NVIDIA Kepler speculation thread

[jaredpace]

AMD is beat on performance/watt and performance/mm2. Finally!

 

[CarstenS]

I wonder what happens at a normal 4xAA and higher res. It might even be behind HD7970. At least in
a few games. Might even be further ahead who knows.

Higher AA should favor the card with higher ROP and memory throughput theoretically. Higher resolution's usually good for architectures with massive fillrates, the memory size and architecture to support it and is not as dependant on a very efficient front-end, as triangle-sizes go up.

 

[doob]

AMD is beat on performance/watt and performance/mm2. Finally!

And in the end both will also will beat the consumer in performance/price.

 

[SimBy]

AMD is beat on performance/watt and performance/mm2. Finally!

I think it's about on par with Pitcarin. Which is impressive.

 

[silent_guy]

17.18 x 18.55 comes out to 318 mm^2. So is that with the heat spreader on? The slide on that site still says it being 294 mm^2.

There are 3 numbers possible:
This one: includes active die, includes scribe around active die, includes some amount of glue around the scribe
Die size as used to calculate dies per wafer: glue not included.
Die size as used to calculate yield: scribe not included. This is usually the number used most internally.

Not unreasonable to go from ~300mm2 to ~320mm2 to go from one to the other.

 

[Gipsel]

Higher AA should favor the card with higher ROP and memory throughput theoretically.

Actually, there is a difference between 4x and 8x MSAA. As AMD ROPs can't handle 8xMSAA at the speed NV ROPs are capable of, the ROP load basically doubles at 8x for AMD cards (but not for nV). The memory bandwidth plays a bigger role at 4x in case of AMD (because it is less limited by the ROP throughput itself), at 8x it doesn't help that much. So at 4x the bandwidth can have a potentially larger effect than at 8x (while in nV's case a higher bandwidth should pay off more at 8x).

 

[CarstenS]

The big question is not why is it so good, it's why Tahiti sucks so much.

Tahiti does not suck. Why do people always say that it sucks? I mean, ok, it could be cheaper, but that's really all that kept me from buying one in the last two months.

 

[tviceman]

The big question is not why is it so good, it's why Tahiti sucks so much.

This is AMD's most compute-focused GPU yet. In order to compete with Nvidia in this regard, they had to take a hit somehow. Nvidia has been doing compute-focused architectures for 3 generations now, whereas Cayman was AMD's first, but not fully-focused, attempt. Nvidia has more experience in this regard.

 

[DarthShader]

On one of the slides they mention "8 Geometry Units". But I see only 4 GPC's?

The Polymorph 2.0 Engines are the "Geometry Units".

Actually, there is a difference between 4x and 8x MSAA. As AMD ROPs can't handle 8xMSAA at the speed NV ROPs are capable of, the ROP load basically doubles at 8x for AMD cards (but not for nV). The memory bandwidth plays a bigger role at 4x in case of AMD (because it is less limited by the ROP throughput itself), at 8x it doesn't help that much. So at 4x the bandwidth can have a potentially larger effect than at 8x (while in nV's case a higher bandwidth should pay off more at 8x).

And all the game benchmarks given are a 8x AA. Can we infer 4xAA and noAA performance from that?

 

[CarstenS]

Actually, there is a difference between 4x and 8x MSAA. As AMD ROPs can't handle 8xMSAA at the speed NV ROPs are capable of, the ROP load basically doubles at 8x for AMD cards (but not for nV). The memory bandwidth plays a bigger role at 4x in case of AMD (because it is less limited by the ROP throughput itself), at 8x it doesn't help that much. So at 4x the bandwidth can have a potentially larger effect than at 8x (while in nV's case a higher bandwidth should pay off more at 8x).

Actually, what I'm seeing is different. Comparing the latest High-End cards that both have 384 bit of memory interface and where Nvidia has the advantage of 50% more ROPs, AMD looses less performance when going from 4x MSAA to 8x MSAA.

