At 08:41 AM 8/11/2006, Geoff Jacobs wrote:
Mark Hahn wrote:
>> http://www.eweek.com/article2/0,1759,2000867,00.asp
>>
>> 20% is a lot, both in terms of consumption and cooling capacity. I'm
>
> the article says "up to 15%" at the facility level - did you get 20%
> by figuring a savings at the rac
At 09:38 PM 8/10/2006, Mark Hahn wrote:
http://www.eweek.com/article2/0,1759,2000867,00.asp
20% is a lot, both in terms of consumption and cooling capacity. I'm
the article says "up to 15%" at the facility level - did you get 20%
by figuring a savings at the rack level, too?
afaikt, the artic
Um, I haven't looked closely at Woodcrest lately, but everyone does
remember that on a write, you have to fetch the cache line that you are
writing to, right? So, if you have a 10 GB/s memory system, the most a
copy should be able to do is:
read the source at 3.3 GB/s
read the destination at 3.3
Heh...
Stuart Midgley wrote:
> actually, latency determines bandwidth more than memory speed... see my
> rant(?) from a little over a year ago
>
> http://www.beowulf.org/archive/2005-July/013294.html
Someday the english language will be a complete impediment to
understanding - Matt Austern
We
Hi,
Unlikely that a single core is the problem.
Woodcrest 3Ghz we know little technical details though.
A few scenario's for memory reads to L1
as i didn't see public technical data from intel yet.
Most likely it has 1 port @ 2 cycles,
as its 'father' the Banias (hmm bit bitter that near this
Pardon my commenting on a discussion that's a little over my head, but
if I understand the architechtural differences between Opteron and
Pentium/Xeon, wouldn't the latency difference between them be
attributed at least in part to the on-die memory controller?
On 8/14/06, Stuart Midgley <[EMAIL P
Pardon my commenting on a discussion that's a little over my head, but
if I understand the architechtural differences between Opteron and
Pentium/Xeon, wouldn't the latency difference between them be
attributed at least in part to the on-die memory controller?
On 8/14/06, Stuart Midgley <[EMAIL P
actually, latency determines bandwidth more than memory speed... see
my rant(?) from a little over a year ago
http://www.beowulf.org/archive/2005-July/013294.html
prefetch etc. help, but the limiting factor is still latency. Hence
the opterons have significantly higher real memory bandwidt
Hi Stu:
Stu Midgley wrote:
> sorry, forgot to reply all... don't you hate gmail's interface sometimes?
>
>
> What is the memory latency of the woodcrest machines? Since memory
> latency really determines your memory bandwidth.
Hmmm... not for large block sequential accesses. You can prefetch
sorry, forgot to reply all... don't you hate gmail's interface sometimes?
What is the memory latency of the woodcrest machines? Since memory
latency really determines your memory bandwidth.
If Intel hasn't made any improvements in latency then the limited
number of out-standing loads in the x8
Joe Landman wrote:
>4-threads
>
>Copy:6645.4125 0.0965 0.0963 0.0976
>Scale: 6994.6233 0.0916 0.0915 0.0917
>Add: 6373.0207 0.1508 0.1506 0.1509
>Triad: 6710.7522 0.1432 0.1431 0.1433
>
>I may have
Jason Holmes wrote:
Joe Landman wrote:
I have it on good authority that with the other chipset (we have a
Blackford here), we should see higher numbers. Not exceeding the
Opteron 275 though.
We have both a greencreek based (the one with the snoop filter.. I guess
they're calling it the 50
Joe Landman wrote:
I have it on good authority that with the other chipset (we have a
Blackford here), we should see higher numbers. Not exceeding the
Opteron 275 though.
We have both a greencreek based (the one with the snoop filter.. I guess
they're calling it the 5000x now) and a blackf
Yes very dissappointing bandwidth.
Probably the simple logics is that in a dual machine, 2 memory controllers
is always
faster than 1.
http://www.cs.virginia.edu/stream/stream_mail/2005/0020.html
2x 2.8Ghz opteron
-
Function Rate (MB
Mark Hahn wrote:
numbers. It seems that the fortran version of the program with
pathscale 2.3 (the numbers I reported above) reports different numbers
than the C version of the program with pathscale 2.3:
my numbers were from a random stream binary I had around,
so I don't really even remembe
Mark Hahn wrote:
kinda sucks, doesn't it? here's what I get for a not-new dual-275 with
8x1G PC3200 (I think):
Function Rate (MB/s) RMS time Min time Max time
Copy:5714.6837 0.0840 0.0840 0.0841
Scale: 5821.0766 0.0825 0.0825 0.
Dear *,
please allow me to post a final CALL FOR PAPERS for the EuroPVM/MPI
conference (www.pvmmpi06.org):
A new type of session introduced with EuroPVM/MPI 2006 are the "Late and
Breaking Results": 2-page abstracts on the latest results and
approaches can be submitted until September, 1st,
Does anyone know of any mailing lists or forums that discuss the issues
surrounding design/use/operation of clusters and other HPC resources within a
university environment? Would there be any interest in participating in
such a forum?
I may be wrong, but I think a lot of the people on this l
Hi Bill
Bill Rankin wrote:
Hey gang,
Sorry for the slightly OT post, but I figure that clusters have been
such a staple of the higher ed HPC diet for so many years that this
group is probably the best to ask.
Does anyone know of any mailing lists or forums that discuss the issues
surroundi
numbers. It seems that the fortran version of the program with pathscale 2.3
(the numbers I reported above) reports different numbers than the C version
of the program with pathscale 2.3:
my numbers were from a random stream binary I had around,
so I don't really even remember which compiler f
Hey gang,
Sorry for the slightly OT post, but I figure that clusters have been
such a staple of the higher ed HPC diet for so many years that this
group is probably the best to ask.
Does anyone know of any mailing lists or forums that discuss the
issues surrounding design/use/operation of
Mark Hahn wrote:
Here's a 4 core Woodcrest system with 4 sticks of 2 GB 667 MHz FB-dimms:
Function Rate (MB/s) Avg time Min time Max time
Copy: 4813.2590 0.2663 0.2659 0.2675
Scale: 4826.2181 0.2656 0.2652 0.2662
Add:5244.0135 0.3664
Here's a 4 core Woodcrest system with 4 sticks of 2 GB 667 MHz FB-dimms:
Function Rate (MB/s) Avg time Min time Max time
Copy: 4813.2590 0.2663 0.2659 0.2675
Scale: 4826.2181 0.2656 0.2652 0.2662
Add:5244.0135 0.3664 0.3661 0
Here's a 4 core Woodcrest system with 4 sticks of 2 GB 667 MHz FB-dimms:
Function Rate (MB/s) Avg time Min time Max time
Copy: 4813.2590 0.2663 0.2659 0.2675
Scale: 4826.2181 0.2656 0.2652 0.2662
Add:5244.0135 0.3664 0.3661 0
*** ***
*** EuroPVM/MPI 2006 ***
*** www.pvmmpi06.org ***
***
> On Tue, Aug 08, 2006 at 06:41:14PM -0400, Mark Hahn wrote:
>
>> first GPUs, now NICs? http://www.bigfootnetworks.com/ appears to be
>> an actual linux-based coprocessor marketed as an offload NIC for gamers.
>> weird, but true! from their whitepaper, I can't tell whether their
>> performance nu
26 matches
Mail list logo
