K8l

Toggle sidebar Toggle sidebar
R

RandMcnally

Distinguished
Nov 28, 2006
347
0
18,780
Just something I thought was sort of important. Everyone keeps bringing up the fact that AMD's solution for the native quad core will be more expensive because of smaller yields, and I have something to keep in mind. This is just a representation, and I know that the 65 nm yields are not necessarily up to the 90 nm standards yet, but they may eventually be. So here it is.

http://upload.wikimedia.org/wikipedia/commons/e/e4/Amd_cpu_die_compare.jpg

The die size isn't terribly bigger than a dual core 90nm, and with 300 mm wafers over 200 mm wafers, you can expect this to compensate somewhat for smaller yields, but the main point of me posting this is, I don't see what everyone is talking about. I don't see how these yields will be so much smaller, as everyone makes them out to be. It won't be that much of a difference will it? (I know we don't know the yields of amd, but I don't think we need to) And besides they have 2 months (at least before final production starts -that's speculating) to get yields up.
 
A

ajfink

Distinguished
Dec 3, 2006
1,150
0
19,280
Just something I thought was sort of important. Everyone keeps bringing up the fact that AMD's solution for the native quad core will be more expensive because of smaller yields, and I have something to keep in mind. This is just a representation, and I know that the 65 nm yields are not necessarily up to the 90 nm standards yet, but they may eventually be. So here it is.

http://upload.wikimedia.org/wikipedia/commons/e/e4/Amd_cpu_die_compare.jpg

The die size isn't terribly bigger than a dual core 90nm, and with 300 mm wafers over 200 mm wafers, you can expect this to compensate somewhat for smaller yields, but the main point of me posting this is, I don't see what everyone is talking about. I don't see how these yields will be so much smaller, as everyone makes them out to be. It won't be that much of a difference will it? (I know we don't know the yields of amd, but I don't think we need to) And besides they have 2 months (at least before final production starts -that's speculating) to get yields up.
Click to expand...

Most people aren't horribly worried considering the amount that AMD is investing in expanding their fab capacity.
 
D

DavidC1

Distinguished
May 18, 2006
494
67
18,860
You do not get direct cost advantages of 300mm2 over 200mm2. It happens as manufacturing improves over time and prices go down. K8L also has significantly bigger die than K8. 291mm2 is nothing to scoff at when it'll be also a mainstream CPU. Intel's Kentsfield for mainstream has similar total die size but they have cost advantages of being dual die.

Penryn/Wolfdale with dual core is supposed to have a 119mm2 die, they'll be able to make a quad core under 240mm2 if its a dual die. If they make a native quad core, Intel will likely optimize the die size to be even smaller than that.

BTW:
Penryn/Wolfdale with 6MB L2 and dual core has ~410 million transistors. Yorkfield will have approximately double that. Probably a little less but close to 800 million transistors.
 
dasickninja

dasickninja

Splendid
May 16, 2006
7,557
0
30,810
If you have 939 and only have $200 to spend, if you have AM2 and not enough money to make an entirely new system, or if you just plain prefer to buy AMD (ain't nothing wrong with that).
 
moparman390

moparman390

Distinguished
Feb 21, 2006
206
0
18,680
K8L will probably be worth buying when it comes out, but that won't be for a while yet. Right now Core 2 is king. Quad cores are really a waste of money right now for most, they are really expensive and really don't help Joe Average all that much. If you are not in to content creation (video editing and the like) a dual core will do you just fine. By the time K8L comes out (probably summer) I think we will see enough multithreaded games and other applications to make a quad core worth while, as long as you don't have to spend about $900 on the chip alone.
 
Q

qcmadness

Distinguished
Aug 12, 2006
1,051
0
19,280
The die size of a native quad with 2 Meg L3 cache is 283 mm^2, compared to a 90 nm dual core of 183 mm^2, it is significantly larger. AMD stopped making the 90 nm 2x1 meg chips because they were a monsterous 230 mm^2....

