News: The 7800gtx isn't really the G70....

It IS a new revision! These are *not* old high-binned chips, these are a limited run, unknown-yield, high clocked custom production with both silicon and logic revisions. Yes, there is a slightly more robust PCB, faster memory, and a better cooler and slightly more voltage, but these are definitely not the same cores people have in their GTX-256s.

BTW--this is to EVERYBODY--silicone is for boobies, silicon is for tech.

I believe that Vapor is wrong on this one. Nvidia does have great yields on G70. And as Maxxx mentioned before for 512MB version there are cherry picked ones - collected from the first day of it`s production begining. Just check OC results on ORB - there are many results with core going all the way to 700MHz.
512MB version IS just faster GDDR and improoved PCB for it`s - because this new 1.1ns chips have different pins quantity.

Quote:

---

Originally Posted by Vapor

BTW--this is to EVERYBODY--silicone is for boobies, silicon is for tech.

---

LMAO :rofl: can I quote you? wanna use this in my sig....way too funny!

read anadtech or bit-tech.net reviews dont remember which says tsmc re designed a little bit the core to boost the performance besides they are using 1.1ns ram and 512mb perhaps a better pcb too so it is safe to say is not the same as 7800gtx is a slighty better card or better a new revision


/conspiracy thinking:

perhaps these are the real g70 (what they had in mind in early stages) and the 7800gtx 256 is the nv4x (dont remember the exact number) as ati took so long to release their x1xxx series nvidia used the Nv4x and saved the g70 for later is the best i can think of

maybe the engineers just got lazy and didn't feel like ridding all the traces of "nv" from the product? this seems like it could be some sort of typical legacy traces of product due to developer laziness to me... maybe.. the architechture of these 110nm chips is for sure different from the old series of chips... so I have a hard time thinking this is a scandal of some sort...

EDIT: after all, they can call it whatever they want. but none the less, a damn cool find. :)

Sounds too good to be true.. I will wait untill I see the ATi killer..

Quote:

---

Originally Posted by Cooper

I believe that Vapor is wrong on this one. Nvidia does have great yields on G70. And as Maxxx mentioned before for 512MB version there are cherry picked ones - collected from the first day of it`s production begining. Just check OC results on ORB - there are many results with core going all the way to 700MHz.
512MB version IS just faster GDDR and improoved PCB for it`s - because this new 1.1ns chips have different pins quantity.

---

Yes, they do have great yields....but have you seen a single core in a 256 that does 550MHz in the ROP/shader domain at 1.45V on aircooling, artifact free? Okay, there's a SMALL handful that can....but how about 600Mhz? Nope. How about at 1.55V and aircooling? Nope. The fact that these do it SO easily and that the 256s can't do it at all and there's no grey area between should point to the fact that these are custom-run, unknown-yield, high-clocking parts.

I always thought that NV47 = G70 in terms of naming?

It kind of makes sense: NV40 is nvidias new beginning, and G7 is the 7th generation, the 0 being the first revision in that generation.

:confused:

Geez...can this thread die already? it's just some nutty theory... :stick:

Quote:

---

Originally Posted by LOE

Stick the "old rev" G70 chip into a new design PCB and a double performance cooler and you might see 550 Mhz@1.45V.. the PCB design itself can be very limiting in terms of stability. Like the difference btwn ordinary NF4 boards that do ~250FSB and DFI boards that to 330 FSB with the VERY SAME CHIP... that's ~30% boost from a better design - funny that 130% of the 256MB GTX equals aprox 560 MHZ :)

I think it's like AMD's low power CPU - they are picked, there's nothing special bout them, they just got better in the process...

http://www.xbitlabs.com/articles/vid...0gtx512_3.html

I read A2 rev on both ships, one is N (normal?!?) the other is U (Ultra), there is no difference in TNT2 pro and Ultra, no diff in GF2 NU and U, and so goes for all generations. Everything points to a picked chip with better designed card. I think it's not such a big secret and if it was a different chip nvidia would say it is, and they didn't. That's logic, nothing is for sure... Now, the process might've got better with time, first chips couldn't hit such high freqs, but process can be improved without modifying the core itself... both cores say A2

I don't know who is owned, but somebody just got pwned :banana:

---

The new PCB isn't that much better, it's largely tweaked for the more demanding memory and has a few small tweaks for 1.45V with a much higher current requirement of the new cores. It's actually a VERY similar PCB overall, with only a few bolsters in select parts. It's not a miracle PCB that suddenly allows 520MHz cores to run at 580+MHz.

In addition, there's another interesting thing about the new GTX 512s that's very telling.....the mem controller can go to much lower temps....a PCB can't do that. These are absolutely different cores.

TSMC can produce a high-yield part for 500MHz (the GTX 256) where all the cores bust at 525MHz and all work at at least 400MHz or they can tweak it slightly (as well as getting some tweaks from nV--mem controller improvements for dealing with the cold and dealing with the higher speed memory) and produce a medium/low-yield part at 600MHz at a price premium. Going for the medium/low-yield part for 600MHz is a better option for nV with the 512 because they can easily recoup the price by charging $200 more and get a much easier supply of the fast parts (and they can get the tweaks in that they need).

BTW, the new cooler isn't that great either--the fan spins too slowly for it to be a great cooler. It's no doubt better than the old one but as you can see from FCG's experience with it, it's not THAT great--he couldn't even get close to 600MHz with 1.55V with it, IIRC.