When milky admins will upgrade server to newest version with ATI applications support?

Probably after it's released. Official BOINC support for ATI is still in development.

Probably after it's released. Official BOINC support for ATI is still in development.

Really ?

http://boinc.berkeley.edu/trac/timeline


Now scroll down that page and start reading at
15:54 Changeset [18846] by davea

boinc_6.10.0_windows_intelx86.exe
boinc_6.10.0_windows_x86_64.exe

---

Join Support science! Joinc Team BOINC United now!

Probably after it's released. Official BOINC support for ATI is still in development.

Really ?

Just because stuff has been committed, it doesn't mean the support is finished. Likely, there'll be bugs.

Also, DA said back in February that ATI support was likely a month away. Rom said the same thing yesterday. We'll see how that turns out.

---

Well done Crunch3r. Truly awesome job! I look forward to playing with this tonight once I get home from w#$^.

@ Crunch3r
Good job I noticed a big improvement on the Cuda. My computer has been a little sluggish with Cuda running(Collatz) noticed a big improvement in my computers performance immediately. Kool Thanx
;-p

---

>We'll see how that turns out.
I'll keep running Cluster's app for the moment.

---

Cheers,

PeterV

.

@ Crunch3r
Congrats on getting some well deserved recognition and your name up in lights. I'm sure your GPU detection code advanced things considerably. The official release may actually happen before Christmas now. :)

I'll keep running Cluster's app for the moment.

It doesn't exclude each other. Quite the opposite ;)

Howdy !

I installed 6.10.0 to give it a try. It did not detect my Radeon 4850
(XFX Brand) for some reason but, it is working just fine.

Crunching away.

Bill



8/19/2009 12:50:43 PM Starting BOINC client version 6.10.0 for windows_intelx86
8/19/2009 12:50:43 PM log flags: task, file_xfer, sched_ops
8/19/2009 12:50:43 PM Libraries: libcurl/7.19.4 OpenSSL/0.9.8k zlib/1.2.3
8/19/2009 12:50:43 PM Data directory: D:\Program Files\Boinc\Application Data
8/19/2009 12:50:43 PM Processor: 2 GenuineIntel Intel(R) Core(TM)2 CPU X6800 @ 2.93GHz [x86 Family 6 Model 15 Stepping 5]
8/19/2009 12:50:43 PM Processor: 4.00 MB cache
8/19/2009 12:50:43 PM Processor features: fpu tsc sse sse2 mmx
8/19/2009 12:50:43 PM OS: Microsoft Windows XP: Professional x86 Edition, Service Pack 3, (05.01.2600.00)
8/19/2009 12:50:43 PM Memory: 2.00 GB physical, 9.85 GB virtual
8/19/2009 12:50:43 PM Disk: 417.90 GB total, 288.70 GB free
8/19/2009 12:50:43 PM Local time is UTC -5 hours
8/19/2009 12:50:43 PM Can't load library nvcuda.dll
8/19/2009 12:50:43 PM No usable CAL devices found
8/19/2009 12:50:43 PM No coprocessors
8/19/2009 12:50:43 PM Einstein@Home Found app_info.xml; using anonymous platform
8/19/2009 12:50:43 PM Milkyway@home Found app_info.xml; using anonymous platform
8/19/2009 12:50:43 PM Not using a proxy

20/08/2009 5:04:26 AM Can't load library nvcuda.dll
20/08/2009 5:04:26 AM ATI GPU 0: RV770 (CAL version , 1024MB, 1088GFLOPS)
20/08/2009 5:04:26 AM ATI GPU 1: RV770 (CAL version , 512MB, 1088GFLOPS)

GPU 0 is a 4870
GPU 1 is a 4850

To: boinc_projects, boinc_dev

BOINC now supports ATI GPUs.
The client detects them and manages their work queue as it does for NVIDIA GPUs.
This will be in the 6.10 client, which should be released in a month or so.
The server code now also accommodates ATI GPUs.

In addition, the coprocessor model has been extended
to allow jobs to use a fraction of a GPU.
For example, a job that will at most half of a GPU's processors
and at most half of its RAM should be specified as using 0.5 of the GPU
(this is done in the app_plan() function; the fraction in general
will depend on the characteristics of the particular GPU).
The BOINC client will try to run multiple jobs on a GPU
if their usage fractions total 1 or less.

