ATI's next gen process details [S|A]

Toggle sidebar Toggle sidebar
Wreckage

Wreckage

Banned
Jul 1, 2005
5,529
0
0
It sounds like Charlie is trying to sugar coat that there is something wrong with Northern Islands.
 
Z

zagood

Diamond Member
Mar 28, 2005
4,102
0
71
It's going to be too cool, too small, too fast, and completely manufacturable.
 
Schmide

Schmide

Diamond Member
Mar 7, 2002
5,745
1,036
126
Wreckage said:
It sounds like Charlie is trying to sugar coat that there is something wrong with Northern Islands.
Click to expand...

Did you read the article? It was all about TSMC delaying 28nm and AMD coming out with a stop gap measure to get a refresh.
 
crazylegs

crazylegs

Senior member
Sep 30, 2005
779
0
71
Wreckage said:
It sounds like Charlie is trying to sugar coat that there is something wrong with Northern Islands.
Click to expand...

Thats amazing... We clearly didn't read the same article.

Because that sounded to me like TSMC's roadmap for a new node is screwed... something that a drunk hobo could probably tell you giving their recent 40nm issues *cough cough Fermi had a real smooth production and release to market right?*

Having little to know confidence in TSMC's 28nm production hitting deadlines, AMD/ATI are coming up with a sensible plan (unlike Mr. I Have No Plan B...) and adapting their designs to fit the available production method.

IF this is true... then i'm guessing NI will be released @ GF and AMD can wave bye bye to TSMC...
 
Maximilian

Maximilian

Lifer
Feb 8, 2004
12,604
15
81
Sounds like a decent plan, rather than make everyone wait for the new architecture it gives us a hybrid in the meantime whilest TSMC sorts out its 28nm process for the completely new architecture.
 
F

Fox5

Diamond Member
Jan 31, 2005
5,957
7
81
I could see northern islands being late.
But a new, half way part between northern islands and evergreen? Basically a bigger die evergreen? I seriously doubt that ati's engineering team is that far ahead of nvidia's that they can really run that many simultaneous designs without delays.

Nvidia is running with fermi, minor updates of g9x, and tegra.
ATI is doing evergreen, northern islands, southern islands (if this article is correct), providing graphics ip for the snapdragon family (I doubt it would go in house at qualcomm), and likely at least one new console gpu.

If ati can balance all that, there's no reason nvidia couldn't make a suped up 40nm g9x to compete in the mid range.

I COULD see ATI possibly doing one more member of the evergreen series, a bigger chip to place at the high end.
 
happy medium

happy medium

Lifer
Jun 8, 2003
14,387
480
126
Fox5 said:
I could see northern islands being late.
But a new, half way part between northern islands and evergreen? Basically a bigger die evergreen? I seriously doubt that ati's engineering team is that far ahead of nvidia's that they can really run that many simultaneous designs without delays.

Nvidia is running with fermi, minor updates of g9x, and tegra.
ATI is doing evergreen, northern islands, southern islands (if this article is correct), providing graphics ip for the snapdragon family (I doubt it would go in house at qualcomm), and likely at least one new console gpu.

If ati can balance all that, there's no reason nvidia couldn't make a suped up 40nm g9x to compete in the mid range.

I COULD see ATI possibly doing one more member of the evergreen series, a bigger chip to place at the high end.
Click to expand...

What is this?

There is a smaller femi part tapping out as we speak and will be avalable in June last I heard.
 
cbn

cbn

Lifer
Mar 27, 2009
12,968
221
106
Southern Islands sounds a lot more interesting than a Evergreen refresh.
 
D

Dark4ng3l

Diamond Member
Sep 17, 2000
5,061
1
0
Well at least ATI has some room to increase die size at 40nm unlike nvidia who really can't add a single mm2 unless the process becomes MUCH better fast.
 
SunnyD

SunnyD

Belgian Waffler
Jan 2, 2001
32,675
146
106
www.neftastic.com
OCGuy said:
Same reason why there are only 480 for the release version of the GTX480.

:(


Can we all agree that TSMC is blowing it for us?
Click to expand...

In all fairness Nvidia really didn't help any by creating a die size of 500mm^2+. The bigger the die, the higher the inherent defect rate, the lower the yield. It's that simple.
 
V

Voo

Golden Member
Feb 27, 2009
1,684
0
76
Fox5 said:
ATI is doing evergreen, northern islands, southern islands (if this article is correct), providing graphics ip for the snapdragon family (I doubt it would go in house at qualcomm), and likely at least one new console gpu.
Click to expand...
Haven't they released all evergreen variants already? And Southern islands and the console GPU are more or less existing things they just have to adapt a bit - probably a bit more work for the console, but all in all it's not as if they designed a new GPU there.

If they wanted to release NI this year on 32nm then they are probably ahead in schedule and have some time to work on SI..
 
S

SlowSpyder

Lifer
Jan 12, 2005
17,305
1,002
126
Wreckage said:
It sounds like Charlie is trying to sugar coat that there is something wrong with Northern Islands.
Click to expand...

I didn't see anything that would hint at that. The 5870 has 1600SP's, 32 ROP's 80TMU's compared to the previous generations high end (for AMD) parts having half that number. I'm sure Northern Islands will have a similar increase... they're not likely going to double up everything and keep the same process. They were able to do a 2.5x increase over the 38xx cards moving to the 48xx cards due to the fact that the 38xx were already such small GPU's. Cypress is already a decent sized GPU at 330mm2.


OCGuy said:
Same reason why there are only 480 for the release version of the GTX480.

:(


Can we all agree that TSMC is blowing it for us?
Click to expand...


While having a single foundry for 40nm and other more advanced manufacturing processes certainly isn't a good thing, Nvidia can't blame anybody but themselves for Fermi's shortcomings.

AMD released the 4770 on 40nm early on as a feeler for 40nm. Nvidia released the GT 240 when Fermi was already supposed to be out. I have a feeling they've learned a little something about doing their homework on a new process before jumping to a 500mm2+ GPU.

Radeon 4770: April 09
GeForce GT240: November 09 (I believe this was Nvidia's first 40nm part, though I do remember them possibly having some 40nm mobile parts in the works around that time too)
 
Last edited:
B

Blue Shift

Senior member
Feb 13, 2010
272
0
76
On the "Howitzer vs Pen Knife" comment... I find myself agreeing with Charlie here, believe it or not.

The only real way that nVidia can compete on the 40nm process past releasing a 512-sp fermi, a dual-gf102 part, or optimizing their part and decreasing leakage for minor gains would be to abandon Fermi's touted general-computing potential, and build something that's actually designed with gaming in mind.

ATI, however, can just build a larger chip based on Evergreen or NI... Problem solved.
 
RussianSensation

RussianSensation

Elite Member
Sep 5, 2003
19,458
765
126
I think 4-6 months ago we started reading about TSMC's manufacturing problems with transitioning to 32nm and 28nm. Months ago it was already suspected that as a result of the node issues, 6xxx series was either going to have to come out on 40nm in 2010, or ATI would need to wait until Q1 2011 at the earliest for 28nm design.

Last night I played Dirt2 @ 1980x1080 2AA/16AF on my 4890. No problems at all, although 4AA wasn't as smooth as I wanted to. Looks like 4890 will be waiting a while before her retirement. :\

In any case, 40nm for a more advanced 5870 can't be good news. However, if they keep the die < 450mm^2, the more mature 40nm transistors in 6 months+ may need slighly lower voltage to run similar clock speeds (i.e., see Core i7 920 130W --> Core i7 860 95@ on same 45nm process).
 
Last edited:
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom