Page 2 of 4

Re: GPUs.txt

Posted: Sun Apr 08, 2012 2:29 pm
by digibrat
That beta Core16 seems to at least do something, avg GPU load is now 94%
Thanks Jimmy,
hopefully they'll solve it soon :)

Re: GPUs.txt

Posted: Sun Apr 08, 2012 2:37 pm
by 7im
Are there any data integrity checks when anyone can put anything inside this file?! We don't want another -forcegpu data corruption problem!

Re: GPUs.txt

Posted: Sun Apr 08, 2012 3:36 pm
by itsmekirill
Can someone official confirm that running core 16 on GTX 680 is ok? It seems to work ok but I was under the impression that it was for ATI cards.

Re: GPUs.txt

Posted: Sun Apr 08, 2012 3:43 pm
by jimmyyang207
itsmekirill wrote:Can someone official confirm that running core 16 on GTX 680 is ok? It seems to work ok but I was under the impression that it was for ATI cards.
:wink:

Code: Select all

10:23:36:WU00:FS00:0x16:Completed  25000000 out of 25000000 steps (100%).
10:23:36:WU00:FS00:0x16:Finished fah_main status=0
10:23:36:WU00:FS00:0x16:Successful run
10:23:36:WU00:FS00:0x16:DynamicWrapper: Finished Work Unit: sleep=10000
10:23:46:WU00:FS00:0x16:Reserved 172272 bytes for xtc file; Cosm status=0
10:23:46:WU00:FS00:0x16:Allocated 172272 bytes for xtc file
10:23:46:WU00:FS00:0x16:- Reading up to 172272 from "00/wudata_01.xtc": Read 172272
10:23:46:WU00:FS00:0x16:Read 172272 bytes from xtc file; available packet space=786258192
10:23:46:WU00:FS00:0x16:xtc file hash check passed.
10:23:46:WU00:FS00:0x16:Reserved 21480 21480 786258192 bytes for arc file=<00/wudata_01.trr> Cosm status=0
10:23:46:WU00:FS00:0x16:Allocated 21480 bytes for arc file
10:23:46:WU00:FS00:0x16:- Reading up to 21480 from "00/wudata_01.trr": Read 21480
10:23:46:WU00:FS00:0x16:Read 21480 bytes from arc file; available packet space=786236712
10:23:46:WU00:FS00:0x16:trr file hash check passed.
10:23:46:WU00:FS00:0x16:Allocated 544 bytes for edr file
10:23:46:WU00:FS00:0x16:Read bedfile
10:23:46:WU00:FS00:0x16:edr file hash check passed.
10:23:46:WU00:FS00:0x16:Allocated 32040 bytes for logfile
10:23:46:WU00:FS00:0x16:Read logfile
10:23:46:WU00:FS00:0x16:GuardedRun: success in DynamicWrapper
10:23:46:WU00:FS00:0x16:GuardedRun: done
10:23:46:WU00:FS00:0x16:Run: GuardedRun completed.
10:23:47:WU00:FS00:0x16:+ Opened results file
10:23:47:WU00:FS00:0x16:- Writing 226848 bytes of core data to disk...
10:23:48:WU00:FS00:0x16:Done: 226336 -> 201889 (compressed to 89.1 percent)
10:23:48:WU00:FS00:0x16:  ... Done.
10:23:48:WU00:FS00:0x16:DeleteFrameFiles: successfully deleted file=00/wudata_01.ckp
10:23:48:WU00:FS00:0x16:Shutting down core 
10:23:48:WU00:FS00:0x16:
10:23:48:WU00:FS00:0x16:Folding@home Core Shutdown: FINISHED_UNIT
10:23:48:WU00:FS00:FahCore returned: FINISHED_UNIT (100 = 0x64)
10:23:48:WU00:FS00:Sending unit results: id:00 state:SEND error:OK project:8023 run:19 clone:1 gen:2 core:0x16 unit:0x000000036953ee304f4dce64cfa48e22
10:23:48:WU00:FS00:Uploading 197.66KiB to 171.67.108.144
10:23:48:WU00:FS00:Connecting to 171.67.108.144:8080
10:23:50:WU00:FS00:Upload complete
10:23:50:WU00:FS00:Server responded WORK_ACK (400)
10:23:50:WU00:FS00:Final credit estimate, 1200.00 points
10:23:50:WU00:FS00:Cleaning up

Re: GPUs.txt

Posted: Sun Apr 08, 2012 3:47 pm
by 7im
The is a core 16 for each. Look at the directory structure of where the core was downloaded from and to as noted in the log.

Re: GPUs.txt

Posted: Sun Apr 08, 2012 9:37 pm
by digibrat
seems i spoke too soon :-(

Code: Select all

19:34:10:WU01:FS00:0x16:Completed   6750000 out of 25000000 steps (27%).
19:34:10:WU01:FS00:0x16:mdrun_gpu returned 52
19:34:10:WU01:FS00:0x16:NANs detected on GPU
19:34:10:WU01:FS00:0x16:
19:34:10:WU01:FS00:0x16:Folding@home Core Shutdown: UNSTABLE_MACHINE
19:34:10:WU01:FS00:FahCore returned: UNSTABLE_MACHINE (122 = 0x7a)
19:34:10:WARNING:WU01:FS00:Too many errors, failing
19:34:10:WU01:FS00:Sending unit results: id:01 state:SEND error:FAILED project:8023 run:9 clone:2 gen:0 core:0x16 unit:0x000000016953ee304f4dce40c0507d95
19:34:10:WU01:FS00:Connecting to 171.67.108.144:8080
19:34:11:WU01:FS00:Server responded WORK_QUIT (404)
19:34:11:WARNING:WU01:FS00:Server did not like results, dumping
19:34:11:WU01:FS00:Cleaning up
I'll try another WU, see if that fails as well...

