Posts by Skip Da Shu
log in
1) Message boards : Number crunching : ATI 9.12 / Linux64 / BOINC problem... (Message 9684)
Posted 2685 days ago by Skip Da Shu
I have 2 HD5870 on Gentoo Linux system.
I have had to configure it as crossfire otherwise boinc only see one gpu.


Well that is the answer I was after, thanx much.

2) Message boards : Number crunching : ATI 9.12 / Linux64 / BOINC problem... (Message 9664)
Posted 2687 days ago by Skip Da Shu
I am missing something... probably something I've even read before... but not coming to me now...

I've got a machine running Ubuntu v10.04 x64. With fglrx installed from the repos a HD4850 works fine.

With fglrx removed and catalyst v10.8 installed from the AMD site it works fine with a HD5830 in it once you add gfarmerfr's fix into the startup script:
xhost local:boinc &> /dev/null


When I plugged in a 2nd HD5830 boinc is not seeing it:

12-Sep-2010 21:50:45 [---] Starting BOINC client version 6.10.58 for x86_64-pc-linux-gnu
12-Sep-2010 21:50:45 [---] Config: use all coprocessors
.
.
.
12-Sep-2010 21:50:45 [---] OS: Linux: 2.6.32-24-generic
12-Sep-2010 21:50:45 [---] Memory: 3.86 GB physical, 3.72 GB virtual
12-Sep-2010 21:50:45 [---] Disk: 33.00 GB total, 27.71 GB free
12-Sep-2010 21:50:45 [---] Local time is UTC -5 hours
12-Sep-2010 21:50:46 [---] ATI GPU 0: ATI Radeon HD5800 series (Cypress) (CAL version 1.4.792, 1024MB, 1915 GFLOPS peak)
12-Sep-2010 21:50:46 [Collatz Conjecture] Found app_info.xml; using anonymous platform


Any ideas on what to check?

lspci shows both cards:
01:00.0 VGA compatible controller: ATI Technologies Inc Device 689e
01:00.1 Audio device: ATI Technologies Inc Cypress HDMI Audio [Radeon HD 5800 Series]
02:00.0 VGA compatible controller: ATI Technologies Inc Device 689e
02:00.1 Audio device: ATI Technologies Inc Cypress HDMI Audio [Radeon HD 5800 Series]

3) Message boards : Number crunching : Optimized Applications (Message 9489)
Posted 2698 days ago by Skip Da Shu
What's the dividing line for determining install/run optimized app:

collatz_2.04_x86_64-pc-linux-gnu__cuda31.tar.bz2

vs

collatz_2.02_x86_64-pc-linux-gnu__cuda22.tar.bz2

Thanx, Skip
4) Message boards : Number crunching : Optimized Applications (Message 9488)
Posted 2698 days ago by Skip Da Shu
What's the dividing line for determining install/run optimized app:

collatz_2.04_x86_64-pc-linux-gnu__cuda31.tar.bz2

vs

collatz_2.02_x86_64-pc-linux-gnu__cuda22.tar.bz2

Thanx, Skip
5) Message boards : Number crunching : Does project still generate v2.01 WUs for Linux x64/ATI GPU crunching? (Message 9428)
Posted 2701 days ago by Skip Da Shu
I found it but I'm fairly certain this is a deviation from 'the old days'...

I had the following (to get it to run 2 WUs at a time):
<coproc>
<type>ATI</type>
<count>0.5</count>
</coproc>


I changed it to:
<coproc>
<type>ATI</type>
<count>1.0</count>
</coproc>


And started getting work!

So how do you get it to run 2 WUs on the GPU at the same time nowdays since the -n parm is "depreciated"?
6) Message boards : Number crunching : Does project still generate v2.01 WUs for Linux x64/ATI GPU crunching? (Message 9427)
Posted 2701 days ago by Skip Da Shu
It requested 76032.86 seconds of GPU work.

