NEW CORE for gpu client Thread. Updated regularly.

Printable View

Show 100 post(s) from this thread on one page

11-03-2008, 03:22 PM
OldChap

Thanks MtM, but I'm happy to stay Pure. They are a good bunch even if most don't stick with folding (and they call me grumpy!)
11-03-2008, 03:49 PM
Marvin_The_Martian

Ow but we're a good bunch to :buddies:

But hey, you don't have to be on our team to get my respect, if there is ever anything I can do for you just let me know :)
11-03-2008, 04:12 PM
road-runner

Quote:

Originally Posted by OldChap

Gpu core 1.19 available

http://www.stanford.edu/~friedrim/FahCore_11.exe.

edit: Back to 1.09 speeds for me

Slower than 1.15 for sure... Now where the heck did I put the bookmark of the manual download page, I need 1.15 back...
11-03-2008, 08:40 PM
Kingcarcas

Pip old chap! I see you decided to join us :D
11-04-2008, 12:38 AM
Marvin_The_Martian

Quote:

Originally Posted by road-runner

Slower than 1.15 for sure... Now where the heck did I put the bookmark of the manual download page, I need 1.15 back...

PM Kondik if you can't find it :)

Hate other people getting quoted for stuff I post here only because someone has a problem reading it seems :shrug:
That's not aimed at you RR :up:

Edit* I'll put what I posted before in this post just for 'completeness'

Quote:

Originally Posted by Marvin_The_Martian ;)

Hey people a new core is out, version 1.19 you can download it from 'link pasted in first post by Mike' :up:

We informed a mod that the thread title needed a change so we could put all the new cores into this thread but that hasn't been done yet so I posted this in a new thread. When a mod changes the title he can merge this post with the thread as well to keep the clutter down.

I been folding on this core for some time now, and my eue's have dissapeared as snow for the sun. The new core is basicly a 1.09 core from late summertime, so no optimizations from the 1.15 and 1.17/1.18 cores gone into it. This is because on the larger proteins the single precision nature of the gpu's together with the optimizations caused the results to be unstable, hence the many eue's people got.

I would urge especially the people with 8600 boards who haven't been able to do a couple of wu's in a row succesfully to download the new core. People with gt200 cards who have a much better succes rate with the 1.18 core can decide to keep that core for the time being, though it's probable that when Stanford decides it's time for larger proteins to be folded that the new core will become an auto download for those wu's.

Just to note, before this core, I ran my 9600gt at 1900 shaders and was a tad below 4K most of the time, with this new 1.19 core and no oc shaders 1600 ( beta testing core's and an oc you're not sure has influence on the eue rate is a :nono: imo ) I'm down to a tad over 2K, running wu's I can't directly compare though. The thread on the folding forums is growing though and more people are reporting ppd with the new core so you can take your info from there.

Thanks for looking :wave:
11-04-2008, 12:46 AM
Kondik

RR i have the 1.15 saved , I used the new 1.18 and 1.19 but they are ... ummm ... my VGA does 1% around 50 ~ 55 sec , with 1.18 and 1.19 it's over 64 seconds ...
my ppd went down by a solid amount, back to 1.15
11-04-2008, 06:12 AM
road-runner

Quote:

Originally Posted by Kondik

RR i have the 1.15 saved , I used the new 1.18 and 1.19 but they are ... ummm ... my VGA does 1% around 50 ~ 55 sec , with 1.18 and 1.19 it's over 64 seconds ...
my ppd went down by a solid amount, back to 1.15

Thanks, I was not thinking when I posted that. I had 1.15 on my other rigs, just opened the folder and got a copy...
11-06-2008, 06:55 PM
coo-coo-clocker

Now I know the answer to this is probably in 10 different threads on here, but I'm feeling tired and lazy right now, too much folding fluid this evening I guess...

For dual GPU folding on a C2D rig, I know I cannot use AC to "do the magic" to balance the two fah cores across the two CPU cores. I went in and manually assigned them, but I also recall that when each WU finishes, that instance of the fah core shuts down and a new starts up upon the new WU arrival. How does one go about making a lasting fix on this?
11-06-2008, 06:57 PM
Marvin_The_Martian

Quote:

Originally Posted by coo-coo-clocker

Now I know the answer to this is probably in 10 different threads on here, but I'm feeling tired and lazy right now, too much folding fluid this evening I guess...

For dual GPU folding on a C2D rig, I know I cannot use AC to "do the magic" to balance the two fah cores across the two CPU cores. I went in and manually assigned them, but I also recall that when each WU finishes, that instance of the fah core shuts down and a new starts up upon the new WU arrival. How does one go about making a lasting fix on this?

