-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
The first "proper" Kepler news Fri 17th Feb?
- Thread starter dave_beast
- Start date
Soldato
It'd be highly disappointing if it didn't.
Me too. Makes me wonder what the GK112 with the 512-bit memory bus will pack.
What you'd be disappointed if the GTX 660 didn't equal the 7970??? I think Nvidia are in dreamland at the moment, never in their wildest dreams would they have expected this!
Caporegime
Haven't they rebadged the GK104 as the high end now?
Soldato
The 660? That's not going to happen.
I had the "680" in mind and that's not the proper chip I want to see.
Associate
Where does it say that this is going to happen?
The 660 is the 680
Gx104 part is the x60 in Nvidia's naming scheme, Nvidia have found they can compete with the 7970 with the GK104, ie the 660 (Nvidia will just name it 680)Full kepler single GPU part isn't coming until August
Note all of the above is based on the current rumours but looking very likely to be true.
Soldato
People were saying it was the 670ti in house but that was more rumour churning.
Why are they calling it the 680? Because of the performance? What are they going to call the full fat version? 680Ti or something daft?
Apparently AMD have confirmed the 7990 for April to counter the 680. But AMD should be more concerned who's writing their crossfire drivers at the moment.
Why are they calling it the 680? Because of the performance? What are they going to call the full fat version? 680Ti or something daft?
Apparently AMD have confirmed the 7990 for April to counter the 680. But AMD should be more concerned who's writing their crossfire drivers at the moment.
Soldato
What are they going to call the full fat version? 680Ti or something daft?
I suspect that the "full-fat" Kepler will receive the GTX780 name.
So let me get this right, people are saying that GK104 was so good that nVidia decided to call it a GTX680 instead of what it was originally planned to be, i.e a GTX660Ti.
Yet they're late to the party because of yield problems?
Something doesn't add up there.
Yet they're late to the party because of yield problems?
Something doesn't add up there.
Caporegime
Of course maybe people should look at the name too - the GK100 seems to have never made it but the GK104 did. The big Kepler is the GK110 which indicates the MK2 Kepler series just as the GF110 was the MK2 Fermi card and the GF100 was the MK1 Fermi card.
It could be a simple case of the GK104 being the biggest Nvidia Kepler GPU that Nvidia is producing for the next few months,as the GK100 might have had issues. I really doubt the GK100 being canned so late in development just because AMD did not produce a faster HD7970 for gaming. After all the GK100 would be of great important in professional and commercial markets as the basis of the the next generation Tesla and Quadro cards.
Maybe they were expecting an R600/RV670 and G92 situation, but realised the GK104 could compete with or maybe exceed the HD7970,so hence the comments they made.
It could be a simple case of the GK104 being the biggest Nvidia Kepler GPU that Nvidia is producing for the next few months,as the GK100 might have had issues. I really doubt the GK100 being canned so late in development just because AMD did not produce a faster HD7970 for gaming. After all the GK100 would be of great important in professional and commercial markets as the basis of the the next generation Tesla and Quadro cards.
Maybe they were expecting an R600/RV670 and G92 situation, but realised the GK104 could compete with or maybe exceed the HD7970,so hence the comments they made.
Last edited:
It's sounding like this rev of Kepler will be about the same as the 7970, especially if you take into account the 7970s inherent overclock headroom vs Keplers turbo boost etc.
The full Kepler (110) piece seems to be due later in the year, although at that point AMD could have the new Tenerife part available, either as the first of the 8000 series, or, I think more likely as the new top end 7XXX card.
In either case I suspect prices and performance is going to be similar from both camps for the next 12months so pick the flavour you prefer and enjoy. Neither is likely to be a "no brainer" by the sound of it.
I've gone for 7970s for the overclock headroom (really, there's absolutely no reason not to be running at the very least at 1000/5700) and crossfire for me works better than Sli (which I need for 2560x1440). For others Nvidia 3d, drivers etc make a better choice.
Either way I think we've got some decent if not astounding GPUs this time and it'd be pretty hard to make a "bad" decision either way.
The full Kepler (110) piece seems to be due later in the year, although at that point AMD could have the new Tenerife part available, either as the first of the 8000 series, or, I think more likely as the new top end 7XXX card.
In either case I suspect prices and performance is going to be similar from both camps for the next 12months so pick the flavour you prefer and enjoy. Neither is likely to be a "no brainer" by the sound of it.
I've gone for 7970s for the overclock headroom (really, there's absolutely no reason not to be running at the very least at 1000/5700) and crossfire for me works better than Sli (which I need for 2560x1440). For others Nvidia 3d, drivers etc make a better choice.
Either way I think we've got some decent if not astounding GPUs this time and it'd be pretty hard to make a "bad" decision either way.
Soldato
Top 7 series card is 7990.
If they do have a 2k+ shader GPU they might just use that to fight the GK110.
If they do have a 2k+ shader GPU they might just use that to fight the GK110.
Why are they calling it the 680? Because of the performance? What are they going to call the full fat version? 680Ti or something daft?
Yes exactly due to performance, hence why Nvidia couldn't believe their luck.
Soldato
AMD will have Tenerife ready within a few days of Kepler launching.
They need to maintain a clear lead as it makes all of their cards desirable from a buyer's standpoint.
A lot of people will buy a card from the market leader.
They're just waiting for Kepler to release, then they will do what they have to do to gain a clear lead over it and release it IMO.
It may not be luck tbh. No one knows exactly how much is locked out on Tahiti and waiting in the wings to be Tenerife.
For all we know the Tenerife card could absolutely smash the 680 to bits. If it does then given that Nvidia are currently on the back foot it could prove bad for Kepler.
If Nvidia give out the 104 with everything it is capable of and AMD hit back with something faster that screws Nvidia for another few months. Any lead Nvidia gain will be very short lived IMO.
Last edited:
I had a feeling AMD would effectively jump a generation ahead of Nvidia at some point but I expect it to be a couple of years away and not this year.
Soldato
It's impossible to speculate on exactly what AMD's 28nm (stupid names aside) is capable of tbh. I mean we have definitely had a taster ! but given they were first to the punch it was clear from the start they did not release it all.
Nvidia got lucky with Fermi IMO. By the time the 580 was launched they were forced to give it all out. Everything it was capable of was there, and thankfully AMD couldn't manage to beat it.
This time around though I think it's going to be rather different. The fact that the 104 was going to be launched as a 670ti and barely manage to beat the 7950 speaks volumes. I would imagine what happened after was them simply pushing up the clocks to see how far they could go to make a stable card.
Thus, I don't think there is going to be a lot left in the tank for overclocking with Kepler 104s. I could be wrong of course, but I just don't see it. I think Nvidia have already had to give it a bloody hard squeeze.
Still, at least those who were waiting for the 104 will at least get everything it has to offer without having to buy two or three revisions of it
