Article: Niagara II: The Hydra Returns
By: Tom W (twerges.delete@this.hotmail.com), September 8, 2006 9:42 am
Room: Moderated Discussions
David Kanter (dkanter@realworldtech.com) on 9/4/06 wrote:
---------------------------
>
>Niagara II is a fully integrated server on a chip that is a natural extension of
>Niagara I. It is projected to deliver 2x the performance of its predecessor, and
>includes quite a few improvements aside from increasing the throughput. Our article can be found at:
>
>http://www.realworldtech.com/page.cfm?ArticleID=RWT090406012516
>
Excellent article.
BTW, I predicted 8 months ago that Niagara II would have two ALUs per core (and one FPU) and the ability to execute instructions from 2 threads each cycle.
http://realworldtech.com/forums/index.cfm?action=detail&id=60907&threadid=60800&roomid=11
http://realworldtech.com/forums/index.cfm?action=detail&id=60941&threadid=60800&roomid=11
"I'd have to guess that the Niagara2 cores will have some kind of primitive SMT... Niagara2 [could] execute 2 instructions from 2 threads PER CORE... A simple SMT core could always assign instructions from different threads to the two ALUs."
[ASIDE]
In response to my speculation, Paul DeMone politely disagreed, however Jon Anderson hinted that I should stop posting: "No wild theories to explain unsubstantiated data is necessary from anyone... I suggest waiting for more real data..."
But now it turns out that what I speculated about Niagara2 is in fact the case. It's funny that a SUN employee would dismiss correct speculation as "wild" and "unsubstantiated"...
[END ASIDE]
...Anyway, it seems likely that Sun's claims of 2x performance over the first Niagara are reasonable. It appears that Niagara2 will be able to execute a theoretical maximum of 16 instructions (from 16 threads) per clock @ 1.4GHz. Of course, I realize that MIPS (millions of instr per sec, not the SGI CPU) is a silly index of performance, however in this case there's good reason to believe that Niagara2 will achieve results reasonably close to the theoretical maximum much of the time.
...I've been a Niagara booster for awhile, and the more I think about it, the more I become convinced that the "throughput" approach is the best strategy that Sun could have devised.
---------------------------
>
>Niagara II is a fully integrated server on a chip that is a natural extension of
>Niagara I. It is projected to deliver 2x the performance of its predecessor, and
>includes quite a few improvements aside from increasing the throughput. Our article can be found at:
>
>http://www.realworldtech.com/page.cfm?ArticleID=RWT090406012516
>
Excellent article.
BTW, I predicted 8 months ago that Niagara II would have two ALUs per core (and one FPU) and the ability to execute instructions from 2 threads each cycle.
http://realworldtech.com/forums/index.cfm?action=detail&id=60907&threadid=60800&roomid=11
http://realworldtech.com/forums/index.cfm?action=detail&id=60941&threadid=60800&roomid=11
"I'd have to guess that the Niagara2 cores will have some kind of primitive SMT... Niagara2 [could] execute 2 instructions from 2 threads PER CORE... A simple SMT core could always assign instructions from different threads to the two ALUs."
[ASIDE]
In response to my speculation, Paul DeMone politely disagreed, however Jon Anderson hinted that I should stop posting: "No wild theories to explain unsubstantiated data is necessary from anyone... I suggest waiting for more real data..."
But now it turns out that what I speculated about Niagara2 is in fact the case. It's funny that a SUN employee would dismiss correct speculation as "wild" and "unsubstantiated"...
[END ASIDE]
...Anyway, it seems likely that Sun's claims of 2x performance over the first Niagara are reasonable. It appears that Niagara2 will be able to execute a theoretical maximum of 16 instructions (from 16 threads) per clock @ 1.4GHz. Of course, I realize that MIPS (millions of instr per sec, not the SGI CPU) is a silly index of performance, however in this case there's good reason to believe that Niagara2 will achieve results reasonably close to the theoretical maximum much of the time.
...I've been a Niagara booster for awhile, and the more I think about it, the more I become convinced that the "throughput" approach is the best strategy that Sun could have devised.
Topic | Posted By | Date |
---|---|---|
Hot Chips 18: Niagara II Article Online | David Kanter | 2006/09/04 09:01 PM |
Hot Chips 18: Core 2 tidbits | anonymous | 2006/09/04 09:15 PM |
Hot Chips 18: Core 2 tidbits | Andi Kleen | 2006/09/05 11:29 AM |
Hot Chips 18: Core 2 tidbits | anonymous | 2006/09/05 12:33 PM |
Hot Chips 18: Core 2 tidbits | Andi Kleen | 2006/09/05 12:48 PM |
Hot Chips 18: Core 2 tidbits | anonymous | 2006/09/05 03:40 PM |
Hot Chips 18: Core 2 tidbits | anonymous | 2006/09/06 06:20 AM |
Hot Chips 18: Niagara II Article Online | Garius Bias | 2006/09/05 01:35 AM |
Hot Chips 18: Niagara II Article Online | David Kanter | 2006/09/05 12:08 PM |
Hot Chips 18: Niagara II Article Online | Chuck | 2006/09/06 03:59 AM |
Hot Chips 18: Niagara II Article Online | David Kanter | 2006/09/06 08:50 AM |
Hot Chips 18: Niagara II Article Online | Tom W | 2006/09/08 09:42 AM |
Hot Chips 18: Niagara II Article Online | David Kanter | 2006/09/08 11:18 AM |