GPU PPD in Folding at Home

[iwajabitw]

Now have 10490 / 381178

 

[Markfw]

I really don't know why yours are failing out at stock. Like I say, I have 11 cards, and Tony has 2, and all basically are working ok (yes, I did have the one failure)

 

[iwajabitw]

This system is using Win 10 Pro, but all drivers I update separately from Asrock or MSI/Nvidia. Don't know if that would have anything to do with it.

 

[TennesseeTony]

9190 / 774k
9209 / 601k
13500 again / 672k
9152 / 639k
11709 / 677k
10494 / 654k
9660 / 543k but short ~30m. 677 Base/ 12321 Est Credit

So far my tasks seem to track well with the data found here, except lower, because those users have about a 200core/800mem extra overclock compared to my OC'd cards.

 

[Markfw]

And after adding that and mine, here is the latest:

 

[TennesseeTony]

I guess we are still suffering the ongoing effects of the IoT attacks of late. 80MB task download from Stanford took 17m20s. My internet (still better than dial up though!) has really sucked this week.

80MB task 9213 / 586k but only at 4%. Long task, so hopefully I'll catch it's progress later b4 it's done. Kinda aggravated at the loss of 17m of crunch to download a lower scoring task.

Checking my logs I see that most tasks are in the 768k to 3.7MB range, with some 18-20MB files, then several of these 80MB monsters.

The spreadsheet is filling out nicely Mark, looks good!

 

[Markfw]

Thanks Tony ! I am only going to keep this up for about a week or so. Then I think we have some good numbers to tell people about what card they may want to buy, and what to expect.

BTW, check the thread on the bad work unit. I think the SLI must be the issue, I can't find any other reason. And no matter what even Stanford says, SLU may NOT work, we will see.

 

[TennesseeTony]

One round for each 980Ti I have, for comparison to yours, Mark. It's going to be a bit warm when I get back home:

BumbleBee Clocked @ ~1275
9193 / 539k
9210 / 508k

WUSS Clocked @ ~1275
13500 / 537k
13500 / 552k

Devastator Clocked @ ~1380
11401 / 647k
9211 / 600k
10495 / 499k Thermal limit-only 1200MHz

...back to the drawing board for Devastator.

 

[TennesseeTony]

I really have other things to do, but...

And the 980 fleet, one round each:
System is a AMD FX8320E at 3.2GHz, 8 'core' but many say it's 8 threads.

3x980@~1340GHz

11707 / 318k 3%
13500 / 399k 5%
10495 / 370k 5%

And with that I'm out of the house for a while. Looks like I'm on track for 6M+/- PPD for the race. And a huge electric bill in January.

 

[TennesseeTony]

1070 -last report for these guys for a bit, GPUGrid now has Pascal test tasks, so they are being repurposed.
11401 / 678k
9213 / 589k

980
~same ppd and tasks as in the previous post, finishing and shutting down.

980Ti
Only 2 new ones since last update, finishing.
11707 / 549k
9175 / 579k

Breaks over, back to hanging drywall on a ceiling...alone. :/

 

[TennesseeTony]

Missed one
980Ti @1380
9207 / 574k

 

[Kiska]

System is a AMD FX8320E at 3.2GHz, 8 'core' but many say it's 8 threads.

Well its physically 8 cores, but the way that AMD has done the architecture, its more like 4 cores, worth of performance. Its because each 'module' has 2 cores, with 1 Floating Point unit, 2 Integer core each core has its own scheduler, and all that.

 

[crashtech]

GTX 1060 @ 1.81GHz
PCRG / PPD
9660 / 329327
9657 / 328106
9151 / 318979
10496 / 358635

R9 Fury @ 1.11GHz
PCRG / PPD
9037 / 466872
9036 / 413800
9193 / 431497
9189 / 475333
11406 / 530764

R9 380 @ 1.04GHz
PCRG / PPD
11709 / 138937
9211 / 123245


I will submit more if they would be helpful

Edit: I don't know how to catch them all, the ppd estimate is not logged anywhere, is it?

Edit 2: I'm adding results as they become available, if anyone noticed, a result from the Fury was stricken as there was an overclocking failure at 1.14GHz. It appears to be stable at 1.11GHz.

 

[TennesseeTony]

If you were really really dedicated you could search for the final points per task in the log, but having said that I don't seem to recall it saying what PCRG those points were for.

Thanks for the numbers! Certainly good to have some variety in the chart! And well done, figuring out the problems you had going on.

 

[crashtech]

Would the PPD calculation would be:

PPD = 24(points per task)/(elapsed time for task in hours)

?

 

[TennesseeTony]

No no, it's much simpler than that:

final_points = base_points * max(1, sqrt( k * deadline_length / elapsed_time))

PPD = 14.4 * base_points * max(1, sqrt( 14.4 * k * Expiration / TPF)) / TPF

From: https://folding.stanford.edu/home/faq/faq-points/

 

[crashtech]

I suppose I might just say that my contribution to this project is incalculable...

I do know that the PPD this 1060 has been registering is well below what I was led to believe.

 

[Markfw]

I suppose I might just say that my contribution to this project is incalculable...

I do know that the PPD this 1060 has been registering is well below what I was led to believe.

I do know that you have reported 2 ppd numbers, and one of the 2 is higher than my 980ti, which normally just edges out or equals a 1070. (328 vs 299) one of the worst units out there. (for ppd)

 

[crashtech]

Well, I don't know what to say. I don't really want to include screenshots unless we think it warrants it. I'm just copy pasting the numbers that appear on the screen. The ppd estimate does tend to fluctuate, but I have no impetus to inflate the numbers. My GPU buying decisions have been partly based on this document, perhaps the data it is compiled with includes a high number of OCed GPUs.

Edit: Clearly I misunderstood what Mark was saying! Many contentious years on message boards have probably made me overly defensive. What I can glean instead is that I ought to be ecstatic about my 1060 purchase, and if its performance holds, I'll be buying more.

 

[TennesseeTony]

I think Mark is saying that the 1060 beat a 980Ti, which is impressive! That's more along the output of a 1070, which is supposed to be the equal of the older 980Ti.

And yes, those two 1060 users are highly overclocked.

Also are you using a passkey/completed 10 tasks already?

 

[crashtech]

I have a passkey, got it the first day. I haven't tried overclocking this 1060 yet. It's a blower card, and I wonder if it would become objectionable if the clocks/volts were cranked up. Maybe for December I will tolerate it, haha.

 

[iwajabitw]

If the driver fix works I currently have.

9679 / 200524
9152 / 284713

 

[iwajabitw]

10490 / 382801

 

[iwajabitw]

11709 / 362067

 

[iwajabitw]

7:11 CST

11406 / 371058
11401 / 404915