Related Articles

Introducing some new benchmarks

There always comes a time when the existing benchmark tools are no longer sufficient. This is one of those times.

Our old friend, Sandra 2004 is simply not up to providing reliable, repeatable results with the newest high speed multi core processors. We are therefore adding Sandra 2007, and will be phasing out running Sandra 2004 over the next few reviews. It has served us well, but it is time to update it.

As we are updating Sandra, we've also added a few more sandra benchmarks - so for now, instead of just processor and memory bandwidth, we will also include MMX, memory latency, and inter-core bandwidth.

We have also been getting a LOT of requests to start using Quake 4 for benchmarking; therefore from this article we are adding Quake 4 to processor, memory and motherboard benchmark suites. We are considering retiring our old faithful Doom 3 - but that decision is not final yet.

Quake 4

Quake 4 turns out to be well suited (when run at low resolutions) to show the potential effects of higher speed processors assuming that the GPU's are not the speed limitation - which we do, by running at low resolution. These tests were run with "multi cores" enabled; and future reviews will show more results as we build up our database of results.

We show a 57.1fps gain (50.7%) from a 70.9% increase in processor speed. That's pretty good scaling!

Sandra 2007 CPU Tests

We can see a very clear advantage for the overclocked setting; as a matter of fact we get a 45.5% increase in integer performance, and a 45.1% increase in floating point performance.

Sandra MMX Tests

I thought with Intel and AMD continously tweaking their SSE/SSE2/SSE3 circuitry, we should try running the MMX tests again.

Here, we showed quite strong scaling,  with an 84.8% improvement in integer SSE performance, greater than the increase in clock rates, therefore some of the gain must be from additional memory bandwidth at the higher speed.

Sandra Latency Tests