Message boards :
Questions and problems :
Benchmarking and redistributing work
Message board moderation
Author | Message |
---|---|
Send message Joined: 1 Aug 08 Posts: 2 |
Is there a likelyhood that the methods used to benchmark PC's will be updated in the future? I ask as the current methods used are not very accurate. A short sample WU (or several from different types of WU's) would be very much more accurate to test how performance is changed by changing memory timings/frequency/amount/etc or altering the way the windows subsystem handles memory (and I guess the linux people would profit as well). Secondly some users such as myself have PC's which are on a network but are not allowed to access the internet. Is there a server/client program I can use which will list the resources of client PC's on a network and utilize them as if they were part of the server PC? Somewhat like a Beowulf cluster I would assume. Also hoping for a GPU client soon ;) |
Send message Joined: 29 Aug 05 Posts: 15480 |
Is there a likelyhood that the methods used to benchmark PC's will be updated in the future? I don't think so, but I have asked the developers. Secondly some users such as myself have PC's which are on a network but are not allowed to access the internet. Is there a server/client program I can use which will list the resources of client PC's on a network and utilize them as if they were part of the server PC? Somewhat like a Beowulf cluster I would assume. The [trac]wiki:Superhost[/trac] is under investigation. Also hoping for a GPU client soon ;) 6.2 has got the code in it, but it isn't working yet. The next client, 6.4 will have full support for GPU and Cell computing. Of course it then still depends on the project, as not all projects will have an application capable of using this. To be honest, I got a new Sapphire HD3850 yesterday, I wonder how well it would do. It's a double precision GPU. :-) |
Send message Joined: 1 Aug 08 Posts: 2 |
Many thanks for the info! As an interesting addition to the Superhost, might it be worthwhile to expand the capabilities to include other programs so it could be used by users wishing to use idle resources on their network for other types of workloads? Rendering, calculation, and pretty much anything that is heavily CPU (or GPGPU) related may benefit from this. |
Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License,
Version 1.2 or any later version published by the Free Software Foundation.