Mon 30 Aug 2010 09:37:50 PM CDT Collatz Conjecture [sched_op_debug] Starting scheduler request
Mon 30 Aug 2010 09:37:50 PM CDT Collatz Conjecture Sending scheduler request: To fetch work.
Mon 30 Aug 2010 09:37:50 PM CDT Collatz Conjecture Requesting new tasks for GPU
Mon 30 Aug 2010 09:37:50 PM CDT Collatz Conjecture [sched_op_debug] CPU work request: 0.00 seconds; 0.00 CPUs
Mon 30 Aug 2010 09:37:50 PM CDT Collatz Conjecture [sched_op_debug] ATI GPU work request: 76032.86 seconds; 1.00 GPUs
Mon 30 Aug 2010 09:37:52 PM CDT Collatz Conjecture Scheduler request completed: got 0 new tasks
Mon 30 Aug 2010 09:37:52 PM CDT Collatz Conjecture [sched_op_debug] Server version 611
Mon 30 Aug 2010 09:37:52 PM CDT Collatz Conjecture Message from server: No work sent
Mon 30 Aug 2010 09:37:52 PM CDT Collatz Conjecture Project requested delay of 303 seconds
Mon 30 Aug 2010 09:37:52 PM CDT Collatz Conjecture [sched_op_debug] Deferring communication for 5 min 3 sec
Mon 30 Aug 2010 09:37:52 PM CDT Collatz Conjecture [sched_op_debug] Reason: requested by project
7) Message boards : Unix/Linux : Opt ATI app on Linux x64 (Message 9420)
Posted 2701 days ago by Skip Da Shu
My original thought was that this:

Initialize step array on GPU (256 MB)
predicted runtime per iteration is 43 ms (33.3333 ms are allowed), dividing each iteration in 2 parts
borders of the domains at 0 2048 4096
along with a GPU load in the 82~87% range might indicate some tuning was in order but...

I changed the app_info.xml entry that controls how many GPUs to use to 0.5 so it'd run two at a time. WU times went up quite a bit, but not double and GPU is now 93~98%. Was looking good the other day but no more WUs since then. :-(

If somebody thinks there's some tweaking that should / could be done to <cmdline> please advise. I was gonna play with 'f20' setting but no WUs so that's on a back burner now.

Thanx, Skip
8) Message boards : Number crunching : Does project still generate v2.01 WUs for Linux x64/ATI GPU crunching? (Message 9419)
Posted 2701 days ago by Skip Da Shu
Title says it all... haven't been able to get any for a bit and the Nvidia carded machine (same OS) is steady getting v2.02 WUs.

Thanx, Skip
9) Message boards : Number crunching : Running multiple GPUs (Message 9412)
Posted 2701 days ago by Skip Da Shu
...installed Evga Precision to allow the fans to be run at 100% (not an option with linux unfortunately)...


To set fanspeed to 66%:
aticonfig --pplib-cmd "set fanspeed 0 66"


or in a script with a launcher on my desktop to set it:

#!/bin/bash
#
# Filename : setfan_ati
# Date : 21.8.2010
# Author : skip at skipsjunk dot net
# Options : none yet
# Usage : setfan %percent duty cycle for fan%
#
# Sets fan speed on card 0 (1st) ati card with aticonfig
#
#V.0.1a, 21.8.2010 S.R.Guenter
#

clear
echo " "
echo "Enter the % duty cycle you want the fan to run at"
echo -en "(valid 10 thru 100) or default [auto]: "

read fan
if [ -z "$fan" ]; then
fan="auto"
fi

echo " "
if [ "$(lsmod|grep fglrx)" ] ; then
aticonfig --pplib-cmd "set fanspeed 0 $fan" &> /dev/null
else
echo -e "fglrx not loaded!"
sleep 1
fi

echo " "
echo -en "Done, Press ENTER to end"
read

#
# End


Now as to the cooling solutions, the bottom 'not recommended' pic appeals to me, but click the link called "crunchers" at the end of my sig ;-)
10) Message boards : Number crunching : 1 WU - 2 ATU-GPUs - only 1 is crunching (Message 9411)
Posted 2701 days ago by Skip Da Shu
Only dnetc grabs all gpus to crunch 1 task. Other projects dont support this feature and I hope they wouldn't. It will only open a can of worms IMO. DNETC took quite a while to make their app stable but there are still some people having problems with it. That app is only ideal if you have the same gpu with the same clocks. I bet when you crunch at DNETC, the time reduced by the 4650 in helping the 5870 is very miniscule. I've tried it with a 4350 and a 5670, the results didn't impress me. Of course my cards are much slower than yours.

Running on a machine with a GTX-275 and a GTS-250 I saw the same.

On a machine with two GTS-250s I had another issue which is another reason why I DON'T run DNETC on my machines with 2GPUs in them.

