Welcome to MilkyWay@home

Posts by Haris Dublas

1) Message boards : Number crunching : 6990 - Who Is Going To Be First? (Message 46532)
Posted 11 Mar 2011 by Haris Dublas
Post:
Getting 2 6950 then mod to 6970 is probably better.
2) Message boards : Number crunching : GPU Requirements [OLD] (Message 46480)
Posted 7 Mar 2011 by Haris Dublas
Post:
as almost all of HD 6000 series have single precision...



As did the HD5xxx series before it and the HD4xxx series before that again.

It was only the top end cards in each series that supported double precision.

I.E:

HD3870 and HD3850

HD4890, HD4870 and HD4850;

HD5970, HD5870 and HD5850;

The HD6990, HD6970 and HD6950.

All cards below those mentioned (with I am sure some exceptions) only support single precision.


You forgot the 4700 series.
3) Message boards : Number crunching : Message from server: double precision GPU required (Message 46434)
Posted 4 Mar 2011 by Haris Dublas
Post:
update:

1 of the 30+ MW@H GPU tasks that were downloaded was worked on by my 5870 GPU for ~45-sec and got to 50% complete, then stopped and went back into the "waiting to run" state. an hour or so later (maybe longer), it restarted and finished in another ~45 sec. the giveaway was that my GUI lag increased severely while the GPU was crunching. anyone have any idea why my 5870 GPU is only crunching MW@H tasks once every hour or so for only 45-sec at a time?


Check your BOINC Manager preferences, at the processor usage tab, make sure the "Use GPU while computer is in use" box is checked, then put 0 on the "While processor usage is less than __ percent (0 means no restriction).
4) Message boards : Number crunching : Message from server: double precision GPU required (Message 46433)
Posted 4 Mar 2011 by Haris Dublas
Post:
After attaching the dummy plug, did you extend your desktop to the 5870? BOINC will not recognize it if you dont extend the desktop.

I've tried this setup before with a 785G and 4850 and it works, you can even OC the card a bit more. My 4850 if used as primary only OC to 650, anything higher will trigger the gpu recovery (momentary black screens and mouse lock-ups). Set as secondary and IGP as primary, it can reach 680 without any problems (any higher will produce invalid wus).

If you're ok with the 5870 as primary, you can use the MW opt app, find "cmdline" in the app_info.xml file and insert "f50 b1" (or "f100 b2", "f50 b3", etc.). Do a trial and error until you find your comfort zone. Of course this will lower the gpu utilization a bit, you can adjust it by adding w (w.80 if the gpu utilization drops 80-95%, this will vary from card to card). Since the admins are planning to do away with the apps made by Gipsel in the near future, I hope they can make the future apps as customizable as his. This also applies to Collatz, Cluster Physik (Gipsel) was also the one responsible for the ati apps there.
5) Message boards : Number crunching : Message from server: double precision GPU required (Message 46413)
Posted 3 Mar 2011 by Haris Dublas
Post:
If its not your primary GPU, then you have to connect a monitor or a dummy plug. Your directx is fine, since there's no official Microsoft release of the dx10 for winxp. Headless boxes are fine as long as the gpu is the primary one. If the gpu is the second or third, you will need to attach a monitor or dummy plug.

All my hosts were running xp up until I stopped crunching MW about 2 months ago due to frequent server problems and validation issues (invalid cpu result became valid and the real valid gpu result became invalid). Now that the cpu wus are different than the gpu wus so the problem is gone.
6) Message boards : Number crunching : nVidia GPUs and double precision (Message 46348)
Posted 25 Feb 2011 by Haris Dublas
Post:
Or get a 440, its basically a 430 (a little bit higher clocks) with ddr5. It will perform better in Collatz than the 430 when MW have troubles. The OEM one have 112 shaders while the retail one is only a pumped up 430 (96 shaders).
7) Message boards : Number crunching : killo-o-watt readings ... I thought they'd be higher.! (Message 46333)
Posted 21 Feb 2011 by Haris Dublas
Post:
Probably your WUs are suspended at that time you measured (check boinc settings). My athlon 620 idles @ 74w (82w OC'ed) then adds about 20w per core @ full load.

