First photographs from NV40

Just a side note: All pics are indeed from the same photograph. The author just selected different regions and gave them different res.
 
991060 said:
pakotlar said:
this article states that the nv40 has 220 million transistors. uhoh :oops:

not a surprise given the massive die size :LOL:

What is going on here. uttar, aliruos, vegetto? I thought 175mil was the final count?

Btw, it's not like 175 mil won't make for a huge die. p4 c's on the same process, and sporting 55mil takes up 119mm^2 if memory serves me. These things are going to be friggin hot.

On a side note, while looking at the nv40 board pictures, I was particularly struck by the dual molex connectors. A friend of mine once told me that it was speculated that had it not been for AMD we would still be on 1st generation p4's running at 1.7ghz. The dual molex connectors reminded me of the power of competition. Nvidia and ATI are pushing themselves to the limits of modern engineering. Sigh, this is a great day in age. :D
 
Can't the 170 and 210 million transistors be the same chip?
I mean, it has been said that r420 is 160 but that 'wasn't comparable with NV' because ATI uses 'number of active logic transistors' while NV uses 'total transistors' to get 210.
Couldn't 170mil be the 'active logic' count of the 210mil chip?
 
arrrse said:
Can't the 170 and 210 million transistors be the same chip?
I mean, it has been said that r420 is 160 but that 'wasn't comparable with NV' because ATI uses 'number of active logic transistors' while NV uses 'total transistors' to get 210.
Couldn't 170mil be the 'active logic' count of the 210mil chip?
I was thinking the same thing. a couple more weeks and we will get a better picture of whats true/false.

later,
epic
 
I believe 205-210M was the final count.
A 175M part definitively did tape-out; whether it was ever meant for retail avaibility, I've got no clue though. And there also were designs of a 150M NV40; again, whether this part was just experimental or not, I've got no clue.

Uttar
 
Uttar said:
I believe 205-210M was the final count.
A 175M part definitively did tape-out; whether it was ever meant for retail avaibility, I've got no clue though. And there also were designs of a 150M NV40; again, whether this part was just experimental or not, I've got no clue.

Uttar

Then what are the differences between these revisions besides the trans count?
 
that card doesn't have the power reg circuit to feed a 100+W card. it is 5 times smaller then the engineering nv35 board. if it really is nv40 it should be one of the 300 MHz out earlier this year, but it looks like an almost final design.
 
Triple po... Hey, Mindriot, i was trying to delete them.

Meh, at least you're getting a picture

Old (above) and current (below) NV35 reference cards

5900_05.jpg
 
vb said:
that card doesn't have the power reg circuit to feed a 100+W card.
Of course, we can't see the back of the card. Maybe all the linear regs (and/or DCDC converters) are there, instead of on the front.

But definately, that little linear reg isn't going to handle the current for that card.
 
Back
Top