Re: GPUs.txt

Posted: Sun Apr 08, 2012 10:12 pm
by clawlor
Could you please whitelist the Radeon HD 7770 - device ID: 1002-683D please? In GPU-Z the name shows up as "AMD Radeon HD 7700 Series" if you need that too.

Re: GPUs.txt

Posted: Sun Apr 08, 2012 10:16 pm
by Swiftdeathz
Here is a GPU-Z screenshot of my Sapphire based AMD Radeon 7870 card which after tying to add it manually to the GPUs.txt file it still can't download any WU's.

Image

Uploaded with ImageShack.us

Re: GPUs.txt

Posted: Mon Apr 09, 2012 1:18 am
by itsmekirill
digibrat wrote:seems i spoke too soon :-(

I'll try another WU, see if that fails as well...
I had that problem as well, until I reset clocks to stock.

Re: GPUs.txt

Posted: Mon Apr 09, 2012 9:07 am
by friedrim
You need v2.24 of Core_15 to run on the new NVIDIA boards (Keplers); this version is currently only in beta.

Re: GPUs.txt

Posted: Mon Apr 09, 2012 9:57 am
by jimmyyang207
friedrim wrote:You need v2.24 of Core_15 to run on the new NVIDIA boards (Keplers); this version is currently only in beta.
How to assign v2.24 of Core_15? thx for help :D

Re: GPUs.txt

Posted: Mon Apr 09, 2012 4:30 pm
by 7im
Join the beta team, then use the beta team setting. client-type = beta

Note that forum support for clients using the beta setting is limited to beta team members, but since anyone can apply for the team, it's more of a formality. It's not a secrets thing, simply an organizational thing. Beta stuff is kept to separate section mostly for house-keeping purposes. You'll find yourself being ignored if you report bugs in the beta versions of FahCores because the people working on those cores are unlikely to read this topic.

Re: GPUs.txt

Posted: Mon Apr 09, 2012 6:56 pm
by bruce
jimmyyang207 wrote:I found a solution to make GTX 680 work :)
Just add "client-type:beta" to gpu slot, FAHClient will download core16 and start folding.
Core16 works on GTX 680, but low PPD? :?
See what 7im just said.

Getting impatient and testing things that are not yet working is frustrating, isn't it.

Complaints about beta WUs and beta FahCores WILL BE IGNORED unless you accept the terms and conditions of joining the beta team.

To join, follow this link: viewtopic.php?f=2&t=10364

Re: GPUs.txt

Posted: Mon Apr 09, 2012 9:32 pm
by 7im
Swiftdeathz wrote:Here is a GPU-Z screenshot of my Sapphire based AMD Radeon 7870 card which after tying to add it manually to the GPUs.txt file it still can't download any WU's.
Ticket # 883

Ran a diff on the GPUs.txt downloaded today, with AMD driver device list from 03/28/12...

Devices in driver list, not in GPUs.txt...

Vendor ID: 1002
PCI IDs "Description"

675F "AMD Radeon HD 6510 Series"
9805 "AMD Radeon HD 6250 Graphics"
9645 "AMD Radeon HD 6410D"
9809 "AMD Radeon HD 7310 Graphics"
9808 "AMD Radeon HD 7340 Graphics"
683D "AMD Radeon HD 7700 Series"
683F "AMD Radeon HD 7700 Series"
6819 "AMD Radeon HD 7800 Series"
6818 "AMD Radeon HD 7800 Series"
9649 "AMD Radeon(TM) HD 6480G"
958A "ATI Radeon HD 2600 X2 Series"

Re: GPUs.txt

Posted: Tue Apr 10, 2012 1:40 am
by bruce
7im wrote:Are there any data integrity checks when anyone can put anything inside this file?! We don't want another -forcegpu data corruption problem!
That's not possible. We cannot add arbitrary WUs to the whitelist. Only Stanford can authorize a new entry to the copy of GPUs.txt that they maintain. For a GPU to work, (1) it has to be on the official list and (2) the client has to know about it.

In the past, the official list might be updated but you still couldn't run that GPU because they weren't ready to release an updated client. Now you can download a fresh copy of GPUs.txt without Stanford having to release a new version of the client. Manually editing GPUs.txt is a waste of time, but if there has been an update, downloading a fresh copy will allow you to use the new entries.

There were four steps and now there are three. (1) Somebody needs to open a ticket to get it added. (2) Stanford has to add it to their copy and (3) you can refresh your copy [and (4) you don't need a new release of V7.]

Someday soon, there will be an updated version of V7 which will perform step 3 semi-automatically, leaving one less step, but that may not be soon. Step 2 will ALWAYS be required.