RV350 Die size clue?

[Sabastian]

to help in B3D's quest to find the die size I've found out that the M10 has 60 million transistors.

That is much smaller then I suspected. AFAIK RV350 is supposed to be used in the new mobile part M10. That ought to be giving them some pretty good yeilds on the .13um process waffers. They ought to be able to price that very competitively. Now it is just a matter of seeing them in action.

 

[mboeller]

to help in B3D's quest to find the die size I've found out that the M10 has 60 million transistors.

WOW, small; the NV31 has 80 million transistors (as everyone knows)

So with 60 x (0,13²/0,15²) = 45 Mio Transistors

So an chip in 0,15µm-tech with 45 Mio Trans. would have the same die-area. WOW thats as much as the NV34 has...surprise surprise. So except for the new 0.13µm-process they should be able to sell the RV350-chip at nearly the same price as Nvidia the NV34-chips; or?

 

[RussSchultz]

To clarify: approximately the same area.


Also:

So except for the new 0.13um process (which includes more expensive wafers, more expensive masks, and reportedly lower yields) they should be able to sell the RV350 chip at nearly the same price.

 

[Dave Baumann]

You can lop a little bit more out of the M10 size as well, as the desktop version won't have the power regulation features and some other stuff.

 

[Joe DeFuria]

So except for the new 0.13um process (which includes more expensive wafers, more expensive masks, and reportedly lower yields) they should be able to sell the RV350 chip at nearly the same price.

Whether or not they can sell RV350 it at the same price NV34 really isn't the point, IMO. It's that they should be able to sell it at a lower price than the NV31, which also has to deal with the same expensive wafers,masks, and lower yields...

 

[Joe DeFuria]

You can lop a little bit more out of the M10 size as well, as the desktop version won't have the power regulation features and some other stuff.

Damnit...What's this OTHER STUFF you keep teasing us with!?

 

[RussSchultz]

So except for the new 0.13um process (which includes more expensive wafers, more expensive masks, and reportedly lower yields) they should be able to sell the RV350 chip at nearly the same price.

Whether or not they can sell RV350 it at the same price NV34 really isn't the point, IMO. It's that they should be able to sell it at a lower price than the NV31, which also has to deal with the same expensive wafers,masks, and lower yields...

Except he said NV34.

 

[Joe DeFuria]

Yes, and I'm just saying that comparing it to the NV34 isn't really the point. And I'm also saying that I agree with your statements on other issues related to fab cost than simply die size.

 

[Sabastian]

You can lop a little bit more out of the M10 size as well, as the desktop version won't have the power regulation features and some other stuff.

Jesus, how many transistors are on the thing? The little bugger is going to be running at 400mhz!! Just what "other stuff" are you refering too? I never thought it would be as small as 60million transistors but you are saying approx how many more less?

 

[RussSchultz]

It seems the RV350 will be approximately the same cost to manufacturer (from the info we have), though likely somewhat cheaper, depending on the packaging. (Flip chip is kinda pricy; don't know what NVIDIA uses as packaging)

 

[mboeller]

Yes, and I'm just saying that comparing it to the NV34 isn't really the point. And I'm also saying that I agree with your statements on other issues related to fab cost than simply die size.

I have compared the RV350 with the NV34, because at the moment Nvidia has an big marketing advantage with the DX9-capable NV34 chip.
But if ATi can produce an cheap version of the R9600 cards maybe with an downclocked RV350 at the same price point then this marketing advantage is gone, and ATi can deliver cards to all mainstream markets ($99 - $199) with one chip instead of two, and so they have an even bigger advantage here.
Nvidia has to maintain drivers for 3 new chips at the moment (NV34, NV31, NV30) and due to the different architectures it is IMHO not possible to do it with one adapted driver only. So with an RV350 to cover all of the mainstream market this could turn out to be in ATi favour.

 

[bdmosky]

Nvidia has to maintain drivers for 3 new chips at the moment (NV34, NV31, NV30) and due to the different architectures it is IMHO not possible to do it with one adapted driver only.

Hmm... so you're saying this "unified driver" that they've had going on now for many, many years won't work because they have 3 chips? Give me a break.

 

[rwolf]

Yes, and I'm just saying that comparing it to the NV34 isn't really the point. And I'm also saying that I agree with your statements on other issues related to fab cost than simply die size.

I have compared the RV350 with the NV34, because at the moment Nvidia has an big marketing advantage with the DX9-capable NV34 chip.
But if ATi can produce an cheap version of the R9600 cards maybe with an downclocked RV350 at the same price point then this marketing advantage is gone, and ATi can deliver cards to all mainstream markets ($99 - $199) with one chip instead of two, and so they have an even bigger advantage here.
Nvidia has to maintain drivers for 3 new chips at the moment (NV34, NV31, NV30) and due to the different architectures it is IMHO not possible to do it with one adapted driver only. So with an RV350 to cover all of the mainstream market this could turn out to be in ATi favour.

I think in their GDC release they said $100 and up. The 9200 is for below $100. I have also heard suggested $150.00.