How about if you're playing at greater than 1080 resolution?
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Little update on 28nm production
- Thread starter opethdisciple
- Start date
How about if you're playing at greater than 1080 resolution?
Associate
- Joined
- 4 Jul 2009
- Posts
- 1,136
- Location
- United Kingdom
Don't forget simulators like ARMA 2, IL-2, DCS. I regularly see 1.5gb reached on my 580 playing these.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
No not really, you have to realise what is meant by volume production and ramping on volume.
capacity on a new process will be VERY small, but still plenty of them, thing is AMD might sell 1 million high end parts, but 20mil midrange/low end parts, thats whats meant by volume.
Thats also why 98% of the time you start off with the high end rather than low end, because a new process will have limited volume, more problems and the plant won't be anywhere near max capacity in terms of pushing wafers through the line, slow and steady to start.
From all the info currently available it would seem that high end 7xxx parts will launch and be available Sept/Oct kind of time frame, theres no reason to suspect poor availability for the product yet. Mid end probably very late Q4 or early Q1 2012, and low end after that.
Nvidia, we won't be seeing anything till Jan at the earliest, hopefully they'll have learnt their lesson and working towards the process better this time and there won't be the added spins Fermi needed, but I also don't exactly have faith Nvidia won't need extra respins, if it does that would mean another 1-2 month delay on top. I think early Jan, or maybe Jan at all is probably overly optimistic judging by the difference in time between AMD/Nvidia tape outs, assuming both need 2 respins they should launch roughly as far apart as they taped out, which for a late Sept AMD launch would probably mean late Feb Nvidia launch.
Actually it could be very interesting to see if either company shove a 28nm low end part out the door before the high end parts, when the 5870 launched and Fermi was no where to be seen, Nvidia was buying up every wafer it could for 40nm ultra low end 2xx parts which used up a lot of the problematic capacity. This time around if neither has anything at all at 28nm till the high end, then AMD should have relatively free use of all available capacity till Nvidia can actually put kepler into production so AMD could arguably have massively massively massively more cards available at launch than for the 5870 launch.
Though it all depends how much capacity TSMC has planned for, but basically it looks like production for 28nm will be much more available than at the 40nm launch.
Last edited:
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
It is, its massive, TSMC are the biggest foundry on the planet, they make everything from what is now 28nm, up to 130nm minimum and possibly a LOT more.
Glofo make products from also 28nm up to 600nm, yes 600nm.
LInked from another forum where I asked a Glofo guy
This is Glofo's rough capacity right now(though Malta is only just being installed and realistically Dresden is set to have some 28nm capacity before Malta I think).
Now keep in mind what 2-3 % is, and realise TSMC are at the moment quite a lot bigger. Then realise TSMC opened 40nm with 9k wafers per month, yes, just 9k, AND they had MASSIVE problems so were realistically massively below 9k wafers per month output on from around a month after the 5870 launch, to around Jan/Feb the next year.
3% of the not to distant future of Glofo would be 11190 wafers per month, or WAY more than AMD need for the high end alone, and probably twice what TSMC had due to problems on the 40nm launch, and that was shared between AMD and Nvidia (uneven split going Nvidia's way by all accounts).
Now TSMC are bigger, so 3% of a bigger number is, bigger, and its looking likely that Nvidia won't be just using half their capacity up on crap low end stuff just to stop AMD making more high end cards, meaning more availability. Assuming TSMC don't have another massive screw up(and Nvidia don't send someone to drop a bag of dust inside their clean room), then its likely capacity is miles more.
TSMC's 40nm capacity, now, years later, with zero supply problems for Nvidia, AMD, while also making millions upon millions of APU's for AMD and chips for other people, will probably barely be above 15% of their total capacity. 3% for the high end is WAY more than is required.
http://www.tsmc.com/download/1Q11_installed_capacity.pdf
gah, finding numbers and comparing them is pretty hard, thats TSMC's current capacity, the figures in the last column are wafers over the entire year, in thousands, standardised to 8" wafer capacity(200mm vs 300mm most people talk about these days). They make 13.6 million equivilent 8" wafers in 2011, 3% of that is a pretty huge number. 13.6mil/2.25 to get 300mm wafer capacity equivilent is 6.04million a year, or 503,703 wafers per month capacity, 3% of that is 15k wafers per month, which frankly, is probably approaching 3 times the capacity they had for 40nm launch, thats still low, because by say mid 2012 that will probably be 70-100k a month, but that will be for Nvidia, AMD, their entire range, and other companies and products aswell.
