NEW CORE for gpu client Thread. Updated regularly.

[gallardo]

Didn't read all 4 pages so dunno if it was posted, but my F@H won't get the new core.
I delete the core_11.exe and restart fah, but it just downloads core_11 again. I did this 3 times already...

[p2501]

The new client has yet to affect my 9600gso's. I'm getting the exactly same ppd (5057.56) as before.

Is this core targeted towards better GPUs?

Definitely not, it's for all nV cores. My GSO saw some 8% gain on project 5800 as I wrote here. Maybe your FahMon is acting up again? Did you set it to L3F for computing the PPD?

Didn't read all 4 pages so dunno if it was posted, but my F@H won't get the new core.
I delete the core_11.exe and restart fah, but it just downloads core_11 again. I did this 3 times already...

That's absolutely correct, it still is core 11. You need to look carefully:

[06:55:17] Core required: FahCore_11.exe
[06:55:17] Core found.
[06:55:17] Working on queue slot 06 [October 17 06:55:17 UTC]
[06:55:17] + Working ...
[06:55:17] - Calling '.\FahCore_11.exe -dir work/ -suffix 06 -priority 96 -checkpoint 15 -verbose -lifeline 3288 -version 620'

[06:55:17] 
[06:55:17] *------------------------------*
[06:55:17] Folding@Home GPU Core - Beta
[06:55:17] Version 1.15 (Mon Oct 13 11:11:30 PDT 2008)

It's all good!

[Sparky]

yep what he said, you've gotten the new core, same filename, just a new version

[coo-coo-clocker]

The new client has yet to affect my 9600gso's. I'm getting the exactly same ppd (5057.56) as before.

Is this core targeted towards better GPUs?

how many SPs on that GPU? Perhaps improvement is relative to number of SPs?

[p2501]

how many SPs on that GPU? Perhaps improvement is relative to number of SPs?

It's got 96SP's, the new core definitely has an impact.

[Shadowtester]

The general impression of the discussion over on the Stanford Folding forum is that with the new code they are trying to optimize and make overclocking of the card not as necessary for maximum ppd. As most can see by having to reduce their overclocking to remain stable while still seeing any improvement in ppd. The GPU2 client and folding is a ever evolving thing I doubt we have seen the final evolution yet.

[Marvin_The_Martian]

Well since I removed the flag getting me those trouble projects it's been running fine again, even with the new core. Glad Voelz decided to pull them, but even then I had to remove the flag to ger rid of them still showing up ( must have been one of the latest ).

[Shadowtester]

Well since I removed the flag getting me those trouble projects it's been running fine again, even with the new core. Glad Voelz decided to pull them, but even then I had to remove the flag to ger rid of them still showing up ( must have been one of the latest ).

Yea those projects even with the beta flag have been pulled except for ones they have already put out and had comeback EUE so they are some still floating around. Lately I have been getting a pretty steady diet of 5800 wu's with the flag and I still get a few EUE's maybe 1 out of 10 wu's folded average which I do not think it too bad its heck of a lot better than the 40-50% I was getting when the 57xx projects were in being pushed full force.

[gallardo]

Right.... it's all good!

Didn't think to check the log file...

[[XC] Synthetickiller]

I'm trying to update today and the client wants to download a new core, even though I manually replaced the old core w/ the new one along w/ the dll files. It will connect to the server and download, but says that there is a "core download error" every time.

Any ideas?

Edit:

One of my GSOs is working fine. Downloaded and updated the core. It's getting 5057.56ppd STOCK. I just OC'd it and am awaiting the new ppd. My other GSO in the first slot will not update. I even tried copying the core from the other folder. No luck.

[Sir_Loin]

To upgrade, stop the client, delete only the old core, and let it auto download the new core. You other files should've been fine.

Of course, server issues on the other side isn't beyond the realm of possibility.

[[XC] Synthetickiller]

To upgrade, stop the client, delete only the old core, and let it auto download the new core. You other files should've been fine.

Of course, server issues on the other side isn't beyond the realm of possibility.

I had to copy and paste everything to a new folder. It wouldn't let me delete the actual core.

Anyways, its working now.

[MikeB12]

I'd advise everyone to check their logs for UNSTABLE_MACHINE instances...

I went thru all 4 gpu clients, both logs:
FAHlog.txt
FAHlog-Prev.txt

and did a find for "UNSTABLE_MACHINE"
8 logs, 5 days and 11 eue's...
these cards run 24/7, and crank out like 10 wu's a day each.
so I know that's not alot, but still... a 5-10% eue rate..

so I went down a notch on all my cards shader oc.
8800gt's and 9800gt is now 1728
9600gso is now 1674

see if it helps any... dunno, it may just be the new core. but I don't remember seeing this many before with the old core.

