[QUOTE=TheJudger;257415]Hi Christenson,
I would bet that a few per day on a cheap GPU is still more than a single core of a current highend CPU can do... And since you've a lot spare cycles on your system (high average wait) you might want to try to run Prime95 with low priority on the same CPU(-core) once I've released mfaktc 0.17.

Oliver[/QUOTE]

That core would have to do LL work....the memory on the work machine regularly gets filled up with hogs like Autocad Viewer and Altium PCB designer, requiring about every other day reboot of Windows. A five-year old Pentium IV isn't really my idea of high end, either.

But I could run all day if mfaktc didn't slow down the screen (and therefore me) so much when I was working.

I just looked, and I don't have any data handy on how long it takes to do the mfaktc tests on the GEForce 210 that is directly comparable to the 1 core of the PhenomII x6 CPU at home...but it is at least comparable. Now if my results got turned in automagically, then I'd know from the results queries....

"Interesting" trial of mfaktc
 

Interesting, as in the old curse.

I downloaded v0.16p1 and, to my great surprise, it built perfectly after typeing "make" in the src directory. It's so unusual for CUDA code to just build without fuss, including my own code, that I'm very impressed. Well done!

Following the instructions, the mersenne server gave me a couple of exponents to test. All went well until roughly 50% through the second exponent when screen updating went crazy. It's hard to describe, but unfortunately the attachment manager won't let me upload a full resolution screenshot because my screen is too big, so the one attached has been reduced from 1920x1200 to 640x400 which gives and idea of what it looks like.

Enough information leaked through the mess that I was able to let the trial complete but the X11 server needed to be restarted to clean up the screen.

I'm not at all sure what's wrong but the usual suspects are the nvidia device driver, the gpu memory, the gpu-cpu PCI comms, the X11 server and its component libraries and, of course mfaktc itself. Somehow I rather doubt the last is at fault. More likely, it seems to me, is that it is stressing something else in the system beyond its elastic limit.

Paul

Hi Paul,

looks interesting...
What kind of GPU? It is known that mfaktc drives the GF100/GF110 chip very hard. Can you monitor the temperature of your GPU?

Can you reproduce this? Does it occur at the same point?
For a wrong memory access in mfaktc your screen does not look weird enough, I would expect "more colors", etc...

Oliver

[QUOTE=TheJudger;258405]Hi Paul,

looks interesting...
What kind of GPU? It is known that mfaktc drives the GF100/GF110 chip very hard. Can you monitor the temperature of your GPU?

Can you reproduce this? Does it occur at the same point?
For a wrong memory access in mfaktc your screen does not look weird enough, I would expect "more colors", etc...

Oliver[/QUOTE]It's a 768MB GTX460 from nvidia fitted into a AMD1090T-based ssytem running a fully up to date Fedora14. The "NVIDIA X Server Settings" app claims to measure the GPU temperature but wasn't running at the time.

So far I've made no attempt to reproduce the errors.

My best guess is that the X11 server doesn't talk to the nvidia driver correctly.


Paul

[QUOTE=TheJudger;258405]Hi Paul,

looks interesting...
What kind of GPU? It is known that mfaktc drives the GF100/GF110 chip very hard. Can you monitor the temperature of your GPU?

Can you reproduce this? Does it occur at the same point?
For a wrong memory access in mfaktc your screen does not look weird enough, I would expect "more colors", etc...

Oliver[/QUOTE]

Maybe blow an additional 45 CFM into the box to provide the gpu enough fresh air.

Reporting results.
 

I'm doing something wrong. The results.txt file contains[code]no factor for M56399117 from 2^68 to 2^69 [mfaktc 0.16p1 barrett79_mul32]
no factor for M56399237 from 2^68 to 2^69 [mfaktc 0.16p1 barrett79_mul32]
[/code] but after uploading that file to [url]http://mersenne.org/manual_result/[/url] the server reports[code]
No factor lines found: 0
Mfaktc no factor lines found: 0
Factors found: 0
P-1 lines found: 0
LL lines found: 0
Mlucas lines found: 0
Glucas (G29) lines found: 0
Glucas lines found: 0
MacLucasFFTW lines found: 0
CUDALucas lines found: 0
ECM lines found: 0[/code]

Paul

[QUOTE=TheJudger;258405]Hi Paul,

looks interesting...
What kind of GPU? It is known that mfaktc drives the GF100/GF110 chip very hard. Can you monitor the temperature of your GPU?

Can you reproduce this? Does it occur at the same point?
For a wrong memory access in mfaktc your screen does not look weird enough, I would expect "more colors", etc...

Oliver[/QUOTE]A second run failed to show reproduce the effect. At the start of the run the temperature was 32C. A few minutes in it settled down in a bimodal pattern with temperatures averaging 55C and 58C according to whether the "avg. rate" column contained ~158M/s or ~75M/s. The corresponding fan speeds were reported at 44% and 48% respectively. The absolute maximum values were 59C and 48%. The tests took 222m20s and 22m15s for exponents 56430991 and 56431079.

The gpu looks to be running cool enough to me and I still suspect the driver and/or X11 server.

Oh well, interesting experiment but it's unliikely I'll be running mfaktc 24/7 because of my other interests in CNT.

Paul

[QUOTE=xilman;258413]I'm doing something wrong.[/QUOTE]My guess would be that the manual results form doesn't like pre-release versions?

[QUOTE=xilman;258413]I'm doing something wrong. The results.txt file contains[code]no factor for M56399117 from 2^68 to 2^69 [mfaktc 0.16p1 barrett79_mul32]
no factor for M56399237 from 2^68 to 2^69 [mfaktc 0.16p1 barrett79_mul32]
[/code] but after uploading that file to [URL]http://mersenne.org/manual_result/[/URL] the server reports[code]
No factor lines found: 0
Mfaktc no factor lines found: 0
Factors found: 0
P-1 lines found: 0
LL lines found: 0
Mlucas lines found: 0
Glucas (G29) lines found: 0
Glucas lines found: 0
MacLucasFFTW lines found: 0
CUDALucas lines found: 0
ECM lines found: 0[/code]Paul[/QUOTE]Ah, it worked when all four results were submitted.

Very strange. Something took a dislike to me the first time around. No matter. It appears that I've 4.2387 GHz days to my credit. For an expenditure of 0.0622 days, the GPU seems to have an effective clock rate of 68GHz.

Paul

[QUOTE=James Heinrich;258419]My guess would be that the manual results form doesn't like pre-release versions?[/QUOTE]

George told me once that it should accept and record any mfaktc version string.
mfaktc 0.16p1 is [B]not[/B] a pre-release. 0.16p1 is 0.16 patch 1. Pre-releases have another string, e.g. 0.16-pre2.

Oliver

[QUOTE=TheJudger;258423]mfaktc 0.16p1 is [B]not[/B] a pre-release[/QUOTE]Yes, sorry, my mistake. What I meant was that it was an "unusual" version string. But it seems that on a second try the results were accepted, so I guess PrimeNet was just being grumpy. :smile: