New core = significant production drop GPU

[chaosdsm]

Mod Note: Moved topic as it was not about a problem with a specific WU.

Running a GTX-460 first generation card with the full width 256bit memory bus - i.e. the "good" 460's... I noticed recently that on the 5187 point GPU work units my production suddenly took a nose-dive. After diggin into the details, I discovered that I'm now running a newer core & that seems to be the cause of the drop.

Previous results running core version 2.22 with card clocked at factory settings of 715MHz core / 1430MHz shaders / 1800MHz memory
p7623 = 7minutes 26seconds per frame for 10,048.4 ppd averaged over 4 completed work units.
p7624 = 7minutes 33seconds per frame for 9,893 ppd averaged over 5 completed work units.
p7625 = 7minutes 33seconds per frame for 9,893 ppd averaged over 4 completed work units.

Since updating to core version 2.25 & after overclocking the card to 800MHz core / 1600MHz shaders / 2000MHz memory at an attempt to compensate for lost production...
p7623 now gets 9minutes 2seconds per frame for 8,268ppd - average after 28 completed frames
p7625 at 715/1430/1800 clocks = 9minutes 21seconds per frame, at 800/1600/2000 clocks = 9minutes 10seconds per frame - GPU speeds changed half-way through this work unit.

[chaosdsm]

There's nothing else using GPU cycles. I shut down & restarted PC. Only change since the last round of 5187 point GPU work units is in the Core Version. Couple snippets from the 7623 wu logs:

Core Version 2.22

Code: Select all

15:01:17:WU02:FS00:Connecting to assign-GPU.stanford.edu:80
15:01:17:WU02:FS00:News: Welcome to Folding@Home
15:01:17:WU02:FS00:Assigned to work server 171.64.65.105
15:01:17:WU02:FS00:Requesting new work unit for slot 00: RUNNING gpu:0:"GF104 [GeForce GTX 460]" from 171.64.65.105
15:01:17:WU02:FS00:Connecting to 171.64.65.105:8080
15:01:18:WU02:FS00:Downloading 122.11KiB
15:01:18:WU02:FS00:Download complete
15:01:18:WU02:FS00:Received Unit: id:02 state:DOWNLOAD error:OK project:7623 run:546 clone:1 gen:3 core:0x15 unit:0x00000003664f2dd14fe4fb821487bb8f
15:01:18:WU02:FS00:Downloading project 7623 description
15:01:18:WU02:FS00:Connecting to fah-web.stanford.edu:80
15:01:19:WU02:FS00:Project 7623 description downloaded successfully
15:07:28:WU02:FS00:Starting
15:07:28:WU02:FS00:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/Users/chaosdsm/AppData/Roaming/FAHClient/cores/www.stanford.edu/~pande/Win32/AMD64/NVIDIA/Fermi/Core_15.fah/FahCore_15.exe -dir 02 -suffix 01 -version 701 -lifeline 4784 -checkpoint 15 -gpu 0
15:07:28:WU02:FS00:Started FahCore on PID 16936
15:07:28:WU02:FS00:Core PID:5872
15:07:28:WU02:FS00:FahCore 0x15 started
15:07:29:WU02:FS00:0x15:
15:07:29:WU02:FS00:0x15:*------------------------------*
15:07:29:WU02:FS00:0x15:Folding@Home GPU Core
15:07:29:WU02:FS00:0x15:Version                2.22 (Thu Dec 8 17:08:05 PST 2011)
15:07:29:WU02:FS00:0x15:Build host             SimbiosNvdWin7
15:07:29:WU02:FS00:0x15:Board Type             NVIDIA/CUDA
15:07:29:WU02:FS00:0x15:Core                   15
15:07:29:WU02:FS00:0x15:
15:07:29:WU02:FS00:0x15:Window's signal control handler registered.
15:07:29:WU02:FS00:0x15:Preparing to commence simulation
15:07:29:WU02:FS00:0x15:- Looking at optimizations...
15:07:29:WU02:FS00:0x15:DeleteFrameFiles: successfully deleted file=02/wudata_01.ckp
15:07:29:WU02:FS00:0x15:- Created dyn
15:07:29:WU02:FS00:0x15:- Files status OK
15:07:29:WU02:FS00:0x15:sizeof(CORE_PACKET_HDR) = 512 file=<>
15:07:29:WU02:FS00:0x15:- Expanded 124527 -> 501826 (decompressed 402.9 percent)
15:07:29:WU02:FS00:0x15:Called DecompressByteArray: compressed_data_size=124527 data_size=501826, decompressed_data_size=501826 diff=0
15:07:29:WU02:FS00:0x15:- Digital signature verified
15:07:29:WU02:FS00:0x15:
15:07:29:WU02:FS00:0x15:Project: 7623 (Run 546, Clone 1, Gen 3)
15:07:29:WU02:FS00:0x15:
15:07:29:WU02:FS00:0x15:Assembly optimizations on if available.
15:07:29:WU02:FS00:0x15:Entering M.D.
15:07:31:WU02:FS00:0x15:Tpr hash 02/wudata_01.tpr:  4143751295 4013527236 2356051826 2185125249 1283299892
15:07:31:WU02:FS00:0x15:GPU device info: vendor=0 device=0 name=<NA> match=0
15:07:31:WU02:FS00:0x15:Working on Protein
15:07:31:WU02:FS00:0x15:Client config unavailable.
15:07:31:WU02:FS00:0x15:Starting GUI Server
15:08:37:WU02:FS00:0x15:Setting checkpoint frequency: 400000
15:08:37:WU02:FS00:0x15:Completed         3 out of 40000000 steps (0%).
15:16:02:WU02:FS00:0x15:Completed    400000 out of 40000000 steps (1%).
15:23:29:WU02:FS00:0x15:Completed    800000 out of 40000000 steps (2%).
15:30:55:WU02:FS00:0x15:Completed   1200000 out of 40000000 steps (3%).
15:38:22:WU02:FS00:0x15:Completed   1600000 out of 40000000 steps (4%).
15:45:48:WU02:FS00:0x15:Completed   2000000 out of 40000000 steps (5%).
15:53:15:WU02:FS00:0x15:Completed   2400000 out of 40000000 steps (6%).
16:00:41:WU02:FS00:0x15:Completed   2800000 out of 40000000 steps (7%).
16:08:08:WU02:FS00:0x15:Completed   3200000 out of 40000000 steps (8%).
16:15:34:WU02:FS00:0x15:Completed   3600000 out of 40000000 steps (9%).
16:23:01:WU02:FS00:0x15:Completed   4000000 out of 40000000 steps (10%).

