Welcome to MilkyWay@home

Posts by Mumak

21) Message boards : Number crunching : Machines runs out of units every few minutes (Message 63020)
Posted 16 Jan 2015 by Profile Mumak
Post:
I'm experiencing the same problem right now. Only getting a few units, sometimes my queues are empty.
22) Message boards : Number crunching : MilkyWay@Home v1.02 (opencl_amd_ati) Compute Errors (Message 62841)
Posted 15 Dec 2014 by Profile Mumak
Post:
The same here too !
23) Message boards : News : New Modfit Runs (Message 62822)
Posted 11 Dec 2014 by Profile Mumak
Post:
Any chance to look at the high error rate of current Modfit units ?
I have no issues with the default MW application, only a few errors during several days.
Today I gave Modfit another chance, but after ~1 hour I have disabled them again - 10 invalids during such a short period and that's probably not the end.
24) Message boards : News : New Separation Modfit Version 1.36 (Message 62605)
Posted 21 Oct 2014 by Profile Mumak
Post:
I'm too seeing a higher error rate of the Modfit since a few weeks (several dozen invalids per day). Had to opt-out...
25) Message boards : News : New Version of Separation Modified Fit (1.32) (Message 62359)
Posted 18 Sep 2014 by Profile Mumak
Post:
I'm too getting lots of ps_modfit errors recently on AMD HD 7950.
I had to opt-out of the Modfit tasks.
26) Message boards : Number crunching : ps_modfit_16TestStars (Message 62313)
Posted 11 Sep 2014 by Profile Mumak
Post:
All the ps_modfit_16TestStars... tasks are failing on my (otherwise) stable machine (2xHD7950). I never had big issues with MW@H tasks, but this seems to be a problem in those units. I have to disable the entire ModFit set because of this.
I just realized this is already reported in the News section..
You can delete this then.
27) Message boards : Number crunching : Server error (Message 61605)
Posted 24 Apr 2014 by Profile Mumak
Post:
Working again..
28) Message boards : Number crunching : Server error (Message 61604)
Posted 24 Apr 2014 by Profile Mumak
Post:
It seems, there's a server error - I can't report nor get any tasks.

Server error: feeder not running

29) Message boards : Number crunching : R9 290 (Message 61484)
Posted 8 Apr 2014 by Profile Mumak
Post:
Maybe posting a few lines from the log (BOINC start with GPU IDs and work request/result) might help to understand the problem.
30) Message boards : Number crunching : Thermal issue with OpenCL (Message 61477)
Posted 7 Apr 2014 by Profile Mumak
Post:
Also, projects like this, which use double-precision load the GPU execution units higher, so that causes additional heat. In my case (HD7950, R9 280X) the GPUs run ~10 C hotter than on other projects with single-precision math.
You might need to undervolt your GPUs if possible. I do this too, otherwise they would run too hot.
31) Message boards : Number crunching : Thermal issue with OpenCL (Message 61377)
Posted 10 Mar 2014 by Profile Mumak
Post:
I don't think it's possible to run the GPU at 120 C. What tool do you use to monitor it?
You might try to undervolt it. I do this with my 7950's which had VDDC=1.25V default, but fortunately have digital PWMs, so it's fairly simple to change the voltages.
32) Message boards : Number crunching : Benchmark results - times wanted for any hardware, CPU or GPU, old or new! (Message 61235)
Posted 25 Feb 2014 by Profile Mumak
Post:
I don't think that it's possible to utilize the NVIDIA GPU well using OpenCL. NVIDIA drivers are just not well optimized for OpenCL and they don't care about it either...
33) Message boards : Number crunching : Benchmark results - times wanted for any hardware, CPU or GPU, old or new! (Message 61227)
Posted 24 Feb 2014 by Profile Mumak
Post:
I agree ;-)
I'm still checking and tuning few things there, so it might change..
BTW, that cooling solution works pretty nice - 68 C max (with a side fan in the case). Without the side fan, ~71 C.
34) Message boards : Number crunching : Benchmark results - times wanted for any hardware, CPU or GPU, old or new! (Message 61224)
Posted 24 Feb 2014 by Profile Mumak
Post:
OK, so the new Gigabyte 280X is up and running:


There are 2 GPUs in the machine:
#0 - HD7770 @ 1020 - Left graph column
#1 - R9 280X @ 1100 - Right graph column

Tasks: http://milkyway.cs.rpi.edu/milkyway/results.php?hostid=534385&offset=0&show_names=0&state=4&appid=
So the slowest times belong to the HD7770.
For the 280X and slower units, I see times 70-75 secs so far...
35) Message boards : Number crunching : Benchmark results - times wanted for any hardware, CPU or GPU, old or new! (Message 61222)
Posted 24 Feb 2014 by Profile Mumak
Post:
and I checked out the HWinfo site but it's not updated as frequently as GPU-Z or SIV64X but thanks for the tips guys.


What?! I released a new version a week ago and will release a new Beta tomorrow ;-)
36) Message boards : Number crunching : Benchmark results - times wanted for any hardware, CPU or GPU, old or new! (Message 61212)
Posted 23 Feb 2014 by Profile Mumak
Post:
I forgot to add, that some GPU monitoring tools tend to round-up the actual GPU clock (over a period of time) on some later AMD families. HWiNFO on the other hand reports the actual precise clock.
37) Message boards : Number crunching : Benchmark results - times wanted for any hardware, CPU or GPU, old or new! (Message 61206)
Posted 23 Feb 2014 by Profile Mumak
Post:
if you want something that'll substantially outperform a 7970/7970GE, you'll have to upgrade to a 290 or a 290X...


But not for MW@H (double-precision), since the 290 (Hawaii) has crippled DP performance. But this is something we all know...

WES - you can use HWiNFO and log all sensor values to see at which exact clock it is running.
38) Message boards : Number crunching : Benchmark results - times wanted for any hardware, CPU or GPU, old or new! (Message 61199)
Posted 23 Feb 2014 by Profile Mumak
Post:
I had a look at your host WES and for the longer 213.76 credit WUs I saw 81-85 secs.
39) Message boards : Number crunching : Benchmark results - times wanted for any hardware, CPU or GPU, old or new! (Message 61185)
Posted 22 Feb 2014 by Profile Mumak
Post:
I'll utilize the CPUs for WCG and T4T.

75-80 C? I hope not so high !
My 7950's with STOCK settings were reaching power limit and overheating on MW. Then I noticed the default max VDDC was set to 1.25V, which is way too high. So I reduced it. Now at full DP load the VDDC is ~1.07V and temps are ~65 C.
40) Message boards : Number crunching : Benchmark results - times wanted for any hardware, CPU or GPU, old or new! (Message 61176)
Posted 21 Feb 2014 by Profile Mumak
Post:
Will see the next week with a new Gigabyte GV-R928XOC-3GD REV 2.0 ;-) Boost 1100 MHz. Also what temperatures will the WINDFORCE 3X cooling maintain under sustained DP load...

Will also get a new Dual Xeon workstation soon. I was hoping it will be fitted with a Tesla (full DP), but it wasn't possible :-(


Previous 20 · Next 20

©2024 Astroinformatics Group