Just for the sake of it, once the GloFo Malta lab is up and running at 60k wpm capacity, they'll have a roughly 273wpm capacity in total(notice some of those are in 200mm), so they are currently on course to be about half the size of TSMC. They've got planning permission for 2 more 60k wpm fabs on the Malta site which would push them up to almost 400k, they're building another fab in Dubai and planning more expansions elsewhere. Then again, TSMC have ANOTHER Gigafab(meaning more than 100k wpm(300mm) capacity going online at some stage this year. The capacity they have is mental frankly.
Last edited:
What nVidia say means SFA
yup trying to find numbers is definitely a real pain, anyway heres my take on it all.
I was looking at this
TSMC's management report
its just the last quarter that is relevant not the years production anyway i think it works out that 2 or 3% is 71680 or 107520 wafers respectively either way it certainly seems a lot but of course that is for all their customers and god only knows how good/bad yields will be to be begin with.
bottom line is their will be production in the last quarter of this year but when it will actually start we just don't know Oct/Nov/DEC who really knows. either way its not going to be a sept launch for either company.
i do agree that Nvidia will be later than AMD with the next gen.
of course theres still plenty of time for TSMC to mess everything up and delay the whole process which will quack everybody up good and proper.
I was looking at this
TSMC's management report
its just the last quarter that is relevant not the years production anyway i think it works out that 2 or 3% is 71680 or 107520 wafers respectively either way it certainly seems a lot but of course that is for all their customers and god only knows how good/bad yields will be to be begin with.
bottom line is their will be production in the last quarter of this year but when it will actually start we just don't know Oct/Nov/DEC who really knows. either way its not going to be a sept launch for either company.
i do agree that Nvidia will be later than AMD with the next gen.
of course theres still plenty of time for TSMC to mess everything up and delay the whole process which will quack everybody up good and proper.
same goes for AMD as well unfortunately.
its all marketing bull from both companies.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
you've got the numbers mixed up slightly in that, when the 5870 launched there were 9k wafers per month(or 27k for the quarter) but thats 300mm capacity, the numbers you're looking at there are adjusted to 200mm capacity or, 2.25x higher than the same numbers adjusted to 300mm capacity(as 28nm will be on 300mm capacity). The last quarter numbers are also only guessed, and don't include fab 15(which is supposed to be another gigafab but not sure its on time to be producing yet but isn't in those numbers). Either way take your 107k number, divide by 2.25, then divide by 3 to get the wafers per month and see its still dramatically higher than the 9k per month they had last time around.
They also had a contamination problem which killed that 9k wpm capacity for a good few months, its been suggested they were barely running at half capacity, from October or so which is why the 5870/5850/5970 launched fine then became some rare for a while. Also keep in mind very few customers run a new process from TSMC, Nvidia and AMD will have almost exclusive use of the capacity available and it will be for very low volume high end market. Essentially two years ago at 40nm launch(or a little after) AMD probably had as little as 2k wafers per month AND had launched and were producing the 5770(a far higher volume part) while Nvidia were using half TSMC's 40nm capacity for 40nm dx10.1 crap cards. This time around we're probably looking at at least 10k wpm, up to 15k, with no contamination problems, no low end volume eating up supply, and maybe almost completely for AMD for a good few months.
Likewise, production is starting WAY before Q4, its likely under production already with numbers ramping up fairly quickly, theres nothing to suggest late Q3 launch for AMD isn't on schedual, if AMD only needed 1/2 spins its likely TSMC is in limited scale production today, for a launch in 2 months or so.
Nvidia aren't delayed, they just taped out 3-ish months after AMD did.
AMD tape out and guessed launch dates haven't been leaked or quoted from AMD press, Nvidia marketing HAVE been leaking stuff to guys like Fud the second Charlie gave a rough date for AMD's chips. Basically no one at AMD directly has said anything, Nvidia have indicretly responded with a bunch of crap since then.
Last edited:
once again AMD is perfect and Nvidia mud.......try removing your rose tinted specs once in a while.