-forceasm boost - test results 100% boost

[ShaunBrewer]

These are some test results that were recorded on one of my dedicated folders.

See below

HW - NF7-S 1700+ clock below

results with 2 instances runnign different clock and also SSE and 3DNOW

Same Gromac - restarted several times. I will do some tests on different WU's at some point

Good case for forceasm!!!

Happy reading

Shaun

Gromac info:

[19:21:45] - Ask before connecting: Yes
[19:21:45] - User name: ShaunBrewer (Team 32)
[19:21:45] - User ID = 6EDA80672C8C6260
[19:21:45] - Machine ID: 1
[19:21:45]
[19:21:45] Loaded queue successfully.
[19:21:45] Initialization complete
[19:21:45] + Benchmarking ...
[19:21:47]
[19:21:47] + Processing work unit
[19:21:47] Core required: FahCore_78.exe
[19:21:47] Core found.
[19:21:47] Working on Unit 05 [October 22 19:21:47]
[19:21:47] + Working ...
[19:21:48]
[19:21:48] *------------------------------*
[19:21:48] Folding@home Gromacs Core
[19:21:48] Version 1.51 (September 25, 2003)
[19:21:48]
[19:21:48] Preparing to commence simulation
[19:21:48] - Ensuring status. Please wait.
[19:22:05] - Assembly optimizations manually forced on.
[19:22:05] - Not checking prior termination.
[19:22:06] - Expanded 980236 -> 6054869 (decompressed 617.6 percent)
[19:22:06]
[19:22:06] Project: 1003 (Run 8, Clone 52, Gen 0)
[19:22:06]
[19:22:06] Assembly optimizations on if available.
[19:22:06] Entering M.D.
[19:22:27] (Starting from checkpoint)
[19:22:27] Protein: p1003_ppg10c_350

Arguments: -advmethods -forceasm

Clock 214x10.5

Console + GUI

09:11:44] Completed 60000 out of 250000 steps (24)
[09:18:50] Writing local files
[09:18:50] Completed 62500 out of 250000 steps (25)
[09:25:59] Writing local files
[09:25:59] Completed 65000 out of 250000 steps (26)
[09:33:09] Writing local files
[09:33:09] Completed 67500 out of 250000 steps (27)
[09:40:20] Writing local files
[09:40:20] Completed 70000 out of 250000 steps (28)
[09:47:32] Writing local files
[09:47:32] Completed 72500 out of 250000 steps (29)
[09:54:46] Writing local files
[09:54:46] Completed 75000 out of 250000 steps (30)

Clock 200x10.5

GUI Only

[19:22:27] Writing local files
[19:22:27] Completed 75000 out of 250000 steps (30)
[19:22:27] Extra SSE boost OK.
[19:30:07] Writing local files
[19:30:07] Completed 77500 out of 250000 steps (31)
[19:37:40] Writing local files
[19:37:40] Completed 80000 out of 250000 steps (32)
[19:45:21] Writing local files
[19:45:21] Completed 82500 out of 250000 steps (33)

Console + GUI

[19:53:18] Writing local files
[19:53:18] Completed 85000 out of 250000 steps (34)
[20:01:09] Writing local files
[20:01:09] Completed 87500 out of 250000 steps (35)
[20:09:00] Writing local files
[20:09:00] Completed 90000 out of 250000 steps (36)
[20:16:52] Writing local files
[20:16:52] Completed 92500 out of 250000 steps (37)
[20:24:46] Writing local files
[20:24:46] Completed 95000 out of 250000 steps (38)
[20:32:41] Writing local files
[20:32:41] Completed 97500 out of 250000 steps (39)
[20:40:37] Writing local files
[20:40:37] Completed 100000 out of 250000 steps (40)

Console + GUI (NOT SSE) (flag removed)

