<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
<title></title>
</head>
<body bgcolor="#ffffff" text="#000000">
Daniel Pfenniger wrote:<br>
<br>
"But while configuring a cluster remember that you can get a better<br>
deal (=speed/cost for your application) by choosing cheaper CPUs in <br>
larger quantities."<br>
<br>
I disagree. Sometimes this is true, and sometimes it is not.<br>
<br>
The CPU cost is a fractional element of the total cost. The total cost
can vary quite a bit.<br>
<br>
Let's do a fictional example, with 60 nodes, one rack, GbE networking:<br>
<br>
Quad core Opteron blades, in a 10 blade 8U chassis.<br>
6 chassis per rack, 60 blades, 120 CPUs, 480 cores<br>
<br>
<br>
1 Motherboard Opteron S1207x2/ split rail power, newer chipset,
video/GbE*2<br>
2 AMD Opteron2347 Quad Core 1.9GHz<br>
8 2GB DDR2 667MHz ECC REG DIMM Assuming 16GB RAM ( 2GB per core)<br>
1 160 GB 7200rpm HDD<br>
Total per node: $ 2,600<br>
<br>
Blade frame/chassis/power cost per node: $300<br>
<br>
Rack, PDU, GbE network switch, cables, etc., etc, per rack of 60
nodes: $6,000<br>
Cost per node: $100<br>
<br>
Total: $180,000<br>
<br>
So, each node costs roughly $3,000 with 1.9GHz CPUs ( 2 per node)<br>
Each CPUs is about $400/3000 of the cost, or 13.4%<br>
<br>
Upgrade to:<br>
2.0GHz: Each node is $3,200, CPU $500 15.7%<br>
2.3GHz: Each node is $4,200, CPU $1000 23.8%<br>
2.5GHz: Each node is $5,150, CPU $1475 28.7%<br>
<br>
Assuming performance is pretty well linear with clock speed in your
applications.<br>
Yes, a big assumption, but one that holds true for most, <br>
unless you are limited by memory bandwidth or network performance..<br>
<br>
Using the above figures, we can see that:<br>
Upgrade from 1.9 to 2GHz: 5.3% performance gain, 6.6% cost increase.<br>
Upgrade from 1.9 to 2.3GHz: 21.1% performance gain, 40% cost increase.<br>
Upgrade from 1.9 to 2.5GHz: 31.5% performance gain, 72% cost increase.<br>
<br>
When one factors power consumption and cooling the curve of
cost/performance certainly gets steeper.<br>
But, the jump from 1.9 to 2.0 is a reasonable one.<br>
Going from 1.9 to 2.0GHz is an example that disagrees with your
statement.<br>
<br>
Going from 1.9 to 2.3GHz it is different:<br>
If we factor in the performance gain ( again, assuming even scaling by
simply adding nodes):<br>
Add a second rack, add 13 more nodes, to gain 26.4% mode performance (
equiv to 1.9 vs 2.3GHz) :<br>
2 more 10 blade chassis, 16U, more network, another rack, PDU, etc.
$8,000<br>
13 nodes @ $3,000 = $39,000<br>
$180,000 + $47,000 = $227,000<br>
227/180 = 26.1% cost for 21% performance gain.<br>
It seems that the gain for more CPUs versus faster holds true for the
jump to 2.3GHz.<br>
With Quad core Opterons it is more cost effective to add nodes versus
faster CPUs comparing 2.3GHz to 1.9GHz.<br>
<br>
Of course you are supporting a second rack, more power, more cooling to
do so.<br>
And, you are maintaining more nodes which is more work, more risk of
failure, etc..<br>
<br>
-----------------------------------<br>
Barnet Wagman wrote:<br>
"Of course I'd rather wait for AMD's quad but that's not an option (I <br>
doubt that they'll be readily available until next year). So I'm <br>
leaning towards the low cost per node solution - one quad processor <br>
(probably a Q6600) per node."<br>
<br>
Huh? What does "readily available" mean to you?<br>
How many machines would you like with these?<br>
Give me a call.<br>
<br>
<br>
<br>
<div class="moz-signature">-- <br>
<font color="blue">With our best regards,</font><br>
<br>
<font color="black">
<i><i>Maurice W. Hilarius Telephone: 01-780-456-9771</i><br>
<i>Hard Data Ltd. FAX: 01-780-456-9772</i><br>
<i>11060 - 166 Avenue <a class="moz-txt-link-freetext" href="email:maurice@harddata.com">email:maurice@harddata.com</a></i><br>
<i>Edmonton, AB, Canada <a class="moz-txt-link-freetext" href="http://www.harddata.com/">http://www.harddata.com/</a></i><br>
<i> T5X 1Y3</i><br>
</i></font>
</div>
</body>
</html>