The_King :
What Nvidia promised the GTX 680 to be and what they gave the public is two different things (GTX 670 Ti) cause the 7970 is not fast has Nvidia expected it to be. I will still stick with my 5970
This is dumb. These product launches are planned out WAY in advance, and nVidia decided to remove the compute support of the 680 long before they knew what the benchmarks would be for the 7970. Remember, these things were in the fab process for 6-12 months now trying to get 28nm to work properly, which means the die has been set for most of that time. It is far more likely that nVidia simply looked at the profit margins of their Quadro and Tesla cards, and then looked at their 500 lineup and realized that there are plenty of power users like me who were more than happy with the 570's CUDA support for Adobe CS5, but who don't really game all that much. If it was not for the 570 for $300 (Paid $275 for mine after sale and rebate
) then I would have had to purchase a $600 Quadro card to do the same workload (granted, a faster driver, and ecc is thrown in there, but still not worth 2x the price for what I do when there is a cheaper option), and nVidia would have gotten an extra few hundred $$ out of me had the 570 been a game-only card. So in short, that is why the compute performance sucks, because they are trying to differentiate their game cards from their professional cards. Plus by making it more purpose built they can save on die space, keep profits higher, temps cooler, etc. I know everyone is waiting for the 'real' performance king with the wider busses and the higher compute rate to come out, but I think this time around that is going to be reserved for the much more expensive professional cards, and for gaming it will not perform much better than the 680 they just released.
If you think about it, it was a brilliant move. Have a multipurpose GPU to get the entry-level professionals like myself hooked on hardware acceleration, get all of the software support from the big production software companies, and then take away the cheap option a few years later when we are all ready to upgrade again. It worked so well that AMD is now taking the same route. Previously AMD has SUCKED for professional use (with the exception of excellent multi-monitor support that could only be rivaled by Matrox... who is a tiny little company, and not really much of a competitor) not so much because their GPUs were bad, but because the could not get community support. But the new 7000 series is absolutely mindblowing for production work which will attract the entry-level "prosumers", and if they can get the support of places like Adobe, Maya, and other big multimedia companies to push their cards then it could be a huge upset for nVidia who has traditionally dominated that market.
I think that this shows a general direction that AMD is headed as a whole; Gamers (well... most gamers) do not have money, and it is very expensive to be 'the best' in that market, and there is not much profit in being the !/$ king. So now we saw Bulldozer, which while a flawed product, has a great vision of architecture for media and many-thread applications... but not really a gaming monster. If they figure themselves out on the next go-around then they could easily win me over from Intel for production work as they would be much cheaper than the SB-E/IB-E series while still providing a high thread count... again, not games. Now we see the GPU side following suit; The 7970 is no slouch for gaming, but to say that games are the focus of the card design is silly. The big focus was on the pro market and making a good production card. If they pull it off then they could find an excellent new niche to save them from ever better integrated graphics on the low end, and companies that do not have a CPU division bleeding money on the high end. And, lets face it, while PC gaming is not going to disappear any time soon, the mass market is in consoles, and it will only get worse when the next gen is released and can really do 1080p gaming right on a TV, which will pull a lot of budget gamers away from the more expensive PC gaming market. So that knocks out GPU sales on the low end, and GPU sales in the middle. What is left is the high end gaming market which has stiff competition with nVidia, and the pro market which they have largely neglected in the past. But with the other 2 traditional cash cows under major threat of drying up, I think AMD graphics is trying to break into the only market that they have left to explore.
In short; No nVidia gave us exactly what they promised: a kick-ass gaming card that yields higher profits while still offering a lower price to consumers and keeping with the 20-30% performance increase we expect over the previous gen high-end chip. This will also fuel the sales of their higher end product lines when the new quadro cards arrive. If nVidia is successful at this model (which AMD has tried for years without success) then it means they get higher profits to sit on for more R&D and patent purchasing power, but now that AMD is following nVidia's old way of doing business there could be much stiffer competition across the board, which will make things much more interesting for the next set of cards to come in 1.5-2 years.
Lastly, the 680 is a refresh of the 560 not the 570 (570 is a dumbed down 580, but still based on the 580), AND this is a firmi refresh/die shrink, not an entirely new architecture. So look at the benchmarks of the 560 vs the 680 to do your comparisons... this is quite simply the largest increase in horsepower between generations EVER. Period. Without exception. End of story. It is absolutely amazing what they managed to do, and while I still stand by my earlier statement of the GK114 potentially not being that much better for games than the GK104 is, I think that it will hold something truly special for the pro markets that just might make all of the executives at AMD cry, and make their engineers bow down in worship. Sure, I could be wrong on this one, but after seeing what they did with the 560, I could only imagine what they manage to do with the GK114 which is based off of the 580/GF114. Not to mention the rumors of AMD and nVidia both getting ready to release duel GPU setups within the next few months.
I am personally quite happy with my little 570 and will not be upgrading for a long time (typically 4 years), but it is really exciting and fun to watch all this new tech come down the pipe.
Can't wait.
Facebook
Twitter
Instagram