 

[Man from Atlantis]

can anyone capture the die shot without watermark

New Adaptive VSync technology


NVIDIA, on the other hand, made ​​for the actual operation of fluency of the game is optimized and to promote the new Adaptive VSync VSync setting are gamers, open the main purpose is to reduce the graphics card output image is generated in different fps too fast or too slow situations, and the output frame rate is locked in with the mainstream display screen 60fps. However, the image of the actual output of the graphics card will still be due to different scenes within the game for differences between the actual output fps lower than 60fps fps, will be directly reduced to 30 below 30fps is reduced to 20fps.


In fact, such cases are quite common, and convert 60fps to 30fps of the screen will significantly slow the beating, which greatly affect the fluency of the game, the new Adaptive VSync just solve the problem, when the Adaptive VSync enabled, the system automatically detection of the actual operation fps than 60 VSync will open a locked 60fps output frames, such as less than 60 will be automatically VSync turn off and running to the actual output frames to reduce all of a sudden dropped to 30fps from 60fps is the emergence of slow beat, can be very useful.


Practical function open: new antialiasing technology TXAA


In addition to the graphics performance, NVIDIA also for the needs of gamers, add a number of new and useful features for a new generation of the GeForce GTX 680, released a new anti-aliasing technology TXAA, a new generation of hardware rendering anti-aliasing technology, actually belongs to the old MSAA's enhanced version provides better anti-aliasing I can reduce the open antialiasing fluency decline, for the time being belonging to the GeForce GTX 680 unique anti-aliasing, but NVIDIA said that the above will be open at a later time to the GeForce GTX 500 or Model.


TXAA temporarily can be divided into two levels: TXAA 1 and TXAA, the former costs only 2x the MSAA similar, but the MSAA is better than the 8x anti-aliasing; As for TXAA 2 effect is more outstanding, but the cost only similar with 4x MSAA. TXAA game itself support to be open to use, support TXAA variety of games will be held during the year stage, although present, failed to test its real performance, but based on the NVIDIA official photos can be seen, antialiasing does, it is worth looking forward to.


In addition to the new TXAA Antialiasing, NVIDIA is also an existing graphics card users to bring good news, the introduction of the original few games before the corresponding FXAA to the NVIDIA Control Panel, users can easily turn the feature on. lead to better anti-aliasing for different games, and enhance the quality of the game screen.

ok i found it

Spoiler

 

[A1xLLcqAgt0qc2RyMz0y]

On one of the slides they mention "8 Geometry Units". But I see only 4 GPC's?

I believe the "Geometry Units" are called "PolyMorph Engine 2.0" and they are 8 of them one per SMX.

http://imgur.com/a/aQmuA#EFjJN

http://imgur.com/a/aQmuA#EXqKU

 

[A1xLLcqAgt0qc2RyMz0y]

No mention of cache sizes?

The L1 cache size is 512 KB. 64KB per SMX. 8 SMX's.

The Uniform, Texture and Instruction Cache sizes are unknown.

http://imgur.com/a/aQmuA#EXqKU

Now I wonder why AMD's boards are longer when Nvidia has faster memory and yet has shorter boards.

Wouldn't the wider memory bus size and more memory chips result in longer boards?

 

[fellix]

Cleaned for thou geekish pleasure :

The command processor in the middle is strangely smeared. Looks like NV doesn't want anyone to peek in the gigathread factory.

 

[jaredpace]

L2 is 768

 

[A1xLLcqAgt0qc2RyMz0y]

L2 is 768

Link please?

 

[Kaotik]

What's up with hiding the BF3 7970 performance? The bar is cut short and number hidden

 

[fellix]

What's up with hiding the BF3 7970 performance? The bar is cut short and number hidden

More like a sloppy graph editing.

 

[A1xLLcqAgt0qc2RyMz0y]

Cleaned for thou geekish pleasure :

The command processor in the middle is strangely smeared. Looks like NV doesn't want anyone to peek in the gigathread factory.

I remember a shot of this naked kepler die that had the individual sections highlighted with labels. Can someone provide a link?

 

[Kaotik]

More like a sloppy graph editing.

Whatever it is, it's hiding 7970 performance completely on purpose, it's not being done on any other graph, and the box entitling game isn't showing on them either.