I been trying to get people to use this

http://sourceforge.net/project/platf...roup_id=238550

As it has built in core logic, you can make sure smp is on the same die, and have two cores for gpu folding/other things.
11-07-2008, 04:46 AM
coo-coo-clocker

Thanks for the link, Marvin. Great program!! :up:
11-07-2008, 06:08 AM
Marvin_The_Martian

Quote:

Originally Posted by coo-coo-clocker

Thanks for the link, Marvin. Great program!! :up:

If you like and play with it, I think the whole team could benefit if you showed us some examples about how to use it and how big if any the benefits are :up:

I couldn't because of my os and hw I'm running, no much sense in starting a thread about it if I can't show the benefits. But it should really help with smp's because you will loose fsb traffic using two cores on the same die I would expect.
11-07-2008, 04:46 PM
catawalks

Well, I was using fahwatch to look through my logs at old WU's I've done, and apparently I did a 5748 and didn't even notice it. Now I've got both 1.15 and 1.19. I guess they renamed core 1.19 to FahCore_13.exe so that it wouldn't overwrite the old core? Any way of telling how many points I got for that WU?

Code:

Os version: Microsoft Windows NT 6.0.6001 Service Pack 1 User: JOELZ0R\Administrator Account: Administrator --------------------------------------------------------------------------------- -FAHMON NAME: GTX260 - Desktop INDEX: 1 LOCATION: C:\Users\Administrator\AppData\Roaming\Folding@home-gpu\ LOCAL CLIENT: True CLIENT: Windows GPU Console Edition - Folding@Home Client Version 6.20r1 FLAGS: -gpu 0 CORE: Folding@Home GPU Core - Beta - Version 1.19 (Thu Oct 23 10:15:05 PDT 2008) StartTime: 08:01:03 StartPercentage: 1 EndTime: 10:38:43 EndPercentage: 100 -VIDEO CARD Caption NVIDIA GeForce GTX 260 Description NVIDIA GeForce GTX 260 AdapterRam 896MB DriverVersion 7.15.11.7824 DriverDate 20081007173300.000000-000 Index 0 InfSection nv_GT2x COMPILER: Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14.00.50727.762 for 80x86 BUILDHOST: amoeba BOARD: Nvidia PROJECT: Project: 5748 (Run 1, Clone 115, Gen 2) QUEUESLOT: Working on queue slot 06 [November 6 03:19:15 UTC] CORESTATUS: CoreStatus = 64 (100) TIMESTAMP: [10:38:56]
11-07-2008, 04:48 PM
MikeB12

if you completed it you got 511.. but if you wanna see your ppd, look at fahmon.. menu-benchmarks.
11-07-2008, 04:56 PM
Marvin_The_Martian

There is also a project browser in fahwatch :) Options, project browser :up: Fahcore_13 getting assigned to everyone was a glitch I think, since it should be a beta only core :)
11-07-2008, 05:12 PM
catawalks

lol...fahmon didn't really have much to say:

Project : 5748 (Unknown)

-- GTX260 - Desktop --

Min. Time / Frame : 1mn 34s
Avg. Time / Frame : 1mn 34s
Cur. Time / Frame : 1mn 35s
R3F. Time / Frame : 1mn 34s
Eff. Time / Frame : 1mn 34s

And nice touch Marvin. I like that project browser.
11-07-2008, 05:16 PM
MikeB12

that's 4696ppd cat.. very good...

my 8800/9800gt's got 3500ppd

sp's obviously make a big difference in larger atom wu's.. guess I need to uncheck "larger wu's" in my clients. if that even matters.
11-07-2008, 05:22 PM
Marvin_The_Martian

Shouldn't matter Mike, gpu currently has no disctinction between wu sizes. Those 12xx atoms you will get even with small wu's ( and btw it's not the atom count but the upload size of the completed wu's ).
11-07-2008, 05:28 PM
MikeB12

Quote:

Originally Posted by Marvin_The_Martian

Shouldn't matter Mike, gpu currently has no disctinction between wu sizes. Those 12xx atoms you will get even with small wu's ( and btw it's not the atom count but the upload size of the completed wu's ).

dam :down:
11-07-2008, 05:39 PM
Marvin_The_Martian

Yeah you think I like getting those big wu's with a 64shaders card? Ow no wait I beta test. I'll always get the largest wu's I think :rofl:
11-07-2008, 07:09 PM
catawalks

Whoa, workin on 5014 now and fahmon is showing its ppd up to 7.5k. Awesome. Keep those 5014's rollin in.
11-07-2008, 09:43 PM
sniperbob

i keep getting stuck with 5506's really throwing my total PPD down a good 2k points using 1.19... oh well, i guess thats what happens in the wonderful world of folding...
11-08-2008, 12:40 AM
DAK1640

Can we still use 1.15???
11-08-2008, 12:43 AM
MikeB12

1 Attachment(s)

just delete your core dak. and let the client dl one. there still sending the optimized core..
11-08-2008, 12:59 AM
DAK1640

can I do it while it's already in the middle of a WU?
11-08-2008, 01:10 AM
MikeB12

sure, just stop the client, delete the core/s, and restart it..

Show 100 post(s) from this thread on one page