Project 762x - Fermi GPU advmethods

[GreyWhiskers]

Since they 7620/7621 were released on Aug 18 to general FAH/advmethods, and then since Dec 8 when 7622 was released, I have gotten nothing but these WUs, with a handful of exceptions of "regular" core 15 Fermi projects.

Starting yesterday, 2012/01/19 3:10AM GMT, I've gotten none of these on advmethods, but the "regular" 680x WUs that were the staple before.

Has the science been accomplished for these WUs? Or, is it just a pause to get the servers organized?

I've been getting a steady stream of WUs, so there is nothing like not having work to do. Just curious when something cuts off so abruptly like this.

[bruce]

I've seen no announcements of a change so my comments are general in nature, not necessarily related to those projects. While it's possible that you'll get an answer from the project owner, that isn't seen as a problem that needs to be solved so my guess is that you won't. in any case, I'll inject half an answer. If no more information is added later, it's better to get half an answer than no answer -- and they still can reply with the specifics if they want to.

Projects shift around due to all sorts of changes. A server may go down or a new server may be provisioned. Project priority may be adjusted whenever it makes sense to the research (Maybe they didn't intend 762x to run exclusively and 680x has gotten behind.) Projects may move from Adv to Fah (i.e.-"normal") causing an abrupt shift in effective priority for clients configured like yours) ... and projects do get competed (though not usually when in Adv status).

[GreyWhiskers]

Thanks, Bruce. I just checked Psummary, and the 762x projects aren't there any more. These were introduced with a great fanfare [see below], both because of the new science, as well as being a huge load on the Fermi GPUs. Many people needed to down-clock their GPUs, and watch their temps and stability with these. These projects seemed to be a good match for the GTX560Ti card, since I at least was able to take them in stride - not worry about the overclock, and get a slight ppd advance over the P68xx WUs that were the staple fare for the Fermi earlier in the year.

Essentially, I was looking to see if the researchers had any reportable results for these demanding projects we had been folding for so long.

I appreciate your response.

... new GPU projects: 7620 & 7621
by tjlane » Wed Aug 03, 2011 7:45 pm

... we've been gearing up here at Stanford to release some new GPU projects that will be larger and more computationally intensive than previous released WUs. The projects will be p7620 and p7621, and represent another advance in science on our end. We hope (and expect) everything will run smoothly, but I wanted to give people a heads up that if you get one of these units, your GPU will be cranking for a while. I've detailed some info about the projects below:

PROJECTS 7620 and 7621 - CheY Folding
For nVidia Fermi GPUs (GPU3) on Win32 machines
Points = 5187
timeout = 12.6
deadline = 17.9
atoms = 1980

Additionally, these WUs will require a new version (v2.20) of Core15, which should be automatically downloaded as needed.

These projects aim to study the folding of CheY, which, if successful, will be the largest protein folded in silico. CheY is well-studied experimentally, and exhibits many interesting scientific features. Due to it's wide study in the lab, we hope that it will provide a test case for FAH methodology, and give us good ideas about how to improve our techniques. This is an exciting project, since it represents something that is really only possible on FAH combined with the power of GPUs.

...

[Edited out some extraneous material in the quote]

[Napoleon]

Just a guess since I can't remember how the gens went, but I presume P762x projects got very popular among SMP+GPU folders with GTX460 cards or better. Stable despite advanced status, about the same PPD and very low CPU overhead, urging people to opt in for e.g smp:8 + P762x instead of smp:6 or smp:7 + P680x.

[Jonazz]

I know they were still active (on the psummary) wednesday. I guess something went wrong or the projects are complete (bt that would be very soon, though).

[derrickmcc]

The last 762x WU I received was on 17/01/12, from WS 171.64.65.105
That WS is still shown as STATUS = full, but is not being assigned any work ( % ASS G is blank) so is the problem with the AS?

