|
[Sponsors] |
July 11, 2006, 15:41 |
Woodcrest benchmark data
|
#1 |
Guest
Posts: n/a
|
Got some benchmark data for the new Intel Woodcrest processor, running CFX 11 Preview 4 64-bit under Linux. This is for the Benchmark.def file in the installation examples directory. The CFX Community Forum has a lots of data for running this case to compare to. Serial: Total wall clock time = 1 min 33 seconds 2-way Local: Total wall clock time = 48 seconds
That's about twice as fast previous best results... nice! Stumpy |
|
July 11, 2006, 16:40 |
Re: Woodcrest benchmark data
|
#2 |
Guest
Posts: n/a
|
Just what I needed ... I'm designing a new cluster. Thanks for the info.
You forgot the most important thing though ... what is the clock rate of your Woodcrests? |
|
July 11, 2006, 16:42 |
Re: Woodcrest benchmark data
|
#3 |
Guest
Posts: n/a
|
Some details on the ram spec and motherboard used would be handy as well ...
|
|
July 11, 2006, 16:54 |
Re: Woodcrest benchmark data
|
#4 |
Guest
Posts: n/a
|
PS: For comparison, my Opteron 275s on Linux 64 bit are almost precisely double your times.
Woodcrests are the bargain of the century considering how much cheaper they are than Opeterons. |
|
July 11, 2006, 16:57 |
Re: Woodcrest benchmark data
|
#5 |
Guest
Posts: n/a
|
To elaborate on the ram spec ... are your FBDIMMs in a quad channel configuration?
|
|
July 12, 2006, 03:49 |
Re: Woodcrest benchmark data
|
#6 |
Guest
Posts: n/a
|
Have you compared the benchmark with CFX 10 on this Woodcrest? Maybe it is to ascribe partial to the new solver. But anyway very interesting.
|
|
July 12, 2006, 06:08 |
Re: Woodcrest benchmark data
|
#7 |
Guest
Posts: n/a
|
Yes I ran it on 10. 10->11 would never result in 100% performance improvement, 5% maybe but not 100%
|
|
July 12, 2006, 07:23 |
Re: Woodcrest benchmark data
|
#8 |
Guest
Posts: n/a
|
That's right. But on my old Smithfield with win32 there was a performance improvement of 30% from cfx10 to 11. But as I told: woodcrest benchmark is interesting at all.
|
|
July 12, 2006, 08:06 |
Re: Woodcrest benchmark data
|
#9 |
Guest
Posts: n/a
|
Perhaps Stumps could re-run on CFX 10.
|
|
July 12, 2006, 08:57 |
Re: Woodcrest benchmark data
|
#10 |
Guest
Posts: n/a
|
A bit more info on the system... CPU clock speed is 3.0GHz It's a 4 socket machine with 8GB RAM
Sorry, thats all - I don't have much access to this system. Stumpy |
|
July 12, 2006, 09:04 |
Re: Woodcrest benchmark data
|
#11 |
Guest
Posts: n/a
|
A 4 socket Woodcrest mobo? AFAIK that doesnt exist... 2 socket max.
I presume you mean 2 sockets x 2 dual core Woodcrests = 4 cores? |
|
July 12, 2006, 10:30 |
Re: Woodcrest benchmark data
|
#12 |
Guest
Posts: n/a
|
Yes, sorry. The OS sees 4 CPU's. I presume it's 2 dual cores. Stumpy
|
|
July 12, 2006, 10:36 |
Re: Woodcrest benchmark data
|
#13 |
Guest
Posts: n/a
|
And how many partitions did you run?
|
|
July 12, 2006, 13:24 |
Re: Woodcrest benchmark data
|
#14 |
Guest
Posts: n/a
|
1 and 2. The case is too small to get meaningfull results with 3 or 4 partitions.
|
|
July 13, 2006, 08:01 |
Re: Woodcrest benchmark data
|
#15 |
Guest
Posts: n/a
|
Have you done any comparison with single-core processor?
|
|
July 13, 2006, 19:01 |
Re: Woodcrest benchmark data
|
#16 |
Guest
Posts: n/a
|
Hi Longbow
There is extensive data on comparative benchmarks on all sorts of CPUs on the CFX Community page, under the solver discussion group. But as Stumpy says, his reported serial speed is half the time of the previously best reported serial run (on a 3+GHz P4 with 2MB cache). Glenn Horrocks |
|
July 13, 2006, 21:40 |
Re: Woodcrest benchmark data
|
#17 |
Guest
Posts: n/a
|
Impressive!!!
By the way, is there any comercial woodcrest workstation or cluster available now? Where can I buy one? |
|
July 14, 2006, 02:12 |
Re: Woodcrest benchmark data
|
#18 |
Guest
Posts: n/a
|
No problem. Every good dealer sales them by now.
|
|
July 14, 2006, 02:15 |
Re: Woodcrest benchmark data
|
#19 |
Guest
Posts: n/a
|
In my opinion the doubled cache might be an explanation for a halv solving time...
|
|
July 14, 2006, 10:51 |
Re: Woodcrest benchmark data
|
#20 |
Guest
Posts: n/a
|
Really?
Good dealer means DELL, HP, SUN? |
|
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Problem in running ICEM grid in Openfoam | Tarak | OpenFOAM | 6 | September 9, 2011 18:51 |
Benchmark data for temperature field in 2D cavity flow | Vladislav | Main CFD Forum | 0 | June 18, 2010 13:33 |
benchmark data | Zarchary | Main CFD Forum | 0 | March 23, 2007 18:30 |
Benchmark flows? DNS data ? | karuna | Main CFD Forum | 0 | June 3, 2006 19:05 |
How to update polyPatchbs localPoints | liu | OpenFOAM Running, Solving & CFD | 6 | December 30, 2005 18:27 |