TH results

shadowscotland · 9 Aug 2006 at 13:24

Not usually in a position to see two of the same WU folding side by side on a PC but today is that day.

[08:05:56] Project: 775 (Run 21, Clone 119, Gen 4)
[08:06:18] Completed 200000 out of 250000 steps (80)
[11:21:18] Completed 225000 out of 250000 steps (90)

[08:14:10] Project: 775 (Run 43, Clone 47, Gen 5)
[08:14:31] Completed 202500 out of 250000 steps (81)
[11:09:14] Completed 227500 out of 250000 steps (91)

As you will notice both are at a simular step, the first starting 10 minutes earlier than the second. You may also notice that the second thread is taking 20min less per 10 steps aka 10% faster.

So is it my usage only affecting one thread (the first one started) and not the second or are Run 21 & Run 43 actually different, quick check of task manager say that mem usage is 31K and 33K (21 and 43 respectively)

so more WU using more memory running faster - reason
It's at 50% and other thread at 45% doh! working at 95% to stop EUE due to OC on stock cooling. sorry people.

but does show that same number WU's are not all the same. Just restarted run 21 now show 40K Mem usage.... :confused:

- brain hurts off for some lunch

Edit - lunchtime folds (no user input)
Project: 775 (Run 21, Clone 119, Gen 4) 19min (reduced from 20) 32K % 45%
Project: 775 (Run 43, Clone 47, Gen 5) 17mins (reduced from 18) 40K % 50%

Would HT bias the computation of the greater usage (%) as 10% faster for only 5% extra processor usage (and more memory usage) does not stack up.

Mattus · 9 Aug 2006 at 14:36

There's often subtle differences between WUs of the same project. It all depends how complicated the interactions between the atoms are - if more atoms come closer together, the movements take longer to calculate. I've noticed it quite a bit, particularly when the QMDs used to be around.

shadowscotland · 9 Aug 2006 at 15:06

Thanks mattus - of late my logs and observations agree - points count doesn't seam to corralate with variation in folding time it would seam. Guess its at atom proximity thing as mattus said and not size dependent.

anyone got any stats/benchmarks on multipule larger WU crunch on same machine

Project : 775 (Unknown)
-- thread 3 --(@50%)
Min. Time / Frame : 17mn 14s
Avg. Time / Frame : 19mn 52s
-- thread 4 --(@50%)
Min. Time / Frame : 15mn 59s
Avg. Time / Frame : 18mn 56s
-- thread 7 -- (@45%)
Min. Time / Frame : 17mn 28s
Avg. Time / Frame : 17mn 29s
-- thread 8 --(@50%)
Min. Time / Frame : 16mn 47s
Avg. Time / Frame : 19mn 19s

Project : 2096 (Unknown)
-- thread 3 --
Min. Time / Frame : 18mn 19s
Avg. Time / Frame : 26mn 15s
-- C:/folding@home5/ --
Min. Time / Frame : 24mn 47s
Avg. Time / Frame : 25mn 02s

Project : 2095 (Unknown)
-- thread 3 --
Min. Time / Frame : 23mn 41s
Avg. Time / Frame : 27mn 29s
-- thread 4 --
Min. Time / Frame : 26mn 02s
Avg. Time / Frame : 27mn 16s

as mentioned on other threads my ppd score and core type don't show up in benchmark logs here at work.

Steevo38 · 14 Aug 2006 at 21:42

Same wu has very different times for me too.

Project : 2096
Core : GBGromacs
Frames : 100
Credit : 178

-- CORE2DUO 1 --

Min. Time / Frame : 8mn 19s - 308.20 ppd
Avg. Time / Frame : 9mn 46s - 262.44 ppd

-- CORE2DUO 2 --

Min. Time / Frame : 10mn 27s - 245.28 ppd
Avg. Time / Frame : 10mn 33s - 242.96 ppd

-- AMD1 --

Min. Time / Frame : 9mn 38s - 266.08 ppd
Avg. Time / Frame : 9mn 39s - 265.62 ppd

-- AMD2 --

Min. Time / Frame : 10mn 35s - 242.19 ppd
Avg. Time / Frame : 10mn 39s - 240.68 ppd

rich99million · 14 Aug 2006 at 21:48

if score and PPD don't show in FahMon have you tried in the FahMon menu...

FahMon -> Download New Projects

sorry if you're already done this and they just aren't on the summary page at this time

shadowscotland · 15 Aug 2006 at 09:23

FahMon -> Download New Projects

twice a weeks for the last three months - no joy as ever (becoming an ongoing joke - if it wasn't for my home rig running FahMon with all the bell and whistles I'd have thought it didn't like me

)

rich99million · 15 Aug 2006 at 10:18

shadowscotland said:
twice a weeks for the last three months - no joy as ever (becoming an ongoing joke - if it wasn't for my home rig running FahMon with all the bell and whistles I'd have thought it didn't like me )

hmm odd - well there is the option of using a proxy but I expect you've not had any luck with that one either

of course if your one at home seems to update fine you could copy the necessary file from there and try copying it into the program folder at work - the file is points.dat

shadowscotland · 15 Aug 2006 at 10:46

have played with the proxy options (as we have a proxy server) but no joy

of course if your one at home seems to update fine you could copy the necessary file from there and try copying it into the program folder at work - the file is points.dat

I'll try that - thanks

shadowscotland · 16 Aug 2006 at 12:04

Move whole Fahmon folder (from home to work pc) deleated redundent clients and added correct ones - RESULT

Top tip rich - my eternal thanks and five gold stars *****

rich99million · 16 Aug 2006 at 13:08

ahhh cool - glad to hear that it worked alright

you should only need the points.dat file for future updates - since FahMon doesn't actually install anything anywhere you can't do any harm

shadowscotland · 16 Aug 2006 at 14:23

Didn't have a points.dat file in the folder or would have just copied that.
Next best option was all file copy transfer (so that's what I did)

only three dat files my folder contains are - prefs.dat, projects.dat & benchmarks.dat

rich99million · 16 Aug 2006 at 14:31

ahhh points.dat was used by the old version (v.1.8 something) - I would guess that projects.dat does the same thing in v2 but probably worth backing it up before trying that one next time

should have remembered I kept the old version handy incase i wanted to check old benchmarks

shadowscotland · 16 Aug 2006 at 14:36

stats junky