NVIDIA to release 10 new cards before NV30?

[Bigus Dickus]

Can't believe I'm going to reply and nitpick this, but it's bugging the hell out of me.

I've already given my opinion, so I won't go through the logic again, but I will point out one or two things:

Before you reply, look at the slide with a resolution of 640x480.

I used this nifty feature in graphics programs called "zoom." You'd be surprised how much easier that is than changing monitor resolution. Blowing a psuedo 5/6/8 up until it covers the whole screen doesn't help, it's still going to be an interpretation.

That makes it NV13, not NV18.

So NVIDIA is going to give the next iteration of a GPU a lower code number? When else in their history have they done this?

Such as? reference to 3's

There are a couple that look identical on first glance, but actually have slightly different pixel arrangements. Look closely... it's like a "where's Waldo" pic.

Actually, my belief is that the NV15 is actually NV16. Looking at the slide at the lower resolution it's absolutely a '6'.

Looks can be deceiving.

Now you're probably saying there isn't a NV16. Agreed, I just think they had a typo, because I think they made the same mistake again with the number of transistors, 26 instead of 25. This makes more sense then saying that the numbers actually look like 5's. Is NV26 also a typo?

Doesn't that seem a bit too coincidental? That they would make the same typo in more than one place? Isn't it a bit more logical to conclude that, due to the inadequate resolution used to capture the image, 5's simply look like 6's? Wouldn't that make all the other numbers fit better with known information (i.e., where you would expect to find NV15, NV25, or 25 million, you actually do).

My original intention on posting was in deciphering the numbers. Before any conclusions can be made the numbers need to be correct.

Unfortunately, the image was simple captured at a resolution insufficient to correctly identify characters. It really looks like resampling/resizing or compression has altered the characters from their original state, which explains why some numbers don't look like they should.

I stand firmly by my opinion, and time will prove me right!

 

[Dave Baumann]

Two things: every site that has talked about these rumoured parts have refered to them as NV18 and NV28, and so do the 30.83 drivers:

%NVidia.Nv18.1% = nv4, PCI\VEN_10DE&DEV_0180
%NVidia.Nv18.2% = nv4, PCI\VEN_10DE&DEV_0181
%NVidia.Nv18.3% = nv4, PCI\VEN_10DE&DEV_0182
%NVidia.Nv18GL.1% = nv4_WSApps, PCI\VEN_10DE&DEV_0188
%NVidia.Nv18GL.2% = nv4_WSApps, PCI\VEN_10DE&DEV_018A
%NVidia.Nv18GL.3% = nv4_WSApps, PCI\VEN_10DE&DEV_018B
%NVidia.Nv28.1% = nv4, PCI\VEN_10DE&DEV_0280
%NVidia.Nv28.2% = nv4, PCI\VEN_10DE&DEV_0281
%NVidia.Nv28GL.1% = nv4_WSApps, PCI\VEN_10DE&DEV_0288
%NVidia.Nv28GL.2% = nv4_WSApps, PCI\VEN_10DE&DEV_0289

Also, NV18 and NV28 has always made sense to me with their names since they are supposed to be AGP8X parts. I think its safe to say the next parts are NV18 and NV28.

 

[AzBat]

I used this nifty feature in graphics programs called "zoom." You'd be surprised how much easier that is than changing monitor resolution. Blowing a psuedo 5/6/8 up until it covers the whole screen doesn't help, it's still going to be an interpretation.

Here on my computer at work I don't have any graphic programs. So for me changing the resolution was my only option.

So NVIDIA is going to give the next iteration of a GPU a lower code number? When else in their history have they done this?

NV11. The NV15 had already been shipped before they released the NV11.

Looks can be deceiving.

Nahhh couldn't be.

Doesn't that seem a bit too coincidental? That they would make the same typo in more than one place? Isn't it a bit more logical to conclude that, due to the inadequate resolution used to capture the image, 5's simply look like 6's? Wouldn't that make all the other numbers fit better with known information (i.e., where you would expect to find NV15, NV25, or 25 million, you actually do).

