The press release from NVidia reads NVIDIA® Tesla™ GPU Computing Processor Ushers In the Era of Personal Supercomputing. NVidia (so far) very carefully fails to mention any hint that their new Tesla branded GPU could possibly be a future competitor to AMD, Intel, or IBM's PowerPC CPU lines.
After all, it's a Graphics Processor Unit, right?
I don't think so. I think the CPU market just entered a revolutionary battle unlike any we have seen in the past. If the claims are true (and they seem to be verified by several leading research centers, see links from the press release) then this is more than an incremental improvement -- it's a major jump in processing technology available this coming August that surpasses anything Intel or AMD has announced into the next two years.
Here's a hint: MRI (Magnetic Resonance Imaging) processing speeds increased on Tesla from 245 to 415 times previous speeds over CPU or older GPU based computing solutions. There's more, but you can read the details yourself over on the NVidia site. (links pop) Performance increases are so good that in many cases scientific computing that took weeks and a cluster of machines can now be done in days or hours on a single machine.
My feeling is that this is NVidia's shot across the bow of the entrenched CPU market leaders. They may not remove the misleading "GPU" designation for a while yet, but make no mistake: this technology has the potential to completely change what we consider a "Desktop PC." And it's about time!
And they've done it right: the first generation of Tesla can be installed onto almost any existing PC with a modern PCI express bus. They've released a free API. There are several open source simulations that you can immediately download. The entry barriers (other than price, tba) are very low.
We moved from slide rules to programmable calculators and were amazed at the changes to our lives. Then we moved to the personal computer -- that box sitting under or near your desk today contains more power than most supercomputers built 12 years ago.
It's time for a jump in processing power to the next order of magnitude.