BD performance under VMware

Wrench00

2[H]4U
Joined
Sep 30, 2003
Messages
3,423
I know that games like single threaded stuff and the windows scheduler is kinda well all over the place.

I would love to see how BD does under VMware. Since the scheduler under VM is very different I wonder how well these chips will do under a virtual environment. If you have a good consolidation ratio you can save huge amount of money on hardware. Also servers don't really care about top performance I wonder how many VM's could be stuck on a single host? The question is is there enough Memory bandwidth supply all those cores? I wish I had a BD server so I could play.
 
But servers do care about power usage. Power in datacenters is expensive and Zambezi/Interlagos is worse than Thuban/Magny Cours. That, and AMD is full of shit with their module = awesome theory, because Interlagos is two Zambezi dies with HT linked... where are the advantages of modules? Memory bandwidth still sucks.

'dozer won't do any better in VMware than Thuban. Sandy will still be a far better option, especially with 2011 bringing proper quad channel (not two channels/die, striped over Hyper Transport).
 
But servers do care about power usage. Power in datacenters is expensive and Zambezi/Interlagos is worse than Thuban/Magny Cours. That, and AMD is full of shit with their module = awesome theory, because Interlagos is two Zambezi dies with HT linked... where are the advantages of modules? Memory bandwidth still sucks.

'dozer won't do any better in VMware than Thuban. Sandy will still be a far better option, especially with 2011 bringing proper quad channel (not two channels/die, striped over Hyper Transport).
Minor correction: Interlagos is two Orochi dies in a G34 package, and Valencia is one Orochi die in a C32 package, whereas Zambezi is one Orochi die in AM3/+ package.

It is possible that Valencia/Interlagos may out perform Lisbon/Magny-Cours in VMware and the like, just have to wait until someone tests it.
 
Minor correction: Interlagos is two Orochi dies in a G34 package, and Valencia is one Orochi die in a C32 package, whereas Zambezi is one Orochi die in AM3/+ package.

Right!

It is possible that Valencia/Interlagos may out perform Lisbon/Magny-Cours in VMware and the like, just have to wait until someone tests it.

It is possible, but judging from large drop in single thread performance, and multi threaded barely faster than Thuban, I am doubtful.
 
I would like to know VM performance under BD vs. SB vs. Thuban. Could some of you guys run VMmark from VMware? I know, I know, synthethic gross, etc. but at least its something relatively unbiased and should give us some idea of how it'll do. I'm itching to upgrade my X3 720 in my little test server here at home, but not sure if it'd be worth it to go with a BD 6xxx 6-core or Thuban, or maybe just grab a low-power i5 SB...
 
Back
Top