Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Facts and assuptions on Hammer arch.

Author: Robert Hyatt

Date: 08:47:45 01/01/03

Go up one level in this thread


On January 01, 2003 at 01:58:01, Vincent Diepeveen wrote:

>On January 01, 2003 at 01:46:20, Robert Hyatt wrote:
>
>>On December 31, 2002 at 23:22:00, Vincent Diepeveen wrote:
>>
>>>On December 31, 2002 at 23:15:53, Dan Andersson wrote:
>>>
>>>>The 8-way chipset was the product of another company. They canned it. And then
>>>>AMD was already on the way with the Hammer. So they could not be bothered to do
>>>>it themselves. And the easy out disappeared when Intel got the Alpha processor
>>>>infrastructure.
>>>>
>>>>MvH Dan Andersson
>>>
>>>thanks, intel really seems to have put those alpha guys to work. That
>>>McKinley blows away the 21264 for me *bigtime*.
>>
>>I would bet you have _never_ touched a 21264.  It was marginally faster for
>>me than mckinley.  Tim had a 600mhz machine that hit 800K with crafty.
>
>Wrong.
>
>Itanium2 blows the 21264 away for me. It's just that crafty cannot
>profit too much from the next major advantage McKinley offers offers
>over the 21264 for IPC:
>
>Look to this table:
>
>McKinley 6 instructions a cycle (bundles)
>Alpha    4 instructions a cycle



Why don't you study Mckinley a bit before popping off?  a VLIW machine does
_not_ directly compare to a 4-way super-scalar processor, which is what the
alpha is.

You pick two numbers, one is larger than the other, and you assume the larger
is better.  Two people buy a pair of pants each.  One has a 32" waist, the
other has a 40" waist.  Which person is _smarter_?

bigger is _not_ the same thing as "better".

Mckinley is definitely good.  It is definitely _not_ as good, clock for clock,
as the 21264.  Future versions might be.





>
>If your thing is a bit less dependant upon caches and more upon IPC
>then you of course also get faster on the McKinley.
>
>Add some knowledge i'd say and you'll see!
>


If you don't fit in cache, you will run slower.  _period_.

So your point totally escapes me...



>>
>>>
>>>Very impressive for a 2nd version of the Itanium.
>>>
>>>Can't wait to see how the I3 is going to be. If such processors get
>>>SMT on chip it is going to be perhaps really very well performing.
>>
>>SMT doesn't fit Itanium very well...
>>
>>
>>>
>>>Imagine they make it 4 threads a cpu. In which case i could run at a
>>>128 processor partition already like 500 processes easily!!



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.