Hello all, just wanted to see what people thought about this build.
The PC will be built by (but he will pay for it, I'm just not charging for the build time) me to a neighbor because he is helping me out with some CAD work (See my Custom Waterblock thread). This will be replacing a 2P Intel Xeon 7040 build.
The build will be used for pro level CAD/CFD, mostly CFD. The CAD models will be imported from a different PC. The CFD software will be: FlowWorks(SolidWorks), CFDesign (Inventor) and Thermal Desktop (Inventor). Because it will be mainly CPU intensive and the actual models won't be created on this PC, there is no real need for a workstation GPU.
The budget for this build is $5000-6500, excluding S&H,etc. Everything will be bought from Newegg, unless there is a huge difference in price for the same product elsewhere (but the site HAS to be reputable like Newegg).
HDD(DATA): 4x Western Digital Caviar Green WD10EADS 1TB SATA2 http://www.newegg.com/Product/Product.aspx?Item=N82E168...
$79 each; $315 total.
These HDDs are going to be in RAID5 for a total of 3GB. The reason For 3TB is the simulations will be saved ~ every 50-100 iterations. Typical simulation will run for ~ 60000 iterations. Each saved iteration point is ~20MB so a full simulation is ~ 25GB. After the sim most of the saved will be compiled in to one result file (~1-2GB).
^Yeah, I know, any recommendations? The reason I didn't get the E5540 is that the extra 530Mhz wasn't worth the extra money. For what this will be used it would only make it about 10 minutes faster. As the simulations are usually run over night, there is no reason for the extra cost.
I was also thinking, would a 4P (Core 2) make sense than a 2P Nehalem for this kind of work? Either way, this is a huge step up from the 2P he was using earlier.
It's a significant clockspeed bump (it should give a 20%+ speedup in processor limited tasks). I don't see how that equates to only 10 minutes, unless the overall task is only 50 minutes or so (in which case why run it overnight?). As for a 4p core 2, no, that isn't the way to go. It would use quite a bit more power, and not give significant gains in performance over a 2P nehalem.
It's a significant clockspeed bump (it should give a 20%+ speedup in processor limited tasks). I don't see how that equates to only 10 minutes, unless the overall task is only 50 minutes or so (in which case why run it overnight?).
The sim software (mainly Thermal Desktop), for some reason don't benefit from speed increase (not sure about the core # increase) after a certain point (in the designs sims he run). The benefit, for example in a Q6600 @ 3Ghz is almost the same as that of a Q6600 @3.6Ghz. Only a few minutes difference on the results for some reason. I think this has more to do with the Auto optimization of the cells,etc at certain clock speeds. For the Q6600 @ 3.6Ghz, the result (time step=~.00000001s) was smaller, meaning higher detail than the Q6600 @3Ghz (time step = ~ .00001s). After a certain point, the accuracy/precision is insignificant for the real world application that are being simulated.
^ There are new stuff out now. It's basically what I have in my OP (except the case was swapped for a HAF 932). The HDDs were also swapped for WD Blacks instead of the Greens due to RAID issues. As for the RAID card, using an IBM SATA RAID card (IIRC, this one: http://www.redbooks.ibm.com/abstracts/tips0738.html ,M5015)
The system was recently (February) upgraded to a 160GB SSD (X25-M G2). The set up was also fully water cooled with 2x Swiftech XTs + 2x MCP655+ 2xMCR 320s at the same time. Total, ~$3500-3600 with RAID card and SSD.