[Marvin_The_Martian]

I'd advise everyone to check their logs for UNSTABLE_MACHINE instances...

I went thru all 4 gpu clients, both logs:
FAHlog.txt
FAHlog-Prev.txt

and did a find for "UNSTABLE_MACHINE"
8 logs, 5 days and 11 eue's...
these cards run 24/7, and crank out like 10 wu's a day each.
so I know that's not alot, but still... a 5-10% eue rate..

so I went down a notch on all my cards shader oc.
8800gt's and 9800gt is now 1728
9600gso is now 1674

see if it helps any... dunno, it may just be the new core. but I don't remember seeing this many before with the old core.

Do you run ****? I asked for my thread here to be deleted because my errors where all related ( excluding two exceptions but they where spread out enough to be coincidental ) to only those projects.

[MikeB12]

no, I've never run that on the gpu clients. just smp. my gpu clients are just normal no flag clients. I saw your thread, but didn't know why it was eueing until shadow mentioned it.. afterwards, I looked in the beta forum and saw what yall were talking about.

[Marvin_The_Martian]

Lol I keep forgetting that... I mean everyone who I know who folded for some time knows the flag, when I started on fm people used it to boost ppd everywhere. Now it's less beneficial, well atleast not always If you pay enough attention you won't get the issues I had ( 11 eue's in 3 days ).

I got banned from the forum once for making an installer which exposed the flag as a normal setup option

Btw, been working..

[MikeB12]

yeah, just thought I'd mention it since 7im slapped me for it recently. I really dont know why some of that stuff is so secretive.. but it's their project and forum, so their rules.. whatcha gonna do.. as the old saying goes..

[Marvin_The_Martian]

Same feeling here, though I must say they changed allot in how they deal with it. Seems allot of newer people don't know the flag. I think two things are responcible, most importantly using it mostly brings trouble but also the added value has dropped on it's own, both responcible for not many teams openly admitting to all using the flag ( which is how I remember it from the 'old' days ). I would rather see them make the flag public but add a check in the client to see if you have the elevated forum rights ( and responcibilities.. I forgot that last week ) so it won't be of any use if you don't have shown the intrest in what the flag is supposedly used for.

[DAK1640]

I'd advise everyone to check their logs for UNSTABLE_MACHINE instances...

I went thru all 4 gpu clients, both logs:
FAHlog.txt
FAHlog-Prev.txt

and did a find for "UNSTABLE_MACHINE"
8 logs, 5 days and 11 eue's...
these cards run 24/7, and crank out like 10 wu's a day each.
so I know that's not alot, but still... a 5-10% eue rate..

so I went down a notch on all my cards shader oc.
8800gt's and 9800gt is now 1728
9600gso is now 1674

see if it helps any... dunno, it may just be the new core. but I don't remember seeing this many before with the old core.

Wow Mike, my 8800GT's run 1812 Shaders, no eue's

[MikeB12]

1836 used to be stable on core 1.09, it's not on core 1.15


my shader stops are 1782 or 1836... I can set it on whatever and it'll just round it off to the next hop.... are you sure it's 1812?

[Marvin_The_Martian]

Lol Dak yeah that's old news, didn't know you wouldn't know

[Marvin_The_Martian]

http://foldingforum.org/viewtopic.php?f=52&t=6552

No showstoppers were found during the beta testing of v1.18, but we do not want to force a core upgrade at this time. If you are having problems with v1.15, please try replacing it with this newer version. It can be downloaded from http://www.stanford.edu/~friedrim/FahCore_11.exe. Thanks.

Does not fix the 5801 issue, more info is in the thread

[p2501]

Did anyone backup their 1.09 to test if 5801 is stable on them like Pande said? I'm not at home atm so I wouldn't know if they're still being sent out.

[Marvin_The_Martian]

Did anyone backup their 1.09 to test if 5801 is stable on them like Pande said? I'm not at home atm so I wouldn't know if they're still being sent out.

Where did he say that? 5801 is taken down because it was unstable with all cores, 5800 went through QA on 1.09 I think you mean and 5801 is almost a copy ( or is a copy? ) so their abit stumped on how it happend.

That's what I can make of it?

[p2501]

Where did he say that? 5801 is taken down because it was unstable with all cores, 5800 went through QA on 1.09 I think you mean and 5801 is almost a copy ( or is a copy? ) so their abit stumped on how it happend.

That's what I can make of it?

Oh I think the link was in the let's-all-complain-about-p5801-thread. I think, but I'll check later. I'm not mixing up things then Vijai or whoever answered to a post complaining about fail at QA generally, saying that those in fact might have been checked with an earlier core then 1.15. His statement was very vague though.