NV48 is NV40 refresh?

[DSC]

http://www.anandtech.com/showdoc.html?i=2068

We've heard two different things from manufacturers. Some manufacturers have told us that NV45 will be available in August, while others have basically said that NV45 won't be out this year, instead we will see NV48 arrive as an AGP refresh to NV40 by Christmas.

What will most likely happen is that NVIDIA will determine the fate of NV45 by the acceptance (or lack thereof) of PCI Express this year. If the acceptance is low enough, NV45 will be pushed out to next year, otherwise we may see it in Q3/Q4. Regardless of what happens, it does seem like NV48 will be NVIDIA's fall refresh product, and that will be an AGP solution

[lIqUID]

aha I know what Ill be getting myself for Xmas LOL.

[Morrow]

Well considering that the nv45 is just a nv40 with PCI-Express support, the nv45 will be out at the latest next month. The nv48 refresh part will probably ship sometime in Q4 of this year.

[Caliche]

geez, the annoucement of another "new" graphics card by Nvidia, but they can't even get their current product to shelves. ROFL.


Caliche

[Morrow]

nope, nvidia has not made any announcement concerning a new graphiccard. This is all rumours, speculation and guestimates.

[GigaDrive]

whatever the refresh of NV40 is, (not the PCI Express version of NV40 but the true refresh) be it NV45 or NV48, it'll be out this fall, or at the latest, early spring. then we should see NV50 in the fall of 2005.

[nutball]

whatever the refresh of NV40 is, (not the PCI Express version of NV40 but the true refresh) be it NV45 or NV48, it'll be out this fall, or at the latest, early spring. then we should see NV50 in the fall of 2005.

Maybe.

[bkswaney]

We should see the NV48 this fall.
Nvidia will need it to fight back against
ati's new cards.
But all ATI's new cards will be is an X800 with SM3.0 support.
I'm just going to get a new GTOC and be happy till spring.
At that time I'll upgrade my cpu and mobo to PCI-E.
At that time I'll add a PCI-E videocard. :)

[schuey74]

We should see the NV48 this fall.
Nvidia will need it to fight back against
ati's new cards.
But all ATI's new cards will be is an X800 with SM3.0 support.
I'm just going to get a new GTOC and be happy till spring.
At that time I'll upgrade my cpu and mobo to PCI-E.
At that time I'll add a PCI-E videocard. :)
I thought ATI would not have SM 3.0 support until the R500, not the R420 refresh (R480). Has some new info been leaked recently that I missed?

[AthlonXP1800]

I thought ATI would not have SM 3.0 support until the R500, not the R420 refresh (R480). Has some new info been leaked recently that I missed?

R420 refresh to have SM 3.0 are completed nonsense. ATI's first chipset to support SM 3.0 will be R500 series next year.

[Annihlist]

I don't know but it seems to me that there is no time for a refresh card this year. Well there is no need for a refresh with the types of performance gains we are seeing. :)

[coldpower27]

Can't Wait for the new cards,

NV41 Apparently is Geforce 6800 with actually 12 Pipes vs 16-4

NV43 Mainstream card with 8x1 architecture hopefully:)

NV44 Budget Card with Shader Model 3.0 Support :D

NV45 A NV40 with HSI on Chip instead of PCB

NV48 The refresh card like NV30 -> NV35 or R300 -> R350 hopefully :) and not just higher clocks like NV35-> NV38 ot R350 -> R360 :(

[MUYA]

NV41 will not IMHO be a 3 quad processing GPU. Doesn't make sense to have a nv40 3 quad GPU and then come out with a 3 quad gpu named nv41. I don't have anything to base this spculation on but, maybe the nv41 will be the 2 quad GPU, and 43, 1 quad and budget models will be half quad. just my 2 cents

[Viral]

The latest speculation states that NV41's purpose is to make a cheaper to produce 6800. Apparently the best way to do this is to make it natively 12x1 and not just disable 4 pipes.

[nutball]

NV41 will not IMHO be a 3 quad processing GPU. Doesn't make sense to have a nv40 3 quad GPU and then come out with a 3 quad gpu named nv41. I don't have anything to base this spculation on but, maybe the nv41 will be the 2 quad GPU, and 43, 1 quad and budget models will be half quad. just my 2 cents

How so? Surely a 3-quad design will be smaller than the 4-quad design, so more per wafer, so cheaper? Given the rumoured restrictions on the number of wafer starts at IBM/TSMC/etc. at the moment, wouldn't supply in the mid-range be better if you're producing a chip that's designed for it rather than resucitating high-end chips with faulty quads?

They can (persumably) play the same trick with 3-quad design that they're playing with 4-quads now, ie. any 3-quad core with a dodgy core can be sold as a 2-quad. So this means that (in theory) they'd be able to sell:

4-quad chips, NV40 fully functioning
3-quad chips, NV41 fully functioning *or* NV40 with 1 quad disabled
2-quad chips, NV43 fully functioning *or* NV41 with 1 quad disabled

[MUYA]

look basically a three quad GPU is nv40? ie GeForce 6800 non ultra ? yes we agree? so to have anotehr 3 quad gpu.. spec for spec give or take a few transisors and calling it nv41 to me seems.... absurd...

now however, I am hearing of a GeForce 6800 LE. 2 quad version which codename GPU that is..i have no idea but if I had to guess, it would be nv41

[nutball]

look basically a three quad GPU is nv40? ie GeForce 6800 non ultra ? yes we agree? so to have anotehr 3 quad gpu.. spec for spec give or take a few transisors and calling it nv41 to me seems.... absurd...

