Welcome to MilkyWay@home

found a small memory error in the code

Message boards : Application Code Discussion : found a small memory error in the code
Message board moderation

To post messages, you must log in.

Previous · 1 · 2

AuthorMessage
Profile Glenn Rogers
Avatar

Send message
Joined: 4 Jul 08
Posts: 165
Credit: 364,966
RAC: 0
Message 27107 - Posted: 5 Jul 2009, 15:15:05 UTC

I wouldnt mind a faster cpu app if there is one to have. Another cruncher here name "ICE" is hosting the optimsed apps maybee get in touch with him to get it on his site.
Glenn
ID: 27107 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Cluster Physik

Send message
Joined: 26 Jul 08
Posts: 627
Credit: 94,940,203
RAC: 0
Message 27114 - Posted: 5 Jul 2009, 15:51:08 UTC - in response to Message 27107.  

Another cruncher here name "ICE" is hosting the optimsed apps maybee get in touch with him to get it on his site.

I know how it works. The Windows/GPU apps there are all my ones ;)
ID: 27114 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Glenn Rogers
Avatar

Send message
Joined: 4 Jul 08
Posts: 165
Credit: 364,966
RAC: 0
Message 27149 - Posted: 6 Jul 2009, 6:30:57 UTC - in response to Message 27114.  

Sorry CP this wasnt directed at you was directed at dobrichev, I know that you know who is hosting your apps
ID: 27149 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Bill592
Avatar

Send message
Joined: 19 May 09
Posts: 30
Credit: 1,062,540
RAC: 0
Message 27181 - Posted: 6 Jul 2009, 21:29:05 UTC - in response to Message 27090.  

Using a CPU is largely energy waste compared to a GPU so I stopped looking to the CPU application.



Hi CP,

That is something I have been wondering about. I wonder if that energy waste is occurring (or reversed) on other projects plus
Milky Way when running Cuda.

Over on Einstein @ Home they have only managed to attain a
1.5 x ( geforce ) to 3.5 x ( Tesla ) speedup compared to the
CPU app. when testing their beta Cuda app.


Some Nvidia cards burn up a lot of power so, for that minor speed advantage, you may end up wasting power running Cuda.


Best Regards,

Bill
ID: 27181 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Paul D. Buck

Send message
Joined: 12 Apr 08
Posts: 621
Credit: 161,934,067
RAC: 0
Message 31649 - Posted: 28 Sep 2009, 2:12:59 UTC - in response to Message 27181.  

Over on Einstein @ Home they have only managed to attain a
1.5 x ( geforce ) to 3.5 x ( Tesla ) speedup compared to the
CPU app. when testing their beta Cuda app.

Which is why it is still in Beta ...

It is not always straight forward to vectorize an application and there are always subtleties that have to be learned ... but, eventually they will likely get a much better behaved application. Heck GPU Grid at the turn of the year used as much as 21% of the CPU just to idle wait for the GPU. They changed and got that down to less than 1% with almost no loss of speed on the CUDA side.

For me, one of the advantages, besides just the speed up is that we can bring more parallel processing elements into the mix and as so I can run more work in parallel. So right now I am running 3 projects on the GPUs and 4 on the CPU of which 3 are NCI ...
ID: 31649 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Cluster Physik

Send message
Joined: 26 Jul 08
Posts: 627
Credit: 94,940,203
RAC: 0
Message 31661 - Posted: 28 Sep 2009, 14:52:44 UTC - in response to Message 27181.  

Over on Einstein @ Home they have only managed to attain a
1.5 x ( geforce ) to 3.5 x ( Tesla ) speedup compared to the
CPU app. when testing their beta Cuda app.

I don't get why people are so obsessed to attribute Tesla cards a higher performance than the normal consumer cards. The GPU hardware is identical. There is nothing why a Tesla card is going to be faster. Tesla cards have usually more (but slower) RAM, so this could enable some more data to be buffered, but aside from this Teslas are usually even a bit slower than a GTX285 (which you can buy as 2GB model with fast RAM) for instance.
ID: 31661 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Previous · 1 · 2

Message boards : Application Code Discussion : found a small memory error in the code

©2024 Astroinformatics Group