[QUOTE=James Heinrich;245054]I don't remember exactly what the job was, but I believe it was something like 2^76-2^77 on something in the M333M range. My 8800 isn't as fast as your 460 ;)[/QUOTE]

Aaaah, that's significantly deeper than me too. I'm in the 100M ish range max I've seen is up to 2^72 (where the 90m figure came from).

-- Craig

Hi!

[QUOTE=aaronhaviland;245019](on a side note, I found it to run a little faster without MORE_CLASSES, on my gts250. Is this right?)[/QUOTE]

Take a look at params.h. :wink:
On a "small" job it can be faster without MORE_CLASSES. Depends on realativ speed of CPU and GPU, too.

Oliver

If anyone cares, I've modified my stats site to accept TF results from mfaktc:
[url]http://mersenne-aries.sili.net/[/url]

Currently all results are forcibly reported under user "[url=http://mersenne-aries.sili.net/stats.php?showuserstats=mfaktc]mfaktc[/url]" and machine name equal to the version (e.g. "0.13-Win 71bit_mul24").

Would is be possible to have mfaktc output a benchmark result similar to Prime95's output? Modified a little, of course, to capture useful system specs. I'm no expert on which values are important for comparison but something like:[code][Mon Jan 04 06:34:03 2011]
GPU name: NVIDIA GeForce 8800 GT
compute capability: 1.1
maximum threads per block: 512
number of multiprocessors: 14 (112 shader cores)
clock rate: 1500MHz
mfaktc v0.13-Win
Best time for 58 bit trial factors: 4.473 ms.
Best time for 59 bit trial factors: 4.473 ms.
Best time for 60 bit trial factors: 4.488 ms.
Best time for 61 bit trial factors: 4.469 ms.
Best time for 62 bit trial factors: 4.489 ms.
Best time for 63 bit trial factors: 8.367 ms.
Best time for 64 bit trial factors: 8.357 ms.
Best time for 65 bit trial factors: 8.219 ms.
Best time for 66 bit trial factors: 8.152 ms.
Best time for 67 bit trial factors: 8.185 ms.
Best time for 68 bit trial factors: 4.473 ms.
Best time for 69 bit trial factors: 4.473 ms.
Best time for 70 bit trial factors: 4.488 ms.
Best time for 71 bit trial factors: 4.469 ms.
Best time for 72 bit trial factors: 4.489 ms.
Best time for 73 bit trial factors: 8.367 ms.
Best time for 74 bit trial factors: 8.357 ms.
Best time for 75 bit trial factors: 8.219 ms.
Best time for 76 bit trial factors: 8.152 ms.
Best time for 77 bit trial factors: 8.185 ms.
Best time for 78 bit trial factors: 8.357 ms.
Best time for 79 bit trial factors: 8.219 ms.
Best time for 80 bit trial factors: 8.152 ms.[/code]

Found two ways to break mfaktc:

[code]Factor=DEADBEEFDEADBEEFDEADBEEFDEADBEEF,1000037,67,68[/code]
mfaktc crashes, not sure why.


[code]Factor=DEADBEEFDEADBEEFDEADBEEFDEADBEEF,900037,67,68[/code]
mfaktc goes into infinite loop complaining about[quote]got assignment: exp=900037 bit_min=67 bit_max=68
WARNING: exponents < 1000000 are not supported! Ignoring this assignment![/quote] but doesn't skip to next assignment in worktodo.txt

(yes, I realize I'm probably doing dumb things out of ignorance, but half the code in programming is validating input :smile:)

I too get the "WARNING: exponents < 1000000 are not supported! Ignoring this assignment!" error, but no crash.
Using mfaktc v0.13-Win64 .

I tried it again, it's reproducable for me. Also using 0.13-win64. Screenshot attached.

If this helps, here's my run.

[QUOTE=James Heinrich;245276]Would is be possible to have mfaktc output a benchmark result similar to Prime95's output? Modified a little, of course, to capture useful system specs.[/QUOTE]

I won't compare to Prime95 TF benchmarks, they report the average time for a block of factor candidates... and AFAIK the blocksize varies between prime95 and prime64... perhaps not a perfect measurement.
Just run a specific exponent (my benchmark exponent is 66362159).

[QUOTE=James Heinrich;245285]Found two ways to break mfaktc:

[code]Factor=DEADBEEFDEADBEEFDEADBEEFDEADBEEF,1000037,67,68[/code]
mfaktc crashes, not sure why.
[/QUOTE]

I know why and it is allready fixed in mfaktc 0.14. :smile:
Just lower SievePrimes below ~80000 and I will work (you'll need to set SievePrimesAdjust to 0 to make sure I won't get bigger again)!
Karl used a lower SievePrimes, thats the reason why it works for him.
Small exponents need smaller SievePrimes. SievePrimes=100k works for exponents > 1.3M.

[QUOTE=James Heinrich;245293]I tried it again, it's reproducable for me. Also using 0.13-win64. Screenshot attached.[/QUOTE]

Thank you for the screenshot, helped to figure out it is a known bug which is allready fixed.

Oliver

[QUOTE=TheJudger;245308]allready fixed in mfaktc 0.14. :smile:[/QUOTE]

I can't find 0.14 anywhere. Is it out?

[QUOTE=TheJudger;245126]Hi!

Take a look at params.h. :wink:
On a "small" job it can be faster without MORE_CLASSES. Depends on realativ speed of CPU and GPU, too.

Oliver[/QUOTE]

So true! Switched to a GTX 460 yesterday (clocked at 875MHz), still seems better without MORE_CLASSES, until I get to the 2^69-2^70 ranges.

Also, I've found that I need to lower SIEVEPRIMES much lower...
I reduced the lower limit and found that it liked to hover around 4500, if I'm not also using the CPU for other intensive tasks. If I am, it likes it around 500.