Intel Woodcrest Performance Preview
Posted on: 05/17/2006 05:00 AM

SiSoft Sandra 2007 (cont'd)

I have been a huge fan of Ace's precompiled Linpack binary for a long time. Recently though, we've started running into the limits of the benchmark as it only supports datasets out to 2MB. Since a lot of the modern processors we look at around here have L caches larger than 2MB, Linpack doesn't let us see how bandwidth scales all the way out to main memory. I think SiSoft saw an opening here, and while their Cache/Memory benchmark isn't Linpack, it does do what we need it to do.

I split this test into two different graphs to better show the scale of the bandwidth as the block sizes increase. There are two very different tests going on here: Cache speed and main memory speed.

It is painfully obvious that Woodcrest's cache is extremely fast. As we move out into main memory, though, the Opteron takes and maintains the lead in the bandwidth department.

Many people have expressed concern over latency penalties incurred by Intel's switch to FB-DIMM technology. In the Bensley Preview I didn't have the a way to test latency in a 64bit operating environment, but that has changed. As if they were reading my mind, Sisoft added a latency benchmark to this new version of Sandra, and here are the results (also split into two graphs):

As you can clearly see, the latency concerns are unwarranted. The Dempsey, Woodcrest and Opteron systems are pretty much neck-and-neck when the block size spills into main memory.

Now that we've gotten the synthetic tests out of the way, let's move on to the "real world" workloads and see if any of the performance trends carry-over.

Printed from (,4.html)