Posts by frankhagen

\n studio-striking\n
1) Message boards : Number crunching : Any Cuda improvements in the pipeline? (Message 820)
Posted 20 Jul 2011 by frankhagen
Post:
I also wanted to share in the credit feast at PG and made a mistake when I paid a lot of money for a Nvidia a month ago.
For a bit more I could have got a HD6980
Fact is now I’m stuck with it and would like to use it here at Moo.


well your second best bet imho would be collatz then..
2) Message boards : Number crunching : Too high CPU core estimation (Message 648)
Posted 17 Jun 2011 by frankhagen
Post:
Interesting that the CPU reservation doesn't really affect how much CPU is being used, just that when it's set too high it keeps BOINC CPU tasks from running.


acutally boinc does NOT monitor CPU usage, it just relies on what the project tells. OMFG!
3) Message boards : News : Credits (Message 476)
Posted 25 May 2011 by frankhagen
Post:
....... you would need WU's with a single packet AND boinc running a seperate WU on each GPU...... not possible now imho..


Impossible to run singles on each GPU, and likely to remain so for foreseeable future.


oh, really?

just not implemented.

the CPU app can run multiple instances and can be configured to use selected cores instead of running in MT-mode.

but that's a task for distributed.net...
4) Message boards : News : Credits (Message 473)
Posted 25 May 2011 by frankhagen
Post:
[quote]Multiple cards work sub-optimally at both, but it seems that's a problem with the dnet client. /quote]

right!

you would need WU's with a single packet AND boinc running a seperate WU on each GPU.


not possible now imho..
5) Questions and Answers : Windows : 20 tasks per day? (Message 391)
Posted 19 May 2011 by frankhagen
Post:
It *IS* surprising to have such a low trigger. Most projects use 100 or 1000.


i don't know how it is set up, but your host details still show that you killed at least 42 tasks in a row. probably a bad idea...
6) Questions and Answers : Windows : inconclusive work (Message 264)
Posted 12 May 2011 by frankhagen
Post:
Hi,

That result lost one packet and therefore didn't do the full work and got granted only partial credit. Dnet client sometimes looses work like this due to some bug in it. :(

-w


yup, i saw it got credit by now.

btw.: they say they have fixed some things in v 519 meanwhile..


7) Questions and Answers : Windows : inconclusive work (Message 260)
Posted 12 May 2011 by frankhagen
Post:
i got one WU which was marked as invalid, but stderr output does not show any strange things: http://moowrap.net/result.php?resultid=50160
what went wrong?
8) Message boards : Wish list : Fixed credits based on the amount of work done (Message 257)
Posted 12 May 2011 by frankhagen
Post:
Hi,

I'm beginning to agree with this change. That new credit system seems a mess and granted credit is still varying way too much for my taste. :(

I should be able to easily switch to static credit based on work in a work unit. This method seems to be supported by the system.

-w


butbutbut..

it's way DA want's it, and we all could save a lot of energy by running our cards in low power mode.. ;)
9) Questions and Answers : Windows : Multi-GPU task takes longer than single-gpu task (Message 250)
Posted 12 May 2011 by frankhagen
Post:
Well, when I run -bench repeatedly, I get wild variations in the results per core. Does the system need to be idled to get an accurate -bench result? Shouldn't be running BOINC during the test?


for sure - you'll want to disable everything which could interfer..
10) Questions and Answers : Windows : Multi-GPU task takes longer than single-gpu task (Message 210)
Posted 10 May 2011 by frankhagen
Post:
Yes, I have SDK 2.4 installed. I also have 11.3 with APP. This is win7 64.


i suggest you run the benchmarks while boinc is shut down.

"dnetc518xxx.exe -- C 1 --benchmark"
"dnetc518xxx.exe -- C 2 --benchmark"

and so on.

oh, and btw.:

you might find something when running

"dnetc518xxx.exe --help"

i can not test this, because i got only NVIDIA's...

