IBM Reveals World's Fastest Microprocessor: z196
Sorry, this new 5.2 GHz processor wasn't designed for your desktop.
Tuesday during the Hot Chips 2010 conference, IBM revealed more details surrounding its 5.2 GHz microprocessor, the z196. The chip isn't designed for consumer consumption, but rather for the company's Z-series of mainframes. Its cost is estimated to be in the hundreds of thousands--perhaps even over a million bucks.
During the conference, IBM said that the processor is based on CISC architecture and contains 1.4 billion transistors on a 512 sq. mm chip fabricated on 45-nm PD SOI technology. The chip provides 64KB L1 instruction cache, 128KB L1 data cache, and 1.5MB private L2 cache per core. It also uses a pair of co-processors for cryptographic operations.
Additionally, a 4-node system uses 19.5MB of SRAM for L1 private cache, 144MB for L2 private cache, 576MB of eDRAM for L3 cache, and 768MB of eDRAM for a level-4 cache.
IBM engineer Brian Curran said Tuesday that the z196 uses 1079 different instructions. 75 of these instructions can be used by millicode (IBM's term for instructions internally executed by the processor) only, and 219 can be executed by millicode. An additional 24 instructions are "conditionally executed" by millicode.
Curran said that the z196 processor will be available sometime in September.
2) Great if you're trying to create a simulation of the entire universe
3) Still not powerful enough to figure out what the hell the Microsoft Jerry Seinfeld TV ads were trying to sell
The mainframe cores are based on the z196 chip, a 5.2 GHz quad-core out-of-order CISC-based z/Architecture processor. They are the highest clock speed microprocessors available at the introduction date and, at least at introduction, the fastest CPUs in the world.[1] These cores can run a variety of operating systems in any combination, including z/OS, z/VM, Linux on System z, z/VSE, and z/TPF.
A z196 node is a two rack system; the "A frame" consisting of the Central Processing Complex, cooling, none or one I/O drawer or up to two drawers and power supply and the "Z frame" with two system Support Elements and up to four I/O drawers or up to two I/O cages.
Comes in five models, named after the number of processors: M15, M32, M49, M66 and M80.
Up to 96 cores segmented to up to 80 processor units, up to 14 system assist cores and 2 spares.
Up to 3 TB of RAIM (Redundant Array of Independent Memory), i.e 3.75 TB of RAM where 750 GB is used for parity redundancy. This function allows the system to recover from entire memory channel failures in addition to chipkills.
Nodes can be either water or air cooled.
Compatible with the previous z9 EC and z10 EC mainframes providing a clear upgrade path.
(c) Wikipedia
2) Great if you're trying to create a simulation of the entire universe
3) Still not powerful enough to figure out what the hell the Microsoft Jerry Seinfeld TV ads were trying to sell
Something tells me, the massive parallel architecture based in this chip is meant to do things your 980x can only dream of. AMD proved clock speed doesn't matter around 5+ years ago. Where were you?
Its not how fast you spin your wheels, its how much power you put to the road.
Don't get me wrong, Windows, at the API and administration level, is a bit stupid and half-baked, but atleast it kind of makes sense(aside from being stupid and ill-conceived). If any Windows fanboys think I'm making that up, go learn VBScript.