At the Four Seasons Hotel, IBM displayed its first machine to make use of the Power6 processor. It was a System p 570 server that used Power5 and Power5+ chips, previously.
One of the major changes that IBM is doing with this most recent generation of Power chips is shifting to smaller manufacturing techniques, having shifted from 90 nanometres down to only 65 nanometres. The shrink is allowing them to add even more components, like a decimal floating point unit and vector math unit. Not only that, but they can increase the speed without the processor generating more heat stating that the Power6 chips, along with the new memory and I/O subsystems, are within five perfect of those of the Power5+ running at 2.2 GHz.
The Power6 has almost double the clock speed of the Power5. This is thanks to IBM's engineers. Instead of shrinking the length of the instruction pipeline, as is what is commonly done with an increase in clock speed, they found a way to keep it the same length, but to make it more efficient.
The added math units are also going to give significant improvement. With the vector processor math unit, the Power6 will have a marked effect on performance while performing floating point calculations. The decimal floating point unit, is said to boost base 10 operations by 2 to 7 times what they were without it. |
|
Reader Comments: 0