Core Version 2.25

Code: Select all

05:01:02:WU02:FS00:Connecting to assign-GPU.stanford.edu:80
05:01:02:WU02:FS00:News: Welcome to Folding@Home
05:01:02:WU02:FS00:Assigned to work server 171.64.65.105
05:01:02:WU02:FS00:Requesting new work unit for slot 00: RUNNING gpu:0:"GF104 [GeForce GTX 460]" from 171.64.65.105
05:01:02:WU02:FS00:Connecting to 171.64.65.105:8080
05:01:03:WU02:FS00:Downloading 122.14KiB
05:01:03:WU02:FS00:Download complete
05:01:03:WU02:FS00:Received Unit: id:02 state:DOWNLOAD error:OK project:7623 run:391 clone:0 gen:64 core:0x15 unit:0x0000006a664f2dd14fe4faa6954c2bf3
05:01:03:WU02:FS00:Downloading project 7623 description
05:01:03:WU02:FS00:Connecting to fah-web.stanford.edu:80
05:01:04:WU02:FS00:Project 7623 description downloaded successfully
05:33:49:WU02:FS00:Starting
05:33:49:WU02:FS00:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/Users/chaosdsm/AppData/Roaming/FAHClient/cores/www.stanford.edu/~pande/Win32/AMD64/NVIDIA/Fermi/Core_15.fah/FahCore_15.exe -dir 02 -suffix 01 -version 701 -lifeline 3880 -checkpoint 15 -gpu 0
05:33:49:WU02:FS00:Started FahCore on PID 1648
05:33:49:WU02:FS00:Core PID:20468
05:33:49:WU02:FS00:FahCore 0x15 started
05:33:49:WU02:FS00:0x15:
05:33:49:WU02:FS00:0x15:*------------------------------*
05:33:49:WU02:FS00:0x15:Folding@Home GPU Core
05:33:49:WU02:FS00:0x15:Version                2.25 (Wed May 9 17:03:01 EDT 2012)
05:33:49:WU02:FS00:0x15:Build host             AmoebaRemote
05:33:49:WU02:FS00:0x15:Board Type             NVIDIA/CUDA
05:33:49:WU02:FS00:0x15:Core                   15
05:33:49:WU02:FS00:0x15:
05:33:49:WU02:FS00:0x15:Window's signal control handler registered.
05:33:49:WU02:FS00:0x15:Preparing to commence simulation
05:33:49:WU02:FS00:0x15:- Looking at optimizations...
05:33:49:WU02:FS00:0x15:DeleteFrameFiles: successfully deleted file=02/wudata_01.ckp
05:33:49:WU02:FS00:0x15:- Created dyn
05:33:49:WU02:FS00:0x15:- Files status OK
05:33:49:WU02:FS00:0x15:sizeof(CORE_PACKET_HDR) = 512 file=<>
05:33:49:WU02:FS00:0x15:- Expanded 124558 -> 501826 (decompressed 402.8 percent)
05:33:49:WU02:FS00:0x15:Called DecompressByteArray: compressed_data_size=124558 data_size=501826, decompressed_data_size=501826 diff=0
05:33:49:WU02:FS00:0x15:- Digital signature verified
05:33:49:WU02:FS00:0x15:
05:33:49:WU02:FS00:0x15:Project: 7623 (Run 391, Clone 0, Gen 64)
05:33:49:WU02:FS00:0x15:
05:33:49:WU02:FS00:0x15:Assembly optimizations on if available.
05:33:49:WU02:FS00:0x15:Entering M.D.
05:33:51:WU02:FS00:0x15:Tpr hash 02/wudata_01.tpr:  2743259063 4003468159 1111672364 4065075476 3381836864
05:33:51:WU02:FS00:0x15:GPU device id=0
05:33:51:WU02:FS00:0x15:Working on Protein
05:33:51:WU02:FS00:0x15:Client config unavailable.
05:33:51:WU02:FS00:0x15:Starting GUI Server
05:34:53:WU02:FS00:0x15:Setting checkpoint frequency: 400000
05:34:53:WU02:FS00:0x15:Completed         3 out of 40000000 steps (0%).
05:43:55:WU02:FS00:0x15:Completed    400000 out of 40000000 steps (1%).
05:52:58:WU02:FS00:0x15:Completed    800000 out of 40000000 steps (2%).
06:02:00:WU02:FS00:0x15:Completed   1200000 out of 40000000 steps (3%).
06:11:03:WU02:FS00:0x15:Completed   1600000 out of 40000000 steps (4%).
06:20:06:WU02:FS00:0x15:Completed   2000000 out of 40000000 steps (5%).
06:29:09:WU02:FS00:0x15:Completed   2400000 out of 40000000 steps (6%).
06:38:12:WU02:FS00:0x15:Completed   2800000 out of 40000000 steps (7%).
06:47:15:WU02:FS00:0x15:Completed   3200000 out of 40000000 steps (8%).
06:56:18:WU02:FS00:0x15:Completed   3600000 out of 40000000 steps (9%).
07:05:21:WU02:FS00:0x15:Completed   4000000 out of 40000000 steps (10%).

[Nathan_P]

Core 2.25 has had several fermi optimisations removed , it is reported that they were removed to help get Kepler folding. There is no indication yet as to whether or not those optimistations will be put back

[7im]

I don't know where that reported "story" came from, but the only change disclosed by PG was an environment setting change. No other changes, scientific method or HW optimization-wize was addressed yet.

Yes, there is a performance change. We don't know why. But this is the core issued to all new client installs. Standard issue for now, until the next one comes.

[patonb]

Okay, I'm running 2.25, and my ppd only drops on certain units.
The 6 and 7 series are poor, the 8's are normal.

[Nathan_P]

I don't know where that reported "story" came from, but the only change disclosed by PG was an environment setting change. No other changes, scientific method or HW optimization-wize was addressed yet.

Yes, there is a performance change. We don't know why. But this is the core issued to all new client installs. Standard issue for now, until the next one comes.

Try here http://foldingforum.org/viewtopic.php?f ... 88#p228802 or

http://foldingforum.org/viewtopic.php?f ... 57#p227747, http://foldingforum.org/viewtopic.php?f ... 35#p221635, http://foldingforum.org/viewtopic.php?f ... 30#p227398.

I think that just about covers it properly, in the absence of a definitive statement i'll take the word of site admin and PG group members as the statement

[bruce]

Core 2.25 has had several fermi optimisations removed , it is reported that they were removed to help get Kepler folding. There is no indication yet as to whether or not those optimistations will be put back

Nothing I said implied that there was any connection between the Kepler changes and the optimization changes ... only that they happened to both happen between 2.22 and 2.25 ... or that the change in optimizations were specifically for fermi. I've seen no data to imply they're related nor any proof that they're unrelated. Can you reference your source? I simply do not know that your assumptions are valid or not.

[chaosdsm]

Okay, I'm running 2.25, and my ppd only drops on certain units.
The 6 and 7 series are poor, the 8's are normal.

Not that I don't belive you, but can you post logs to support this??? I'm seeing the same drop in 8xxx series GPU wu's as well.

Core Version 2.22 project 8054 - card clocked at 715/1430/1800 = 2minutes 36seconds per frame

Code: Select all

19:59:09:WU02:FS00:Connecting to assign-GPU.stanford.edu:80
19:59:09:WU02:FS00:News: Welcome to Folding@Home
19:59:09:WU02:FS00:Assigned to work server 171.67.108.143
19:59:09:WU02:FS00:Requesting new work unit for slot 00: RUNNING gpu:0:"GF104 [GeForce GTX 460]" from 171.67.108.143
19:59:09:WU02:FS00:Connecting to 171.67.108.143:8080
19:59:10:WU02:FS00:Downloading 59.30KiB
19:59:10:WU02:FS00:Download complete
19:59:10:WU02:FS00:Received Unit: id:02 state:DOWNLOAD error:OK project:8054 run:0 clone:730 gen:69 core:0x15 unit:0x0000005c6953ee2f50626ab5d285547d
20:01:12:WU02:FS00:Starting
20:01:12:WU02:FS00:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/Users/chaosdsm/AppData/Roaming/FAHClient/cores/www.stanford.edu/~pande/Win32/AMD64/NVIDIA/Fermi/Core_15.fah/FahCore_15.exe -dir 02 -suffix 01 -version 701 -lifeline 1092 -checkpoint 15 -gpu 0
20:01:12:WU02:FS00:Started FahCore on PID 7580
20:01:12:WU02:FS00:Core PID:6408
20:01:12:WU02:FS00:FahCore 0x15 started
20:01:13:WU02:FS00:0x15:
20:01:13:WU02:FS00:0x15:*------------------------------*
20:01:13:WU02:FS00:0x15:Folding@Home GPU Core
20:01:13:WU02:FS00:0x15:Version                2.22 (Thu Dec 8 17:08:05 PST 2011)
20:01:13:WU02:FS00:0x15:Build host             SimbiosNvdWin7
20:01:13:WU02:FS00:0x15:Board Type             NVIDIA/CUDA
20:01:13:WU02:FS00:0x15:Core                   15
20:01:13:WU02:FS00:0x15:
20:01:13:WU02:FS00:0x15:Window's signal control handler registered.
20:01:13:WU02:FS00:0x15:Preparing to commence simulation
20:01:13:WU02:FS00:0x15:- Looking at optimizations...
20:01:13:WU02:FS00:0x15:DeleteFrameFiles: successfully deleted file=02/wudata_01.ckp
20:01:13:WU02:FS00:0x15:- Created dyn
20:01:13:WU02:FS00:0x15:- Files status OK
20:01:13:WU02:FS00:0x15:sizeof(CORE_PACKET_HDR) = 512 file=<>
20:01:13:WU02:FS00:0x15:- Expanded 60213 -> 264278 (decompressed 438.9 percent)
20:01:13:WU02:FS00:0x15:Called DecompressByteArray: compressed_data_size=60213 data_size=264278, decompressed_data_size=264278 diff=0
20:01:13:WU02:FS00:0x15:- Digital signature verified
20:01:13:WU02:FS00:0x15:
20:01:13:WU02:FS00:0x15:Project: 8054 (Run 0, Clone 730, Gen 69)
20:01:13:WU02:FS00:0x15:
20:01:13:WU02:FS00:0x15:Assembly optimizations on if available.
20:01:13:WU02:FS00:0x15:Entering M.D.
20:01:14:WU02:FS00:0x15:Tpr hash 02/wudata_01.tpr:  3122237291 2525903205 4183294797 1330824829 1674821873
20:01:14:WU02:FS00:0x15:GPU device info: vendor=0 device=0 name=<NA> match=0
20:01:14:WU02:FS00:0x15:Working on Good ROcking Metal Altar for Chronical Sinners
20:01:14:WU02:FS00:0x15:Client config unavailable.
20:01:15:WU02:FS00:0x15:Starting GUI Server
20:02:16:WU02:FS00:0x15:Setting checkpoint frequency: 500000
20:02:16:WU02:FS00:0x15:Completed         3 out of 50000000 steps (0%).
20:04:51:WU02:FS00:0x15:Completed    500000 out of 50000000 steps (1%).
20:07:28:WU02:FS00:0x15:Completed   1000000 out of 50000000 steps (2%).
20:10:04:WU02:FS00:0x15:Completed   1500000 out of 50000000 steps (3%).
20:12:40:WU02:FS00:0x15:Completed   2000000 out of 50000000 steps (4%).
20:15:16:WU02:FS00:0x15:Completed   2500000 out of 50000000 steps (5%).
20:17:52:WU02:FS00:0x15:Completed   3000000 out of 50000000 steps (6%).
20:20:29:WU02:FS00:0x15:Completed   3500000 out of 50000000 steps (7%).
20:23:05:WU02:FS00:0x15:Completed   4000000 out of 50000000 steps (8%).
20:25:41:WU02:FS00:0x15:Completed   4500000 out of 50000000 steps (9%).
20:28:17:WU02:FS00:0x15:Completed   5000000 out of 50000000 steps (10%).

