What market is there for the NV28?

[Dave Baumann]

Xabre is a fairly low end part, with no hardware vertex processor and those tests were done on an application that doesn't really stress the 4X bus. I would wait to see the differences on more high powered solutions (NV28/R300/NV30) before signalling the usfulness (or not) of AGP8x over AGP4x.

 

[alexsok]

Xabre is a fairly low end part, with no hardware vertex processor and those tests were done on an application that doesn't really stress the 4X bus. I would wait to see the differences on more high powered solutions (NV28/R300/NV30) before signalling the usfulness (or not) of AGP8x over AGP4x.

Well sure, I understand that, but the weird thing is that the authors of the article were stating that AGP 8X is indeed not needed as the perfomance boost is none (i think it was on digit-life).

But we should wait for R300/NV30.

 

[Ailuros]

Dave,

Theoretically at least: isn't there a chance that a Xabre might show some increase in performance between 4x and 8x under heavy vertex shader conditions, since they'll get emulated by the cpu?

Sorry for the OT by the way.

 

[alexsok]

Yep, it was on Digit-Life:

http://www.digit-life.com/articles/triplexxabre/index.html

Hmm...

Can't find any comments about AGP 8X there...

Maybe that English article was changed in content from the original Russiah version I read some time ago...

 

[Richthofen]

The NV28 will be nothing more than a little faster GF4Ti4600 with AGP 8x.
I think the intention is to keep ASP's up.

The Radeon9700 ist not that much of a problem but the Radeon9000 is because of the name and the price.
But Nvidia already reacted and that was the right thing for them.
Just put all products one class lower concerning price and launch a new top line to keep ASPs up.
Concerning the NV18 i think that Nvidia is going to clear their GF4MX product line. The NV18 will mark the end of the NV17 i think. Maybe they are going to move some MX versions away from market. They do not need 3 or 4 versions below the 100$ price point. I think the MX460 and 440 will meet end of life and the only one remaining will be the MX420 for mass OEM production and of course the new NV18.

This will hit ATI for sure because they want to try earning money with the Radeon9000 just to balance the costs they will have with their top line. 0.15 and 110 Mio transistors doesn't sound cheap at all.
But with a Ti4200 below 150$ in retail stores will hit ATI a lot.
I think that is the only intention Nvidia has with that.
Normal behaviour for market leaders. Intel does the same all the time.
If you replace your top-line produtcs non-stop you will keep ASP's up.
If not they will get into a free fall at the time the competitor reacts.

 

[Dave Baumann]

Theoretically at least: isn't there a chance that a Xabre might show some increase in performance between 4x and 8x under heavy vertex shader conditions, since they'll get emulated by the cpu?

thats the point isn't it - you're likely to hit a CPU limitation before you hit the limits of the AGP bus. Remember as well, if you aren't using static T&L buffers then untransformed vertex data (i.e. raw vertices being passed to the card to process) can be greater in size than tranformed vertices (i.e. stuff that processed on the CPU) so you are more likely to see the differences on cards that have hardware vertex shading. Then of course you are also more likely to notice the difference when textures overspill the amount of RAM locally to the card - double the AGP bandwidth will help plenty there!

 

[Randell]

hmm I cant see many people outside of OEM's buying an NV28 untill more is known about the Radeon9500 especially what makes it cheaper than the 9700 and therefore its relative price/performance ratio.

 

[Dave Baumann]

I don't think NV28 will be competing against R9000 - R9000 is a lower-end part that appears to be pitech at the higher MX's IMO. NV28 would probably sit inbetween MX and NV30 as the midrange card - this will hurt ATi becuase their current midrange card is Rageon 8500 until such a point that thy can release 9500.

 

[Geeforcer]

hmm I cant see many people outside of OEM's buying an NV28 untill more is known about the Radeon9500 especially what makes it cheaper than the 9700 and therefore its relative price/performance ratio.

IMO, its should be the other way around: "I CAN see many people outside of OEM's buying an NV28 UNTILL more is known about the Radeon9500..."

 

[GraphixViolence]

Now that I think about it, it looks like the whole reason ATI quasi-announced the Radeon 9500 was to hold a trump card against the NV28. We don't know yet how it will stack up performance-wise, but we do know it will have the 9700's anti-aliasing, pixel shaders, etc. - so it should definitely have the advantage from a feature standpoint. Sounds like this will be an interesting match-up in the next few months.

 

[Randell]

hmm I cant see many people outside of OEM's buying an NV28 untill more is known about the Radeon9500 especially what makes it cheaper than the 9700 and therefore its relative price/performance ratio.

IMO, its should be the other way around: "I CAN see many people outside of OEM's buying an NV28 UNTILL more is known about the Radeon9500..."

well they are fools then (IMHO) We will know more about the 9500 in a few months. Its the same concept as as all those who held off buying an 8500 or Gf3Ti500 until the Gf4Ti4200 was available. If there is going to be a big gap in time though between NV28 availablility and 9500 availability then this will hurt ATI for sure.

 

[Tahir2]

Just FYI (particularly alexsok) the AGP4X to AGP8X benchmark comparison was at www.ocworkbench.com

They only ran one benchmark and concluded that AGP8X for the Xabre grpahics was not its major selling point - performance wise anyway.