Nvidia has unveiled new Fermi-based Tesla products, the Tesla 20 series GPUs, which are aimed at the high-performance computing market. Designed for parallel computing, the company said its Tesla 20 series graphics chips reduce the cost of computing by delivering the same performance of a traditional CPU-based cluster at one-tenth the expense and one-twentieth the power consumption.
Nvidia's Tesla 20 series chips are reportedly equipped with features that speed up many applications, such as ray tracing, 3D cloud computing, video encoding, database search, data analytics, computer-aided engineering, and virus scanning. The new GPUs will also combine various parallel computing features that have never been incorporated on a single device before.
Among the features is support for the next-generation IEEE 754-2008 double precision floating point standard, ECC, multi-level cache hierarchy with L1 and L2 caches, and support for the C++ programming language. The Tesla 20 series also supports as much as one terabyte of memory, concurrent kernel execution, fast context switching, 10x faster atomic instructions, 64-bit virtual address space, system calls, and recursive functions.
Nvidia's new graphics lineup will consist of the Tesla C2050 and C2070, as well as the S2050 and S2070. The first two will be workstation-friendly single GPU PCIe 2.0 cards with up to 3GB and 6GB of GDDR5 memory and performance in the range of 520GFlops to 630GFlops. The latter two will be aimed at datacenters, boasting four GPUs, up to 12GB and 24GB of GDDR5 memory, and a performance of 2.1TFlops to 2.5TFlops.
In order of mention, the Tesla 20 series GPUs will cost $2,499, $3,999, $12,995, and $18,995 – all will arrive in the second quarter of 2010.