Nehalem Xeon L5530-based server power consumption

The mainboard does have integrated basic graphics, but of course I wanted to pair it up with a good piece of graphics hardware - in this case, the ATI Radeon 5870, currently the fastest single GPU platform around. Then, to measure the power load as it really is, I put the trusty old MGE 500W power supply with its own LCD display showing the power usage in place.

What was the measured power consumption, then?

The first test was the platform itself using integrated graphics, no mass storage or such complications. Boot to BIOS and mess around with it, and the PSU showed a consistent 92W to 96W total. Again, this is a dual processor, 8-core 16 thread platform. Not bad at all. But then, what happens when you add an A-Data XPG 64GB SSD as well as the ATI HD5870 card, and then boot Vista?

The power consumption jumped to a still respectable 164W at boot time, but then dropped down to about 138W to 145W when running Vista, as the GPU's power saving features kicked in. Then I ran the 3DMark Vantage benchmark. Of course the power draw jumped, but across all the tests the highest power consumption for the whole system I recorded was 189W. This is half the consumption of the similar W5590 dual Xeon configuration and, guess what, 30 per cent less than the similar setup with default speed AMD Phenom II 965 BE processors.

But then, AMD has the 6-core Istanbul at 1.8GHz, which I found to be in the same power envelope as these Nehalems. However, per core, running actual applications the Nehalem will be up to twice as fast, so using 4-core Shanghai chips with higher core clocks in a similar power envelope might be more appropriate.

Performance wise, it's pretty much one third off the W5590 scores across the board, as CPU, cache and memory speeds are all proportionally slower. For illustration, above is the Sandra CPU benchmark result. Not a bad result for a low power compact ATX system.