Re: Bang for the Buck

From: Jarrod Smith <>
Date: Fri 9 Feb 2001 20:13:31 -0600

I agree with the comments to stay away from the VIA chipset. The best bet
price/performance wise is probably still the BX chipset. But if you're
willing to pay a little bit more for a true "server" platform, consider a
solution built around the Serverworks Serverset III chipset. Supermicro has
some very nice motherboards based on this that take dual coppermine FCPGA
CPUs at 133MHz FSB. The boards cost a little more, and require you to use
registered ECC memory which is a bit more expensive than regular 133MHz
SDRAM. But there's much better technology under the hood, and you end up
with better performance and reliability.

Here's what we're buying:

It supports interleaved memory access which benchmarks close to DDR memory
performance (see recent Serverset III review at without
paying for DDR memory. You just have to install DIMMs in pairs. It has
onboard hotswap SCSI, onboard video and fast ethernet, built-in floppy and
CDROM, and a 64-bit, 66MHz PCI slot for upgrading the interconnect (i.e.
Myrinet 2000). And they squeeze all this in a 1U rackmount case with power
supply. To put together a dual 1GHz with 512MB of RAM and a 10G SCSI drive
should run a little over $2000. I think this is hard to beat for a fairly
high-end solution with hot-swap SCSI and rack-mountability in a high-density


----- Original Message -----
From: "Robert Matthew Fesinmeyer" <>
To: <>
Sent: Friday, February 09, 2001 3:50 PM
Subject: Bang for the Buck

> This question is primarily for the x86 hardware geeks on the list.
> My research group will be purchasing a system or systems to be used for MD
> calculations and trajectory analysis. We primarily look at small
> peptide/waterbox systems using sander in Amber6. Our primary system
> currently is a 1 year old dual P3 650 running RedHat 6.2 that cost ~$2000
> built from off-the-shelf parts.
> In looking at new hardware, our one of our interests is, as the subject
> suggests, "bang for the buck". For the $2000 spent last year, we could
> currently purchase a dual 1 GHz P3 machine using PC133 SDRAM (Via
> chipset).
> Is there anything likely to be gained (given our uses) from RDRAM, Xeon P3
> processors, P4 processors, or even Alpha-based systems? My impression is
> that for the cost of each of these improvements, the "better" answer is
> simply purchase an additional dual 1GHz system and run the 1 calculation
> on X machines (or alternatively, run X calculations on X machines).
> Thank you for your comments.
> Robert Fesinmeyer
> _________________________
> R. M. Fesinmeyer
> Department of Chemistry
> University of Washington
Received on Fri Feb 09 2001 - 18:13:31 PST
Custom Search