PDA

View Full Version : A Peek at Faster Power6, Cell Chips


SLippe
12-30-06, 03:26 AM
Judging by details revealed in a chip conference agenda, the clock frequency race isn't over yet.

IBM's Power6 processor will be able to exceed 5 gigahertz in a high-performance mode, and the second-generation Cell Broadband Engine processor from IBM, Sony and Toshiba will run at 6GHz, according to the program for the International Solid State Circuits Conference that begins February 11 in San Francisco.

Chipmakers have run into problems increasing chip clock speed--essentially an electronic heartbeat that synchronizes operations in a processor--because higher frequencies have led to unmanageable power consumption and waste heat.

To compensate, Intel and Advanced Micro Devices have turned instead to the addition of multiple processing cores on each slice of silicon. That's effective when computers are juggling numerous tasks at the same time, but increasing the clock speed means an individual task can run faster.

The first-generation Cell Broadband Engine chip, co-developed by IBM, Sony, and Toshiba, has just appeared in Sony's PlayStation 3 game console and can run at 4GHz. The second-generation chip will run at 6GHz, according to the ISSCC program. In addition, the new chip will have a dual power supply that increases memory performance--a major bottleneck in computer designs today.

For servers, IBM has said its Power6 processor, due to ship in servers in 2007, will run between 4GHz and 5GHz. But in the ISSCC program, Big Blue said the chip's clock will tick at a rate "over 5GHz in high-performance applications." In addition, the chip "consumes under 100 watts in power-sensitive applications," a power range comparable to mainstream 95-watt AMD Opteron chips and 80-watt Intel Xeon chips.

Power6 has 700 million transistors and measures 341 square millimeters, according to the program. The smaller that a chip's surface area is, the more that can be carved out of a single silicon wafer, reducing per-chip manufacturing costs and therefore making a computer more competitive. Power6, like the second-generation Cell, is built with a manufacturing process with 65-nanometer circuitry elements, letting more electronics be squeezed onto a given surface area.

Intel isn't standing idly by, though. In September, Intel showed a glimpse of a prototype chip with 80 cores that can perform a trillion mathematical calculations per second. At ISSCC, the company will shed more details on the design, including an updated speed measurement of 1.28 trillion calculations per second.

The chip measures 275 square millimeters--smaller than the 303-square-millimeter area indicated in September--and runs at 4GHz, according to the program. The chip, which Intel describes as a "network-on-chip architecture," has 100 million transistors and dissipates 98 watts of waste heat. Intel called each core a tile and said each has network switch features to route packets of data.

"It was designed as a research tool to test interconnect strategies for many-core processors," Intel spokeswoman Erica Fields said. Research goals for the project included testing new chip design methods and investigating "how to move terabytes of data rapidly between cores on-chip and between the cores and memory." She added that the prototype can't run conventional software for Intel chips.

Well, any thoughts?

Subtestube
12-30-06, 04:57 PM
It's hard to say how much it matters [to desktop users] really. IBM chips aren't in any commercial desktops, and let's be honest - they aren't going to provide upgraded PS3s with the new Cell. For servers it's interesting, but the whole 'increasing the clock means more on a single task' is only a half truth, as we know. For single threaded apps a Core 2 will slaughter a much higher clocked Prescott, because per task efficiency, memory fetching, and branch disambiguation all play a huge part in the real world performance. Ultimately, interesting, but aside from the server market, I think it'll take some time for this tech to filter down to us.