Possibly, but my brain still says those are 5's.

Unfortunately, the image was simple captured at a resolution insufficient to correctly identify characters. It really looks like resampling/resizing or compression has altered the characters from their original state, which explains why some numbers don't look like they should.

When I first was viewing them I couldn't figure out what format they were using since a QuickTime plug-in was being used to display them. So I couldn't make any conclusions as to the format or if they were using any compression. In writing my reply to this message I decided to look for the source file and I eventually found it. I found it on ON24's site and it's a PNG file. You can see the source file here...

http://www.on24.com/event/618/1/slides/14.png

You should look at the rest of the slides. There are 36 of them(1.png - 36.png). It's the Annual Shareholders Meeting slides from July 11.

I stand firmly by my opinion, and time will prove me right!

Fairly sure aren't we? Well, considering the info that Dave just gave regarding NV28 and NV18, I'll cede and say that the 3's and 8's look the same and likely other numbers may look the same too. How's that for you? Hehehehe Man, it is hard to admit when you are wrong. However, those that remember me they know I always do, well eventually, I promise!

Joe!! Where are you? It's your turn now. I've been out of commission so long I forgot how to put up a fight. Come on back this forum needs you.

OK, I return you to your regularly scheduled broadcast.

Tommy McClain

 

[Basic]

OK combine these known numbers from various reviews with the image:

99H2 - NV10   - GF1      - 23M
00H1 - NV15   - GF2      - 25M
00H2 - NV11   - GF2MX  -
00H2 - NV16   - GF2U    - 25M
01H1 - NV20   - GF3      - 60M
01H2 - NV20Ti- GF3Ti    - 60M
02H1 - NV17   - GF4MX  -
02H1 - NV25   - GF4Ti    - 63M
02H2 - NV18   - ??MX     -
02H2 - NV28   - ??          -
02H2 - NV30   - TNG       - 120M

OK, we only know that the 02H2 parts are supposed to come then, not that they actually will.

If you go by how the known numbers look, then it's quite clear that:
NV28 is 65M or 66M, either of them sounds reasonable.
NV18 is 31M or 81M.

Now, how likely is it that NV18, the next mainstream chip, that still is in the NV1x series (DX7) is three times as large as last chip, and 35% larger than their current largest (DX8) chip? And when all the other chips are ordered in some kind of performance or size order (heightwise), should this superbig NV18 realy be placed next to NV17?

The reasoning about placement works for NV28 too.

Time to do the admitting Tommy?

And Tommy, don't you even have MSPaint at work? It's an exelent little program. Start=>Programs=>Accessories=>Paint

PS
Boy, if me commenting this isn't a proof of my miserable life quality, then I don't know what it would take.
DS

 

[kid_crisis]

I watched the CC webcast live on July 11th, and the numbers were clearly visible on the slide presented there, and Jen Hsun also verbally stated them during the conference call. They are:

NV18: 81 million
NV28: 86 million
NV30: 120 million

The NV18 and NV28 were shown as being released in August-02, and the NV30 was scheduled to be released in the Oct/Nov-02 timeframe.

Clearly the NV28 has more than AGP-8x added to it over the GeForce-4 since it has ~23 million more transistors.

 

[stevem]

NV11. The NV15 had already been shipped before they released the NV11.

NV11 was not a next gen product... It was a lower cost/performance derivative of the NV15 - hence lower product code. We assume that NV28 will also be a derivative of the NV25 series with AGPx8 & possible refinements/features filtered down from the NV30 - as per ATI with 9000 PS/VS optimisations, etc, from 9700.

 

[KimB]

NV18: 81 million
NV28: 86 million
NV30: 120 million

Very interesting. Unless nVidia has completely departed from their previous naming scheme, this fall is going to be very interesting. I'd really like to know that nVidia could have possibly done to an NV1x core to make an 81 million transistor part.

