FAH is like a relay race. Only once you completed a WU a new one can be created.
Therefore, the question is whether your contributing or slowing down the system with a slow CPU/GPU.
When you're folding with an modern average or faster CPU then you're contributing.
When your're folding with a CPU that is slower than the average then you're slowing down the system.
Just adding some CPU power to FAH does not increase the actual computing power. In a worst case scenario a i3 will get a WU while a 3950X is idle because of no WUs available.
CPU vs GPU point rewards
Moderators: Site Moderators, FAHC Science Team
-
- 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: CPU vs GPU point rewards
… but until such time as FAH decides it is elitist and only issue WUs by priority of speed of return then everyone is welcome to join and fold as long as their kit can complete WUs within the deadline??
Yes, at the moment there are more folders than the can be served, so yes, at the moment the fastest way to get the science done would be to prioritise WU issue to folders in order of speed of return of WUs … but next week, next month, next year, when the processes have caught up all the slower folders who have left after being disenfranchised by not being allowed to help and some of the much larger folding pools walk away having done their bit and helped "solve" CV will not be around there may be a whole load of science not being done because only the fastest felt welcome??
Under "normal circumstances" where all folders are able to be served work then X,000s of slower folders pushing through WUs within expiry which otherwise wouldn't get done is still contributing to the science as a whole … priority work might be pushed through (in a way I believe) by allocating high core numbers to certain CPU projects - I don't know if there is the same type of thing for GPUs.
Anyways … I'll leave it to FAH to decide who should/shouldn't be allowed to fold … Until they tell me then I (or anyone else using slower kit) am not welcome I am here to stay …. Sorry.
Yes, at the moment there are more folders than the can be served, so yes, at the moment the fastest way to get the science done would be to prioritise WU issue to folders in order of speed of return of WUs … but next week, next month, next year, when the processes have caught up all the slower folders who have left after being disenfranchised by not being allowed to help and some of the much larger folding pools walk away having done their bit and helped "solve" CV will not be around there may be a whole load of science not being done because only the fastest felt welcome??
Under "normal circumstances" where all folders are able to be served work then X,000s of slower folders pushing through WUs within expiry which otherwise wouldn't get done is still contributing to the science as a whole … priority work might be pushed through (in a way I believe) by allocating high core numbers to certain CPU projects - I don't know if there is the same type of thing for GPUs.
Anyways … I'll leave it to FAH to decide who should/shouldn't be allowed to fold … Until they tell me then I (or anyone else using slower kit) am not welcome I am here to stay …. Sorry.
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)
-
- Posts: 146
- Joined: Sun Jul 30, 2017 8:40 pm
Re: CPU vs GPU point rewards
I understand your points.
Bust most (new) users don't understand how FAH work. There are many comments implying that most users don't know how pausing or not to continue to finish a WU is affecting FAH in a bad way.
Donating CPU power to FAH has not the same effect as donating money to a charity. It can affect folding power in a bad way.
I don't want to discourage anyone I just want to give proper information.
Everyone should decide themselves.
I decided not to fold with my i5 4-thread-CPU as I think it affects folding in a bad way.
Many of us (including myself) fold with a GPU as well. You need one thread for the OS and one for the NV-GPU to not affect GPU folding. Would leave only 2 threads for CPU folding, which is really slow.
However I encourage everyone with a modern fast multi-core (meaning more than 4 cores) CPU to start and continue folding.
Cheers!
Bust most (new) users don't understand how FAH work. There are many comments implying that most users don't know how pausing or not to continue to finish a WU is affecting FAH in a bad way.
Donating CPU power to FAH has not the same effect as donating money to a charity. It can affect folding power in a bad way.
I don't want to discourage anyone I just want to give proper information.
Everyone should decide themselves.
I decided not to fold with my i5 4-thread-CPU as I think it affects folding in a bad way.
Many of us (including myself) fold with a GPU as well. You need one thread for the OS and one for the NV-GPU to not affect GPU folding. Would leave only 2 threads for CPU folding, which is really slow.
However I encourage everyone with a modern fast multi-core (meaning more than 4 cores) CPU to start and continue folding.
Cheers!
-
- 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: CPU vs GPU point rewards
Sorry, my previous may have been tainted by tiredness … and actually like you I am not folding under the current "overloaded" circumstances (other that a test or two) with my slower system or either of my GPUs (mine are actually worse/slower that my CPU slots).