At full load, the 460 consumes 80w more than the 5670 and the phenom II consumes more than an athlon II. Its impossible that my athlon II and 5670 consumes more than your phenom II and 460. That box only have board, cpu, gpu, 2 sticks ram and 1 hdd, no other add-on devices.
8) Message boards : Number crunching : killo-o-watt readings ... I thought they'd be higher.! (Message 46316)
Posted 20 Feb 2011 by Haris Dublas
Post:
I have an Athlon II x4 620 with an ATI 5670 (TDP is 80w lower than the 460). 3 cores @ stock 2.6 ghz (1.25v) running Computing for Clean Water and Collatz running on the gpu measures 155w +/-. OC'ed to 3.3ghz (1.42v) reads 170w +/-. TDP of 4850 is 37w higher than the 5670.

TDP of 460 @ stock speed full load is about 150w so I dont know why your system only measures 96w with 3 cpu cores @ full load.

Your P4/4850 readings sounds about right.

Edit: Used to have a 4850 on that machine but it seemed that the power supply is nearing its last legs. The PS temp is higher than the cpu and gpu when I OC the 4850 to 675 core, so I replaced it temporarily with the 5670 to play safe.
9) Message boards : Number crunching : Milky way blew up (Message 45608)
Posted 18 Jan 2011 by Haris Dublas
Post:
In my case, I think BOINC assumed that the projects were causing the restarts so it detached the projects. I don't know about the current version, I was using 6.10.18 at that time.
10) Message boards : Number crunching : Milky way blew up (Message 45597)
Posted 16 Jan 2011 by Haris Dublas
Post:
I think this a BOINC thing. Last year there were serious power fluctuations in my area and made my boxes restart several times. 2 boxes crunching MW and 1 box crunching Collatz detached from both projects.
11) Message boards : Number crunching : Problem with BOINC Stats (Message 45124)
Posted 19 Dec 2010 by Haris Dublas
Post:
Checked and everything's back to normal again. Must be a glitch somewhere out there. :)
12) Message boards : Number crunching : Problem with BOINC Stats (Message 45109)
Posted 18 Dec 2010 by Haris Dublas
Post:
Any of you guys getting zero MW@H credit on boincstats? Site say userid doesnt exit or zero credit.
13) Message boards : News : a fix for the output file issue (Message 43103)
Posted 23 Oct 2010 by Haris Dublas
Post:
(...)
If I enable CPU + GPU in the prefs, my PC get only GPU WUs.
I need to disable GPU for to get CPU WUs.
Then after my PC DLed a few CPU WUs, I need to enable again the GPU.
It would be possible, to 'optimize' the scheduler?
Maybe 4 WUs/CPU-Core and 12 WUs/GPU (limit of tasks in progress)?


Just curious..

I'm the only one which have probs to get CPU & GPU WUs simultaneously?

(with app_info.xml file)

If I enable CPU & GPU in the project prefs, my Intel Core2 Duo E7600 with GTX260-216 get only GPU WUs. Max. 12 WUs.

If BOINC DLed 12 GPU WUs, I need to disable GPU in the prefs.
BOINC DL then after every UL/report of a GPU result a new CPU WU.
If 10 GPU WUs were ULed/reported, and BOINC DLed 10 CPU WUs I enable again (also) GPU.

Then I have one GPU WU in calculation and one GPU WU prepared for calculation. And 2 CPU WUs in calculation and 8 CPU WUs prepared. If one CPU result ULed/reported, BOINC DL a new GPU WU. If only 2 CPU WUs in calculation, I need to disable again the GPU in the prefs.

How it's at your BOINC - how you do it?

Could be the MW@h scheduler do this better?
Like I mentioned.. 'reached limit of x tasks in progress' for CPU and GPU separated.
S@h have after the 3 day outages always a limit set, CPU and GPU separated. The server/scheduler could do it.
IIRC, the low limit is 40 WUs/CPU-Core and 320 WUs/GPU.

So here at MW@h maybe 2 WUs/CPU-Core and 6 WUs/GPU.

This would be possible?

This would be very helpful for us.

Thanks!


CPU or GPU, limit is 6 wus per cpu core. So if you have dual core, you can only have 12 wus even if you have a 10 days cache. Even if you only crunch on the gpu, the limit will still be based on the number of cores of your cpu.
14) Message boards : Number crunching : Torrents of Invalid WU's (Message 42640)
Posted 6 Oct 2010 by Haris Dublas
Post:
213355053 201854 6 Oct 2010 7:36:33 UTC 6 Oct 2010 7:54:19 UTC Completed and validated 83.27 5.83 0.04 213.76 MilkyWay@Home v0.23 (ati13ati)
213369417 182806 6 Oct 2010 8:06:35 UTC 6 Oct 2010 13:42:27 UTC Completed, marked as invalid 1,294.64 406.41 2.15 0.00 Anonymous platform
213538977 104894 6 Oct 2010 13:47:35 UTC 6 Oct 2010 14:12:59 UTC Completed and validated 0.00 9.05 0.05 213.76 Anonymous platform


The underclocked 3850 wu is invalid while the cpu wu that have 0 secs runtime and 9 secs cpu time is valid.

