Assuming the northern islands were/are the Evergreen refreshes, what makes you think they would be done at GF, not TSMC?
AMD: R9xx Speculation
- Thread starter Lukfi
- Start date
Assuming the northern islands were/are the Evergreen refreshes, what makes you think they would be done at GF, not TSMC?
Groo The Wanderer
Regular
Where did I say NI was 32nm?
-Charlie
-Charlie
I may be mistaken, but I remember a post on yours forum mentioning NI @32nm and your name as source of this information
Groo The Wanderer
Regular
I may be mistaken, but I remember a post on yours forum mentioning NI @32nm and your name as source of this information
There were posts on the forums, but they weren't from me. I don't think I have ever said which way it will go.
-Charlie
Why make a thread when one is already there? :smile:
My bet, caches. Bigger, better, unified. Nv has done a smart thing by doing away with ff FIFO's between pipeline stages. That is where I think a big chunk of improvement will come.
Also, distributed geometry handling. Parallel geometry setup and rasterization seems like a fairly obvious target now.
Whether or not it is worth >2% perf in existing games is somewhat uncertain
rpg.314: Wouldn't it be sufficient to design double-clocked triangle setup? I think Fermi approach is a bit overkill. It can be nice for syntetic tests, but real applications won't utilize it. More complex geometry means more shadows to generate/filter or more intensive HDAO/SSAO + more polygon edges for MSAA. Boosting triangle-rate disproportionally to other recources may not be the best idea...
This will prob be a dumb question but...
Is there no way of frigging that so that you could render all the shader / shadow / SSAO etc in the pre tessellation state and somehow infuse / composite the tessellated geometry silhouette / lighting after?
I suppose there would be some ugly render artifacts right?
During yesterday's earnings conference call, Dirk Meyer said AMD would refresh its entire GPU lineup in H2'2010. So anywere between July and December, but since he said they would refresh everything, I guess they'd have to start in Q3.
a refresh can simply be faster parts though. Doesn't have to be a brand new gpu .
Ninjaprime
Regular
Did the whole 32nm cancelled thing ever get truly cleared up? Might give more insight, 32nm parts could/should be available soon enough, in time for that 2H timetable, unless it was really canned. Although 28nm is primed for year end last I saw so, its like a 6 months window for 32nm there, which seems like they shouldn't even bother. That makes me wonder, will R9xx be 28nm or 32nm refresh, or will it be a 40nm refresh, in which case, how much bigger are they willing to go at 40nm, seeing NV's problems? 400mm? 420mm was R600 and as I recall they said they'd never go that big again.
Essentially, if 32nm is cancelled, they have to wait for 28nm, which will be 2011. I don't think they can really wait that long. Which, again, if 32nm is really cancelled as it appears, leaves them with a 40nm refresh, with a bigger die.
Essentially, if 32nm is cancelled, they have to wait for 28nm, which will be 2011. I don't think they can really wait that long. Which, again, if 32nm is really cancelled as it appears, leaves them with a 40nm refresh, with a bigger die.
Last edited by a moderator:
That's true, but across the entire lineup? Has this ever happened?
Sounds like you know a bit more than we do, care to elaborate? :sly:
Jawed
Legend
Actually, with rumours that AMD "played safe" with Evergreen GPUs, "simplifying them" because of 40nm woes - though I'm a bit sceptical about that, because the first Evergreen chips were in AMD's hands less than 2 quarters later than RV740 should originally have been on the market (March for Evergreen, November-ish for RV740, maybe earlier for Evergreen) - the refresh would be an opportunity to return Evergreen to "where it should have been".
Still, I'm dubious...
Jawed
Ninjaprime
Regular
They could probably cram a 1920 shader, 96 TMU part in less than 400mm @40nm, with some architecture improvements and aim for 225W or even go 6+8 and aim for Fermi's rumored 250+ TDP to clock it up to or above 1GHZ, that would probably be more than enough to take back whatever Fermi wins.
Similar threads
