NVIDIA Kepler speculation thread

More or less?

Those mumblings from 3DCenter were 1024CC for GK104 and 1536CC for GK110.

None of the material I've seen so far on the 3DC homepage was written in any absolute sense but rather in a speculative manner. I'd personally count more than 1k for the 104 and call for bullshit on anything regarding "GK100" or "GK110" speculated so far as hypothetical specifications. There should be leaked material for the 104 but not the 110.

Obviously the stuff from VR-zone are more than just off as you say. But then again when did vr-zone, chiphell and the likes have any real "sources" that I've missed anything? Most of the times just hearsay and if N% should be correct it's more a matter of luck.
 
So. I'm still totally unclear here, GK104 is the mid-high sized chip? (fits with 560 Ti successor)

Is there a big daddy single chip? Is that GK110? Or is GK110 just supposed to be a dual GK104 card?

In other words which codename is supposed to be the highest single end Nvidia Kepler GPU?

Edit: I'm getting a clearer picture from this link http://www.geeks3d.com/20111126/nvidia-kepler-gpus-roadmap-gk107-gk106-gk104-gk110-and-gk112/

They do have the GK104 wrong at 384 bit though. presumably. Which throws the whole chart into question.

But I guess most people think GK104 is the most powerful kepler coming anytime soon?
 
I hope they will cut short with all those name extensions. There are a hundred numbers going from 600 to 700 (or from 700 to 800) - give or take. ;) That should suffice for just about any model variation that comes to mind.
 
So. I'm still totally unclear here, GK104 is the mid-high sized chip? (fits with 560 Ti successor)

Is there a big daddy single chip? Is that GK110? Or is GK110 just supposed to be a dual GK104 card?

In other words which codename is supposed to be the highest single end Nvidia Kepler GPU?

Edit: I'm getting a clearer picture from this link http://www.geeks3d.com/20111126/nvidia-kepler-gpus-roadmap-gk107-gk106-gk104-gk110-and-gk112/

They do have the GK104 wrong at 384 bit though. presumably. Which throws the whole chart into question.

But I guess most people think GK104 is the most powerful kepler coming anytime soon?

From the top of my head NV claimed in a marketing slide that Kepler will launch in their FY13 which if memory serves well ends in April. Also I recall Jensen stating somewhere that they intend to launch the entire product line within a 3 months timeframe. It could be that they'll be able to keep the first, the latter we'll see.

As for GK104 I guess it's more a matter of marketing perspective than anything, You can eventually have a GPU with a 256bit bus yielding higher performance than former performance parts used to, but I don't see how you could claim a genuine high end part with as little bandwidth ***edit: it could probably if it would be a TBDR, but that's likelier than me growing physically another 10" taller.
 
desktop_2012_01_21_16qdvgf.png


https://twitter.com/#!/CDemerjian
 
What does that even mean? I guess it implies what many of us had already suspected, it's not going to creme 7970 in raw performance, but it might be priced extremely well.

I certainly hope so.
 
Obviously the stuff from VR-zone are more than just off as you say. But then again when did vr-zone, chiphell and the likes have any real "sources" that I've missed anything? Most of the times just hearsay and if N% should be correct it's more a matter of luck.
VR-Zone was spot on with G80 specs (too accurate to be luck). Don't know about anything else.
 
What does that even mean? I guess it implies what many of us had already suspected, it's not going to creme 7970 in raw performance, but it might be priced extremely well.

I certainly hope so.

If they're planning a blitzkrieg based on low-ball pricing where are they going to get volumes from to meet the demand? GK104 probably isn't that much smaller than Tahiti.
 
a 256bit 2GB GTX 660TI matching the HD7970...is such a thing even possible? Are we talking about AMD 1st gen SIMD vs Nvidia...4th gen woes? Vrzone has been pretty accurate of late...a $299 GTX 660TI perf @ 95-100% of HD7970 would if anything...shows HD7970 launch & pricing was a poorly thought out strategy..im excited....an nvidia card to take the mid-high perf/$ segment..once again..? 8800gt redux?
 
a 256bit 2GB GTX 660TI matching the HD7970...is such a thing even possible? Are we talking about AMD 1st gen SIMD vs Nvidia...4th gen woes? Vrzone has been pretty accurate of late...a $299 GTX 660TI perf @ 95-100% of HD7970 would if anything...shows HD7970 launch & pricing was a poorly thought out strategy..im excited....an nvidia card to take the mid-high perf/$ segment..once again..? 8800gt redux?

Although I don't like Nvidia that much, I really do hope that this is how things will play out. It will bring much needed price wars. The 79XX series is ridiculously priced. That would teach them a lesson for befriending us with 48XX/58XX series and then anally fist loving as with the 69XX and especially 79XX.

"Made from gamers, priced for gamers" my ass.
 
What does that even mean? I guess it implies what many of us had already suspected, it's not going to creme 7970 in raw performance, but it might be priced extremely well.

I certainly hope so.
Or it does match it or surpass it , but the thing that make it stands out is not the performance , but other metrics .
 
The 79XX series is ridiculously priced.
Really? 3GB HD 7970 is as expensive as 3GB GTX 580. It offers 25-30% higher performance, lower power-consumption and some new features. In fact GTX 580 is inadequately expensive, so it's Nvidia's turn to bring a better priced product.

If their fastest product (GTX 660, or how will they call it) will be slower than HD 7970 (that means slightly faster than GTX 580), they won't have any other option than to sell it significantly cheaper. Because any product with GTX580-like performance for GTX580-like price would be unsalable. We are talking about 1-quater distant future, TSMC's 28nm manufacturing capacity will be much bigger at that time (TSMC promised about +170k wafers compared to Q4), so pricing can be very different at that time.

Anyway, this situation seems to be very similar to late 2009. When AMD launched HD 5800, Nvidia promised better pricing, better availability, better technologies, better performance... they would promise anything including stopping global warming to ruin competitors launch, while they had nothing in hands just like now :)
 
Back
Top