Feeds

Nvidia fires off Fermi, pledges radical new GPUs

Three billion-transistor HPC chip, anyone?

Top 5 reasons to deploy VMware with Tegile

Nvidia last night introduced the new GPU design that will feed into its next-generation GeForce graphics chips and Tesla GPGPU offerings but which the company also hopes will drive it ever deeper into general number crunching.

While the new chip is dubbed 'Fermi', so is the architecture that connects a multitude of what Nvidia calls a "Streaming Multiprocessor". The SM design the company outlined yesterday contains 32 basic Cuda cores - four times as many found in previous generations of SM - each comprising one integer and one floating-point maths unit. It is able to schedule two groups of 32 threads - a group Nvidia calls a "warp" - at once.

Nvidia Fermi chip

Nvidia's Fermi: each of the 16 green strips is...

The networked cores connect to 64KB of shared L1 cache, also used by four Special Function Units (SFUs) which handle complex maths formulae such as sines and cosines.

Fermi itself packs in 16 SMs - that's 512 Cuda cores in total - which tap into shared 768KB L2 cache and can reach out to a maximum of 6GB of GDDR 5 memory over a 384-bit interface and with ECC support

This is only the first Firmi GPU design. It's aimed at science and engineering GPGPU apps rather than game graphics, so future Fermi-based GeForce chips will likely sport less complex layouts. GT300, Nvidia's next GPU core, will be derived from Fermi, but don't expect it to show off all the superlatives Nvidia has been claiming for the Fermi chip.

Stream Multiprocessor

...one of these 32-core Stream Multiprocessors

The first Fermi design uses three billion transistors. It supports 64-bit addressing, double that of the previous generation GT200 architecture, the one used in GeForce GTX 200-series GPUs. It can handle 256 double-precision floating point ops per clock and 512 single-precision ops, up from GT200's 30 and 240, respectively.

And the key fact Nvidia didn't mention: when the Fermi-sourced GeForces will go on sale. ®

Beginner's guide to SSL certificates

More from The Register

next story
Ellison: Sparc M7 is Oracle's most important silicon EVER
'Acceleration engines' key to performance, security, Larry says
Oracle SHELLSHOCKER - data titan lists unpatchables
Database kingpin lists 32 products that can't be patched (yet) as GNU fixes second vuln
Lenovo to finish $2.1bn IBM x86 server gobble in October
A lighter snack than expected – but what's a few $100m between friends, eh?
Ello? ello? ello?: Facebook challenger in DDoS KNOCKOUT
Gets back up again after half an hour though
Troll hunter Rackspace turns Rotatable's bizarro patent to stone
News of the Weird: Screen-rotating technology declared unpatentable
prev story

Whitepapers

Forging a new future with identity relationship management
Learn about ForgeRock's next generation IRM platform and how it is designed to empower CEOS's and enterprises to engage with consumers.
Storage capacity and performance optimization at Mizuno USA
Mizuno USA turn to Tegile storage technology to solve both their SAN and backup issues.
The next step in data security
With recent increased privacy concerns and computers becoming more powerful, the chance of hackers being able to crack smaller-sized RSA keys increases.
Security for virtualized datacentres
Legacy security solutions are inefficient due to the architectural differences between physical and virtual environments.
A strategic approach to identity relationship management
ForgeRock commissioned Forrester to evaluate companies’ IAM practices and requirements when it comes to customer-facing scenarios versus employee-facing ones.