GF FX 5700 Specifications & Date Out ! But ...

[David G.]

Sooo ... what's this ?

Isn't this suposed to be an improved mid-end solution ?

I mean , improved graphics performance for us users and improved yields for nVIDIA ?

OK , so now , this VPU will be made by IBM , so they'll pay only the good chips , will have better yields as IBM's tech is better than TSMC's , and will be cheaper to built as 0.13 micron tech allows more chips per wafer .

Everything is good but , what I want to know is , will there be any changes from the 5600 architecture or there will be only highyer clocks ?

How overclockble will it be ? Will it be a 600 Mhz for VPU wonder like ATi's 9600 or will be just as close to the limit as the 5800 was when launched ?

Isn't there going to be a huge performance gap between the mid-end line and the NV40 line which is said to have more like 300 million transistors ?

http://www.clubic.com , http://www.3dchips-fr.com

 

[gkar1]

....the NV40 line which is said to have more like 300 million transistors ?

This is the first time anyone's ever heard of this.

EDIT: I see now, you're referring to the blurb from the inq

 

[Blitzkrieg]

Is that graph from nvidia? If it is funny how they rate performance using 3dmark03

 

[KimB]

Looks like the 5700 will essentially be half a GeForce FX 5900, but at a higher clock speed. Should be nVidia's fall product, meant for the performance crowd. That roadmap points to nVidia not releasing another product besides this 5700 before the NV40 (it looks like nVidia is going to try to release this one by December).

 

[Richthofen]

Sooo ... what's this ?

Isn't this suposed to be an improved mid-end solution ?

I mean , improved graphics performance for us users and improved yields for nVIDIA ?

OK , so now , this VPU will be made by IBM , so they'll pay only the good chips , will have better yields as IBM's tech is better than TSMC's , and will be cheaper to built as 0.13 micron tech allows more chips per wafer .

Well Nvidia also pays only for the good dies @TSMC.
There is nothing special here with IBM.
Maybe IBM is a littlebit cheaper but i doubt that.
What i think is that NV36 is half of NV35. So there are some small advantages over NV31.
The yield @IBMs 0.13 is probably better. Nvidia also has some 0.13 experience now. So i think NV36 is yielding better than NV31 overall.
That means cheaper wafers.
Additionally NV36 will be clocked higher.
So i would say its a refresh of the mainstream sollutions and there will be different versions.
The Ultra version will be the upper mainstream.

I don't think Nvidia will stop NV31 production because of NV36 in the next couple of months. Why?
Because it's safe to say that there still is not enough 0.13 capacity overall counting TSMC and IBM to satisfy the 0.13 micron DX9 demand.
I thinks we still will see both products.

 

[Dave H]

Looks like the 5700 will essentially be half a GeForce FX 5900

Err, don't you mean a 5900 without the second TMU per pipe? You don't mean to suggest it's a 2x2, do you?

 

[Psikotiko]

Where's NV38?

 

[Arun]

NV30 and NV35 are 4x2.
NV31 and NV36 are 2x2.

NV30 was a mess to figure out based on rumors.
NV31 and NV35 were easy to figure out based on rumors.

NV36 is impossible to figure out based on rumors, at least for now.

NV38 is gonna be released VERY soon, but it isn't even on all nVidia internal roadmaps yet, and on the only ones where it is, the dates are complete BS


Uttar

 

[reznor02]

now thats !2! cards surrounded by confusion

the nv36 is impossible to decypher WHEN and SPECS

now the nv38 has become a confusing IF?

i think we need some solid proof on these, this picture seems overrated and just too easy to mislead.

 

[Joe DeFuria]

NV38 is gonna be released VERY soon, but it isn't even on all nVidia internal roadmaps yet, and on the only ones where it is, the dates are complete BS

Assuming the NV38 is coming to fruition, that can really only mean that it is a reactionay product...like the GeForce4 4200. I can't see it being anything else than the NV35 with different clocks.

The question is, what is it "reactionary" to? Either

1) Presence of R360
2) Lack of forthcoming NV40

or both?

 

[DegustatoR]

Assuming the NV38 is coming to fruition, that can really only mean that it is a reactionay product...like the GeForce4 4200.

What was so reactional in Ti4200? It was the lowest chip in GF4 lineup, kinda a way of selling all those NV25s that didn't run on 275 and 300MHz.

I can't see it being anything else than the NV35 with different clocks.

You're most probably right here. I expect something like 500-550MHz core for NV38, somewhat faster DDR1 and that would be all "new features".

The question is, what is it "reactionary" to? Either

1) Presence of R360
2) Lack of forthcoming NV40

or both?

Presence of R360? Where? And i thought it was renamed R380 some time ago...

Well, NV38 started as R380 counterpart, to hold the parity between top chips. As for NV40, NV _is_ on a schedule with it for now, but even they don't expect broad availability till Q1'04 (that means March i'd say). So NV38 will hold the ground till NV40 arrives in any way. Does it make NV38 "reactionary"? Yes -- to R380. No -- to NV40.

BTW, lets not forget: _R380_ (R360 back then) was truly reactionary to NV35 introduction. So it was ATI who started this "+20MHz core in three months" race.

 

[Bolloxoid]

What was so reactional in Ti4200? It was the lowest chip in GF4 lineup, kinda a way of selling all those NV25s that didn't run on 275 and 300MHz.

Nvidia didn't originally plan to release the Ti 4200 at all. According to their plans, we would have been buying MX460s at Ti 4200's price. Luckily the competition forced them to change their thinking.

 

[micron]

Nvidia didn't originally plan to release the Ti 4200 at all. According to their plans, we would have been buying MX460s at Ti 4200's price. Luckily the competition forced them to change their thinking.

This was my understanding also....

 

[WaltC]

Assuming the NV38 is coming to fruition, that can really only mean that it is a reactionay product...like the GeForce4 4200. I can't see it being anything else than the NV35 with different clocks.

The question is, what is it "reactionary" to? Either

1) Presence of R360
2) Lack of forthcoming NV40

or both?

Is it possible "nv37" is actually what had been assumed to be "nV38"...?

Also, I think 1 & 2 above are just two sides of the same coin. IE, lack of nv40 wouldn't be a concern to nVidia if there would be no R360 to contend with (assuming there is an R360, of course.) In fact, many things would be much different in the nVidia product lineup today if there had been no R3xx...

 

[Dave Baumann]

Nvidia didn't originally plan to release the Ti 4200 at all. According to their plans, we would have been buying MX460s at Ti 4200's price. Luckily the competition forced them to change their thinking.

This was my understanding also....

4200 was announced from day 1 for GeForce 4:

http://www.beyond3d.com/articles/gf4launch/index7.php

It was always scheduled to be later than the rest. With these parts being the high volume ones you really want to understand where your yields are going to be to stick in the flag in the ground for clock speeds, so its not surprising to wait a couple of months to get a good representative sample - same reason 9800 (np) took so long to be finalised.