That setup means that you can't run (efficiently) 2 GPU projects on the machine because one GPU can sit idle (for hours if the other project is GPUGrid). When a DNETC WU is up to run it waits for the 2nd GPU to finish it's other project WU so one GPU sits idle.

If you ask me that ... hmmmm what's the technical term here... oh yea, SUX!
11) Message boards : Number crunching : ATI and Nvidia on Same Machine? (Message 9409)
Posted 2702 days ago by Skip Da Shu
Anyone have the mixed GPUs running under Linux?

I've got all the parts to build a dedicated dual slot cruncher 'cept for vid cards. However I just pulled a HD4850 from a machine and I could also liberate a GT8800 or a GTS-250 from another machine.

Don't recall ever seeing anyone mix drivers under Linux so I'm curious if it's even possible.

Thanx, Skip
12) Message boards : Unix/Linux : Opt ATI app on Linux x64 (Message 9408)
Posted 2702 days ago by Skip Da Shu
Tuning Opportunities:

I wanna ask about the following (test result) but have to get to work now...be back l8r.

shmget in attach_shmem: Invalid argument
17:51:02 (25346): Can't set up shared mem: -1. Will run in standalone mode.

Running Collatz Conjecture (3x+1) ATI GPU application version 2.01 by Gipsel (Linux64, CAL 1.4 - R1.5 multi-GPU Support)
Reading input file ... done.
Checking 4294967296 numbers starting with 2361185725354183731560
CAL Runtime: 1.4.736
Found 1 CAL device

Device 0: ATI Radeon HD5800 series (Cypress) 1024 MB local RAM (remote 247 MB cached + 875 MB uncached)
GPU core clock: 855 MHz, memory clock: 1000 MHz
1120 shader units organized in 14 SIMDs with 16 VLIW units (5-issue), wavefront size 64 threads
supporting double precision

Initializing lookup table (16384 kB) ... done
Setting BRT to use GPU 0
Starting WU on GPU 0
Copy lookup table to GPU memory (16384 kB)
Initialize step array on GPU (256 MB)
predicted runtime per iteration is 43 ms (33.3333 ms are allowed), dividing each iteration in 2 parts
borders of the domains at 0 2048 4096
Resuming from checkpoint ... done
needed 1516 steps for 2361185725355416900338
2242246260675 total executed steps for 4294967296 numbers

WU completed.
17:51:10 (25346): called boinc_finish
13) Message boards : Unix/Linux : Opt ATI app on Linux x64 (Message 9407)
Posted 2702 days ago by Skip Da Shu
Environment:
Ubuntu 10.04, x64
BOINC v6.10.58 (from PPA, Debian package install)
ati-driver-installer-10-7-x86.x86_64 (glxinfo broken so I can't prove at the moment)
single HD5830 card

App
collatz_2.01_x86_64-pc-linux-gnu__ati14_r1.5_mgpu.tar.bz2

Think I followed the README but got "download failed". I think it was trying to download the libbrook* libs it found in the app_info.xml.

I found this. Moved the '64' libs to the project directory and changed the 'non-64' libs in /usr/lib64 to be symbolic links instead of copies of the '64' libs. Also changed them in my /test folder and re-ran the test... all works in the stand-alone test.

Just started boinc back up and it downloaded 4 WUs and is crunching on one:
Sun 29 Aug 2010 07:23:02 PM CDT Collatz Conjecture Starting task collatz_1282112751_844469_0 using collatz version 201

So it's working.

I am curious why the '64' libs need to be in the project folder though. Why can't it locate them in /usr/lib64? Also is there some sort of link I could put in the project directory to point it to copy in /usr/lib64 instead?

Next Post "Tuning Opportunities"

14) Message boards : Number crunching : Supported Platforms (Message 4173)
Posted 2973 days ago by Skip Da Shu

Also any idea why this showed up today?

Mon 30 Nov 2009 01:50:10 PM CST Collatz Conjecture Message from server: No work sent
Mon 30 Nov 2009 01:50:10 PM CST Collatz Conjecture Message from server: Your computer has no NVIDIA GPU
Mon 30 Nov 2009 01:50:10 PM CST Collatz Conjecture Message from server: ATI Catalyst 9.2+ needed to use GPU


Are you sure you aren't getting that message when you already have the max WUs downloaded?

