honestly if this leak is any true, if the prices are true, it brings very little as far as performance per dollar. I expect the prices to already creep down slowly as the year progresses anyway.....
If those specs are true, it's way better than the 780. 780 is only 17-22% faster than 7970GE for $650. This shows a card 35% faster for $600.
At the same time, how does one increase die size from
365mm2 to 420mm2 (15%) and performance increases by 35% without a major GPU clock speed bump? If it was possible to increase the number of ROPs by 50%, number of SPs/TMUs by 12.5% and die size only goes up 15%, how come it took AMD more than 1.5 years to do this? Sounds too good to be true. :hmm:
If AMD decides to release a 420mm2 28nm card, I can see delays to the launch of 20nm parts. I don't see the point of launching these parts in Q3 2013 just to follow them up with 20nm in Q1 2014. Either this rumor is fake and AMD will stick to their guns that there will be no new flagship card above 7970GE in all 2013 (based on TPU interview), or 20nm is facing delays and AMD has been purposely releasing wrong information about them continuing with HD7970GE in 2013 without any faster parts to keep NV in the dark.
Given that RR is very conservative, and he already stated that moving down to lower nodes is too risky and very expensive, it was somewhat hard to believe those rumors of AMD having 20nm parts by Q4 2013/Q1 2014. With AMD you never know what to believe.
Without making a judgement on this particular rumor, I agree AMD would be foolish to leave their lineup unchanged through the end of the year.
Depends. See NV has a very large market for Tesla/Quadro cards, while AMD doesn't. Consider the cost of designing and fabricating a new line-up of 28nm chips and the possible profits may not be worth the hassle. NV has already gone through break-even analysis and it obviously made sense for them since GK110 was planned to service different markets from the very beginning. Now NV simply allocates failed GK110 chips across Tesla/GTX780. AMD doesn't have this luxury. It would also be foolish for AMD to spend half a billion dollars on getting these updated 28nm chips out, using a large number of their engineering resources and then not have sufficient funds and resources to fight Maxwell. Right now, NV only has 2 chips faster than HD7970GE and the rest of NV's line-up is nothing special. If AMD messes up, the entire 20nm Maxwell line-up could walk all over AMD's if AMD doesn't manage its resources effectively. If AMD has no new 28nm parts for the next 8-9 months, it will not be as bad as getting completely dominated because you didn't have enough resources for 20nm generation. AMD has to be very careful here. The real battle is 20nm because 28nm generation is way past the half way point and GTX780/Titan will hardly matter for overall sales. Chasing those cards for a short period of time doesn't make sense if AMD is starved for resources which are imperative for 20nm generation. However, if AMD needs another 12 months to launch 20nm parts, well then it makes sense for this major 28nm refresh.
If I were AMD I'd spend $0 on 28nm parts and spend all my $ on 20nm because I expect Maxwell to be a monster. It is a brand new compute architecture from NV and has been in design for 3-4 years. AMD will need to counter it with a 2-year-old GCN -- good luck! Considering NV was able to surpass GCN with a slightly massaged Fermi architecture, dating all the way to 2010, AMD better not underestimate how great Maxwell can be. GTX780/Titan are not a threat in the context of what NV has coming next...