Milkyway@home will be offering an ATI app fairly soon.
I encourage other projects to investigate this possibility also.
Let me know if any questions or problems.

-- David

---

I have 6.10.0 running on a laptop and it seems to be running better with a couple of CUDA WUs going further than the crashes I have been having recently. So that's good. So far running OK with 6.10.0.

I tried it with an ATI GPU and it started off well, grunching two GPU MW and crunching one CPU CPDN. But then it wouldn't download any more MW. I haven't had much time to have a good look at it and hope to give it another go soon. But it's great to see the ATI detection - nice one Crunch3r.

---

21-8-2009 16:54:25 Local time is UTC +2 hours
21-8-2009 16:54:26 Can't load library nvcuda.dll
21-8-2009 16:54:26 ATI GPU 0: RV770 (CAL version , 1024MB, 1088GFLOPS)
21-8-2009 16:54:26 ATI GPU 1: RV770 (CAL version , 1024MB, 1088GFLOPS)
21-8-2009 16:54:26 Milkyway@home Found app_info.xml; using anonymous platform
21-8-2009 16:54:26 SETI@home Found app_info.xml; using anonymous platform
21-8-2009 16:54:26 Not using a proxy
21-8-2009 16:54:26 Version change (6.4.7 -> 6.10.0)

looks good to me it finds both my XFX 4850 cards immediate
I cant run any other project on my cpu's since i have a really hard time getting my ati's enough units to run all the time.
Basically i get only 1 unit every minute while i should get at least 2 or 4 so i guess its time for MW admins to set the cpu limit to 9 or so
Instead of the really low 6

---

Its new, its relative fast... my new bicycle

I tried 6.10.0 for a while and noticed some errors being reported. As with other late 6 versions, the "CPU time" starts for all work immediately upon download and evidently some of the tasks were timing out. I saw several error out with "Maximum elapsed time exceeded" messages. I guess Crunch3r's off the hook for this one though. ;)

the only issue i have is that i am constant not getting any new work :

21-8-2009 17:10:38 Milkyway@home Sending scheduler request: To fetch work.
21-8-2009 17:10:38 Milkyway@home Not reporting or requesting tasks
21-8-2009 17:10:43 Milkyway@home Scheduler request completed: got 0 new tasks

---

Its new, its relative fast... my new bicycle

21-8-2009 16:54:25 Local time is UTC +2 hours
21-8-2009 16:54:26 Can't load library nvcuda.dll
21-8-2009 16:54:26 ATI GPU 0: RV770 (CAL version , 1024MB, 1088GFLOPS)
21-8-2009 16:54:26 ATI GPU 1: RV770 (CAL version , 1024MB, 1088GFLOPS)
21-8-2009 16:54:26 Milkyway@home Found app_info.xml; using anonymous platform
21-8-2009 16:54:26 SETI@home Found app_info.xml; using anonymous platform
21-8-2009 16:54:26 Not using a proxy
21-8-2009 16:54:26 Version change (6.4.7 -> 6.10.0)

looks good to me it finds both my XFX 4850 cards immediate
I cant run any other project on my cpu's since i have a really hard time getting my ati's enough units to run all the time.
Basically i get only 1 unit every minute while i should get at least 2 or 4 so i guess its time for MW admins to set the cpu limit to 9 or so
Instead of the really low 6

Whst is your cache set at? Having a cache set higher thsn the MW 3 day deadline can result in that behavior.

---

Calm Chaos Forum...Join Calm Chaos Now

hmm checked i have it set at 2 days extra work
and connect at 0,00
in the last 30 minutes i got 2 units from MW server

---

Its new, its relative fast... my new bicycle

the only issue i have is that i am constant not getting any new work :

21-8-2009 17:10:38 Milkyway@home Sending scheduler request: To fetch work.
21-8-2009 17:10:38 Milkyway@home Not reporting or requesting tasks
21-8-2009 17:10:43 Milkyway@home Scheduler request completed: got 0 new tasks

Increase your MW resource share to over 50%.....

resource share set at 1000 %
all other projects set at 5% and set to inactive and no new work
editted the app_info.xml to n1 f20 w1.1
lol gpu load 50 and 61 % xD

---

Its new, its relative fast... my new bicycle