Nvidia has broken records with its latest high performance graphics processor, the Tesla M2090, achieving the fastest-ever performance in a scientific computing test.

The M2090 is equipped with 512 CUDA parallel processing cores and delivers 665 gigaflops of peak double-precision performance. This high computing power is aimed at high-volume data centres, as well as scientific computing.

Pitted onto a biomolecular simulation program as a veritable speed test, four M2090 GPUs together with four CPUs broke the record for the fastest AMBER (the simulation program) performance to date, reaching 69 nanoseconds of simulation per day, a notable step above the previous 46 ns/day set previously.

The M2090 is currently heading out in servers like the HP ProLiant SL390 G7 4U which incorporates up to eight of the GPUs alongside two CPUs, offering the highest GPU-to-CPU density on the market.

Nvidia is also bringing out new consumer graphics cards and drivers to up gaming performance in the wake of 3D gaming.

Nvidia today announced the arrival of the GeForce GTX 560 – the kid brother to the GTX 560 Ti – and the R275 driver to max out current gaming performance. The GeForce GTX 560 GPU starts at US$199.

The R275 driver, available to download from GeForce.com, acts as a patch to boost performance of current Nvidia GPUs for real gaming improvements.

Crysis 2 runs at a six percent improvement in performance, while Bulletstorm runs at an extra 15 percent improvement with the updated driver.

The update also adds greater support for 3D visuals including 3D applications and programmable profiles, improved scaling and improved resizing for HDTVs.