Possible reason for Nv30 delay?

[jjayb]

All this talk about nv30 being delayed because of tsmc .13 problems. Just noticed this in the beyond 3d interview:

The developer documentation was quoting vertex throughput rates of 1.5 times GeForce4's throughput on a clock for clock basis, which would equate to 3 Vertex Shaders.

The 1.5 number is old, because we actually have 3 times the vertex performance of Ti4600, at 500MHz.

Maybe the delay is actually due to Ati surprising them with the r300? After seeing the speeds of the r300 they found it necessary to up the clock to 500mhz to compete? Yeilds too low to make enough 500mhz parts in time? What was it originally supposed to be clocked at if it wasn't supposed to be 500mhz?

 

[Ichneumon]

My guess would be somewhere around 350-400MHz... or whatever would have allowed them to build boards without needing that massive cooler.

 

[ERP]

Well from that quote you'd have to assume that the original planned clock speed was 250MHz. Since chips are targetted to run at a specific speed, I find it pretty hard to believe that Nvidia doubled the final speed in response to R300.

 

[Johnathan256]

I am anxious to know what speed the NV35 will run at upon it release. Nvidia can slap on a 256-bit memory interface but clocking that chip any higher could be a challenge. They might have to incorperate a factory peltier element.

 

[Joe DeFuria]

Well, with the current very large and complex NV30 PCB design (if that is the actual reference design, and not a pre-alpha hack), I'd say "slapping on" a 256 bit memory interface would also be a challenge!

 

[ben6]

The original numbers I saw back in June were for 400MHz

 

[megadrive0088]

I am anxious to know what speed the NV35 will run at upon it release. Nvidia can slap on a 256-bit memory interface

oh hell yeah. 8)

 

[RussSchultz]

Well, with the current very large and complex NV30 PCB design (if that is the actual reference design, and not a pre-alpha hack), I'd say "slapping on" a 256 bit memory interface would also be a challenge!

What makes you say it is "very large and complex pcb design" that might be a "pre-alpha hack"?

It looks like a clean board design to me. It's larger to make room for that heatsink thing, but otherwise looks clean.

It has less layers than a ATI board (I believe the ATI board has 8, NV30 has 6).

 

[Nappe1]

It has less layers than a ATI board (I believe the ATI board has 8, NV30 has 6).

umm... I might be wrong but afaik:
Ti4200 has 6-layers, Ti4400 / Ti4600 / Parhelia has 8-layers and Radeon 9700Pro has 10-layers...

and it it could sound feasible NV30 having 8-layers.

can anyone confirm these? I want to know this too...

 

[RussSchultz]

You're probably more right than I am.

I still don't see how that makes the board a hack, or messy, or a poor effort.

 

[arjan de lumens]

I thought the NV30 board, as pictured, had an inordinately clean, if a bit large, design. Perhaps necessary to get the 1GHz memory bus working.

 

[BRiT]

AFAIK, the ATI R300 has 8 layer pcb. Only the early alpha boards had 10 layers.

--|BRiT|

 

[Nappe1]

AFAIK, the ATI R300 has 8 layer pcb. Only the early alpha boards had 10 layers.

--|BRiT|

quite possible...
I haven't seen real numbers since the launch.

RussSchultz: okay, it was a kind of half and a half. now both are happy.

 

[BRiT]

Here's the most current link I could find confirming ATI R300 boards using 8 layer PCBs:
http://www.beyond3d.com/forum/viewtopic.php?p=55955#55955

FYI, just to clear that up - talking with Brian Skelton from Sapphire (who are the company that manufactures most of the ATI boards for ATI and other vendors) he stated that the production R300 boards are 8 layer boards, not 10.

--|BRiT|

 

[jvd]

the delays were because nvidia messed up... messed up royaly ... hows that ? end of discusion ?

 

[Sabastian]

AFAIK, the ATI R300 has 8 layer pcb. Only the early alpha boards had 10 layers.

--|BRiT|

Ah ha! I knew I had picked up the 10 layer PCB thingy somewhere. (Sorry about that I just had to get it out of my system. )