Core Version 2.25 project 8054 card clocked at 800/1600/2000 = 3minutes 10seconds per frame

Code: Select all

11:02:33:WU00:FS00:Connecting to assign-GPU.stanford.edu:80
11:02:33:WU00:FS00:News: Welcome to Folding@Home
11:02:33:WU00:FS00:Assigned to work server 171.67.108.143
11:02:33:WU00:FS00:Requesting new work unit for slot 00: RUNNING gpu:0:"GF104 [GeForce GTX 460]" from 171.67.108.143
11:02:33:WU00:FS00:Connecting to 171.67.108.143:8080
11:02:34:WU00:FS00:Downloading 59.51KiB
11:02:34:WU00:FS00:Download complete
11:02:34:WU00:FS00:Received Unit: id:00 state:DOWNLOAD error:OK project:8054 run:0 clone:444 gen:191 core:0x15 unit:0x000000e46953ee2f50626a9c76566687
11:11:50:WU00:FS00:Starting
11:11:50:WU00:FS00:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/Users/chaosdsm/AppData/Roaming/FAHClient/cores/www.stanford.edu/~pande/Win32/AMD64/NVIDIA/Fermi/Core_15.fah/FahCore_15.exe -dir 00 -suffix 01 -version 701 -lifeline 1556 -checkpoint 15 -gpu 0
11:11:50:WU00:FS00:Started FahCore on PID 3612
11:11:51:WU00:FS00:Core PID:5624
11:11:51:WU00:FS00:FahCore 0x15 started
11:11:51:WU00:FS00:0x15:
11:11:51:WU00:FS00:0x15:*------------------------------*
11:11:51:WU00:FS00:0x15:Folding@Home GPU Core
11:11:51:WU00:FS00:0x15:Version                2.25 (Wed May 9 17:03:01 EDT 2012)
11:11:51:WU00:FS00:0x15:Build host             AmoebaRemote
11:11:51:WU00:FS00:0x15:Board Type             NVIDIA/CUDA
11:11:51:WU00:FS00:0x15:Core                   15
11:11:51:WU00:FS00:0x15:
11:11:51:WU00:FS00:0x15:Window's signal control handler registered.
11:11:51:WU00:FS00:0x15:Preparing to commence simulation
11:11:51:WU00:FS00:0x15:- Looking at optimizations...
11:11:51:WU00:FS00:0x15:DeleteFrameFiles: successfully deleted file=00/wudata_01.ckp
11:11:51:WU00:FS00:0x15:- Created dyn
11:11:51:WU00:FS00:0x15:- Files status OK
11:11:51:WU00:FS00:0x15:sizeof(CORE_PACKET_HDR) = 512 file=<>
11:11:51:WU00:FS00:0x15:- Expanded 60425 -> 264278 (decompressed 437.3 percent)
11:11:51:WU00:FS00:0x15:Called DecompressByteArray: compressed_data_size=60425 data_size=264278, decompressed_data_size=264278 diff=0
11:11:51:WU00:FS00:0x15:- Digital signature verified
11:11:51:WU00:FS00:0x15:
11:11:51:WU00:FS00:0x15:Project: 8054 (Run 0, Clone 444, Gen 191)
11:11:51:WU00:FS00:0x15:
11:11:51:WU00:FS00:0x15:Assembly optimizations on if available.
11:11:51:WU00:FS00:0x15:Entering M.D.
11:11:53:WU00:FS00:0x15:Tpr hash 00/wudata_01.tpr:  4199784630 4020228428 2150077217 3382075114 612275809
11:11:53:WU00:FS00:0x15:GPU device id=0
11:11:53:WU00:FS00:0x15:Working on Good ROcking Metal Altar for Chronical Sinners
11:11:53:WU00:FS00:0x15:Client config unavailable.
11:11:53:WU00:FS00:0x15:Starting GUI Server
11:12:59:WU00:FS00:0x15:Setting checkpoint frequency: 500000
11:12:59:WU00:FS00:0x15:Completed         3 out of 50000000 steps (0%).
11:16:09:WU00:FS00:0x15:Completed    500000 out of 50000000 steps (1%).
11:19:19:WU00:FS00:0x15:Completed   1000000 out of 50000000 steps (2%).
11:22:29:WU00:FS00:0x15:Completed   1500000 out of 50000000 steps (3%).
11:25:39:WU00:FS00:0x15:Completed   2000000 out of 50000000 steps (4%).
11:28:49:WU00:FS00:0x15:Completed   2500000 out of 50000000 steps (5%).
11:31:59:WU00:FS00:0x15:Completed   3000000 out of 50000000 steps (6%).
11:35:09:WU00:FS00:0x15:Completed   3500000 out of 50000000 steps (7%).
11:38:20:WU00:FS00:0x15:Completed   4000000 out of 50000000 steps (8%).
11:41:30:WU00:FS00:0x15:Completed   4500000 out of 50000000 steps (9%).
11:44:40:WU00:FS00:0x15:Completed   5000000 out of 50000000 steps (10%).

I could see a 1 to 2 % decrease, butIMO, a new bench / point value with the new core is needed if decreases continue to be this significant.

[patonb]

Here's from hfm, I;m running smp -8 and its my daily/work syastem, i7 950, gtx 560ti evga fpb (850mhz clock.) Though I haven't run a gpu for awhile with 2.22 core, but ppd is what i had before.
If its 1-2% to me its within an allowable margin.

