Folding Forum

This project is a test of a novel means to significantly accelearate Folding@home Molecular Dynamics calculations by 100x without any additional increase in hardware. Thus, this method has great promise to push Folding@home way beyond what it can do now, although there is much work to do to test it to make sure that the results are scientifically valid and useful.

These are interesting Uniprocessor projects that actually are very lucrative. I've been assigned a few of these on the Pent 4/HT Uniprocessor, but none on the Toshiba laptop. On uniprocessor assignments, the Toshiba actually gets better performance than the Pent 4/HT on the same projects (~115 ppd for the toshiba, ~85 ppd for the HP), even though the HP has a much higher clock and has HT. More than 90% of points on the HP, though, come from the ATI 4670 AGP GPU.

Question on these assignments. Is there something in the config parameters for the Toshiba that keeps if from getting any of the p70xx WUs?

HP Uniprocessor P4 HT system - This IS assigned 70xx WU.

17:43:50:******************************* System ********************************
17:43:50: OS: Microsoft Windows XP Home Edition
17:43:50: CPU: Intel(R) Pentium(R) 4 CPU 3.20GHz
17:43:50: CPU ID: GenuineIntel Family 15 Model 2 Stepping 9
17:43:50: CPUs: 2
17:43:50: Memory: 2.00GiB
17:43:50: Free Memory: 1.03GiB
17:43:50: Threads: WINDOWS_THREADS
17:43:50: GPUs: 2
17:43:50: GPU 0: ATI:2 RV730 Pro AGP [Radeon HD 4600 Series]
17:43:50: GPU 1: RV710/730
17:43:50: CUDA: Not detected
17:43:50: On Battery: false
17:43:50: UTC offset: -7
17:43:50: PID: 4780
17:43:50: CWD: C:/Documents and Settings/All Users/Application Data/FAHClient
17:43:50:Win32 Service: false

Toshiba Pent M Uniprocessor - Has NOT been assigned any 70xx WUs.
16:49:19:******************************* System ********************************
16:49:19: OS: Microsoft Windows XP Home Edition
16:49:19: CPU: Intel(R) Pentium(R) M processor 1.73GHz
16:49:19: CPU ID: GenuineIntel Family 6 Model 13 Stepping 8
16:49:19: CPUs: 1
16:49:19: Memory: 1.99GiB
16:49:19: Free Memory: 1.17GiB
16:49:19: Threads: WINDOWS_THREADS
16:49:19: GPUs: 0
16:49:19: CUDA: Not detected
16:49:19: On Battery: false
16:49:19: UTC offset: -7
16:49:19: PID: 1456
16:49:19: CWD: C:/Documents and Settings/Admin/Application Data/FAHClient
16:49:19:Win32 Service: false

I've noticed some strange patterns of assignments on my uniprocessor clients, too, but I can't explain them either. Those clients which to get these projects are still getting a variety of assignments, not anything consistent, though.

The Gromacs A4 core (which is used in these projects) requires SSE2 support. I don't know if this is the reason your Toshiba Pent M uniprocessor has not been given any of these projects.

@Izaguirre. Thanks for the note.

Both the laptop (Pentium M) and the desktop (Pentium 4/HT single core) support SSE2.

And, both have <extra-core-args v='-forceasm'/> on to ensure that SSE2 is enabled.

The laptop, but not the desktop, has <client-type v='advanced'/>. the Projects 70xx were just released, I think, to FAH - not advanced, and not Beta.

I just saw the quotes below in another thread (Re: Flag to stop big bigadv on 12 core machines.) That led me to think whether v7 saw 1.99 GiB on the Toshiba laptop, but saw 2.00 GiB on the HP uniprocessor, and that difference is preventing the Toshiba from being assigned Projects 7000-7028. I installed "2 GB" in both computers aftermarket.

So - a couple of questions:
a. Is it likely that these projects do, indeed, have a 2.0 GiB threshold?
b. Is there any way without contravening the FAH "Best Practices" to add a flag into either the V7 GUI or into the config xml file to tell it the Toshiba has 2.0 GiB?

Grandpa_01 wrote:

the animal wrote:Setting your client to report 1.99GB RAM should do it, since I believe these projects require 2GB.

Thanks I did not think of that 1 it should work.

"Toshiba Laptop"
16:49:19:******************************* System ********************************
16:49:19: OS: Microsoft Windows XP Home Edition
16:49:19: CPU: Intel(R) Pentium(R) M processor 1.73GHz
16:49:19: CPU ID: GenuineIntel Family 6 Model 13 Stepping 8
16:49:19: CPUs: 1
16:49:19: Memory: 1.99GiB
16:49:19: Free Memory: 1.17GiB
16:49:19: Threads: WINDOWS_THREADS
16:49:19: GPUs: 0
16:49:19: CUDA: Not detected
16:49:19: On Battery: false
16:49:19: UTC offset: -7
16:49:19: PID: 1456
16:49:19: CWD: C:/Documents and Settings/Admin/Application Data/FAHClient
16:49:19:Win32 Service: false

"HP Uniprocessor"
17:43:50:******************************* System ********************************
17:43:50: OS: Microsoft Windows XP Home Edition
17:43:50: CPU: Intel(R) Pentium(R) 4 CPU 3.20GHz
17:43:50: CPU ID: GenuineIntel Family 15 Model 2 Stepping 9
17:43:50: CPUs: 2
17:43:50: Memory: 2.00GiB
17:43:50: Free Memory: 1.03GiB
17:43:50: Threads: WINDOWS_THREADS
17:43:50: GPUs: 2
17:43:50: GPU 0: ATI:2 RV730 Pro AGP [Radeon HD 4600 Series]
17:43:50: GPU 1: RV710/730
17:43:50: CUDA: Not detected
17:43:50: On Battery: false
17:43:50: UTC offset: -7
17:43:50: PID: 4780
17:43:50: CWD: C:/Documents and Settings/All Users/Application Data/FAHClient
17:43:50:Win32 Service: false

GreyWhiskers wrote:So - a couple of questions:
a. Is it likely that these projects do, indeed, have a 2.0 GiB threshold?
b. Is there any way without contravening the FAH "Best Practices" to add a flag into either the V7 GUI or into the config xml file to tell it the Toshiba has 2.0 GiB?

A: That isn't easy to answer. We used to be able to tell some things like that about a lot of projects, but that method has been broken for a while now and getting something like that fixed is behind a lot of more critical things.

B. If you want to reduce the RAM that the server thinks you have to exclude some WUs, I'd say that's within the OK category of best practices. If you want to increase the amount of RAM that the server thinks you have to include some other WUs, that may not be possible, and if it does happen to be possible, it would fall squarely in violation to the best practices policies.

These are my personal interpretations of an area that Best Practices may not be 100% clear and thus are subject to reinterpretation or to explicit clarification in a revision to the BP document.

To increase 1.99 to more than 2.00, you could always buy more RAM as long as the mb allows it. You might be able to do something about it if you can switch from on-board video using shared RAM to a video board with it's own RAM. I suppose 2.00 GB minus N K for a video buffer might be called 1.99, where N is a small number (depending, of course, whether you're really talking about GB or GiB.