Barcelona Architecture: AMD on the Counterattack

Reagan

Distinguished
Jan 3, 2007
14
0
18,510
Excellent article. [Thanks for posting the link].

Really the only article so far that says what the diff's tween C2D and Barcelona arch's are.

Look for a benchmark program that tries to execute pathological, worst-case behaviour for load-reordering. It would be interesting to see how much it effects C2D 8)

If I were an engineer at AMD I would be begging my managers to license eDRAM technology from their friends at IBM. This move could hurt Intel. Even if AMD is a year behind at 45nm, if they get eDRAM technology into their caches, Intel would be in some trouble. Mark my words. Instead of 2MB L3 cache, start thinking about a 32MB L3 cache. [AMD's TLB's could double again, branch predictor buffer's could double, etc., and still have room left over ... and less silicon means lower price means more sales].

Barcelona has a lot of little improvements that will make this an Enterprise Server shoe-in.

Look for Intel to push their 45nm advantage. [And keep your eyes on Nikon's EUV].

--M
 

4745454b

Titan
Moderator
I only understand maybe 2% of that, but all I can say is "boom!" I worry now not if K8L/K10 will be faster, but if AMD will be able to make enough of these things. Heres to hoping for team green.
 

yakyb

Distinguished
Jun 14, 2006
531
0
18,980
very nice article makes me want to wait for Agena seems to be along the lines of
'Yeah its gonna be better, of course' w/o any numbers tho its hard to back up a statement like that.
the writer seems to know what he is talking about, anyone know who the source is?
with the barcelona core (for opteron) scheduled to emerge half way thru the year when will this translate into the desktop variety, im itching to build a quad setup, and was going to build in the next month or 2 (r600) but ofc i would rather get this if it is better. i see it as buying a fx50 the month before c2d was released (i.e plain stupid!)

bring on some benchmarks please!!
(oh and how will this thing OC? no point in a small improovement over qx6700 if i can OC it to better performance)

what about the chipsets im sure the radeon department is working hard to produce an awesome feature set for this chip that will optimise it and allow for maximum benefits of combining the brands 3 areas (CPU MOBO GPU)

again bring on some benchmarks please!!
 

knightrous

Distinguished
Jan 27, 2006
271
0
18,780
Sounds very interesting. There seems to be some more promise in this rebuild of the K8 uArch...

This kinda reminds me of motors in cars :) One company has the money to build a new motor (uARch) from scratch (Intel), while the other can only afford to improve on what they have (AMD). The bigger company increases displacement, uses different manufacturing technologies like Alloy block and head (65nm) while the other can only keep going with the cast steel block an Alloy head. The smaller company, not able to build a new motor from scratch, looks at what they have to work with, and looks at what they can improve to squeeze that bit more performance out of it to keep in the game. So they optimise there turbos to be more effcient, snappier, make install an intercooler to allow more air to be forced into the motor, they put in variable valve timing to get more torque down low yet still keep up high power with no loss.

The big company uses brute force, the small company uses leverage of optermisation.... Now imagine if you gave the small guys the funding that the big guys have... 8)
 

SSS_DDK

Distinguished
Jan 28, 2007
136
0
18,680
Great article!! After the "AMD sucks in ..." and "AMD kicks ass with..." threads, this is the first interesting and comprehensive (for non CPU architects) thread...
 

Gustafarian

Distinguished
Jan 26, 2007
48
0
18,530
Now imagine if you gave the small guys the funding that the big guys have...

Then they would do the same thing! :lol:

Looks interesting. I hope this returns the high end crown to AMTI so we can have a bit of competition.
 

yakyb

Distinguished
Jun 14, 2006
531
0
18,980
One company has the money to build a new motor (uARch) from scratch (Intel), while the other can only afford to improve on what they have (AMD).

C2D isnt entirely New it was an evolution of P3 through PM was it not?
 

DoubleE460

Distinguished
Oct 4, 2006
36
0
18,530
Put it this way... without the PM there would be no C2D.
I had a PM 740 and it was just awsome... I knew already then what two of those would be capable of.

As example I had 26s in SuperPi with a PM740 @ 2.7 amd a Asus P4C800 mobo.
 

Dade_0182

Distinguished
Apr 3, 2006
1,102
0
19,280
Put it this way... without the PM there would be no C2D.
I had a PM 740 and it was just awsome... I knew already then what two of those would be capable of.

As example I had 26s in SuperPi with a PM740 @ 2.7 amd a Asus P4C800 mobo.
Nice 8)
 

ajfink

Distinguished
Dec 3, 2006
1,150
0
19,280
It looks like the server variants aren't going to be the really showy processors. A 2.9Ghz Kuma? Imagine the decimation that would lay out against a current 6000+.
 

jamiepotter

Distinguished
Oct 8, 2006
375
0
18,780
Can I ask a real noob question? The article mentions an on-die northbridge. Does this mean that there won't be a northbridge on the mobo, and that it will all be done in the CPU?

Apologies if that's a really dumb question, but I confess I'm ignorant!
 
No AMD CPUs have had full northbridges on the motherboard since the last Socket A Semprons stopped being sold a few years ago. All Athlon 64, Opteron, and the socket 754 and AM2 Semprons have most of the traditional northbridge functions on-die. Thus an AMD K8 chipset can often be just a single chip, which is the southbridge.
 

TRENDING THREADS