Yeash. I have 1500ish pendings, and of those, literally hundreds have been inconclusively validated by at least two people.

What gives?

---

Executive Director GPU Users Group Inc. -
brad@gpuug.org

Yeash. I have 1500ish pendings, and of those, literally hundreds have been inconclusively validated by at least two people.

What gives?

LoL, You're processing very very fast, that's what.

---

"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.

Yeash. I have 1500ish pendings, and of those, literally hundreds have been inconclusively validated by at least two people.

What gives?

LoL, You're processing very very fast, that's what.

You are also running the new installer which is more accurate than before. Because of that you are not quite similar to other results. This is a good thing but in a very few cases two older results will gang up on you and throw an invalid result your way. Usually though, all three will validate with the older ones giving results on both sides of yours.

---

PROUD MEMBER OF Team Starfire World BOINC

Yeash. I have 1500ish pendings, and of those, literally hundreds have been inconclusively validated by at least two people.

What gives?

LoL, You're processing very very fast, that's what.

You are also running the new installer which is more accurate than before. Because of that you are not quite similar to other results. This is a good thing but in a very few cases two older results will gang up on you and throw an invalid result your way. Usually though, all three will validate with the older ones giving results on both sides of yours.

I'll also add, perryjay & Slavac, that I'm watching the newer cards like those 560ti's & the GTS 450, very closely to determine what kinks to iron out next.

At present I do believe they, and some other cards, do have some quirks that'll need to be compensated for.

You're running a slightly slower prelim diagnostic build for me perryjay. Any signs of the downclock you were getting ? If not I might get Slavac sorted with the same build while I continue going through everything & watching x38g behaviour. Weird this stuff never turns up in testing ::/ .

Jason

---

"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.

No down clock yet. I was having a problem with my computer running a bit sluggish but a reboot ( and deleting MSI Afterburner) seems to have cured it. This version doesn't seem to be noticeably slower than the 0.38g. I'm still chugging along quite well thank you.

Before anyone hollers, I don't know if it was MSI Afterburner or something else I did around the same time but what ever. I just got rid of the last thing I installed. Could have been a problem with the way I installed it in the first place too.

---

PROUD MEMBER OF Team Starfire World BOINC

Is there a way to just see the inconclusives in the pendings? If not, that sure would be nice.
Thanks
Dan

---

I'm still chugging along quite well thank you.

Keep doing that :). The stability change points to unoptimised(stock) code portions having some issues (on newer hardware). Looks like it's time to take the gloves off & rewrite the last of nVidia's code...

Jason

---

"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.

Dan. That would be nice. I've been thinking about making a suggestion over in the wishlist section of the Questions and answers forum.


Jason, sounds like you have your work cut out for you. Let me know how I can help if you need me.

---

PROUD MEMBER OF Team Starfire World BOINC

...Jason, sounds like you have your work cut out for you. Let me know how I can help if you need me.

Don't worry, just keep doing what you're doing.

jason

---

"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.

It almost seems like a there's a very stupid self
protection routine running on the card.
...

See how this fares.

There are several, mostly in the drivers, and in part it is those I'm watching to see that the improved boincapi code is doing its job. Indeed the original issue that caused downclocking on a wider range of cards appears to be mitigated, though, as you point out, some newer cards have found some other loophole.

At this stage it looks like dialling back some code in a diagnostic build has helped perryjay's 450, though it'll need watching for a few days to be certain.
If you like you could try the same diagnostic build (x39c) found at:

http://lunatics.kwsn.net/12-gpu-crunching/x38g-reports.msg39214.html#msg39214

Jason

---

"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.

Still running great here. Made it through the night with no problems. No sign of anymore down clocking yet.

---

PROUD MEMBER OF Team Starfire World BOINC

It almost seems like a there's a very stupid self
protection routine running on the card.
...

See how this fares.

At this stage it looks like dialling back some code in a diagnostic build has helped perryjay's 450, though it'll need watching for a few days to be certain.
If you like you could try the same diagnostic build (x39c) found at:

http://lunatics.kwsn.net/12-gpu-crunching/x38g-reports.msg39214.html#msg39214

Jason

I'm also running x39c Started @ 11:20 am est June 25/11
EDIT: 2 wu at a time, factory overclock @ 850MHz. I might try 3 at a time again after this has run for a week or so.

Thanks

---

</Tazz>

Well, it looks like this new build is going to get a pretty good workout. Glad we got a few more testers.

Tazz, you might want to check your setup again, I'm seeing you as still running the V0.38g. (I found I had to add .exe to the end when I substituted the 0.39c build in my app info so it came out looking like <name>Lunatics_x39c_win32_cuda32.exe</name>. Could that be what is happening to you?)(Bold text.exe added by me just to show my change.)

---

PROUD MEMBER OF Team Starfire World BOINC

Well, it looks like this new build is going to get a pretty good workout. Glad we got a few more testers.

Tazz, you might want to check your setup again, I'm seeing you as still running the V0.38g. (I found I had to add .exe to the end when I substituted the 0.39c build in my app info so it came out looking like <name>Lunatics_x39c_win32_cuda32.exe</name>. Could that be what is happening to you?)(Bold text.exe added by me just to show my change.)

Just finished the first wu. Well, the first half was crunched by 38g and the last half by 39c. Completed, validation inconclusive. Up against a 32f, resent to a cuda23.

http://setiathome.berkeley.edu/workunit.php?wuid=766593410

Also noticed driver 267.59 showed 993 MiB and 275.33 shows 961 MiB Go figure.

EDIT: This will be the first wu crunched only with x39c.
http://setiathome.berkeley.edu/workunit.php?wuid=766623597

---

</Tazz>

Well, it looks like this new build is going to get a pretty good workout. Glad we got a few more testers.

Wouldn't mind getting in on the testing. That would be my first time doing so, should I do anything special, are you testing for something or just general testing the new build?

It is a test build for those of us that seemed to still be getting down clocking with the new V0.38g app. It gives a bit more information in the stderr of the compleed work so as to be easier to find any problems we encounter. Hop on and join the crowd if you would like to. Since Jason has made this test available to the public I guess he wouldn't mind.

Just to add, My system started getting sluggish again. Didn't down clock but everything was slow. To completion times started to climb and even my fans slowed. I shut down Firefox, just to see but it didn't help so I shut down Thunderbird too. Still slow. Then I shut down connected client and exited BOINC. When I restarted everything was back up to speed. My problem could be just my little 450 doesn't like running three WUs at a time 24/7 and likes to take a break now and then. I think I will go back to two at a time and see if it does it again.

---

PROUD MEMBER OF Team Starfire World BOINC

... My problem could be just my little 450 doesn't like running three WUs at a time 24/7 and likes to take a break now and then. I think I will go back to two at a time and see if it does it again.

Sounds plausible. Newer drivers & Cuda apps use a lot more kernel resources, though 4Gig system RAM (yielding ~2Gig kernel space depending on OS) for 3 tasks should be heaps. Possibly as I refine the suspects indicated by teh more successful run, the issue could go away as well.

Jason

---

"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.

I've cut back to two at a time now. I'll let you know how it goes.

Yeah, it was strange but it did seem like it took it about a day to build up to where it needed to take a nap. Could be just the way I keep things open that is adding to it. We shall see.

---

PROUD MEMBER OF Team Starfire World BOINC

Hmmm, I just noticed an inconclusive workunit where BOTH results were crunched using x38g:

http://setiathome.berkeley.edu/workunit.php?wuid=766872407

Flopcounter for both results is identical, but the spike count is different (4,1,0,0) vs (3,1,0,0).

Can this be caused by the different GPUs being used (GTX 580 vs GT 240) ?

Tom

My Guess is, the GT 240 will win! :D LOL

Helli