NVIDIA Kepler speculation thread

[tunafish]

I could load the first page, then it all went down. Could someone who has read the SMX part explain the changes? Is there hotclock?

 

[AnarchX]

Seems the Chinese expect AMD to reduce HD 7900 prices by 200 USD.

The GPU Boost is up to 1,1GHz for selected application like Heaven 3.0 or Battlefield 3.

 

[Man from Atlantis]

Damn im mobile:/ slower in mw3 preorder cancel

 

[Cuthalu]

It loads ok for me so far. 680 is cooler and perform better than 7970.

Even though the site is mainly in Chinese, you can still see slides like this:

Thanks for linking directly to hkepc, now this thread loads at a fantastical 1.6KB/s. Good times.

 

[CarstenS]

Simple - overclocking.

Even under control of Powertune at default HD 7970 has been measured consuming up to 234 watts.

 

[oramay]

I could load the first page, then it all went down. Could someone who has read the SMX part explain the changes? Is there hotclock?

It's on page 3 GPU Boost http://www.hkepc.com/7672/page/3#view Kepler dynamically adjust clock speed to use the TDP allowed.

Thanks for linking directly to hkepc, now this thread loads at a fantastical 1.6KB/s. Good times.

The site is really fast for me. Never realized it's slow loading from overseas.

 

[fellix]

The SM configuration is like the GF104's one but on steroids -- two Fermi SMs literaly fused in to one "SMX" and then some even more SPs on top!
It's all about ALU/Compute density in Kepler, as it seems.

 

[oramay]

Seems the Chinese expect AMD to reduce HD 7900 prices by 200 USD.

The GPU Boost is up to 1,1GHz for selected application like Heaven 3.0 or Battlefield 3.

The 200 isn't prefixed by USD. So I think they mean HKD200 which is about USD26.

 

[Love_In_Rio]

It seems a good chip but as it seems results would be alike to a overdrived 7970.

 

[tunafish]

The site is really fast for me. Never realized it's slow loading from overseas.

Could you please download all the slides, and upload them to, for example, imgur.com?

Hotlinking them does nothing when they always stop loading after the first 5 rows of pixels...

 

[no-X]

Even under control of Powertune at default HD 7970 has been measured consuming up to 234 watts.

9 watts over TDP value are quite insignificant compared to e.g. GTX 480 going even 70 watts over official TDP value.

 

[CarstenS]

no-X, nobody's argueing that. It's whether or not 6+8-pin is there for a reason (other than using the card out of spec, i.e. overclocking).

 

[SimBy]

From what I could see it dynamically overclocks to around 1100 MHz in some games and benchmarks.
Base clock is 1006 Mhz. Package size is 18.55 x 17.18 mm. 3.54B trannies.

All benchmarks are at 1920x1080 8xAA and biggest advantage I've seen was around 20%.

HD 7970 at 1.2GHz would be on par or slightly ahead.

 

[leoneazzurro]

It seems a good chip but as it seems results would be alike to a overdrived 7970.

Performance/mm^2 seems indeed quite high, but these are also some of the games where Nvidia fared better in the past (and not at the highest resolution available). No word on real power consumption and overclockability (it is even possible?).

 

[oramay]

Here are the slides and photos for people who have problems loading.
http://imgur.com/a/aQmuA

 

[bdmosky]

Only DirectX 11, not 11.1?

 

[Love_In_Rio]

Here are the slides and photos for people who have problems loading.
http://imgur.com/a/aQmuA

Thanks!.

 

[Vardant]

Even if the chip could actually reach a little higher clocks, the 2x6pin plugs won't allow for a lot of headroom in power draw.

So anything above a 5-10% overclock will be a big surprise, imo.

You must have missed the info about them hitting 1400+ then

 

[fellix]

Here are the slides and photos for people who have problems loading.
http://imgur.com/a/aQmuA

Thank you, sir, but: "Imgur is over capacity!"... d'oh!

 

[silent_guy]

Even if the chip could actually reach a little higher clocks, the 2x6pin plugs won't allow for a lot of headroom in power draw.

Q: what's the difference between a 6 pin power connector and an 8 pin power connector?
A: 2 additional ground pins

That's it. It makes now difference in the amount of power the PSU can deliver to the board. It does make some difference how clean that power can be delivered, but for something like over locks, where you basically assume non-worst conditions(*) anyway, it should be such a big deal.

(*) one of the gating things about all these power discussions, 'backed up' by graphs from reviews, is that they don't prove anything unless you have multiple samples, negative sigma included. Trying to prove a point with hand-picked reviewer's samples is, well, pointless.