Could be. I'll take a look at that. Thanx, Skip
15) Message boards : Number crunching : Supported Platforms (Message 4141)
Posted 2974 days ago by Skip Da Shu
9.2 is from Feb this year ;)


Yes, I saw that. Does this mean Collatz can no longer run on 8.12?
16) Message boards : Number crunching : Supported Platforms (Message 4136)
Posted 2974 days ago by Skip Da Shu
What's this mean (from opt app page):

collatz_2.01_x86_64-pc-linux-gnu__ati14_r1.2.tar.bz2 (Linux 64-bit ATI) 2.02 nVidia/CUDA 2.2 11/29/2009 12:12 CST


ati in file name but says it's CUDA?


Also any idea why this showed up today?

Mon 30 Nov 2009 01:50:10 PM CST Collatz Conjecture Message from server: No work sent
Mon 30 Nov 2009 01:50:10 PM CST Collatz Conjecture Message from server: Your computer has no NVIDIA GPU
Mon 30 Nov 2009 01:50:10 PM CST Collatz Conjecture Message from server: ATI Catalyst 9.2+ needed to use GPU


Are we not supporting v8.12 anymore?
17) Message boards : Number crunching : nVidia releases 191.07 drivers (Message 3157)
Posted 3009 days ago by Skip Da Shu
I don't know whether they fixed the issues in 191.03, but they have released a new version 191.07 which you can download here

I have not tested it myself, so use at your own risk. Also, please keep an eye on your machine for a while if you do upgrade so as not to trash too many WUs if it doesn't work well.


Windoze only.
18) Message boards : Number crunching : Radeon HD 5870 Reviews now online. (Message 2764)
Posted 3024 days ago by Skip Da Shu

Guess I forgot to mention I also updated the GPU detection ;)
But it appears to be held back by the low memory frequency. I think it would be capable to complete a task within 300 seconds otherwise.

I suspect immature driver support!


Using v8.12 driver I get the same thing on my windozeXP Pro 64 machine running MW with HD4850. CCC says the current memory clock is at 930MHz(lowered from default of 993). GPU-Z says memory is at 928.1MHz.
......................

CAL Runtime: 1.3.145
Found 1 CAL device

Device 0: ATI Radeon HD 4800 (RV770) 512 MB local RAM (remote 28 MB cached + 512 MB uncached)
GPU core clock: 680 MHz, memory clock: 250 MHz
800 shader units organized in 10 SIMDs with 16 VLIW units (5-issue), wavefront size 64 threads
supporting double precision
....................

However with the same drive and same HD4850 card my wife's MW WUs report the correct default memory frequency. ??
....................

CAL Runtime: 1.3.145
Found 1 CAL device

Device 0: ATI Radeon HD 4800 (RV770) 512 MB local RAM (remote 28 MB cached + 512 MB uncached)
GPU core clock: 670 MHz, memory clock: 993 MHz
800 shader units organized in 10 SIMDs with 16 VLIW units (5-issue), wavefront size 64 threads
supporting double precision
.......................

If I can remember I'll set my 4870 back to stock memory clock and see what it reports on those MW WUs. I can't say on Collatz because I'm running it on a Linux machine which reports differently.

PS: Didn't seem to report any different mem clock even with card set back to defaults. Don't know why the WinXP machine reports correct mem clock but the WinXP64 one doesn't.
19) Message boards : Number crunching : Linux CUDA Application Installation (Message 2473)
Posted 3031 days ago by Skip Da Shu
Can you please confirm that, at least for Ubuntu/Debian install with v6.10.6...

The symlink should point to the downloaded libcudart.so.2.2 not the libcudart.so.2 that's in /usr/lib64/ (a link itself).
eg.
cd /var/lib/boinc-client/projects/boinc.thesonntags.com_collatz
sudo ln -sf libcudart.so.2.2 libcudart.so.2


This seems to have gotten my setup to start working at least to the extent that I have a WU actually running right now.

What remains to be seen is will this GTX-260 be bitten by the "Cuda 2.2 FFT" bug that required me to drop GPUgrid from this card/machine. Has anyone reported this problem on this app?

Will install v6.10.11 as soon as this running WU finishes
20) Message boards : Number crunching : Supported Platforms (Message 313)
Posted 3097 days ago by Skip Da Shu
Is Linux CUDA in the pipeline?


It has been discussed. There are a zillion things on the to-do list and that is one of them.


CUDA or CAL on Linux 64b and you'll be my hero! Where do we paypal the bribes to?


Next 20

Main page · Your account · Message boards


Copyright © 2018 Jon Sonntag; All rights reserved.