Does AMD K9 exist?

Toggle sidebar Toggle sidebar
M

mailrapid

Distinguished
Oct 12, 2007
13
0
18,510
I once browsed some information that due to the bad pronunciation of K9, AMD only has K8 and K10.
Is that right?

BTW, what's the structure of K10?
 
Some believe that K9 turned out to be the dualcore K8 chips, while others believe there was a seperate K9 design which was scrapped when they shifted to dualcore.

However, most, including myself, believe that there was never a K9 as it sounds like canine (i.e. - "dog") which AMD didn't want to be known for - for obvious reasons.

In any case, a K9 was never released.

As far as the structure of the K10 (Barcelona) - read the sticky:

http://www.tomshardware.com/forum/223161-28-collection-data
 
there was an AnandTech article that AMD aimed K9 to be a heavy Integer calculation processor with up to 4 cores but decided to design a processor more suitable to their "customers" mainly aimed on the server side of things. K9 was scrapped.

So now we end up with K10/Barcelona which is a more balanced processor very good on floating points and not bad on the others, clearly i think AMD's target market and their customers are happy with K10 so far.
 


Was there not a time when there was a comparison of the K8L which is now known as the K10 and the K8 showing a very absolute similarity between the dies? I look at the die shots now and see 4 K8's and not a completely revamped u Arc.

Just curious when that discussion was discontinued and everyone accepted the K10 as a new u Arc.


Word, Playa.
 
Spud,

Good to see both you and Exit around.

Dunno when 'acceptance' of K10 as a new core started.....its not...its a highly modified K8 as proven by Anands analysis of K10 Uarch (back around march). I think acceptance started when people started drinking Henri Richards green kool aid.
 


So the conditions here are worse than projected after the exodus of certain members. How have you kept your sanity through all this "opinion's" are facts mentality?

Word, Playa.
 


lol, Ive kept my sanity because Ive been in and out of town for work and very busy with few opportunities to write or respond the past few months. Im only on tonight because Im updating my FW/AV OS software on a crappy/slow hotel connection, and Im not scheduled for anything until 1700 tommorrow, only because I just got to this location today. I could have writen 3 or 4 books worth of stuff pointing out flawed logic and hypocritical counter points over the past few months, but Ive barely had time to edit the really bad stuff out of the posts let alone participate
 



Nice to see you back around Spud. :sol:

K10 is still just thought of as a revision of K8, not a whole new uarch. Just a K8 with:
1) Upgraded FP to 128 bits and added the ability to load two 128-bit values per cycle.
2) Better branch prediction and out-of order execution.
3) HT 3 support.
4) Better power management - including individual core management.
5) IMC tweaks as well as shared L3.

So I would call K10 just a "tweaked" K8 - not a whole new uarch. Judging by the clockspeeds, they may be hitting the same wall as the K8 - We'll need to see what the can do with the desktop (Phenom) K10s.


@Turpit -> Do you think this will be enough to get some flames started? 😀
Good to see you too.
 

Depends on view, how much of an improvement is enough to be called new arch? Because the way you see it, C2D is same old highly modified Pentium M, which in itself is modified P III.
 


You've done better than I. I think I misplaced my sanity a while back. I won't call it lost yet, because I still have hopes of recovering it. This is unlike my marbles, which were lost a long time ago. Of course, when it comes to AMD, I think they lost their marbles, sanity, and enything else that even comes close to the thought of rational thinking a long time ago. And I say this while using an AMD powered machine.

Long live the 939!!!


 


Die shots don't look similar.


Word, Playa.
 


Yes they were, because someone at Intel finally got the clue and realized that the PIII Uarch was superior to that of P4s, though on paper P4 looked fantastic, in practice it failed to live up to predictions.

You can call Core 2 a modified PIII if you wish, personally, I thats how I think of it. I only care that people are accurate in what they call things, not fanatical or fanciful. In the case of K10, its a modified K8.
 
Actually, all current processor architecture are all based on X86, which dated back to Intel's 8086. I suppose you can call every x86 design a "modified" 8086 😛
 
That dog had crap floating point ... unlike the darleks that could levitate.

If the Darlecks are Intel then AMD are probably Cybermen.

This bodes well for the Dr ... who probably works for Sun.

LOL
 
That works really well, AMD have a weakness similar to that of the Cybermen. The Cybermen's weakness is gold, AMD's is money. The Darlecks are out to destroy everything, just like Intel with their price cuts.
 


http://en.wikipedia.org/wiki/AMD_K9 wikipedia suggests it is : K9 appears originally to have been an ambitious 8 issue per clock cycle core redesign of the K7 or the K8 processor core[1]. At one point, K9 was the Greyhound project at AMD, and was worked on by the K7 design team beginning in early 2001, with tape-out revision A0 scheduled for 2003. The L1 instruction cache was said to hold decoded instructions, essentially the same as Intel's trace cache.

The existence of a massively parallel CPU design concept for heavily multi threaded applications has also been revealed, as a planned successor to K8. This was reportedly canceled in the conceptualization phase, after about 6 months work[2].

At one time K9 was the internal codename for the dual-core AMD64 processors as the brand Athlon 64 X2 [3][4], however AMD has distanced itself from the old K series naming convention, and now seeks to talk about a portfolio of products, tailored to different markets. [5]
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom