ExtremeTech NVIDIA & ATI Q&A

[Dave Baumann]

ExtremeTech are canvasing for technical question suggestions that they will submit to both NVIDIA's David Kirk and ATI's Bob Drebin (thats Bob, not Frank) for answers in two articles.

If you can think of something tasty plonk your Q's here:

http://www.extremetech.com/article2/0,1558,1742163,00.asp

 

[Wunderchu]

thanx for the heads up

 

[Trawler]

Nice beaver!

 

[Reverend]

[edit] whoops, I thought Dave posted in the wrong forum! sorry!

 

[Unknown Soldier]

Fek .. what did I miss

US

 

[mcsven]

Nice beaver!

Thanks, I've just had it stuffed.

 

[Geo]

The NV/David Kirk one is up:

http://www.extremetech.com/article2/0,1558,1744817,00.asp

I found this observation startling:

I think that High Dynamic Range Lighting is going to be the single most significant change in the visual quality over the next couple of years. It's almost as big as shading.

Could that be right? He supports it with some observations on HDR display tech penetrating the mass market in the near term.

 

[Trawler]

This may be a little telling...

It's not clear to me that an architecture for a good, efficient, and fast vertex shader is the same as the architecture for a good and fast pixel shader. A pixel shader would need far, far more texture math performance and read bandwidth than an optimized vertex shader. So, if you used that pixel shader to do vertex shading, most of the hardware would be idle, most of the time. Which is better—a lean and mean optimized vertex shader and a lean and mean optimized pixel shader or two less-efficient hybrid shaders? There is an old saying: "Jack of all trades, master of none."

 

[Geo]

This may be a little telling...

It's not clear to me that an architecture for a good, efficient, and fast vertex shader is the same as the architecture for a good and fast pixel shader. A pixel shader would need far, far more texture math performance and read bandwidth than an optimized vertex shader. So, if you used that pixel shader to do vertex shading, most of the hardware would be idle, most of the time. Which is better—a lean and mean optimized vertex shader and a lean and mean optimized pixel shader or two less-efficient hybrid shaders? There is an old saying: "Jack of all trades, master of none."

Well, yes, we've been hearing murmurings of this being their position, right? A little before or after he also makes the point that unifying the programming interface is a differnt matter, and a much more obvious win, which also supports what I think you are pointing at, and what we've heard others say they understand the NV position to be. Nice to have Kirk on the record. "It's not clear to me" is apparently engineer-speak for "As far as I can tell it is a rotten idea --show me some facts to try to convince me otherwise."

 

[Reverend]

"Jack of all trades, master of none"

Wow, I don't think he realized he just described the different philosophies between that of his company and that of ATI's...

 

[ChrisRay]

"Jack of all trades, master of none"

Wow, I don't think he realized he just described the different philosophies between that of his company and that of ATI's...

Care to clarify? Just curious what your implying

 

[Geo]

"Jack of all trades, master of none"

Wow, I don't think he realized he just described the different philosophies between that of his company and that of ATI's...

Meaning NV would rather push the feature envelope whether the performance is there or not ("jack of all trades"), while ATI waits (see the famous Richard Huddy note in the powerpoint) until it can have good enuf performance ("master") to be "useful" before incorporating it?

 

[Fodder]

Meaning ...

I see it the other way round. R300 is a very general, solid, standard sort of architecture. NV3x and NV4x rely more on specific features like ultrashadow and SM3.0.

So, it's Jack of All Games, or Master of Doom.

 

[Trawler]

In relation to Rev's comment, my first thought was FP24 vs FP16/FP32...

 

[rwolf]

I would ask ATI about their work with fast-14 technology...

 

[Tim Murray]

Fast-14 is apparently ATI's version of Gigapixel.

 

[rwolf]

Fast-14 is apparently ATI's version of Gigapixel.

I don't understand.

Fast-14 allows you to mix CMOS and dynamic logic technology so you can have faster switching transistors with lower power consumptions yet the ease of designing cmos transistors.

What does that have to do with Gigapixel (I am thinking Gigapixel bought by 3DFX)?

 

[Fodder]

I don't understand.

Obviously NVIDIA's next generation part will use Gigapixel technology to raise efficiency to 110%, while ATI will effortlessly be able to churn out GHz chips thanks to Fast-14.

 

[rwolf]

I don't understand.

Obviously NVIDIA's next generation part will use Gigapixel technology to raise efficiency to 110%, while ATI will effortlessly be able to churn out GHz chips thanks to Fast-14.

Tile based rendering?

 

[Bjorn]

This may be a little telling...

It's not clear to me that an architecture for a good, efficient, and fast vertex shader is the same as the architecture for a good and fast pixel shader. A pixel shader would need far, far more texture math performance and read bandwidth than an optimized vertex shader. So, if you used that pixel shader to do vertex shading, most of the hardware would be idle, most of the time. Which is better—a lean and mean optimized vertex shader and a lean and mean optimized pixel shader or two less-efficient hybrid shaders? There is an old saying: "Jack of all trades, master of none."

It'll be interesting to see if Ati really goes the whole way with their "unified" architecture and that there won't be any specific vertex or pixel shader units, just shader units. Cause it definitely seems that Nvidia won't go down that road.

And considering that these architectures will be used in the next gen consoles, well, there will be a VERY interesting battle to say the least.