|
[Sponsors] |
November 12, 1999, 08:10 |
CFD on Compaq XP1000
|
#1 |
Guest
Posts: n/a
|
Hello everybody,
does anybody run his CFD Jobs on a Compaq XP1000 machine ? I am very interested if this machine is in real life applications as fast as benchmark tests (SPEC) do promise !! Thanks, Tim |
|
November 13, 1999, 20:47 |
Re: CFD on Compaq XP1000
|
#2 |
Guest
Posts: n/a
|
(1). I remember a few months back, Dr. Kenji Takeda from University of South Hampton (www.soton.ac.uk/~ktakeda)mentioned the Alpha CPU in his multi-processor super-computing environment. (2). In his web page, there are also CFD animations. You can visit his web page or e-mail him about the specifics related to the Alpha CPU speed. (3). From the SPEC results on Internet, It is of order three times faster than other CPU for PC. (4). Based on my three PC upgrades experience, the speed gain is fairly consistent with the publicly available speed data. (from 486/33, to pentium/90, pentium/266, and to AMD/k6/400. Don't remember the exact speed of the old CPU, so these numbers are approx.) I used to run my own 3-D Navier-Stokes codes on my PC, so there is no surprise. (5). The other place you can visit is the company called "Microway", which has been selling high performance multi-processor PC for a while. I think, they also have Alpha CPU system available. (6). Let us know your finding later.
|
|
November 15, 1999, 03:36 |
Re: CFD on Compaq XP1000
|
#3 |
Guest
Posts: n/a
|
We have several Compaq XP1000 and we are very satisfied about them. It's probably the best and fastest processor on the market. We use them for CFD extensively and love their speed. The processor truly delivers expected specs. Also, its price/performance is unbeatable.
|
|
November 15, 1999, 06:06 |
Re: CFD on Compaq XP1000
|
#4 |
Guest
Posts: n/a
|
Dear Michel,
could you comment your experience with the graphical performance of that Workstation. Is it satisfying as well ? Is it possible to have a multiprocessor machine of that type ? Tim |
|
November 15, 1999, 08:11 |
Re: CFD on Compaq XP1000
|
#5 |
Guest
Posts: n/a
|
We are using PowerStorm 4D51T, opengl accelerated, and it also works nicely. We use it for wireframe, color or gouraud rendering and it is more than enough for our CFD analysis. You just need to install some patches to avoid unexepected reboot of your machine.
I know there exists a two-processors architecture for that processor, more I don't know. |
|
November 15, 1999, 09:35 |
Re: CFD on Compaq XP1000
|
#6 |
Guest
Posts: n/a
|
Michel,
are you using WinNT or Unix on that machine ? If you are using WinNT, what are your experiences concerning multi-tasking possibilities of WinNT. (e.g. grid generation while running a CFD job) Thanks, Tim |
|
November 15, 1999, 10:48 |
Re: CFD on Compaq XP1000
|
#7 |
Guest
Posts: n/a
|
We only use UNIX on these machines. If you want do to multitasking, you better use Unix. It's certainly also more stable, and you can easily use this computing power from other machines.
|
|
November 15, 1999, 21:29 |
Re: CFD on Compaq XP1000
|
#8 |
Guest
Posts: n/a
|
I have used Fluent5 and 4.5 on a dual Pentium machine with no problems regarding stability. Multitasking has not been a problem either (both multiple simulations and simulation/grid generation). However, because we have an academic license we have not been able to try the parallel version.
|
|
November 22, 1999, 15:56 |
Re: CFD on Compaq XP1000
|
#9 |
Guest
Posts: n/a
|
We are running Fluent software on a variety of architectures: SGI Octane (R10000/195MHz) and Origin (R10000/200MHz), PII/450, and DEC Alpha 21264 (we have a dual cpu system from Microway http://www.microway.com with two 500MHz cpu's). Here's some benchmark results for 20 iterations of a ~300k node problem:
1.) Octane: 18 min 2.) PII/450: 14.5 min 3.) Alpha (Tru64 Unix): 7 min 4.) Alpha (NT): 12 min There are some problems with Fluent on the alpha, but my tech support guy tells me they will be resolved with 5.2.3. Microway will give you a *much* better price than Compaq and is one of the only vendors who will supply Tru64 Unix and NT. We have Tru64. There's a significant performance penalty for using NT. |
|
November 22, 1999, 16:27 |
Re: CFD on Compaq XP1000
|
#10 |
Guest
Posts: n/a
|
Interesting numbers. Thanks for sharing them with us. What did you run on the PII/450, NT or Linux?
Looking at those numbers I think that the smartest choice for raw CPU power per $ is to buy a standard PC equipped with a 700 MHz Athlon CPU (should be about 50% faster than the PII/450). It is probably significantly cheaper per CPU than the 21264 alphas, even if you buy them from Microway or similar. |
|
November 22, 1999, 17:25 |
Re: CFD on Compaq XP1000
|
#11 |
Guest
Posts: n/a
|
Our PII/450MHz runs NT.
I am pretty excited about the Athlon performance that has been published. At the same clock speed, they seem to outperform PIII's by about 15%, and they have higher clock speeds available (I'm not sure about this now that the PIII/700 is out.) From my perspective, though, hardware is cheap. Let me give you some numbers to consider. A dual alpha/667MHz from Microway (Compaq charges significantly more) with 2GB RAM costs about $24k. I'm not sure about Athlons, but a similarly equipped dual PIII/550 from Dell with 2GB RAM would cost about $16k (this price is a little old). Memory is a big chunk of this cost. If you think about how much the software and the people to run it cost, then saving that $8k is a poor choice if it affects productivity. High-end CFD software costs over $10k/cpu/yr (depending on the license, but this is an estimate based on my experience). In our problems, (we are a vendor of combustion equipment) we sometimes use Fluent to determine what equipment we will ship to a customer or how we will configure that equipment. We regularly run CFD problems for 100 or more hours. Shaving 20% (20 hrs) off of that definitely helps our project turnaround time because we can shave a day off of the time required to make a decision. If we then have to make a new calculation, it can start a day earlier and finish two days ahead of schedule, etc. Good parallel performance (which Fluent definitely has) is a really big benefit. |
|
November 23, 1999, 03:26 |
Re: CFD on Compaq XP1000
|
#12 |
Guest
Posts: n/a
|
Mike,
thanks for this interesting facts ! Am I right that your test case did run on a single alpha processor, didn't it ? Tim |
|
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
CFD Wiki - We Need More Help! | Jonas Larsson | Main CFD Forum | 0 | September 26, 2005 08:11 |
Can we quantify the fruits of CFD? | Brady Brown | Main CFD Forum | 14 | December 15, 1999 09:42 |
Since Last June | John C. Chien | Main CFD Forum | 3 | July 12, 1999 09:38 |
Which is better to develop in-house CFD code or to buy a available CFD package. | Tareq Al-shaalan | Main CFD Forum | 10 | June 12, 1999 23:27 |
public CFD Code development | Heinz Wilkening | Main CFD Forum | 38 | March 5, 1999 11:44 |