It would be very weird for them to keep manufacturing the 1080Ti for retail after the 2070 and 2080 arrive. Considering that the 1070 measurably outperforms the 980Ti - which admittedly was a bigger than usual performance jump - we should expect the 2070 to at least come close, no?My Crystal Ballshit says Christmas will look like this:
Volta Titan - $1499
X2 Vega - $999
Volta 2080 - $799
Big Vega - $599
Pascal 1080Ti - $599
Mid Vega - $449
Volta 2070 - $399
Small Vega - $329
Pascal 1070 - $299
As I mentioned previously in this thread, someone at Computex (an OEM rep of some kind) mentioned that "the GTX 20-series" would arrive "this holiday season" or some such. It was a small line in a post about something completely different, so of course I can't find it again to provide a source. Of course, OEM reps at trade shows are known to spout a lot of BS.Here we go, Volta consumer already for Christmas when GDDR6 is planned for early 2018 (and that's for shipment). Try to clean a little your crystall balls, at least, they seem quite foggy at the moment.
My Crystal Ballshit says Christmas will look like this:
Here we call that crystalbullshit :>
My Crystal Ballshit says Christmas will look like this:
Volta Titan - $1499
X2 Vega - $999
Volta 2080 - $799
Big Vega - $599
Pascal 1080Ti - $599
Mid Vega - $449
Volta 2070 - $399
Small Vega - $329
Pascal 1070 - $299
The Crystal Ballshit reading skill doesn't seem to be a good suit for you.
Being predictable in the past does not mean it is predictable for the future. I.e. they are targetting a new, custom process (12nm).
Looks like I created something
NV have been doing +~30% year on year for a long time now though, and they're putting enough into R&D that there's every reason to expect that to continue for at least a bit longer. V100 basically demonstrates enough technical scope to get that kind of jump for medium Volta then big Volta. They'll tweak details to make it work.
After that there'll be some new process ready for maybe two more cycles. Then we'll have to see. The limits on process tech are obviously looming.
AMD are doing really well to half keep up, especially given the difference in budgets.
Nope.Nv doing same thing since kepler.You can be 100% sure volta GTX2080 will be faster than TITANXP and they will target with GTX2070 1080TI performance.GTX2060 will be close to GTX1080 performance.Big volta will be +- 50% faster than GTX2080.Yes, but as I've said past performance does not translate in future performance as well.
Nope.Nv doing same thing since kepler.You can be 100% sure volta GTX2080 will be faster than TITANXP and they will target with GTX2070 1080TI performance.GTX2060 will be close to GTX1080 performance.Big volta will be +- 50% faster than GTX2080.
TDP-
Big volta 250w
GTX2080 180w
GTX2070 150w
GX2060 120w
The GTX 960 was, at best, comparable to the GTX 770. The GTX 970 came close to the 780Ti in some benchmarks, beat it in some, and was trounced by it in others. As such, I don't see where you're getting your "trend" of two-tier intergenerational jumps from. Pascal was an aberration here, due to the process jump allowing both higher clocks and more cores on cheaper chips at the same time. Although Volta is 12nm, I sincerely doubt we'll see anything comparable. Single-tier performance improvements (or dramatically larger = more expensive dice) seems far more likely. That V100 is ludicrously large isn't really indicative of anything on the consumer side.Nope.Nv doing same thing since kepler.You can be 100% sure volta GTX2080 will be faster than TITANXP and they will target with GTX2070 1080TI performance.GTX2060 will be close to GTX1080 performance.Big volta will be +- 50% faster than GTX2080.
TDP-
Big volta 250w
GTX2080 180w
GTX2070 150w
GX2060 120w
Ignore marketing names and focus on SKU names.GTX770 was renamed GTX680.GTX780 was BIG SKU big kepler.GTX970 was +- 780TI(BIG SKU full).GTX960 was +- GTX680(x04card).NV always target x04SKU performance from last generation with x06SKU and BIG SKU with x04SKU(full is faster and cutdown is +- same performance as BIG SKU)The GTX 960 was, at best, comparable to the GTX 770. The GTX 970 came close to the 780Ti in some benchmarks, beat it in some, and was trounced by it in others. As such, I don't see where you're getting your "trend" of two-tier intergenerational jumps from. Pascal was an aberration here, due to the process jump allowing both higher clocks and more cores on cheaper chips at the same time. Although Volta is 12nm, I sincerely doubt we'll see anything comparable. Single-tier performance improvements (or dramatically larger = more expensive dice) seems far more likely. That V100 is ludicrously large isn't really indicative of anything on the consumer side.
As for TDPs, the 780Ti and 780 were both 250W cards. The 770 was a 230W card and the 760 a 170W card. The 980 was a 165W card, with the 970 at 145W, the 960 at 120W and the 980Ti at 250W. The 1080 is a 180W card, the 1070 is 150W, and the 1060 at 120W. Again, the 1080Ti is 250W. In other words: they're all over the place except for a hard upper limit at 250W. Other than more powerful cards in the same generation needing more power, you really can't say much based on this at all.
Not sure if he means a delay in the past or if there will be a further delay from this point onwards...https://twitter.com/CDemerjian/status/872445675056898048
Charlie Demerijan from SemiAccurate on TT, about Vega:
''AMD made the right call to delay Vega, trust me here. They took a short term hit for long term gain and credibility. It took guts to do.''
Not sure if he means a delay in the past or if there will be a further delay from this point onwards...
+building up stock to fully allow the market to exhaust its demand or at least enough to make it worth AMD's while. If they can beat 1080 ti on average and come in at $600-650 then Radeon is back in a big way!Reads like waiting till end of July/early August might have been a good play for AMD?
I'm betting it's a combination of HBM2 and drivers. Every month of driver development is probably pretty noticeable.
Explain Vega 11 in 2018.... A midrange chip of the Vega lineup that late? And that doesn't raise any eyebrows? Lol ok
My question has nothing to do with HBM2 or GDDR.What's hard to explain about it? Assuming HBM2 is limited or expensive (or both), it doesn't make sense to put it on a midrange product when you can't make much profit from it. If it does use GDDR6 then they have to wait for that as well and it probably means they need to design a memory controller to interface with it because they can't just reuse what they have from big Vega.
Polaris is good enough to hold them over until then and NVidia seems content to leave 1060 prices in place so there's no much need from AMD other than the middle gap between Polaris and Vega that I think the 1070 is going to slide into nicely.
Again, I'm not asking about what memory Vega 11 uses. That's not relevant.My prediction: Vega HBM will be up against renamed Pascal 20XX cards for holidays 2017. Smaller Vega with GDDR of some flavor and mainstream Volta show up 1H 2018.