Hacker News Reader

Comment by j-g-faustus | Hacker News Reader

Here's an article describing what it takes to reach those speeds: http://www.tomshardware.com/reviews/5-ghz-project,731.html

The article is a few years old, and the clock limits have changed, but the principles of reaching them are the same.

As for replicating it at large in the public market, I don't anticipate that liquid nitrogen feeds will be ubiquitous anytime soon :)

View on HN

The top IBM POWER7 CPU is clocked at 5ghz and isn't cooled with liquid nitrogen.

As I said, it's a few years ago. Today you can get 5 GHz with Sandy Bridge on air (although it might entail testing a few dozen CPUs to find one that can reach that high).

But the OP was referring to speeds up to 10 GHz, those still need something special. Here's a current Guiness record, 8.4 GHz using liquid helium: http://www.youtube.com/watch?v=UKN4VMOenNM

Other materials might potentially clock higher, an IBM research prototype chip reached 500GHz: http://news.bbc.co.uk/2/hi/technology/5099584.stm

Most POWER7 cores are clocked in the 3.5-3.7 ish range and high density blade models at 2.4 to 3ghz. There is one model that can hit 4.25ghz but this is only achieved by shutting down half the cores on the chip to limit TDP.

The POWER6 hit 5Ghz but was an in-order architecture, which is far more limited than out-of-order capable architectures. One of the reasons the POWER7 is so much faster than the POWER6 at lower clock speeds.

That's because Power7s are more aggressively pipelined than Intel chips. Intel usually has around 20 FO4s of gate logic between clock latches, while IBM tries to keep it down to around 12 FO4s.

http://en.wikipedia.org/wiki/FO4

Reply on news.ycombinator.com