When a die approaches 200 mm^2 it is considered large.

It will be more expensive than Intel's MCM approach bar none.... it is simple statisitics of poisson distributions for defect density on a wafer.

Jack
Click to expand...

It is even larger than the first K8 sample - ClawHammer / SledgeHammer :?
 
Mandrake_

Mandrake_

Distinguished
Oct 7, 2006
355
0
18,780
Will Tigerton, Tulsa's successor, feature a large L3 cache as well? I get the feeling Tulsa has a huge L3 cache to help hide the fact that it's a Netburst based processor. :wink: Bolting on a huge L3 cache might be a crude way of tackling Paxville's awful performance, but it seems to have worked pretty well.

Meanwhile, K8L is going to frag Intel. It will take Intel three years to catch up. Intel's Q2'08 BK is set in stone. Sharikou has said so himself.

K8L frag starts in April

40% will open a major gap that will take Intel three years to catch up.

It's a cruel world for Intelers out there. Intel's BK in 2Q08 is unavoidable due to AMD's massive capacity ramp.

Once K8L is out, Clovertown will be sold at $150 and Woodcrest will be $95. QX6700 will be $150 and Conroe will be $87. Quad FX will be $999 per pair (two K8L quads). AMD will flood the hell of the market with dual core K8L chips, fragging Conroe by 40%. People think AMD will be slow to transition AM2 chips to K8L core, think again.

For anyone who plans to buy a server there, he needs to consider how to wisely spend his money. Buy a expensive Clovertown machine, and find it's about half the speed of the K8L, or buy a cheap Opteron box, and then pluggin the new K8L chip and get massive performance boost. It should really be a no brainer.

Yet, Hector Ruiz promised a killer architecture in 2008. With this architecture, AMD will frag Intel by 3x at least and permanently neutralize the Intel threat.
Click to expand...
 
Mandrake_

Mandrake_

Distinguished
Oct 7, 2006
355
0
18,780
Er... umm, oh yes. I'm very serious. Intel BK by Q2'08 for sure! :lol: 1.5bn profit a quarter in the of a price war to BK in just over a year for sure!
 
C

clairvoyant129

Distinguished
May 27, 2006
164
0
18,680
K8L frag starts April.

It's a cruel world for Intelers out there. Intel's BK in 2Q08 is unavoidable due to AMD's massive capacity ramp.

Once K8L is out, Clovertown will be sold at $150 and Woodcrest will be $95. QX6700 will be $150 and Conroe will be $87. Quad FX will be $999 per pair (two K8L quads). AMD will flood the hell of the market with dual core K8L chips, fragging Conroe by 40%. People think AMD will be slow to transition AM2 chips to K8L core, think again.

For anyone who plans to buy a server there, he needs to consider how to wisely spend his money. Buy a expensive Clovertown machine, and find it's about half the speed of the K8L, or buy a cheap Opteron box, and then pluggin the new K8L chip and get massive performance boost. It should really be a no brainer.

Yet, Hector Ruiz promised a killer architecture in 2008. With this architecture, AMD will frag Intel by 3x at least and permanently neutralize the Intel threat.
Click to expand...
:lol:
 
A

ajfink

Distinguished
Dec 3, 2006
1,150
0
19,280
K8L will probably be worth buying when it comes out, but that won't be for a while yet. Right now Core 2 is king. Quad cores are really a waste of money right now for most, they are really expensive and really don't help Joe Average all that much. If you are not in to content creation (video editing and the like) a dual core will do you just fine. By the time K8L comes out (probably summer) I think we will see enough multithreaded games and other applications to make a quad core worth while, as long as you don't have to spend about $900 on the chip alone.
Click to expand...

I think once there is competition for quad-cores we will see a general drop of some of them down into easier-to-afford areas. I don't expect software to change much in six months, though.
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom