Opteron kernel

[Derek Richardson]

Bill,
It had 4 DIMMs, w/ 6 slots on of the motherboard.  I am going to go 
confirm which banks correspond to what shortly, just in case IBM put it 
together w/ an imbalance.  Yes, we have done the 2 job scenario ( that's 
our primary mode of operating, running 2 jobs ( called executive shells, 
or es's ) on a dual-cpu node, w/ anywhere from 10 up to 125 nodes 
participating in a job ), the memory should be DDR333 IIRC.  The BIOS 
doesn't have much in the way of tuning options, more's the pity.
What's the benchmark?  If it's a publicly available one, I probably have 
it installed already, if not, yes, I would appreciate it.
When you have seen it repeatedly, you mean an improper distribution of 
memory to CPU?
Derek R.

Bill Broadley wrote:

>>for testing ) and I assumed to be correct.  But given the poor 
>>performance I have seen ( 2 GHz Opterons coming in at ~15% slower than a 
>>3 GHz P4 on a compute/memory intensive application when most benchmarks 
>>I have seen would imply the inverse ), I will double-check that when 
>>    
>>
>
>I've seen this repeatedly.  
>
>Did each opteron cpu have 2 or 4 dimms attached?
>
>Did you benchmark TWO jobs on the opteron vs TWO jobs on the P4?
>
>Is the memory at least PC2700?
>
>Have you played with the BIOS settings, I've seen significant speedups
>playing with both the node interleaving and the memory interleaving.
>
>I can provide a benchmark that should show 2GB/sec to main memory
>on a single opteron, and 3 GB/sec to a dual opteron if it's properly
>setup.
>
>  
>

-- 
Linux Administrator
derek.richardson at pgs.com
derek.richardson at ieee.org
Office 713-781-4000
Cell 713-817-1197
Madison's Inquiry:
	If you have to travel on the Titanic, why not go first class?


_______________________________________________
Beowulf mailing list, Beowulf at beowulf.org
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf