DuckThor Evil
Legend
I thought that positioning in this chart describes price brackets and not perfomance?
You thought wrong. The 6990 retails for over $700.
Well, there are several versions of this slide. The early one states "performance" for the Y axis. The later one doesn't and specifies, that "...is not necessarily an indication of relative performance". So you can choose which one you'll prefer (link)
7870 is between GTX580 and 6970.
7850 is "around 6950."
Unless the Tahiti LE rumors are true, which I currently don't believe, then we might see a 7870 at ~6950.
Edit- Anyone notice the current pricing of 6970 and 6950? Not a good sign.
We won't see any major price shift until there is competition...
DuckThor Evil
Legend
Thinking about it more, I agree I might have undershoot a bit. Last gen mid range and upper mid range was pretty strong. No reason to expect it to change.
Ehh? it pretty much obsoletes the 580 that I can tell. I'd call that a significant blow.
Unless you're probably referring to the 1.5GB 580's (arguably still kind of obsoleted). Personally I dont think buying such an expensive card with such a low RAM is very viable. And the 3GB 580's are still $550, which is obviously obsoleted.
All that assuming AMD gets it's 7970 stock issues sorted soon as Fudzilla stated they will.
I would be shocked if they're even ordering gf110 wafers still. When the 560's get squeezed then they'll feel it. AMD gave nVidia a free pass by positioning the 7970 above the price conscious category.
How many 580's vs 7970's do you think are being sold today anyway? I would bet the former is still selling pretty well simply due to availability, brand recognition and market penetration. There's a little too much angst over a card that's only been on the market for a week! It might end up being a 6 month lead in the end but for now it's a bit premature. Seriously it's like armageddon whenever the two companies don't launch on the same day, we should know better.
AFAIK GF110 is EOL since December last year.
I could believe that, now they can let the market absorb whatever's left in the channel. Considering they made it through the 40nm generation just fine despite Fermi's problems I think nVidia will survive the 7970 too. The last two heavy blows AMD struck were dealt by the 4870 and 5870, considerably cheaper cards.
Personally I don´t see any reason to upgrade from a GTX580 right now and it´s EOL since the end of last year. There is just too much going for it, still. GK104 is not going to change that and the bigger die is still some months away, sadly.
If AMD really wanted to attract my attention, there should have been more mature drivers right at the beginning and a more satisfying performance at stock. I also won´t pay into that ridiculous trend to implement more and more features just to keep my marketing business happy. But that´s another story.
Since the GTX580 is EOL since last year, I expect the GK104 to be introduced fairly soon. There have indeed been some changes WRT performance targets and launch date, so that part of the story seems to be true. The specs remain in the dark, still, don´t trust some of the recent news.
PS: Not quite sure if anyone realized/observed the rather rare driver releases of NV in this and the last month. It´s awfully quiet on the driver front. Kepler seems to be sucking up quite a large chunk of development time. That´s just my observation though.
If AMD really wanted to attract my attention, there should have been more mature drivers right at the beginning and a more satisfying performance at stock. I also won´t pay into that ridiculous trend to implement more and more features just to keep my marketing business happy. But that´s another story.
Since the GTX580 is EOL since last year, I expect the GK104 to be introduced fairly soon. There have indeed been some changes WRT performance targets and launch date, so that part of the story seems to be true. The specs remain in the dark, still, don´t trust some of the recent news.
PS: Not quite sure if anyone realized/observed the rather rare driver releases of NV in this and the last month. It´s awfully quiet on the driver front. Kepler seems to be sucking up quite a large chunk of development time. That´s just my observation though.
Last edited by a moderator:
It should be noted that the 40nm transition was pretty bad, yields were low and so was volume. AMD was selling everything they could make for months. So the impact on NVIDIA was really as big as 40nm capacity allowed, which happened to be pretty small.
It's hard to tell if the 28nm transition will be similar. In any case, the 7970 may be expensive, but I'm sure the 7950 and 7800s will be just as compelling on their respective price brackets. So it will boil down to capacity, once again.
Well if the 28nm process is kinder to AMD than 40nm was, why should we assume it's any less kind to nVidia?Inherent incompetence?
We shouldn't, but what does it matter how kind or unkind it is if they don't have anything to sell?
A1xLLcqAgt0qc2RyMz0y
Veteran
Please remove the OFF topic posts that are polluting this thread
This thread titled: Kepler speculation thread and has been derailed big time.
Why are all the off topic posts on AMD products/positioning/prices allowed? What exactly do they they have to do with what Kepler is/will be.
I expect a title of a thread "Kepler speculation thread" to be about Kepler.
There are other threads here for the AMD parts so please post to those threads.
Moderators please clean up this thread of the off-topic posts.
Thanks
This thread titled: Kepler speculation thread and has been derailed big time.
Why are all the off topic posts on AMD products/positioning/prices allowed? What exactly do they they have to do with what Kepler is/will be.
I expect a title of a thread "Kepler speculation thread" to be about Kepler.
There are other threads here for the AMD parts so please post to those threads.
Moderators please clean up this thread of the off-topic posts.
Thanks
Back to topic, last i heard from a source was that GK100 isnt due out till mid 2012. NV apparently had working GF117 silicon (a 28nm GF106 essentially) in Q3'11 but given that we havent seen a launch so far (considering its a much smaller chip than Tahiti which has already released), im wondering how true that and the information about GK100 was.
Looks like the speculative comparisons to amd cards got a bit out of hand indeed, I guess that's because there's just not much information about Kepler still.
I really have no idea what to expect (in contrast to GCN when AMD released the whitepaper and some educated guesses including unit count from july 2011 turned out to be quite accurate).
The family is built on tsmc 28nm, that's for sure. Everybody also expects GK100 to be a much larger, but also faster chip than Tahiti, and (in contrast to earlier families) launch somewhat later than GK104, which supposedly will launch around March/April.
Since there was no whitepaper for Kepler well in advance, I take that to mean the changes (at least as far as it would have implications for CUDA) will not be that extensive as they were for Kepler. Virtual memory was promised (so I guess similar to GCN in that area).
Since it seems there's still some debate over whether the chips will have hot clock, discussions about unit counts etc. should be taken with a grain of salt. It is also unclear if gk100 will be quite different internally to the rest of the family, as it was with gf100/gf110 and the rest (though rumors talking about gk104 being a "fermi/kepler hybrid" could indeed point to that, as would the later supposed release date).
So, any halfway trustworthy specifications I missed?
I really have no idea what to expect (in contrast to GCN when AMD released the whitepaper and some educated guesses including unit count from july 2011 turned out to be quite accurate).
The family is built on tsmc 28nm, that's for sure. Everybody also expects GK100 to be a much larger, but also faster chip than Tahiti, and (in contrast to earlier families) launch somewhat later than GK104, which supposedly will launch around March/April.
Since there was no whitepaper for Kepler well in advance, I take that to mean the changes (at least as far as it would have implications for CUDA) will not be that extensive as they were for Kepler. Virtual memory was promised (so I guess similar to GCN in that area).
Since it seems there's still some debate over whether the chips will have hot clock, discussions about unit counts etc. should be taken with a grain of salt. It is also unclear if gk100 will be quite different internally to the rest of the family, as it was with gf100/gf110 and the rest (though rumors talking about gk104 being a "fermi/kepler hybrid" could indeed point to that, as would the later supposed release date).
So, any halfway trustworthy specifications I missed?
Man from Atlantis
Veteran
Latest rumour is there will be no GK104 at all.. The named GK104 is nothing more than GK106.. 3 chips are GK110, GK106 and GK107, launch in April..
