Almost Official - NV28/18 only adds AGP 8x support

[Pav_37]

I noticed @ Warp2Search that they discovered that recent 31.00 drivers appear to show what NV18/28 cards will be named and just by looking @ the names - you can't help to conclude that NV18/NV28 will be nothing more then a Geforce 4 with a agp 8x support. Now I wonder what sort of marketing bullshit will nVidia come up with to sell these - let me see - "Geforce 4 is now twice as fast!!!" - LOL

More info @ http://www.warp2search.net/article.php?sid=6443&mode=thread&order=0

 

[alexsok]

Says absolutely nothing.

 

[Joe DeFuria]

Actually, it says more with some credibility than you have told us about any future nVidia part.

 

[Fuz]

Actually, it says more with some credibility than you have told us about any future nVidia part.

Haha, I second that

 

[alexsok]

We'll see who will laugh in the end!

 

[Fuz]

I am already laughing. The end must be at hand!

 

[Ante P]

Says absolutely nothing.

Why wouldn't it?
I just breezed through the thread at warp, I'm under the impresseion that these descriptions were put there by nVidia?
If so it certainly seems like there's no real improvement except AGP8x, otherwise I think they wouldn've called them GeForce 4 Ti 4800 or something along those lines. As MX 440 "SE" got a new name since it's probably a bit faster (ie near 460 speed or something)

 

[alexsok]

Says absolutely nothing.

Why wouldn't it?
I just breezed through the thread at warp, I'm under the impresseion that these descriptions were put there by nVidia?
If so it certainly seems like there's no real improvement except AGP8x, otherwise I think they wouldn've called them GeForce 4 Ti 4800 or something along those lines. As MX 440 "SE" got a new name since it's probably a bit faster (ie near 460 speed or something)

Higher clock speeds, increased cache sizes, nothing is written there about that. AGP 8X was obvious as it has been mentioned before, but that's not the only improvement.

Not sure about higher clock speeds, but it's a possibility.

 

[Dave Baumann]

Higher clock speeds, increased cache sizes, nothing is written there about that. AGP 8X was obvious as it has been mentioned before, but that's not the only improvement.

Think for a second...

If there was significant differentiation from the other models that would impact (positivly or negatively) performance then it wouldn't be called "GF4 Ti 4x00 with AGP8X". Likewise, the GF4 products are all based off the same chip with the number denoting different core speeds of that chip; again if they were going to run at different speeds they would also be named differently.

Also, when 'refreshing' a part it is certianly normal for NVIDIA not to go around phutzing with various elements of the design as there is an increased margin for error - they know the NV25 design works so why jeopardise that by changing other elements (such as caches) than just the part they want to change (AGP interface).

 

[multigl2]

Higher clock speeds, increased cache sizes, nothing is written there about that.

so maybe higher clock speeds, but increased cache sizes? Not bloody likely... these chips are designed to be so effecient based on the size and performance of off everything else in the chip. Increasing cache sizes would offset that balance and could decrease performance if all you do is increase cache sizes.

 

[alexsok]

Also, when 'refreshing' a part it is certianly normal for NVIDIA not to go around phutzing with various elements of the design as there is an increased margin for error - they know the NV25 design works so why jeopardise that by changing other elements (such as caches) than just the part they want to change (AGP interface).

I certainly agree with u there Dave and that does sound logical.

But if that turns out to be true, then the only purpose Nvidia are getting out the NV18/NV28 is to replace the NV17/NV25 chips and also reduce costs of the chips themselves. Oh, and to add AGP 8X

There is one rumour that Nvidia have pushed the release date of NV18/NV28 together with NV30, cause the NV25 chips are apparenly selling pretty well and there is no need to replace them yet.

Another question remains: Why the additional transistors? I mean, the increase is huge! 81million on NV18 and 86million on NV28!

So the NV28 will probably be targeted at the mainstream market until a mainstream NV30 solution is released, that's my assumption anyway...

multigl2 - sure, but if for example they increase the clockspeeds of both the memory & the core and increase the cache sizes? The balance could remain the same, but then again, maybe not...

I sure hope there is not much time left till the announcement...

 

[multigl2]

there is still a lot of debate over what those numbers really were. I for one remain skeptical, because, well, thats what I do best And If you increase cache sizes on the NV25 to NV28 and increase clock speed on the same process (.15) you probably cutting the cost of the chip

So the NV28 will probably be targeted at the mainstream market until a mainstream NV30 solution is released, that's my assumption anyway...
thats a fairly straight forward and good assumption, because if yeilds for the NV30 are ~15% on .13 then you can bet a good deal they won't be a "mainstream" product.

