I think you misread my post.
"His point was that a lot of people on our forum have already written-off Kepler because it's "late", expect AMD to refresh HD7970 shortly without problems (like Q2), while expecting 28nm Kepler to be delayed FAR into 2012, with performance increase. My view is not in-line with that sentiment."
If the above road map is correct it looks like we have some time to wait for some shiny high end NVidia parts. For myself I won't be upgrading, till NVidia shows their cards or a 7970 refresh comes out. I will be recommending the 7970 as an upgrade for those running 5870s who want more performance now (and have the cash).
Ya, it looks like unless there is more concrete information, Kepler's HD7970 competitor might still be ways off, perhaps 2-3 quarters away. Too hard to say since NV is completely tight lipped and sources are all over the place, some saying Q1-Q2, others Q3-Q4.
the gtx 480 have ~10% less bandwidth than a gtx 580, that is ~15% faster.the same way a 7970 have ~30% more bandwidth and is ~25% faster.
to double the performance, they will have to double the bandwidth,
it's impossible even using 512bit-rate
That's a poor example since it assumes the same is true for other videocards.
HD6970 is nearly 75-100% faster than HD4870/HD4890 in demanding games and it barely has 35-40% more bandwidth. We can also see the same situation for NV. GTX580 only as 35% more memory bandwidth than a GTX280 (192 vs. 142), but is 70-100% faster, depending on the game.
You can't just assume that to double the performance, you have to double the memory bandwidth.
It depends on the specific architecture and in fact specific bottlenecks on an individual card basis. For example, HD4870 series had more than enough bandwidth (in fact too much). Your argument also assumes that GTX580 uses all of its 192 GB/sec bandwidth efficiently. It could also be the case that GTX580 has way too much bandwidth given its specs, much like the HD4890 and HD5830 were.
Bragging rights and enthusiast mindshare. Also, people get tired of waiting, and buyers who generally lean toward Nvidia but don't necessarily dislike AMD will eventually buy the shiny new card.
Has a lot less value than you think. If you look at the consumer market segment for graphics cards, you'll realize this quickly.
The market share for graphics cards above $199 is only 14%. That means the real battle takes place at $199 and below. For now HD7970 is only important to enthusiasts. Obviously, for our forum that's a very large fraction of buyers in % terms, but for the entire market, the market share for $550 graphics card is likely 2-3% of the overall graphics segment.
Does AMD have any desktop HD7000 cards under $200 launched yet? The $149-199 market segment, occupied by GTX560 and GTX560 Ti is an astounding
51% of desktop discrete graphics market share. Last time I checked GTX560 and GTX560Ti are very competitive vs. HD6870 and HD6950.
There is no need to launch ASAP. Graphics cards are sold all over the world. Just because you announced the launch of a product, you still have to consider the following factor:
1) Ability to deliver large volumes of that product to the market quickly (are there going to be supply constraints in the channel, manufacturing problems, etc.);
2) Ability to distibute inventory across the world in a timely manner (you might have plenty of volume in US and Canada but little to no volume in Europe and Asia for months);
3) The global economy and market timing of launching in Q1 (the least important quarter of the year for technology). With Europe struggling between little to no growth and a full fledged recession, having a lead in the $450+ market segment isn't really a game changer; and,
4) Consumers are going to be all spent after the holidays. So expect first 1-3 months of the year to be pretty slow. That's why it's FAR more critical to launch leading graphics cards in the $0-199 price segment, not in the $300+ segment.
We have already seen this exact scenario with HD5xxx/HD6xxx vs. Fermi series as the most recent example. Despite AMD launching
6 months ahead (and on top of that, also launching prior to the holiday season -- That's a substantial advantage!!) and it still did almost nothing to dent NV's market share on the desktop, which was quickly regained to almost 60%.
The launch of HD7970 has done little to change anything other than for people who buy $500+ graphics cards, especially since you can't even buy it yet and we have no idea how large the volumes will be. For enthusiasts, this card is great. However, NV has competitive cards in every price level up to $400. When AMD launches a full scale top-to-bottom HD7000 line-up from $0-$550 price level with better cards than NV and in ample supply, only then would NV need to start sweating.
You realize NV has record mobile design wins with Kepler too compared to Fermi? Even without any silicon, they are already doing better than they were doing with Fermi prior to launch.
It's the same situation every generation. People think that whoever launches the fastest card first has some "magical advantage". What matters is the product composition of your next generation line-up and the ability to deliver a top-to-bottom next generation lineup quickly. $550+ cards might matter on the forums, but not in the real world. Generations are fought over a 1.5-2 year period now, not over 2-3 months. Also, you aren't even taking into account the possibility of a price cut. Drop GTX580 to $380 and suddenly a $550 HD7970 doesn't look so great.
This post isn't meant to tarnish the launch of HD7000 series in any way. If NV launches first and AMD doom and gloom posts appeared, I would have responded in the exact same way. Just replace NV with AMD letters in my post.
I'll even provide a counter-example in favour of AMD. NV had the fastest single GPU from March 2010 (GTX480) to now (GTX580). Did that stop AMD from having an excellent line-up in the form of HD5000 and HD6000 series? No, it did not. If you think just because HD7970 smokes GTX580, that suddenly people are going to stop buying NV cards at $299 and below, you are strongly mistaken.