Code: Select all

 Project ID: 6800
 Min. Time / Frame : 00:01:55 - 9,751.9 PPD
 Avg. Time / Frame : 00:01:55 - 9,751.9 PPD

 Project ID: 6801
 Min. Time / Frame : 00:01:54 - 10,216.4 PPD
 Avg. Time / Frame : 00:01:54 - 10,216.4 PPD

 Project ID: 6802
 Min. Time / Frame : 00:01:54 - 10,042.1 PPD
 Avg. Time / Frame : 00:01:56 - 9,869.0 PPD

 Project ID: 6803
 Min. Time / Frame : 00:01:55 - 9,954.8 PPD
 Avg. Time / Frame : 00:01:56 - 9,869.0 PPD

 Project ID: 6804
 Min. Time / Frame : 00:01:54 - 10,042.1 PPD
 Avg. Time / Frame : 00:01:55 - 9,954.8 PPD

 Project ID: 7623
 Min. Time / Frame : 00:07:12 - 10,374.0 PPD
 Avg. Time / Frame : 00:07:15 - 10,302.5 PPD

 Project ID: 7626
 Min. Time / Frame : 00:07:30 - 9,959.0 PPD
 Avg. Time / Frame : 00:07:30 - 9,959.0 PPD

 Project ID: 8018
 Min. Time / Frame : 00:05:45 - 14,417.5 PPD
 Avg. Time / Frame : 00:05:46 - 14,375.9 PPD

 Project ID: 8043
 Min. Time / Frame : 00:02:18 - 14,944.7 PPD
 Avg. Time / Frame : 00:02:18 - 14,944.7 PPD

 Project ID: 8044
 Min. Time / Frame : 00:02:15 - 15,276.8 PPD
 Avg. Time / Frame : 00:02:16 - 15,164.5 PPD

 Project ID: 8045
 Min. Time / Frame : 00:02:18 - 14,944.7 PPD
 Avg. Time / Frame : 00:02:18 - 14,944.7 PPD

 Project ID: 8054
  Min. Time / Frame : 00:02:18 - 14,944.7 PPD
 Avg. Time / Frame : 00:02:18 - 14,944.7 PPD
 Cur. Time / Frame : 00:02:20 - 14,731.2 PPD
 R3F. Time / Frame : 00:02:19 - 14,837.2 PPD
 All  Time / Frame : 00:02:19 - 14,837.2 PPD
 Eff. Time / Frame : 00:02:22 - 14,523.7 PPD

14:52:29:WU00:FS01:Starting
14:52:29:WU00:FS01:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/ProgramData/FAHClient/cores/www.stanford.edu/~pande/Win32/AMD64/NVIDIA/Fermi/Core_15.fah/FahCore_15.exe -dir 00 -suffix 01 -version 702 -lifeline 4708 -checkpoint 15 -gpu 0
14:52:29:WU00:FS01:Started FahCore on PID 7080
14:52:29:WU00:FS01:Core PID:5932
14:52:29:WU00:FS01:FahCore 0x15 started
14:52:29:WU00:FS01:0x15:
14:52:29:WU00:FS01:0x15:*------------------------------*
14:52:29:WU00:FS01:0x15:Folding@Home GPU Core
14:52:29:WU00:FS01:0x15:Version                2.25 (Wed May 9 17:03:01 EDT 2012)
14:52:29:WU00:FS01:0x15:Build host             AmoebaRemote
14:52:29:WU00:FS01:0x15:Board Type             NVIDIA/CUDA
14:52:29:WU00:FS01:0x15:Core                   15
14:52:29:WU00:FS01:0x15:
14:52:29:WU00:FS01:0x15:Project: 8054 (Run 0, Clone 569, Gen 114)
14:52:29:WU00:FS01:0x15:
14:52:29:WU00:FS01:0x15:Assembly optimizations 

[7im]

I don't know where that reported "story" came from, but the only change disclosed by PG was an environment setting change. No other changes, scientific method or HW optimization-wize was addressed yet.

Yes, there is a performance change. We don't know why. But this is the core issued to all new client installs. Standard issue for now, until the next one comes.

Try here http://foldingforum.org/viewtopic.php?f ... 88#p228802 or

http://foldingforum.org/viewtopic.php?f ... 57#p227747, http://foldingforum.org/viewtopic.php?f ... 35#p221635, http://foldingforum.org/viewtopic.php?f ... 30#p227398.

I think that just about covers it properly, in the absence of a definitive statement i'll take the word of site admin and PG group members as the statement

No, it doesn't cover anything.

