The shortage was only for Cypress. Juniper, the cheaper version was aplenty even in the beginning.I remember 40nm history, and how much AMD and NV were burned by it. I would expect them to learn from it, and NV is showing it by postponing Kepler launch as we speak, but you think AMD havent learned?
Regarding asking a high price..
Here both 6970 and 570 are selling from around 2200dkk, while the barely 15% faster 580 is starting at around 3100dkk (and the high end 3gb versions more like 4k) - simply because many want the fastest (single gpu) card.
The 28mm "7970" should have no problem soundly beating gtx580 (which would then have to drop, when loosing the king-of-the-hill status), so I don't see why they can't sell it in the 3700+ range (6990 is currently starting at 4400) - keep profit high and sales low + get the "bragging rights". I mean, why not do it if everything but production capacity is ready?
Regarding GF 28mm that's definitely a no-no when running nothing but test structures now - 32nm SOI should be possible, but they are most probably production constrained on Llano+BD already.
Here both 6970 and 570 are selling from around 2200dkk, while the barely 15% faster 580 is starting at around 3100dkk (and the high end 3gb versions more like 4k) - simply because many want the fastest (single gpu) card.
The 28mm "7970" should have no problem soundly beating gtx580 (which would then have to drop, when loosing the king-of-the-hill status), so I don't see why they can't sell it in the 3700+ range (6990 is currently starting at 4400) - keep profit high and sales low + get the "bragging rights". I mean, why not do it if everything but production capacity is ready?
Regarding GF 28mm that's definitely a no-no when running nothing but test structures now - 32nm SOI should be possible, but they are most probably production constrained on Llano+BD already.
AMD had nothing to learn, other than they had to double up vias for certain ASICs... and they can't trust TSMC to follow up on their volume promises.I remember 40nm history, and how much AMD and NV were burned by it. I would expect them to learn from it, and NV is showing it by postponing Kepler launch as we speak, but you think AMD havent learned?
The numbers they quoted don't add up....
I was talking about Q3 numbers, which they reported as 4% and that is exactly what it is... Q4 they were trying to release a whole line, with Nvidia trying to take up as much as they could, Q4 RV740 was dead but they had Cypress and Juniper in mass production and were trying to ramp up Redwood and Cedar at the sametime, let alone their entire mobile lineup...
AMD shipped 1mil DX11 GPUs before the end of '09... with less than 9% of TSMC's 40nm volume.
Don't know much about firearms? WTF is a double-pump gun?If thats true, it would mean AMD is about to shoot in its leg, with double-pump gun, no less
This is ridiculous... single cards matter in '09, dual cards dont. But in 2011, dual cards matter not single. Make up your mind.
AMD can push the advantage and what they have learned over the years, from Intel, is that they need to.
Agreed, here in Euroland-Ger HD 6970 sells for under 300 EUR, so asking 50% more for a real high-end killer is definitely NOT out of the question.
Wrong thread, but for me the real question is just how quickly AMD will be able to bring out it's <$200 line up, now that a 4 month window between them and nv has opened up. If nv ends up needing one more spin than amd - which is quite plausible, it could be a repeat of the dx11 launch all over again.
We are talking about Kepler and high-end SI on the new node, and their approx launch
No, they (should) have learned not to jump on not mature new node with high-end GPUs, especially when AMD dont have to.
You yourself quoted that site, what you dont like now?
Its true most of the production in '09 happened before Q4, same applies to '11 - if you want to disregard 9% vs 2-3%, how about Q3 and 4% vs 0-1% ratio? If anything, its worse than it was in 2009.
"TSMC initiates production using 28nm process technology in Q4, then the ramp from 0% to 3% of revenue in the first quarter will be pretty fast. Still, it may not be fast enough to fulfill the demand towards the chips made using leading-edge process technology"
http://www.xbitlabs.com/news/other/...ns_to_Ship_28nm_Chips_Commercially_in_Q4.html
As I said, AMD would make a mistake if they try to launch high-end SI this year, but its not uncommon for a company to screw up, just look at NV in the last couple of years
DarthShader
Regular
Absolutely, given the performance projections that top bin GCN card would around the current dual cards. Low volume part, high margins, the main line of cards isn't canibalised, with AMD having the fastest single card. Lots of potential win if launched this year. The rest can come in 2012.
@Harison: Your links are from February. And now we have July.
It depend on contract AFAIK, for example AMD is buying wafers (most likely), while NV sometimes pays per good chips instead of just wafers (also rumoured, no one here knows contract details, only Dave maybe? :smile
.Regardless, not mature node usually have good yields for low-end, decent for mid range, but pretty bad for big chips (58xx yields werent good, and Fermi's simply terrible).
My question is - why AMD would allocate wafers for (probably) poor-yield high-end GPUs, especially when supply is very low, when they dont have to force fast launch anyway? I would like to see them as soon as possible as much as anyone here, GCN preferably
I just dont see it happening in 2011, sorry.You are absolutely welcome to provide July links :smile:
It depend on contract AFAIK, for example AMD is buying wafers (most likely), while NV sometimes pays per good chips instead of just wafers (also rumoured, no one here knows contract details, only Dave maybe? :smile
Regardless, not mature node usually have good yields for low-end, decent for mid range, but pretty bad for big chips (58xx yields werent good, and Fermi's simply terrible).
My question is - why AMD would allocate wafers for (probably) poor-yield high-end GPUs, especially when supply is very low, when they dont have to force fast launch anyway? I would like to see them as soon as possible as much as anyone here, GCN preferably
Actually it was the other way around last time I heard about it - AMD/ATI paying for working chips only, while nV buys whole wafers
I heard about AMD and GF regarding paying for working chips, but what concerns TSMC, I only remember NV in this context. To quote Char... "the person who shouldnt be named"
"The rumor is that TSMC, as a way to slide Nvidia money through backchannels to prevent it hopping to GlobalFoundries, is charging Nvidia per good chip while yields are below 60%."Strictly speaking NVIDIA started with 'per good chip' when they moved to IBM for NV40. But at this point they've been 'per wafer' for many many years except maybe for some early process ramps (e.g. 40nm as implied by Charlie - I really wouldn't describe that as 'sliding money through backchannels' though). And AFAIK AMD is actually the only company of the two which used 'per good chip' pricing even on relatively mature processes in the 110-55nm generations (which is what Kaotik must be thinking of here).
DarthShader
Regular
Here's a round up of what is flying around:
http://vr-zone.com/articles/rumour-...-amd-southern-islands-in-september/12799.html
Now add the fact that 6XXX codenames have been revealed in 10.8 driver and we get 7XXX codenames in 11.7 preview...
Can we get back to Kepler now?
1H 2012 28nm GPU launch from AMD doesn't jive with the March SI tape-out, at least not if it goes smoothly...
Latest as in the newest-plan-to-launch, or latest as in definitely-by-then?