 

[Ichneumon]

NV18: 81 million
NV28: 86 million
NV30: 120 million

Very interesting. Unless nVidia has completely departed from their previous naming scheme, this fall is going to be very interesting. I'd really like to know that nVidia could have possibly done to an NV1x core to make an 81 million transistor part.

I hope they made it a Dx8 part.

 

[Dave Baumann]

I hope they made it a Dx8 part.

If its consistent with NVIDIA's conename scheme then an NV1x part will still be predomanently based off the NV10/15/11 architecture line, so I doubt it would be DX8. However, I also doubt it would have gorwn by 50M transistors if it wasn't DX8...

 

[Ilfirin]

Hmmm, the article I posted before is no longer at nVidia's site.

They put the old one (that was just a presentation form of the CineFX Overview) back in it's place.

I guess that 'nVidia Confidential' that was stamped on every slide wasn't just a joke

Good thing I always download the papers before I read them.. ;p

 

[Bigus Dickus]

Man... no real R300 reviews yet, NV30 information at an infinitesimal trickle...

You know we're bored when we spend this much time staring at individual pixels and trying to guess what number it is.

I still don't understand the logic of producing two new cards that are almost the same transistor count. The argument could be made that ATi is doing the same with the 9700/9500, but these are the next generation chips, not refreshes of current chips.

Also, I wouldn't think an 80+ million transistor chip would make a "value" card. It would be difficult to get the price down to be competitive.

 

[Ilfirin]

Man... no real R300 reviews yet

Ah what? There have been dozens of r300 reviews. Of course Anandtech's are always the only ones that are halfly decent:
http://www.anandtech.com/video/showdoc.html?i=1656

 

[Pete]

He certainly has the most graphs, but he also seems to have a halfway grudge against ATi. Avoid his comments and you'll be fine.

 

[Bigus Dickus]

Man... no real R300 reviews yet

Ah what? There have been dozens of r300 reviews. Of course Anandtech's are always the only ones that are halfly decent:
http://www.anandtech.com/video/showdoc.html?i=1656

review, not preview.

 

[Ilfirin]

Man... no real R300 reviews yet

Ah what? There have been dozens of r300 reviews. Of course Anandtech's are always the only ones that are halfly decent:
http://www.anandtech.com/video/showdoc.html?i=1656

review, not preview.

How is that article not a review? It has a detailed explanation of the card, it's features and specs along with 6 benchmarks in various different games (including Ut2003).

If you mean it is because the card might not be at the same clock speed as the retail card - ATI still hasn't decided upon that. Besides, a review of a card with a lower/higher clockspeed of the retail card is still a review. What marks a preview (IMO) is when there are no benchmarks.

 

[Bigus Dickus]

I tend to think of anything done before the board has been finalized and can be tested in shipping form with shipping drivers to be a preview.

Yes, I agree that the R300 is a bit different because some benchmarking was done instead of just a technical breifing, but the benchmarks were somewhat "limited." The purpose of the previews we've seen has still been to highlight the new technology of the card, discuss its architecture and capabilities, etc.

I'll call it a proper review when a retail board and drivers is sent to a reviewer, and they are allowed to test on any system configuration, games, driver settings, etc. Also, reviews (the good ones) should include some CPU scaling, and of course image comparisons (which we've seen none of).

Until then, I'll treat them as previews like we've always had with some teaser numbers.

 

[BoardBonobo]

I agree with Bigus Dickus there, to me a review is written around a piece of hardware that you can go out an buy after you've read it. We've been given a whole stack of technonlogy previews and some indicators of how the card performs under controlled tests.

And I have to agree about the pixel staring problem! But I also stand by my remastered chart. If I was wrong then so b it, but I get the feeling I wasn't.

 

[Ilfirin]

Yes I agree with him too, just showing the other side of it.

It's more like a review on the pre-release card than a preview of the retail card.

 

[Bigus Dickus]

It's more like a review on the pre-release card than a preview of the retail card.

Heh, I kind of like that one.