This is getting out of hand, I'm gonna move my host to collatz until this problem is fixed.
15) Message boards : Number crunching : crunching with an ATI Sapphire HD3850-512 (Message 42553)
Posted 4 Oct 2010 by Haris Dublas
Post:
My Sapphire 3850-512 works fine with all driver versions I tested it with. 9.3, 10.2, 10.4, 10.6 and the latest 10.9.

Now OC'ed 890 core, 2 wus done at 900-905 secs. Probably 40k RAC. Can reach 900 core but have to increase voltage from stock 1.214v to 1.254v. Temps @ 900 is 95c so I settled on 890 @ 88c. 2 tasks takes about 886-891 secs @ 900 core.

16) Message boards : Number crunching : Client detached (Message 42447)
Posted 29 Sep 2010 by Haris Dublas
Post:
If you use BAM (Boincstats Account Manager), it's probably it. Check your settings or just disable BAM like I did to avoid the hassle.

Btw, this also happens if your computer restarts frequently and not properly shutdown each time. Check the stability of your power supply or the power from the wall outlet.
17) Message boards : Number crunching : CPU worth it for MW@H? (Message 42446)
Posted 29 Sep 2010 by Haris Dublas
Post:
Here's a sample of my app_info. Download a cpu opt app then replace the nbody things.

<app_info>
<app>
<name>milkyway</name>
</app>
<file_info>
<name>astronomy_0.23_ATI_x64.exe</name>
<executable/>
</file_info>
<file_info>
<name>brook64.dll</name>
<executable/>
</file_info>

<app_version>
<app_name>milkyway</app_name>
<version_num>23</version_num>
<plan_class>ati13ati</plan_class>
<avg_ncpus>0.05</avg_ncpus>
<max_ncpus>1</max_ncpus>
<flops>1.0e11</flops>
<coproc>
<type>ATI</type>
<count>0.5</count>
</coproc>
<cmdline>b-1</cmdline>
<file_ref>
<file_name>astronomy_0.23_ATI_x64.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>brook64.dll</file_name>
</file_ref>
</app_version>

<app>
<name>milkyway_nbody</name>
</app>
<file_info>
<name>milkyway_nbody_0.03_windows_x86_64.exe</name>
<executable/>
</file_info>

<app_version>
<app_name>milkyway_nbody</app_name>
<version_num>3</version_num>
<file_ref>
<file_name>milkyway_nbody_0.03_windows_x86_64.exe</file_name>
<main_program/>
</file_ref>
</app_version>

</app_info>
18) Message boards : Number crunching : CPU worth it for MW@H? (Message 42426)
Posted 28 Sep 2010 by Haris Dublas
Post:
Try the optimized 64bit cpu app, I crunched cpu tasks on my p4 3.6 long ago and I think it took less than 6 hrs to finish one.

http://www.arkayn.us/milkyway/
19) Message boards : Number crunching : Wow! CPU WUs done in just 160 secs! (Message 42393)
Posted 27 Sep 2010 by Haris Dublas
Post:
He was (is) only crunching cpu tasks, no gpu tasks. Each of his tasks before were finished in 160 secs average. It would have been understandable if he was crunching on a fast gpu but he was not. All tasks that needed a quorum were invalid. That 119k RAC was done on single host, a 2.33ghz dual xeon 3065 without a gpu. Its not even overclocked. Have you seen a cpu almost as fast as a 4890 (just a speculation since i dont have a 4890 and a 5830 is faster than that)?

Checking that host now, he's averaging about 15-16k secs per task. Thats a lot of difference compared to 160 secs, dont you think? Seems his RAC going back down now, 20k at the moment and might settle at 3-3.5k.

Why does someone with an Anonymous profile cheat? Or was it just a glitch on his hardware or software?
20) Message boards : Number crunching : GPU detection problem with both ATI and nVidia cards installed (Message 42272)
Posted 20 Sep 2010 by Haris Dublas
Post:
If you dont use your computer for gaming or heavy graphics, set the 8500gt as your primary display and the 5870 as secondary. So when you use your computer, the crunching speed of the 5870 wont be affected much.

With both nvidia and ati, dont install CCC. Use a third party utility to control your clocks, like MSI Afterburner or Rivatuner.


Next 20

©2021 Astroinformatics Group