AMD: Volcanic Islands R1100/1200 (8***/9*** series) Speculation/ Rumour Thread

Sure but RV670 was tiny! Also, not a very efficient design.
The same relative increase as between RV670 and RV770 but starting from the basically twice as large Tahiti would end up with about twice the RV770 size, a bit above 500mm². Would be definitly doable from that point of view. And GCN is still a relatively new architecture (RV670 was already the second try with VLIW5 as a shrinked and slightly modified R600, RV770 didn't just increase the amount of shaders, it also changed the internal organization quite a bit). While GCN looks well crafted in general, I'm sure AMD was able to find a few rough edges to smooth for VI. They also had some time for it. The emphasis this time can't be just pure performance of course, but the power consumption has to be kept in check, too. So it will most probably be a smaller jump than a factor of 2.5 in raw shader performance (and often almost factor 2 in games) in the same process as with the RV770.
 
V.I. (and Kaveri) delays console related?

Volcanic Islands, Kaveri and the console chips are arriving at about the same time, so ...

1. They were on similar design/engineering/fabrication timelines.

2. AMD verified they are selling the console chips, not licensing them - all the IP are belong to AMD?

3. The console chips will determine the primary game coding for the next several years - architectural correspondence will be advantageous.
4. V.I. and Kaveri were both delayed about a year from original timelines.

5. AMD has intensely focused on gaming as a primary area they can leverage their technology to their advantage.

6. AMD is deeply involved in working to optimize the major game engines for AMD architectures in general and AMD HSA APUs in particular.

Seems logical to me AMD, who would have still been working on the Kaveri and V.I. architectures when they got the contracts for the console chips, decided to modify the Kaveri and V.I. architectures based on what was being put on the console chips, for optimum correspondence with the console chips, which would optimize gaming performance and ease of porting, and which would have also put them on the same design/engineering/fabrication timeline as the console chips.
 
4. V.I. and Kaveri were both delayed about a year from original timelines.

I would have to agree with all of it except the VI delay and that the console chips were taped out and in production much sooner than the others. Personally it seems like VI might be slightly ahead of schedule, though in the grand scheme might have been pushed back a bit by other product's delays.

My two reasons for thinking VI might be ahead of schedule is:
1. No real SI refresh. Whether that was due to market conditions, competitive landscape/positioning, etc, who knows but I don't think I was the only one that expected more than 7970ghz and a late "official" 7990.

2. 20nm delays. Either they felt or were told that 20nm wasn't going to be on time, according to TSMC's roadmaps from over a year ago. So either they learned a lot from Cayman about backporting or they decided Hawaii would be on 28nm from the start.
 
Last edited by a moderator:
Because they needed more time to prepare series with this relatively big performance increase and of course, because 20 nm is too expensive now and it is not worth it, so they want to push out the transition even despite the possibility that 20 nm production works just normal.
 
If they can sell FirePros for thousands of $ and graphics cards for $500-600+, costs shouldn't be an issue. Especially with the relatively low supply in the beginning.
AMD sells the 7970 with a 365mm2 GPU for basically $300 with all the bundled games. I doubt that they don't make at least some money despice the low price. Even if 20nm were twice as expensive, if they raise prices to $600, they would still make money and gain market share.
 
Now that is fighting talk! :cool:
I suppose I should be clear and say I'm referring to per-unit efficiency: units being area and power.

Some aspects of RV770: e.g. introduction of GDDR5 and Z-rate doubling aren't going to have analogues in the next chip.

So I'm expecting "done right on mature 28nm". It seems to me that 28nm was unkind to Tahiti, specifically. Or if you prefer it had to be conservative to stand a chance of getting out the door.
 
In term of price/perf, they likely will. But Nvidia can always adjust prices. The question is: will they?

If they can sell FirePros for thousands of $ and graphics cards for $500-600+, costs shouldn't be an issue. Especially with the relatively low supply in the beginning.
AMD sells the 7970 with a 365mm2 GPU for basically $300 with all the bundled games. I doubt that they don't make at least some money despice the low price. Even if 20nm were twice as expensive, if they raise prices to $600, they would still make money and gain market share.

My wish is to release R9D700 (or R9 D970, whichever way they prefer) at 499$, pushing all other cards' prices south too.

This will indeed change the entire market and Nvidia's sales will be hurt big time :D
 
Do you honestly think AMD would erode it's brand again by selling a card that is 30-40% faster than Titan for $499??? Where can I get what you smoke? :D
 
Do you honestly think AMD would erode it's brand again by selling a card that is 30-40% faster than Titan for $499???
Why on earth would that erode their brand? Rather, it would be a tremendous boost to their brand!

Titan at its current price is underperforming and overpriced. Even something 30-40% faster, at a thousand bucks it's still going to be hella overpriced.
 
Why on earth would that erode their brand? Rather, it would be a tremendous boost to their brand!

Titan at its current price is underperforming and overpriced. Even something 30-40% faster, at a thousand bucks it's still going to be hella overpriced.

And still it is sold. As are Porsches, Mercedes, Apple etc.
You can build a brand on value and always be perceived as the second and cheap option. This is what AMD currently is. Or you can sell excellent products at higher prices. ATI sold the X800XT - X1900XT(X) for $500-600. Their market position and market share was excellent, their brand strong. And then their brand image took a deep plunge due to the price dumping and is only now slowly recovering.
 
And still it is sold. As are Porsches, Mercedes, Apple etc.
You can build a brand on value and always be perceived as the second and cheap option. This is what AMD currently is. Or you can sell excellent products at higher prices. ATI sold the X800XT - X1900XT(X) for $500-600. Their market position and market share was excellent, their brand strong. And then their brand image took a deep plunge due to the price dumping and is only now slowly recovering.

I would argue that the main reason their brand image took a dive was R600. The price dumping was just a consequence of the latter.
 
Back
Top