The first link is a forum staffer stating an obvious anecdotal observation. It contains nothing about what exactly was changed, only that performance changed, and we knew that already.
Second link is that same forum person who clearly stated his comment was "strictly a guess" so no new facts there either.
Third link is, hey, from a PG member. But the thread is about core_16, and the optimization mentioned was linked to core_15, but not what version of that core, and not what hardware was affected. We don't know if that change came out in v2.21. v2.22, v2.23, or any version for that matter. Yes, that could be v2.25, but correlation is not causation.
Fourth link is from PG again. Hurray. Unfortunately, it mentions a change in the logic to an environmental variable that NO ONE uses. And if you don't use the variable, it doesn't affect you.

So again, performanced changed, we don't know why. It will likely change again.

[chaosdsm]

Logs don't forget or get confused, at least not normally... eVGA 560Ti FPB only 1,600ppd better than GTX-460??? I don't think so... I think you want to remember it doing about the same, in reality, that card should have been runing at least 16,000ppd on the 8054 work unit under core 2.22. Hell my GTX460 at 800/1600/2000 nearly matches your 560Ti under the old 2.22 core if you're correct, and that's with a 50MHz core dissadvantage...

Code: Select all

13:18:31:WU02:FS00:Connecting to assign-GPU.stanford.edu:80
13:18:31:WU02:FS00:News: Welcome to Folding@Home
13:18:31:WU02:FS00:Assigned to work server 171.67.108.143
13:18:31:WU02:FS00:Requesting new work unit for slot 00: RUNNING gpu:0:"GF104 [GeForce GTX 460]" from 171.67.108.143
13:18:31:WU02:FS00:Connecting to 171.67.108.143:8080
13:18:32:WU02:FS00:Downloading 59.36KiB
13:18:32:WU02:FS00:Download complete
13:18:32:WU02:FS00:Received Unit: id:02 state:DOWNLOAD error:OK project:8054 run:0 clone:2032 gen:38 core:0x15 unit:0x0000002f6953ee2f50626b28741e9d8b
13:21:05:WU02:FS00:Starting
13:21:05:WU02:FS00:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:/Users/chaosdsm/AppData/Roaming/FAHClient/cores/www.stanford.edu/~pande/Win32/AMD64/NVIDIA/Fermi/Core_15.fah/FahCore_15.exe -dir 02 -suffix 01 -version 701 -lifeline 1092 -checkpoint 15 -gpu 0
13:21:05:WU02:FS00:Started FahCore on PID 13896
13:21:05:WU02:FS00:Core PID:11832
13:21:05:WU02:FS00:FahCore 0x15 started
13:21:05:WU02:FS00:0x15:
13:21:05:WU02:FS00:0x15:*------------------------------*
13:21:05:WU02:FS00:0x15:Folding@Home GPU Core
13:21:05:WU02:FS00:0x15:Version                2.22 (Thu Dec 8 17:08:05 PST 2011)
13:21:05:WU02:FS00:0x15:Build host             SimbiosNvdWin7
13:21:05:WU02:FS00:0x15:Board Type             NVIDIA/CUDA
13:21:05:WU02:FS00:0x15:Core                   15
13:21:05:WU02:FS00:0x15:
13:21:05:WU02:FS00:0x15:Window's signal control handler registered.
13:21:05:WU02:FS00:0x15:Preparing to commence simulation
13:21:05:WU02:FS00:0x15:- Looking at optimizations...
13:21:05:WU02:FS00:0x15:DeleteFrameFiles: successfully deleted file=02/wudata_01.ckp
13:21:05:WU02:FS00:0x15:- Created dyn
13:21:05:WU02:FS00:0x15:- Files status OK
13:21:05:WU02:FS00:0x15:sizeof(CORE_PACKET_HDR) = 512 file=<>
13:21:05:WU02:FS00:0x15:- Expanded 60272 -> 264278 (decompressed 438.4 percent)
13:21:05:WU02:FS00:0x15:Called DecompressByteArray: compressed_data_size=60272 data_size=264278, decompressed_data_size=264278 diff=0
13:21:05:WU02:FS00:0x15:- Digital signature verified
13:21:05:WU02:FS00:0x15:
13:21:05:WU02:FS00:0x15:Project: 8054 (Run 0, Clone 2032, Gen 38)
13:21:05:WU02:FS00:0x15:
13:21:05:WU02:FS00:0x15:Assembly optimizations on if available.
13:21:05:WU02:FS00:0x15:Entering M.D.
13:21:07:WU02:FS00:0x15:Tpr hash 02/wudata_01.tpr:  1952388665 2585542387 985888651 2076040847 4043457155
13:21:07:WU02:FS00:0x15:GPU device info: vendor=0 device=0 name=<NA> match=0
13:21:07:WU02:FS00:0x15:Working on Good ROcking Metal Altar for Chronical Sinners
13:21:07:WU02:FS00:0x15:Client config unavailable.
13:21:07:WU02:FS00:0x15:Starting GUI Server
13:22:08:WU02:FS00:0x15:Setting checkpoint frequency: 500000
13:22:08:WU02:FS00:0x15:Completed         3 out of 50000000 steps (0%).
13:24:31:WU02:FS00:0x15:Completed    500000 out of 50000000 steps (1%).
13:26:54:WU02:FS00:0x15:Completed   1000000 out of 50000000 steps (2%).
13:29:17:WU02:FS00:0x15:Completed   1500000 out of 50000000 steps (3%).
13:31:40:WU02:FS00:0x15:Completed   2000000 out of 50000000 steps (4%).
13:34:03:WU02:FS00:0x15:Completed   2500000 out of 50000000 steps (5%).
13:36:26:WU02:FS00:0x15:Completed   3000000 out of 50000000 steps (6%).
13:38:49:WU02:FS00:0x15:Completed   3500000 out of 50000000 steps (7%).
13:41:12:WU02:FS00:0x15:Completed   4000000 out of 50000000 steps (8%).
13:43:35:WU02:FS00:0x15:Completed   4500000 out of 50000000 steps (9%).
13:45:57:WU02:FS00:0x15:Completed   5000000 out of 50000000 steps (10%).
13:48:20:WU02:FS00:0x15:Completed   5500000 out of 50000000 steps (11%).
13:50:44:WU02:FS00:0x15:Completed   6000000 out of 50000000 steps (12%).
13:53:06:WU02:FS00:0x15:Completed   6500000 out of 50000000 steps (13%).
13:55:29:WU02:FS00:0x15:Completed   7000000 out of 50000000 steps (14%).
13:57:52:WU02:FS00:0x15:Completed   7500000 out of 50000000 steps (15%).
14:00:15:WU02:FS00:0x15:Completed   8000000 out of 50000000 steps (16%).
14:02:38:WU02:FS00:0x15:Completed   8500000 out of 50000000 steps (17%).
14:05:01:WU02:FS00:0x15:Completed   9000000 out of 50000000 steps (18%).
14:07:24:WU02:FS00:0x15:Completed   9500000 out of 50000000 steps (19%).
14:09:47:WU02:FS00:0x15:Completed  10000000 out of 50000000 steps (20%).
14:12:10:WU02:FS00:0x15:Completed  10500000 out of 50000000 steps (21%).
14:14:33:WU02:FS00:0x15:Completed  11000000 out of 50000000 steps (22%).
14:16:56:WU02:FS00:0x15:Completed  11500000 out of 50000000 steps (23%).
14:19:19:WU02:FS00:0x15:Completed  12000000 out of 50000000 steps (24%).
14:21:42:WU02:FS00:0x15:Completed  12500000 out of 50000000 steps (25%).