[20:47:36] - Ask before connecting: Yes
[20:47:36] - User name: ShaunBrewer (Team 32)
[20:47:36] - User ID = 6EDA80672C8C6260
[20:47:36] - Machine ID: 1
[20:47:36]
[20:47:36] Loaded queue successfully.
[20:47:36] Initialization complete
[20:47:36] + Benchmarking ...
[20:47:38]
[20:47:38] + Processing work unit
[20:47:38] Core required: FahCore_78.exe
[20:47:38] Core found.
[20:47:38] Working on Unit 05 [October 22 20:47:38]
[20:47:38] + Working ...
[20:47:39]
[20:47:39] *------------------------------*
[20:47:39] Folding@home Gromacs Core
[20:47:39] Version 1.51 (September 25, 2003)
[20:47:39]
[20:47:39] Preparing to commence simulation
[20:47:39] - Looking at optimizations...
[20:47:39] - Files status OK
[20:47:40] - Expanded 980236 -> 6054869 (decompressed 617.6 percent)
[20:47:40]
[20:47:40] Project: 1003 (Run 8, Clone 52, Gen 0)
[20:47:40]
[20:47:40] Assembly optimizations on if available.
[20:47:40] Entering M.D.
[20:48:01] (Starting from checkpoint)
[20:48:01] Protein: p1003_ppg10c_350
[20:48:01]
[20:48:01] Writing local files
[20:48:01] Completed 100000 out of 250000 steps (40)
[20:48:01] Extra 3DNow boost OK.
[21:04:01] Writing local files
[21:04:01] Completed 102500 out of 250000 steps (41)
[21:19:59] Writing local files
[21:19:59] Completed 105000 out of 250000 steps (42)
[21:35:58] Writing local files
[21:35:58] Completed 107500 out of 250000 steps (43)

[JetMech]

Unfortunately on most systems the lockups occur deep into the process. I have lost with work as high as 95% completed. That's what makes it all so heart breaking.

[ShaunBrewer]

I am currently rebuilding/re-installing most/all of my PC's.

Once they are up and running in a stable configuration I will experiment with overclock and stability.

However form my experiences so far I am getting somewhere between 1 in 10 to 1 in 20 failure rate. So for the boost is worth the odd failure

With no lockups.

I am changing heatsinks fans and processors currenly running 4 1700+ at various clocks (all tbred 1 is an A)

Shaun

[JetMech]

I have started replacing my AMD rigs with P4s. I'm hoping the reduction in loss will offset the higher frame times. At this very moment I'm testing the internet capabilities of a new addition (P4 2.6C on Abit IC7 - Max3) which will move my main to farm only status. Had two lockups today that don't seem to be CPU related as the mouse/keyboard aren't working but the work is still being done. I'll be overclocking when I figure out how to lock the Memory at 200.

[Audioaficionado]

If Pande Group can't get a handle on the SSE/Tbred-Barton issue, they stand to lose a huge amount of production. They might just lock out SSE for AMDs (I hope not) if things don't change soon. The bottom line is quality over quantity.

[Deathknight]

Not sure I see why they would disable it. Even with the lockup issues sse provides a huge performance boost. As it is you need to manually enable the feature in the first place...

[dz]

I've been running SSE for at least 5 WU's on my 2600+ t'bred B, no issues yet. Using -advmethods -forceasm

[Audioaficionado]

It seems to be the luck of the draw.

Some people have nothing but trouble from the start

Some were OK but now are starting to have troubles with the newer WUs

Some never have had any problems.

I had no troubles whatsoever with my last 50% overclocked NF7-S & XP1700 B and SSE/FAH. I hope it's the same with this Barton version I'm building.

I can't afford P4 layers when the processor alone costs more than an entire AMD layer.

[Fast420A]

Originally posted by Audioaficionado


I can't afford P4 layers when the processor alone costs more than an entire AMD layer.

I know what you mean Audioaficionado! There's 2 things I want, an AMD dually and a Nice P4 with the 800Mhz FSB but when I can knock out 500 - 700 PPW layers for $150 it's hard to build one! :D

[Audioaficionado]

I'm gonna get one of those nice SB75P Shuttle XPCs next I think. It will be nice to have one HT rig that can exceed 1000ppw. I'll sell the nForce2 to offset the cost. I got $1k for the last one I built.

That duallie will come late next year when Xeons or Opertrons get reasonable. Once I secure my two personal rigs, then I'll work on farming. If AMD is still SSE/FAH crippled, I'll switch over to Intel exclusively.

BTW I think you're going to easily beat me to 25k as I'm going to max out at ~1100ppw once I finish up my nForce2 rig In a couple of days. The gains from the PIII rig I got are being neutralized by increased gaming on my daughter's shuttle. That nForce2 rig I sent out hasn't ever sent out anything as far as I know.