INTEL and AMD dual core server price/performance analysis

AMD has taken 38.2% of the 4 way server market in Q3 2005. Currently, AMD's 4 way dual core server ( 4 Opteron 880) is about 40% faster than INTEL's 4 way dual core servers (4 Xeon MP 7040) in integer performance, and 200% faster in floating point performance (AMD 3x of INTEL).

What about price?

I went to dell.com and configured:

PowerEdge 6850

*)Dual-Core Intel® Xeon™ Processor 7040, 3.0Ghz, 2x2MB, 667MHZ FSB, Quad Proc
*)8GB DDR2 400MHZ(8X1GB), Single Ranked DIMMs
*) 2x 73GB 10K RPM Ultra 320 SCSI Hard Drive
*) floppy, no cd-rom
*) 1470 Watt power supply, 200-240 V only

Unit Price: $27,512.00

Then I went to hp.com and configured:

HP DL585, I choose the Opteron 875 CPU, because with 4 Opteron 875s, the Dl585 has a SpecInt_rate score of 128, which is still 21.9% faster than 4 Xeon MP 7040 score of 105 in integer performance.

*) 4 x Opteron 875
*) 8GB REG PC3200 4/(2X1024)
*) 2x HP 72.8GB Pluggable Ultra320 SCSI 10,000 rpm
*) floppy, cd-rom
*) Integrated Smart Array 5i Plus Controller with Battery Backed Write Cache
*) Redundant 870W hot pluggable power supply

System price: $21,379

So, by going with AMD, you pay $6133 less, and get 21.9% better integer performance and over 60% better floating point performance.

AMD: Price/peformance = 128/21379
INTEL: Price/peformance = 105/27512

AMD/INTEL: 1.57

In other words, HP (AMD) has a 57% price/peformance advantage over DELL(INTEL) when you consider integer performance. In floating point, AMD advantage is off the chart.

Now, consider the power consumption, the DELL box consumes 600 watts more, assuming $0.18/kwh for electricity, that is 600*24*365*0.18/1000 = $946 /year more. Considering the HVAC expense to keep this much heat out and the damaging effects to components such as capacitors and HDDs due to long term continuous exposure to 1470 watts of heat...you get the picture.

Updated: INQ suggests that DELL lovers should ask for 50% discount on DELL servers based on price/performance. It's your money.