p2671 -- all Gen 17

[bruce]

There is no magic. The PPD calculations are based on the frame times so it's either half speed or it isn't. The only question is how many frames you or FahMon uses to figure out an average frame time.

*** Time per frame x 100 = Time per WU. . . . Convert to decimal days per WU.
*** Divide points for that WU by days per WU to get PPD.

[alpha754293]

There is no magic. The PPD calculations are based on the frame times so it's either half speed or it isn't. The only question is how many frames you or FahMon uses to figure out an average frame time.

*** Time per frame x 100 = Time per WU. . . . Convert to decimal days per WU.
*** Divide points for that WU by days per WU to get PPD.

No I realized that. But see...on computenode/CPU1: it's still about 8.75 minutes / frame and it's showing a PPD of around 3100 (which is about average normal for a "-smp 4" on my system.)

But on computenode/CPU2: the average time per frame is the same, but it's showing half the PPD.

And it's using the last 3 frames to calculate it, and no matter how many times I reload the clients or update it or whatever; it's still giving that.

computenode/CPU1: 3171.85 PPD

Code: Select all

[04:05:09] Completed 202500 out of 250000 steps  (81%)
[04:13:52] Completed 205000 out of 250000 steps  (82%)
[04:22:35] Completed 207500 out of 250000 steps  (83%)
[04:31:19] Completed 210000 out of 250000 steps  (84%)

computenode/CPU2: 1559.10 PPD

Code: Select all

[04:04:24] Completed 232501 out of 250001 steps  (93%)
[04:13:15] Completed 235001 out of 250001 steps  (94%)
[04:22:07] Completed 237501 out of 250001 steps  (95%)
[04:30:59] Completed 240001 out of 250001 steps  (96%)

my other system was showing something similiar (report about 1800 PPD despite the frame rates, and nominally for an a2 WU, it's about 3600 PPD).

[uncle_fungus]

I suspect you're using a non-current version of FahMon, which accounts for the discrepancy in frame times. The frame count calculator in FahMon used to get confused by WUs which had strange numbers of steps (like 250001), and ended up rounding the calculation in an incompetent manner. This was fixed in v2.3.99.

[alpha754293]

I suspect you're using a non-current version of FahMon, which accounts for the discrepancy in frame times. The frame count calculator in FahMon used to get confused by WUs which had strange numbers of steps (like 250001), and ended up rounding the calculation in an incompetent manner. This was fixed in v2.3.99.

Ahhhhh...that explains it. ok. thanks.

[alpha754293]

P.S. tee doesn't capture ALLL of the console_out stuff.

(for example, the FLOP assessment at the end -- it doesn't capture that.)

So, I might end up having to switch it to that other method if I want to try and make sure it captures it all.

Just a quickie FYI.

the G17s seemed to have otherwise finished normally.

There's one WU that I'm working on right now still outstanding (been busy working on other stuff), and one P2671G18 that the computenode is working on now.

[bruce]

P.S. tee doesn't capture ALLL of the console_out stuff.

I'll bet that's because there are two different console outputs stderr and stdout. You can either switch to uncle_fungus's method with tail or try this:

Code: Select all

./fah6 2>&1 | tee -a foldinglog.txt