[Beowulf] opteron 248 vs dual-core opteron 275

Mark Hahn hahn at physics.mcmaster.ca
Sun Dec 4 10:24:51 PST 2005


> on a one per-core basis. Will this performance be better on a 
> 2proc./4cores node or on two 2proc/2node configuration. I am talking 

threading is completely irrelevant.  the only question is: under your 
workload, will half the memory bandwidth per core hurt performance?
this is not answerable in general, since workloads can vary from 
not to completely dram-dominated.

> On the other hand, it seems clear to me that heat dissipation of the 
> dual-core configuration will be much better. Can anyone confirm that. 

you don't get something for nothing.  DC does amortize some hardware,
so there is a power savings there.  but I think most people just look
at the TDP of SC and DC, and ignore the fact that the SC TDP is for a 
higher-clocked chip.  the actual dissipation of a SC at the same clock
of a DC (ie, a couple steps down from top) will be lower than the SC TDP.

> Does the dual-core cpu dissipate approx. the same heat as a single-core one?

no, not at the same clock.

> Are there many problems on the use of dual-core application by linux OS, 
> or is it really a no-problemo thing as I have read?

there was a bug having to do with misunderstanding of whether the TSC on
DC was synced or not.  but certainly no functional problems.

> BTW, we use our cluster mostly for computational chemistry calculations, 
> mainly gaussian, turbomole, molcas and such.

I'd expect these to be mostly dram-limited, and thus _not_ especially
favorable for DC.  worse yet is where the app is scratch-disk-limited...




More information about the Beowulf mailing list