Nvidia has unveiled a new product family, Tesla, which is Nvidia’s first real move into the none graphics processing business. Tesla is a HPC (High Performance Computing) platform and it takes advantage of the G80 GPU and its 128 stream processors.
Nvidia is claiming that it is “Industry's first massively multi-threaded architecture with a 128-processor computing core”, which is fair enough as they’re not using the word CPU.
There are three products in the Tesla family, all of which have different target markets. First up is the C870 and you’ll most likely look at this and think it’s just a standard G80 based graphics card. Well, it is, but you can’t use it as a graphics card.
The C870 might fit in a standard x16 PCI Express slot, but it doesn’t have any connectors around the back, since it’s not intended for graphics. It has no less than 1.5GB of memory and it shares this with the Quadro cards. Nvidia claims it will “transforms a standard system into a personal supercomputer with over 500 gigaflops of peak floating point performance”
Well, this is the entry level product and it’s expected to retail for around US$1,499. Not exactly cheap, but if it does what Nvidia claims, then this is a cheap price to pay for all that extra computing power.
You can find out more about the C870 here