A 3-quad NV40 is currently a 222M transistor chip with many many of those millions of transistors turned off. Are you saying you think NV41 is also a 222M transistor chip?

From your perspective as an end-user you don't care if your 3-quad chip is a slightly broken 4-quad chip, or a full working 3-quad chip. If the latter is cheaper to make than the former, then NVIDIA do care because it will impact their bottom-line.

This is particularly important if NVIDIA expect 6800 NU to sell very well. Maybe they've run the numbers and don't think they can fulfill demand for 3-quad chips simply by relying on off-casts from 4-quad chip production.

Or are you saying that 3-quad chips are pointless, and that they will fade from the market given time?

[MUYA]

And when did I ever say the 3 quad geForce 6800 were pointlesS? where ? I cannot see it anywhere.

I am saying....a 3 quad GPU whether it is one with 1 quad processer (collectively) is lasered off of the 222m transistor nv40 or it is fabbed with JUST 3 quad processors (roughly 180M), yet they are clocked the same and do the same thing ie 325 Mhz core, paired with XXX mhz memory and their fillrate is XXXXXXX then it's basically the same thing. For me it just doesn't make sense to have 2 product names which do the same thing. It's about naming schemes and that is all and imo it just doesn't makes sense to have 2 GPUs though born of different dies but yet functions and perform the same to have 2 different names. Do u get what I am trying to say. Now maybe u missed my 6800 LE bit...I am hearing from a few that there is rumblings of a 6800 LE which is supposed to be a 2 quad GPU and IMO the nv41.

[nutball]

And when did I ever say the 3 quad geForce 6800 were pointlesS? where ? I cannot see it anywhere.

OK, OK, I'm just trying to get to the bottom of what you're saying!!!

I am saying....a 3 quad GPU whether it is one with 1 quad processer (collectively) is lasered off of the 222m transistor nv40 or it is fabbed with JUST 3 quad processors (roughly 180M), yet they are clocked the same and do the same thing ie 325 Mhz core, paired with XXX mhz memory and their fillrate is XXXXXXX then it's basically the same thing.

Agreed.

For me it just doesn't make sense to have 2 product names which do the same thing. It's about naming schemes and that is all and imo it just doesn't makes sense to have 2 GPUs though born of different dies but yet functions and perform the same to have 2 different names. Do u get what I am trying to say.

You seem to be talking about codenames (NV40, NV41) though, not product names. Now if NVIDIA come out and call NV41-based cards something other than 6800NU, then I'll agree with you, it's daft. What they *could* do is sell both castrated NV40 and NV41-based cards under the same marketing name. Because essentially it's the same thing, as you say. Jo Punter more than likely wouldn't care, so long as she's getting her 3-quads worth.

Marketing name != Internal codename is what I'm saying.

Now maybe u missed my 6800 LE bit...I am hearing from a few that there is rumblings of a 6800 LE which is supposed to be a 2 quad GPU and IMO the nv41.

Yes, I saw that comment. A 2-quad chip makes sense too. I'd be very suprised if in the end we don't see 1-, 2-, 3- and 4-quad chips from NVIDIA, each in different speed flavours. It's a marketeers dream, and NVIDIA is a marketeering company.

[MUYA]

Yes I am on about codenames.... 6800 Non Ultra is still NV40. And the game here is not confuse anyone so if a 180M transistor 3 quad GPU came out, it would still IMO be called nv40 just becaue they have the quad processor disabled nv40 right now in teh 6800 non u. The only way I can thikn of in which NV would call a 3 quad GPU nv41 would be, if it was speed binned (well no refresh) but u see what I am trying to get at. Ie something thats clocked on par with x800 Pros? 400+ mhz instead of paltry 325Mhz :D i think..guess we will find out in q4 or mayabe late q3 this year ;)

[coldpower27]

Calling GPU's with different no. of quads different names is a good idea, but I would still call NV40 4 Quad GPU's however, Geforce 6800 right now, are NV40 with only 3 Quads functioning as you already know. It does make sense to make a actual card with only 3 Quads based on the NV4x architecture and name it different. GPU's with different actualy transistor size tend to be named differently from what I have seen.

AMD has done this itself, initial Athlon 64 3000+ with 512KB cache were still Clawhammer since they had that large 192mm2 die size, it wasn't till Newcastle core where they actually were 512KB chips with 144mm2 die size.

ATI itself did this, making Radeon 9500's with R300's but with 4 pipes enabled. They later replaced it with chips that were actually 4 Pipes in hardware the RV350. Though it's performance would never be on par with R300 thanks to the fact that RV350 is a 4 Pipe 2 Vertex Engine, vs R300's 8 Pipe 4 Vertex Engine

As it stands right now all NV40 are 6800's, with just the 6800 normal having 3 Quads with 1 Quad lasered off/disabled. But physically it's still the large 222 mill transistors die, which isn't in anyway good for nVidia from a production standpoint.

Reducing GPU die sizes whenever possible is good manufacturing practice. ATI is lucky with an 160 mil part they can get away with just disabling/lasering 1 or 2 Quads and use R420 for all their high end cards and not have to make another GPu type.

But this is all speculation of course :D

[ragejg]

:p you just gave me a headache, coldpower... :p