Intel: GPUs aren't 100 times faster than CPUs, just 14 times. Nvidia: Oh no!

Devin Coldewey

Devin Coldewey is a Seattle-based writer and photographer. He has written for the TechCrunch network since 2007. Some posts he’d like you to read: The Dangers of Externalizing Knowledge | Generation i | Surveillant Society | Choose Two | Frame Wars | The User’s Manifesto | Our Great Sin His personal website is coldewey.cc. → Learn More

Thursday, June 24th, 2010


This is pretty funny. You’ve probably seen some of the propaganda over the last year or so about how GPUs are orders of magnitude faster than CPUs at certain tasks, due to their parallel processing engine. Intel got tired of hearing about it, I guess, and decided to debunk the myth. They set out to disprove the notion that a GPU can be 100 times faster than a CPU. They kind of did it, but I think this is what is termed a Pyrrhic victory.

Interestingly, they chose a rather old GPU and a comparatively new CPU to compare: an Nvidia GTX280 and a Core i7 960. Maybe they chose on price parity? Whatever the case, they found that while indeed the GPU advantage was not as great as suggested, it was significant: 2.5 times faster on average and up to 14 times faster in certain situations.

Nvidia took the opening for all it was worth:

It’s a rare day in the world of technology when a company you compete with stands up at an important conference and declares that your technology is only up to 14 times faster than theirs.

To be fair: we can’t really expect unbiased judgment from either side, and the tests in the paper are too complex to be boiled down to a “oh it’s this much faster” talking point. I just think the drama is hilarious.

[via TG Daily]

blog comments powered by Disqus