When (hopefully not if) it gets back to the point where all folders are getting WUs on demand I may fire them up again as at that point even the slower throughput will be a net gain to the science.
When (hopefully not if) it gets back to the point where all folders are getting WUs on demand I may fire them up again as at that point even the slower throughput will be a net gain to the science.
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: CPU vs GPU point rewards
It's a sad fact that the science that FAH needs to do is partitioned into two classes of work: CPUs and GPUs. The same project cannot be done by the other hardware.
If a project has a relatively large number of atoms, it can be very efficient running on a GPU with a large number of shaders. If a protein has a more moderatel number of atoms, it will not be able to make effective use of all of the shaders and the throughput will drop if it happens to be assigned to your GPU. If a protein has a relatively small number of atoms, the productivity on ANY GPU will be quite poor, but that same protein can be done effectively on a GPU.
That also means that the projects being assigned to CPUs are much less complex that the ones that effectively "fill up" the capabilities of a powerful GPU. In each of those three categories, we're talking about assigning project with different scientific complexities. One hour of challenging work for a "big" GPUs is going to be worth more than one hour of less challenging work with either fewer shaders or with some of those shaders remaining idle. One hour of even less challenging work running inefficently on any GPU or running efficiently on a CPU will be benchmarked for lower baseline points, with or without considering the QRB.
If a project has a relatively large number of atoms, it can be very efficient running on a GPU with a large number of shaders. If a protein has a more moderatel number of atoms, it will not be able to make effective use of all of the shaders and the throughput will drop if it happens to be assigned to your GPU. If a protein has a relatively small number of atoms, the productivity on ANY GPU will be quite poor, but that same protein can be done effectively on a GPU.
That also means that the projects being assigned to CPUs are much less complex that the ones that effectively "fill up" the capabilities of a powerful GPU. In each of those three categories, we're talking about assigning project with different scientific complexities. One hour of challenging work for a "big" GPUs is going to be worth more than one hour of less challenging work with either fewer shaders or with some of those shaders remaining idle. One hour of even less challenging work running inefficently on any GPU or running efficiently on a CPU will be benchmarked for lower baseline points, with or without considering the QRB.
Posting FAH's log:
How to provide enough info to get helpful support.
How to provide enough info to get helpful support.
Re: CPU vs GPU point rewards
Please be careful with such statements. There are deadlines for a reason. As long as your C/GPU meets those deadlines you are contributing.foldinghomealone2 wrote:FAH is like a relay race. Only once you completed a WU a new one can be created.
Therefore, the question is whether your contributing or slowing down the system with a slow CPU/GPU.
When you're folding with an modern average or faster CPU then you're contributing.
When your're folding with a CPU that is slower than the average then you're slowing down the system.
Just adding some CPU power to FAH does not increase the actual computing power. In a worst case scenario a i3 will get a WU while a 3950X is idle because of no WUs available.
If the deadlines are not strict enough, I imagine those running a project will adjust them.
Re: CPU vs GPU point rewards
You can slow down FAH's system by letting your hardware sleep or by folding on very slow hardware, but you can't slow it down more than the Timeout. If you fail to return a WU you've downloaded before the timeout, FAH will reassign the same WU to someone else.
Yes, the deadlines are not strict enough, but there are good reasons why FAH doesn't like to unnecessarily duplicate WUs when it can be completed promptly by one person's hardware.
Yes, the deadlines are not strict enough, but there are good reasons why FAH doesn't like to unnecessarily duplicate WUs when it can be completed promptly by one person's hardware.
Posting FAH's log:
How to provide enough info to get helpful support.
How to provide enough info to get helpful support.
-
- Posts: 146
- Joined: Sun Jul 30, 2017 8:40 pm
Re: CPU vs GPU point rewards
I fully agree on the first part, but we have different definitions for 'promptly'.bruce wrote:Yes, the deadlines are not strict enough, but there are good reasons why FAH doesn't like to unnecessarily duplicate WUs when it can be completed promptly by one person's hardware.
On a Ryzen 2700 which I consider an average modern CPU the TPF for p13781 is < 1 min.
However, timeout is 5 days (===> TPF = 72 mins), deadline is 7 days.
Processing of a WU can be delayed by 5 days or by the factor 50 to 100 on really slow CPUs (or bad user behavior) compared to average CPUs.
(values taken from beta-forum for the project viewtopic.php?f=66&t=33278)