slow, will make it before the Timeout date.Darmain wrote:...I thought it worth a try to see if I could make it work for us. If it is going to fail to return in time then that is a disaster and I'll pull the plug on it...
ETA:
Now ↞ Very Soon ↔ Soon ↔ Soon-ish ↔ Not Soon ↠ End Of Time
Welcome To The F@H Support Forum Ӂ Troubleshooting Bad WUs Ӂ Troubleshooting Server Connectivity Issues
Now ↞ Very Soon ↔ Soon ↔ Soon-ish ↔ Not Soon ↠ End Of Time
Welcome To The F@H Support Forum Ӂ Troubleshooting Bad WUs Ӂ Troubleshooting Server Connectivity Issues
-
- Site Moderator
- Posts: 6986
- Joined: Wed Dec 23, 2009 9:33 am
- Hardware configuration: V7.6.21 -> Multi-purpose 24/7
Windows 10 64-bit
CPU:2/3/4/6 -> Intel i7-6700K
GPU:1 -> Nvidia GTX 1080 Ti
§
Retired:
2x Nvidia GTX 1070
Nvidia GTX 675M
Nvidia GTX 660 Ti
Nvidia GTX 650 SC
Nvidia GTX 260 896 MB SOC
Nvidia 9600GT 1 GB OC
Nvidia 9500M GS
Nvidia 8800GTS 320 MB
Intel Core i7-860
Intel Core i7-3840QM
Intel i3-3240
Intel Core 2 Duo E8200
Intel Core 2 Duo E6550
Intel Core 2 Duo T8300
Intel Pentium E5500
Intel Pentium E5400 - Location: Land Of The Long White Cloud
- Contact:
Re: GPU, didn't expect that
Great timing as there's a recent topic that has price and rough figures for you: viewtopic.php?f=38&t=34240Darmain wrote:So, to explore the option, what would be a good graphics card to choose for Folding, without going cost stupid?
In short, you can spend a few hundred bucks and get something decent. Of course, you can go all in and get the Nvidia GTX 2080 Ti... there are heaps of options so hopefully, you can find something that suits your needs
ETA:
Now ↞ Very Soon ↔ Soon ↔ Soon-ish ↔ Not Soon ↠ End Of Time
Welcome To The F@H Support Forum Ӂ Troubleshooting Bad WUs Ӂ Troubleshooting Server Connectivity Issues
Now ↞ Very Soon ↔ Soon ↔ Soon-ish ↔ Not Soon ↠ End Of Time
Welcome To The F@H Support Forum Ӂ Troubleshooting Bad WUs Ӂ Troubleshooting Server Connectivity Issues
-
- Posts: 1996
- Joined: Sun Mar 22, 2020 5:52 pm
- Hardware configuration: 1: 2x Xeon E5-2697v3@2.60GHz, 512GB DDR4 LRDIMM, SSD Raid, Win10 Ent 20H2, Quadro K420 1GB, FAH 7.6.21
2: Xeon E3-1505Mv5@2.80GHz, 32GB DDR4, NVME, Win10 Pro 20H2, Quadro M1000M 2GB, FAH 7.6.21 (actually have two of these)
3: i7-960@3.20GHz, 12GB DDR3, SSD, Win10 Pro 20H2, GTX 750Ti 2GB, GTX 1080Ti 11GB, FAH 7.6.21 - Location: UK
Re: GPU, didn't expect that
I run 24 and 32 core slots (until recently client-type=advanced) which are kept busy 24/7 (even over last few weeks) so larger core counts are at the moment still being "fed" and yes they do return the WUs quicker so faster science and higher points … My simplistic was of rationalising this to myself is that one 16core slot can process two WUs in the same time two 8core slots can process one WU each - but the 16core WUs have each been returned in half the time so are better for the science and accrue higher points - I am sure it isn't that simple but the basic idea is in the right area. There are some Projects that have core slot size limitations for their WUs - recently a few have been <=12 but I believe there are also sometimes have a minimum core count applied.
2x Xeon E5-2697v3, 512GB DDR4 LRDIMM, SSD Raid, W10-Ent, Quadro K420
Xeon E3-1505Mv5, 32GB DDR4, NVME, W10-Pro, Quadro M1000M
i7-960, 12GB DDR3, SSD, W10-Pro, GTX1080Ti
i9-10850K, 64GB DDR4, NVME, W11-Pro, RTX3070
(Green/Bold = Active)
Xeon E3-1505Mv5, 32GB DDR4, NVME, W10-Pro, Quadro M1000M
i7-960, 12GB DDR3, SSD, W10-Pro, GTX1080Ti
i9-10850K, 64GB DDR4, NVME, W11-Pro, RTX3070
(Green/Bold = Active)
Re: GPU, didn't expect that
Thanks Guys. So I am re configuring as WU finish. I'll fix the slots at 12 threads and see how it goes.
Re: GPU, didn't expect that
I had a look at some of these, starting with the best one. Blimey, I knew the cost of Graphics cards was inflated due to Crypocurrency mining but I could buy a boat for that!! I will also have to look at my workstation power supply to see what its rated at. They's a bit thirsty by looks of things. Not dismissing the idea, just need a bit of due diligence.PantherX wrote:Great timing as there's a recent topic that has price and rough figures for you: viewtopic.php?f=38&t=34240Darmain wrote:So, to explore the option, what would be a good graphics card to choose for Folding, without going cost stupid?
In short, you can spend a few hundred bucks and get something decent. Of course, you can go all in and get the Nvidia GTX 2080 Ti... there are heaps of options so hopefully, you can find something that suits your needs
Many thanks, Dave
Re: GPU, didn't expect that
When you added your GPU slot, did you set aside a CPU core for it? It sounded like you had all your CPU cores already assigned to CPU slots; the GPU slot needs one to manage the GPU. If all the CPU cores are assigned to CPU slots, then the GPU slot is competing with the CPU slots for CPU time, which could affect its performance.
-
- Site Admin
- Posts: 7937
- Joined: Tue Apr 21, 2009 4:41 pm
- Hardware configuration: Mac Pro 2.8 quad 12 GB smp4
MacBook Pro 2.9 i7 8 GB smp2 - Location: W. MA
Re: GPU, didn't expect that
Depending on the reason the K600 was configured for you workstation, power usage, cost etc., something in the mid-range such as a GTX 1050 would be enough increase in processing power, and still stay within both power and money limits.
iMac 2.8 i7 12 GB smp8, Mac Pro 2.8 quad 12 GB smp6
MacBook Pro 2.9 i7 8 GB smp3
Re: GPU, didn't expect that
My folding computer has an Intel 14/28 processor and the automatic configuration set up 27 slots for the cpu and 1 for the gpu (RTX 2070). In light of what you said, would I be better off by changing that default configuration?PantherX wrote:Welcome to the F@H Forum Darmain,
2) I would suggest that instead of running a CPU slot with 8 CPUs, up that number to 12, i.e. 12 CPUs per CPU Slot. This is the combination that has the most number of WUs available. Other numbers are 16, 24, 32 & 64. However, there might be some issues getting WUs for 32/64 CPUs.
-
- Posts: 2522
- Joined: Mon Feb 16, 2009 4:12 am
- Location: Greenwood MS USA
Re: GPU, didn't expect that
I would go 24 and then 3, but that is me.
Tsar of all the Rushers
I tried to remain childlike, all I achieved was childish.
A friend to those who want no friends
I tried to remain childlike, all I achieved was childish.
A friend to those who want no friends