NVIDIA Kepler speculation thread

It's a GPU that is comparable to GF104/114 in physical terms (size, power envelope) so perhaps you can think of it in whatever terms you thought of GF104/114.

GK 104 is almost 20 % smaller than GF114 and a bit over 10% smaller than GF104 and with the comments from nVidia that they thought Tahiti was faster and the pics of GTX 670 flying around, there is more than enough reason to believe that GK104 wasn't supposed to power a $499 card. I don't think it's reasonable to assume that it would have been a direct replacement for GTX 560 either, but perhaps something to occupy the $299-399 segment?
 
As I said before, do you really think it was ever Nvidia's intent that their fastest available consumer product would be priced at $299?
 
Like everybody else they probably expected more performance from AMD new-architecture GCN 2048SPs and 384-Bit GPU and also not such a high pricing from AMD.
 
As I said before, do you really think it was ever Nvidia's intent that their fastest available consumer product would be priced at $299?

No. Like I said perhaps $399 for the top GK104, perhaps a quick roll out for the dual chip solution and possibly a bit more aggressive timeline for the big one.
 
No. Like I said perhaps $399 for the top GK104, perhaps a quick roll out for the dual chip solution and possibly a bit more aggressive timeline for the big one.

Even that would have had serious repercussions down the lineup and considering the NV rant about price per transistor it seems more likely that what they are paying for GK104 is closer what they were paying for GF110 than GF114 anyway.
 
Well I dunno... remember that Zotac GTX 680 allegedly running at 2 GHz?
eusa_think.gif
 
Well I dunno... remember that Zotac GTX 680 allegedly running at 2 GHz?
eusa_think.gif

That one was "lost in translation" or some misunderstanding, they obviously meant it's possible to reach 2GHz (which would be 100MHz higher than current record on reference with beefed pwm) on their PCB using LN2 or liquid helium cooling, not air :rolleyes:
 
Yeah, but the slide specifically mentions two frequencies, 2012 MHz only being the "proc clock", which I suppose means the SP clock, while the rest of the chips runs at 1 GHz, just like the regular GTX 680. OC editions of GeForce GTX 560 Ti ran at 900-950 MHz core, which means 1800-1900 MHz "hot clock". At 40nm and air cooling with no problem. I think you are a bit quick to dismiss the info, regardless of today being the 1st of April.
 
Yeah, but the slide specifically mentions two frequencies, 2012 MHz only being the "proc clock", which I suppose means the SP clock, while the rest of the chips runs at 1 GHz, just like the regular GTX 680. OC editions of GeForce GTX 560 Ti ran at 900-950 MHz core, which means 1800-1900 MHz "hot clock". At 40nm and air cooling with no problem. I think you are a bit quick to dismiss the info, regardless of today being the 1st of April.

Yeah, no. There's no hotclocks in GK104.
 
@Lukfi, Again we are the 1st April, there's no hotclock on Kepler, (proc run at the same speed) lol. Nvidia, or Zotac, will not redesign the cores for add hotclock.
 
Surely Photoshoped There's no reason Nvidia introduce back hotclock with Kepler (or revert back to it ) and i dont see why use 2x 6 pin + 1x 8pin, instead of 2x 8pin. ( The hardware for hotclock is not present in 680, this will mean complete new core)

( anyway the image is so blurried on the "GTX letters", the Bezel can come from an Asus DirectCU II )

If the slide was not there and we are not the 1st April, i will maybe think it can be the GTX690, but the slides confirm it is a fake (or faked somewhere ): 1536CC, 365W ... + the photo on the slide differ of the photos taken.
 
Last edited by a moderator:
UnsuccessfulTroll.jpg

(Can't take credit for the pictures, though. A colleague of mine at ExtraHardware made them and posted them, so he could link the post as source for his April Fools' article. Come visit us and don't miss our exclusive SSD overclocking guide! :smile: )

We're obviously fooling noone here, but VRforums might be a different story. The post has so far managed to get a few replies, but keep in mind that it's the middle of the night in Asia. I reckon we can still get a few laughs from the thread.
 
Yeah, but the slide specifically mentions two frequencies, 2012 MHz only being the "proc clock", which I suppose means the SP clock, while the rest of the chips runs at 1 GHz

There's no longer a separate clock domains for the shaders and the rest of the GPU. It is absolutely impossible to run a Kepler chip with a shader frequency that differs from the "GPU frequency".

Especially as one of the doublings of the shaders was achieved by "running both the clocks of the hot clock at the same time" in separate units. Kepler with hotclock just would not work. It would make absolutely no sense.
 
(Can't take credit for the pictures, though. A colleague of mine at ExtraHardware made them and posted them, so he could link the post as source for his April Fools' article. Come visit us and don't miss our exclusive SSD overclocking guide! :smile: )

We're obviously fooling noone here, but VRforums might be a different story. The post has so far managed to get a few replies, but keep in mind that it's the middle of the night in Asia. I reckon we can still get a few laughs from the thread.


You should have post it on Guru3D, im sure it will have work a little bit. Anyway on some.
 
Back
Top