We have a couple of machines at home†:
- Core i7 – 2 cores, hyperthreading, 3Ghz turbo boosting to 3.5Ghz, 16G RAM
- Core i5 – 4 cores, no HT, 3.2Ghz, 32G RAM
I compared them using the POV-Ray 3.7 benchmark with a single worker thread and no other workload than normal background tasks, the i7 system completed in 13m25s average, the i5 system in 12m58s. Now going on clock speed alone I would expect the i5 to have been 6% faster, it was only 3%, so clearly clock-for-clock the i7 does have some advantages, but with boost it should have beaten the i5 in this scenario. The whole point of Turbo Boost is that if you are only using one core it can accelerate it but I’m not seeing that it actually does in any useful way.
Other results, with 4 worker threads the i7 runs it in 6m13s, with 2 threads in 7m1s. So there is some advantage in HT of about 12% even for a compute bound task. But the i5 system runs with 4 threads in 3m35s, nothing beats real actual cores, at a higher frequency! Which is obvious really. Everything else is a gimmick.
No regrets buying the i7 box mind, it will do everything I need it to, and these results are the perfect excuse to really get to grips with something I’ve been meaning to do for years, and have only dabbled in so far, which is offloading compute in a serious way to Azure.
† Well more than a couple!