 

[demalion]

Higher clock speeds, increased cache sizes, nothing is written there about that. AGP 8X was obvious as it has been mentioned before, but that's not the only improvement.

Think for a second...

If there was significant differentiation from the other models that would impact (positivly or negatively) performance then it wouldn't be called "GF4 Ti 4x00 with AGP8X". Likewise, the GF4 products are all based off the same chip with the number denoting different core speeds of that chip; again if they were going to run at different speeds they would also be named differently.

What the heck am I doing agreeing with alexsok, even if slightly? This arguing the point and not the person stuff sure makes me feel strange sometimes .

EDIT: bah, shame on me. Smack me with a wet noodle alexsok, I won't complain...that reads with too much insult compared to the light hearted tone I had in mind, but I can't go back and edit it without distorting things.

Reasons that it wouldn't have a faster clock speed with the info given:

Same model number listed in file (strong reason given past history).

Counterarguments:

Chance that the sale model number is different, despite past history (the awkwardness of the phrase "with AGP 8x" as opposed to a slightly higher numbering supports this).

The positive return from doing this (any perception of better value that can be used to support higher costs) seems extremely significant for the investment (yields have to be better by now, and the chips overclocked fairly well anyways, so even incrementally higher clocking for even supporting a $10 higher cost seems desirable), especially with all the spin resources available.

I'd conclude that there will be more to the product naming than the info present, though I definitely would not say the info says nothing.

Also, when 'refreshing' a part it is certianly normal for NVIDIA not to go around phutzing with various elements of the design as there is an increased margin for error - they know the NV25 design works so why jeopardise that by changing other elements (such as caches) than just the part they want to change (AGP interface).

I support your statements about the cache, but ATi has done clock upgrades with absolutely no fanfare (Radeon SEs), and facing competition I'd expect nVidia to not miss the opportunity to tack on a "+"(Plus), numbering like "4680", or something similar, despite what it is called in drivers, and I'd think that supports some sort of clock increase possibly. Could be wrong, of course...maybe they think the "with AGP 8x" really will be enough, but I don't really buy that...on the other hand, maybe they'd be right?

 

[alexsok]

BTW, it's very reasonable why a gf4ti4600 with AGP 8x is not mentioned in the inf file.

Nvidia will remove it from production (if they didn't do it already), as it's just useless and I think that's pretty logical. Two cards will remain: gf4ti4200 AGP 8x & gf4ti4600 AGP 8x, sounds very reasonable.

 

[kid_crisis]

Again, I'm pretty sure the 81M transistors quoted for NV18 is for the "crush" version, which means it has a Northbridge added onto it (not the addition you were looking for I'm sure).

The 86 million transistors for the NV28 may be for real though, and is a 23 million increase over NV25.

 

[Dave Baumann]

But if that turns out to be true, then the only purpose Nvidia are getting out the NV18/NV28 is to replace the NV17/NV25 chips and also reduce costs of the chips themselves. Oh, and to add AGP 8X

Well, I don't think reducing the cost is necessarily a given, since the parts will ostensibly be larger than what they are replacing. Even if they are at a reduced cost will it even outweigh the engineering and mask costs?

There is one rumour that Nvidia have pushed the release date of NV18/NV28 together with NV30, cause the NV25 chips are apparenly selling pretty well and there is no need to replace them yet.

Personally I think they have done it because nForce2 is AGP8X capable. When nForce2 comes about they know a lot of users will be looking towards getting an AGP8x card to go in that slot - the last thing they want is not to have any of their own GFX cards available that make use of it and people will go to their competitors. Remember the 'SPP' version of nForce2 was designed with descrete graphics in mind. So, really I think these parts are more about nForce2, however nForce2 isn't around just yet...

 

[alexsok]

and is a 23 million increase over NV25.

Exactly!

Why a 23million increase over NV25 and we get only AGP 8x?

 

[Dave Baumann]

The positive return from doing this (any perception of better value that can be used to support higher costs) seems extremely significant for the investment (yields have to be better by now, and the chips overclocked fairly well anyways, so even incrementally higher clocking for even supporting a $10 higher cost seems desirable), especially with all the spin resources available.

Read my previous responce as to what I believe is the real value in these parts.

 

[DemoCoder]

There might be NV30 vertex and fragment shader emulation in these drivers. The Cg beta2 docs said that the next nVidia driver release would include emulation.

 

[KimB]

I'd just like to put my two cents in here.

I got the 31.00 drivers from a leaked nForce driver set. It turns out that the .inf file that came with those drivers, as it stands in that driver set, only supports the nForce display.

It seems all too likely, to me, that those descriptions were placed in there by somebody who edited the .inf file from the drivers in that nForce set.