3 ms·

So what confuses me is that we've seen tech demos (and overclockers) push CPU speed to 4 to 10ghz. Are those gains so artificially made that they just can't be

by tibbon 15y ago

So what confuses me is that we've seen tech demos (and overclockers) push CPU speed to 4 to 10ghz. Are those gains so artificially made that they just can't be replicated at scale in the public market? We can get them to go faster, just seemingly not for the public. It was weird to buy my new Macbook Pro and after 3+ years, it was .2ghz slower than my old one. Now of course it has more cores, better instructions, etc... but it was still a weird thing. Apple's done a great job of explaining the benefits of new ones to the consumers - they don't. I'm not being sarcastic. At the end of the 90's (and still largely today for most PC companies) its all about specs. Apple sold what you can do with each computer instead in a good, better, best format. I bet a large amount of computers sold at the Apple store never have the sales associate mention the clock speed.

beschizza 15y ago
Perhaps because the gains are typically made with the help of cooling systems too unwieldy and unreliable for mass consumption.
j-g-faustus 15y ago
Here's an article describing what it takes to reach those speeds: http://www.tomshardware.com/reviews/5-ghz-project,731.html 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 :)
- apaprocki 15y ago
  The top IBM POWER7 CPU is clocked at 5ghz and isn't cooled with liquid nitrogen.
  - j-g-faustus 15y ago
    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 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 http://news.bbc.co.uk/2/hi/technology/5099584.stm
  - mbell 15y ago
    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.
  - Symmetry 15y ago
    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 http://en.wikipedia.org/wiki/FO4

by tibbon 15y ago

beschizza 15y ago

Perhaps because the gains are typically made with the help of cooling systems too unwieldy and unreliable for mass consumption.

j-g-faustus 15y ago

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

apaprocki 15y ago
The top IBM POWER7 CPU is clocked at 5ghz and isn't cooled with liquid nitrogen.
- j-g-faustus 15y ago
  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 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 http://news.bbc.co.uk/2/hi/technology/5099584.stm
- mbell 15y ago
  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.
- Symmetry 15y ago
  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 http://en.wikipedia.org/wiki/FO4