11) Questions and Answers : Preferences : Client core values (Message 205)
Posted 10 May 2011 by frankhagen
Post:
Can someone please define what each of the values mean? From reading the various threads, it looks like range is -1 through 10, only integer values. Is this correct? I know -1 means auto-select. But what to all the rest of the values mean? Can someone please post the definitions?


you can look yourself by starting the app from a command prompt with "--config".

select 3 and then 1..

for CUDA it's:

RC5-72:-1) Auto select 0) CUDA 1-pipe 64-thd 1) CUDA 1-pipe 128-thd
2) CUDA 1-pipe 256-thd 3) CUDA 2-pipe 64-thd 4) CUDA 2-pipe 128-thd
5) CUDA 2-pipe 256-thd 6) CUDA 4-pipe 64-thd 7) CUDA 4-pipe 128-thd
8) CUDA 4-pipe 256-thd 9) CUDA 1-pipe 64-thd busy wait
10) CUDA 1-pipe 64-thd sleep 100us
11) CUDA 1-pipe 64-thd sleep dynamic
12) Questions and Answers : Windows : CUDA app 1.01 crashes when a core is manually selected (Message 174)
Posted 8 May 2011 by frankhagen
Post:
Hint compare the performance of Core #1 and Core #3 at stock clocks / when overclocked. More than 100% performance increase with 15% higher clocks? By the way: The app should be recompiled with the 3.2 SDK. On several occasions CUDA apps compiled with the 3.1 SDK didn't work correctly: For example PrimeGrid's tpsieve refuses to find a single factor (on a test range with 173 factors) and runs 50% slower when compiled with the SDK 3.1 for Linux.



NOW we're talking.. ;)
13) Questions and Answers : Windows : CUDA app 1.01 crashes when a core is manually selected (Message 170)
Posted 8 May 2011 by frankhagen
Post:
Just tested it. Result: It does not work.


had that on 2 hosts after a change. aborting tasks and a reboot...
14) Questions and Answers : Windows : CUDA app 1.01 crashes when a core is manually selected (Message 168)
Posted 8 May 2011 by frankhagen
Post:
1. I selected a core manually (#9) and the CUDA app crashed.
2. I crunched with the default -1 setting (autoselect) and everything worked.
3. I selected core #10 and the CUDA app crashed.


hi ralf.. ;)


this usually happens when a WU is restarted after changeing somthing - the next one should do fine..
15) Questions and Answers : Windows : High CPU usage (Message 165)
Posted 8 May 2011 by frankhagen
Post:
I'd think core 3 could be best for Stream and core 10 for CUDA clients. Although, you should experiment since this can be different for your setup.


YES and no.

9 uses a full CPU core and pushes the GPU to the limit and is fastest mostly.
10 uses less than 1% CPU and utilizes the GPU above 90%.
11 and your milage may vary.

maybe 2, 5 or 8 will be faster on cards with a high number of cores.

but everyone can test it by running the app from commandline with --benchmark
16) Questions and Answers : Windows : Multi-GPU task takes longer than single-gpu task (Message 156)
Posted 7 May 2011 by frankhagen
Post:
Yeah. The frustrating part is the wasted time, and lack of actual production. It should be doing ~2x the work of the single card machine. In reality, it is doing 1/6th or less.


now you do not ask me to tell you who had a bright idea again.. ;)

of course it's a problem with the app or the wrapper, but it's as silly as it can get..
17) Questions and Answers : Windows : Multi-GPU task takes longer than single-gpu task (Message 151)
Posted 7 May 2011 by frankhagen
Post:
Yep. All tasks on my dual 5870 machine take ~1-2 hours. On my single 5870, they take ~15 minutes.


really strange!

but DA's bright new credit-system has decided to compensate for that and is granting a freaky load of credits for those WU's now..
18) Questions and Answers : Windows : CUDA Failure (Message 139)
Posted 6 May 2011 by frankhagen
Post:
réduice CUDA time units because the last two times were 8000 sec each Envrionment ... unless these units is very long compared to the previous ....


check what's in stderr!

there are packets of different sizes. for my hosts it's about 20% faster..
19) Message boards : Wish list : BOINC stats (Message 138)
Posted 6 May 2011 by frankhagen
Post:
Hi,

Yep, it's up to BOINCstats/BAM/Willy to add us. Personally, I'd like to see that too since I use BAM. :)

Stats export wise things should be good at our end.

-w


unless you contact that guy - he will not..
20) Message boards : Wish list : reduce duration cuda (Message 133)
Posted 6 May 2011 by frankhagen
Post:
simply add <dont_check_file_sizes>1</dont_check_file_sizes> to a cc_config.xml and the client will ignore wrong file sizes.


should work - but i don't think it's a very bright idea to disable it completely if there is another simple way.


Next 20


 
Copyright © 2011-2024 Moo! Wrapper Project