Clearly just a guess, but based off what they did with Maxwell...
- Summer 2016 - Mid Tier - GP104 (970/980 updates; 6GB of GDDR5X; 980 Ti performance; $400-500)
- Fall/Winter 2016 - Low Tier - GP106 (950/960 updates; 4GB of GDDR5; 970/980 performance; $200-300)
- Spring 2017 - Uber Tier - GP100 (Titan X update; 16GB of HBM2; 50% performance increase over Titan X; $1000)
- Summer 2017 - High Tier - GP100 (980 Ti update; 8GB of HBM2; 50% increase over 980 Ti; $650-700)
Moves all current performance/memory down one tier on the hierarchy, representing an upgrade across the board. Quarterly updates keeps the product line fresh. Releasing the Mid Tier and God/High Tier nearly a year apart allows Nvidia to double dip on the enthusiasts who want the latest and greatest.
RS' nitpicking about the 750 Ti aside, this is more or less the pattern we've seen so far for both Kepler and Maxwell. GK104/GM204 first(in terms of broad market availability, not one-off niche GPUs) and then the bigger uarch later on.
I'd just add that we'd likely see GP100 first out of the gate, but for HPC, and not for consumers. The wildcard here is the Titan card. Will NV make an extraordinarily expensive new Titan based off the GP100 in 2016? And instead of the traditional "we'll release the same GPU just a few months later for much cheaper"(like the 780 Ti or the 980 Ti) they might hold off until 2017, giving the GP100 Titan a very long runway compared to previous years.
It'd be interesting to see if the GPU would still sell, from a purely market perspective. RS likes to say that NV has a legion of buyers who are far less price sensitive and will not make a market analysis in a rational way(read: fanboys). Maybe.
But with the decline in dGPUs shipmets overall, we've more or less cut away the "casual"/home PC users. They are now mostly getting iGPUs or they are not getting a PC at all. Those who remain are now powerusers and enthusiasts and as such, their budgets are far less price sensitive. In such an environment, an overpriced Titan XYZ for GP100 might make more sense to keep around for more than a few months.
As always, this depends on AMD. AMD has in the last few years released GPUs who are superior hardware-wise to their NV competitors at their launch, at least if we look at the GK104/GP204 GPUs. Compare the 290 today with the 780, for example.
What held AMD back was poor drivers. Their GPUs didn't really come into their own until at least 6 months down the line. Most people judge by first impressions. And when the launch reviews are out, if AMD isn't clearly ahead, then that perception stays with people. That's just the performance. Now think about the poor reference coolers for both Hawaii and (initially) the Fury X.
If AMD get their
redactedtogether and launch GPUs with non-crap coolers
and non-crappy launch drivers, then NV's strategy would fall on its face. RTG, under Raja, has been making all the right moves since they got more power and autonomy, and I am hoping this will translate into strong 2016 GPU launches. I for one am hoping for a good Greeland plus a great CES 2016 where I can finally take a look at monitors with an AMD equivalent of ULMB which doesn't cost me one of my arms. I could sell my current G-Sync monitor at a profit in the aftermarket since I got it for cheap during BF. I want AMD to succeed but right now, for my needs, only NV can provide.
This is not the AMD sub forum. Limit the discussion to the topic at hand. Also, no profanity allowed in the tech section.
esquared
Anandtech Forum Director