Code: Select all

19:05:56:WU01:FS00:Connecting to assign-GPU.stanford.edu:80
19:05:57:WU01:FS00:News: Welcome to Folding@Home
19:05:57:WU01:FS00:Assigned to work server 171.64.65.105
19:05:57:WU01:FS00:Requesting new work unit for slot 00: RUNNING gpu:0:"GF104 [GeForce GTX 460]" from 171.64.65.105
19:05:57:WU01:FS00:Connecting to 171.64.65.105:8080
19:05:57:WU01:FS00:Downloading 122.16KiB
19:05:59:WU01:FS00:Download complete
19:05:59:WU01:FS00:Received Unit: id:01 state:DOWNLOAD error:OK project:7622 run:12 clone:0 gen:33 core:0x15 unit:0x00000031664f2dd14edd56c226fa216b
19:12:59:WU01:FS00:Starting
19:12:59:WU01:FS00:Running FahCore: "C:\Program Files\FAHClient/FAHCoreWrapper.exe" "C:/Documents and Settings/All Users/Application Data/FAHClient/cores/www.stanford.edu/~pande/Win32/x86/NVIDIA/Fermi/beta/Core_15.fah/FahCore_15.exe" -dir 01 -suffix 01 -version 701 -checkpoint 15 -gpu 0 -service
19:12:59:WU01:FS00:Started FahCore on PID 5584
19:12:59:WU01:FS00:Core PID:6420
19:12:59:WU01:FS00:FahCore 0x15 started
19:12:59:WU01:FS00:0x15:
19:12:59:WU01:FS00:0x15:*------------------------------*
19:12:59:WU01:FS00:0x15:Folding@Home GPU Core
19:12:59:WU01:FS00:0x15:Version                2.22 (Thu Dec 8 17:08:05 PST 2011)
19:12:59:WU01:FS00:0x15:Build host             SimbiosNvdWin7
19:12:59:WU01:FS00:0x15:Board Type             NVIDIA/CUDA
19:12:59:WU01:FS00:0x15:Core                   15
19:12:59:WU01:FS00:0x15:
19:12:59:WU01:FS00:0x15:Window's signal control handler registered.
19:12:59:WU01:FS00:0x15:Preparing to commence simulation
19:12:59:WU01:FS00:0x15:- Looking at optimizations...
19:12:59:WU01:FS00:0x15:DeleteFrameFiles: successfully deleted file=01/wudata_01.ckp
19:12:59:WU01:FS00:0x15:- Created dyn
19:12:59:WU01:FS00:0x15:- Files status OK
19:12:59:WU01:FS00:0x15:sizeof(CORE_PACKET_HDR) = 512 file=<>
19:12:59:WU01:FS00:0x15:- Expanded 124577 -> 501826 (decompressed 402.8 percent)
19:12:59:WU01:FS00:0x15:Called DecompressByteArray: compressed_data_size=124577 data_size=501826, decompressed_data_size=501826 diff=0
19:12:59:WU01:FS00:0x15:- Digital signature verified
19:12:59:WU01:FS00:0x15:
19:12:59:WU01:FS00:0x15:Project: 7622 (Run 12, Clone 0, Gen 33)
19:12:59:WU01:FS00:0x15:
19:12:59:WU01:FS00:0x15:Assembly optimizations on if available.
19:12:59:WU01:FS00:0x15:Entering M.D.
19:13:01:WU01:FS00:0x15:Tpr hash 01/wudata_01.tpr:  1897287577 1146570002 2345656361 3840980695 4156344086
19:13:01:WU01:FS00:0x15:GPU device info: vendor=0 device=0 name=<NA> match=0
19:13:01:WU01:FS00:0x15:Working on Protein
19:13:01:WU01:FS00:0x15:Client config unavailable.
19:13:01:WU01:FS00:0x15:Starting GUI Server
19:14:03:WU01:FS00:0x15:Setting checkpoint frequency: 400000
19:14:03:WU01:FS00:0x15:Completed         3 out of 40000000 steps (0%).

[shad0wfax]

My last 762x WU was downloaded 01/18/2012 at 19:11 hours UTC. I also, have had only 6800 through 6804 WUs since on my Fermi.

I'd like to get another intensive WU to continue long-term stability testing of my overclocked 580. My first 762X (stable) download was 12/27/2011 at 21:18 hours UTC, and I was hoping to get a solid month of the 762X WUs for 24/7 testing. Prior to that I'd been folding but had a NAN or two out of 30 WUs and so I tweaked my OC just a bit and haven't had one since. I'd really like to know if we're going to get more of these high-stress WUs again.

EDIT: As of 01/22/2012 20:53 hours UTC (or perhaps earlier) these WUs are no longer on the Folding@Home Projects Summary list. Perhaps the 6811 WU is intensive? It's still listed on the summaries page but it's an older WU.

[Jonazz]

I'm bettng the newer GPU projects that will be released will be bigger, like the 762x projects. They were testing new methods for bigger proteins so it's unlikely they'll be scaling back to smaller proteins. But you'll have to be patient for new projects.

[tjlane]

Hi All,

Sorry not to post about this, but to answer your questions - these projects had to be stopped temporarily to track down a bug in the work server code on the server that was assigning them. This bug didn't affect these projects, but instead the p763x series. I didn't post because this was supposed to be a minor stop (a day or so) and it turned into a weekend.

As of now (12:30 1.23.12), I am happy to report p762x are being assigned! Hopefully you'll receive one shortly.

Edit: Upon closer reading, I want to answer another question. These projects are nearing completion, but we have no scientific results to announce just yet! We are likely going to be introducing a new set of p762x projects that will deal with circular permutations of CheY, assuming all is looking good with the p762x we've gotten already. I'll be sure to report any science that comes out of these projects, since people seem to be curious - though note that analyzing the massive data sets FAH provides can take a few months .

TJ

[GreyWhiskers]

@tjlane: I appreciate your information. And, in fact, I did just pick up a 7622. Back to the hard folding!!
EDIT: I just reset my SMP setting to SMP-8 for my i7 2600k to take advantage of the much lower CPU usage of the 762x work units compared with other Fermi projects.

[shad0wfax]

Thanks for the update tjlane. I received a 7622 on 01/23/2012 at 10:00 hours UTC so all seems to be well. The 762X really do run much nicer than the 680X series do with SMP on all cores and/or other applications (like games) simultaneously. It's good to have them back and I'm looking forward to the new set when this set finishes.