That's an average of 2min 22sec per frame for my GTX460 @ 800/1600/2000. Extrapolated out, at 850/1700/2000 my GTX460 would be folding the same work unit slightly faster (about 2min 15sec per frame) than your 560Ti FPB at the same speed (under core version 2.22), which is why I'm skeptical of your claim.

BTW, I've had my 460 folding without errors at 875/1650/2150 clocks. Under current core performance, I'm considering going there again...

[chaosdsm]

Related question, I'm seeing people on several team forums saying to just dump core 2.25 and go back to core 2.22 if you're on a Fermi card. Just wondering if PG has an official stance on this subject???

[7im]

Part of the official stance is that the work unit will prompt the client to download an updated fahcore if the current fahcore version is not new enough to process the current WU. The client manages fahcore versions.

The other half of the official stance is that manual manipulation files used by FAH is against the EULA.

Eventually v2.25 will become the minimum required version for all GPU WUs. And then v2.26, then v2.27, etc.

Change fahcore versions at your own risk.

[kiore]

Related question, I'm seeing people on several team forums saying to just dump core 2.25 and go back to core 2.22 if you're on a Fermi card. Just wondering if PG has an official stance on this subject???

I do understand why people would suggest such a thing.. It does appear from a donator point of view that exactly the same science is done but at what appears to be 30% faster speed by manipulating or whatever you choose to call it.. changing the core you run on against what the project installs.. Maybe this is a 'bad' core, the kepler folders will not share this opinion.. or maybe it is the core that the project prefers.. Whatever the answer, it seems that the project is asking us to run 2.25 by updating this and requiring donator input to prevent this change.. I guess the decision by the donator is whether they or the project managers understand best what is best for the project.
Personally I have been disappointed by the sudden reduction in my output this has caused, but as I really don't know the details I personally feel that changing this without being told it is OK by those who really know is against my preference which is to do the best science. Yes I enjoy the points competition and it is a serious motivator, but to cherry pick cores seems not so different to cherry picking work units which I am opposed to. So until told different from those who really know I will stick with whatever core is assigned, just like I will fold whatever unit assigned.

[7im]

It's especially problematic because I see many recommedations that tell people to stop the client, and change fahcore versions in the middle of a work unit. People are just asking to corrupt the WU data.