NVIDIA to release 10 new cards before NV30?

http://www.forum-3dcenter.net/vbull...2495d9017f111b5e9426486e7ad74aa&postid=324971

Thats the slide in question.. I think it'd be rather hard to draw any real conclusions from it

 

Hmmmmmm

It looks to me that when you zoom in the slide says, NV18 (26m), NV20 (60m), NV20-Ti (60m), NV26 (83m), NV28 (86m), NV30 (120m). You will see that when you zoom in, in the words NV30 the 3 looks identical in every pixel to the info for transistor count for the NV 26, and so on.

 

well considering that you have an 8 and a 6 right next to eachother...i dont see how its even remotely possible to confuse the two.

 

Look at the rest of the numbers on the slide. Last I checked there was no NV26. Assuming nvidias nextgen product is the nv30 then instead of an nv18 and nv28 you have an nv13 and nv23. And the "nv26" had 83M transistors apparently. The numbers on that slide are simply not reliable.

 

hmm, this is a tricky situation, I'd all but decided on the 9700 as my next purchase, however, IF the NV28 is as fast, or maybe even faster (remember, the 9700 needs hi-res or AA to really show off its power in current games) and is say $150 cheaper, then YES, I would consider NV28 over the 9700... yep, it's another rehash, but when they're blazingly fast it's hard to resist. (I currently use a GF3, btw)

Let's face it, DX9 games aren't going to arrive anytime soon, or even many games that take full advantage of DX8! It might be better to wait for the .13 version of R300, or NV30. All that the 9700 has going for it right NOW is wonderful AA/hi-res performance, and is also the fastest card in UT2003, but take away AA scores and the gap is much closer, closer than I would've liked. Nvidia could cut that still further with faster core/memory, new drivers, etc.

I agree with Galilee...

$100-$150 cheaper than R300 would make NV28 a very nice card. At a similar price, no way...

 

lol, now i see what you're saying.

 

After fiddling with the picture in PSP, I've come to the conclusion that the NV26 is actually the NV25 with 63M transistors. The NV28 has 66M transistors (AGP 8x logic?) and the NV18 has 31M transistors.

I cleaned it up a bit and this is what I think it shouls look like. But I may be wrong. I often am... sigh

 

Yeah, I think that looks much more accurate! except it still says nv26

 

I didn't bother changing that as it is pretty clear already. I don't think the NV26 ever saw the light of day, but became the NV25 instead.

 

You give the impression that you just ran the image through some filters and the result was the image you linked to, and that's why you think NV18 has 31M transistors.

But it's pretty obvious that you've simply edited the original image to have the figures you wanted, and then added a blur. I'd call that not very honest. Or maybe that's what you ment by "fiddling with the image", and "cleaned it up". In that case I think you should be a bit more clear.

 

Why would you even assume that the NV28 would have parity with the R300. As I understand it (& someone please correct me if I am wrong), the NV28 is a derivative of the NV30. As such would not the 9500 be the "natural" market competitor & not the 9700?

It seems to me that there is a lot of "gun-jumping" going on about these products. I read many posts touting the newly released Matrox product & we saw how that turned out. Also many said that ATI could not make the R300 on a .15 process. Why don't we just wait until somebody, somwhere has silicon in hand before we make autoassumptions...hmmmmkay

ed: NV25 I don't know what happened with your quote

 

Yes, I edited the figures and then expanded the image by 50%. If you do an edge detect or run it through a sharpen mask it doesn't really make it any clearer. The image is quite corrupted. No amount of filters will make it any clearer.

I made the 3s look like 3s instead of 8s. If you add a pixel to the top right of the 60 in the NV20-Ti transistor count it looks like the figure that was under the NV28 transistor count. But we know the NV20 was 60M. But if you take away that spurious pixel from the top right of the transistor count it looks like a 6.

It just seems like the original number was a 6 and got corrupted. I'm not being dishonest. i just changed it to what I think it was originally. Sorry if that wasn't clear.

 

Traditionally speaking with NVIDIA the clue to the series is in the code name - a derirative of NV30 is likely to be named NV3x; NV28 indicates that its from the NV2x line (i.e. GF4/GF4).

 

When you say that the image is edited to what you think it should be, then it's OK with me. I do agree that the numbers seem reasonable (a few meg transisors added to each of NV17->NV18 and NV25->NV28).

Btw, does the vertical axis mean some kind of relative speed? Intresting that NV18/NV28 is on the exact height as NV17/NV25.

 

No I think the vertical axis just represents transistor density. Where did the original of this chart come from?

 

Assuming those numbers are the correct ones, both NV18/28 have the same increase in transistor count, which suggests they both recieved the same new "feature package". I don't know what you can do with 3 million transistors, but I don't expect its gonna be anything mind-boglingly interesting besides AGP8x. NV18 seems to be yet another DX7-value chip...

 

I don't think that looks right either. The way you did it gives you two NV18's, one with 26M and one with 31M. I would have to agree with SanGreal instead...

Look at the slide with a lower res. I viewed it at 640x480 on my 19" monitor and it's apparent that the AGP 8x parts are NV30(120M), NV23(86M) and NV13(31M). The NV26 looks to be 83M.

All IMHO of course.

Tommy McClain

 

I think there are two NV18s, one at 26M trasistors and running on AGP 4x and the other sporting 31M transistors and running on AGP 8x. Ach, my eys hurt now and I working on a crummy laptop which isn't the best tool for the job .

 

I would have to agree with BoardBonobo.

First examine the NV28. It looks at first glance to be 86 million, but then you have to cross check how closely that matches other numbers. The "8" of the supposed 86 isn't anything like any other 8's anywhere else on the slide. Bit-wise, it's much closer to a "6" than anything else. From a thorough check of all the digits, it looks like 8's and 3's are very difficult to distinguish. The "3" of NV30 and the "8" of NV18 look identical, but there are other places where two known 3's don't match identically. It is logical to conclude that numeriacal characters aren't duplicated perfectly from position to position, but vary slightly in their bit makeup. But, the same thorough check reveals that never does a known "8" look anything like a six (with one caveat, the NV17's 28 million, where the 8 does resemble 6's elsewhere, except for the missing bit in the lower right that none of the other 6's have, and it has the corner bit in the upper left that none of the other 6's have). The first digit of the NV28's transistor count does.

Also, let's look at the "6" of the supposed NV26. Looking at other places where there should be a known "5" (such as NV15) we see that 5's look like 6's. This would also explain why it looks like NV26 instead of NV25. The "6" of NV26 is a "5", and it is NV25.

I also can't help but notice the apparent pair of NV16's. As I've alreay explained, the first "6" is most likely a "5" as in NV15. Was there an NV16 (GF3 Ti?)? It looks to me like that "6" is much closer to the "6" of the supposed NV25(6)'s 86 million. So, 66 million transistors would be the most logical conclusion.

Then examine the NV18. The first digit is going to be either a 3 or an 8, based on comparisons with other known digits in other positions. It wouldn't make sense for it to be 81 if the NV25 is only 66. Therefore, it must be 31 million, which is precisely the increase from the NV17 as the NV28 has over the NV25.

Conclusion? NV28 and NV18 can't be much more than NV25 and NV17 with AGP 8x. This really makes sense if nVIDIA is having trouble getting the NV30 out, and rumors about them being stretched thin already. It seems unlikely that they can get a next-gen DX9 part, a seriously beefed up GF4 (more vertex shaders?), and a mysterious "value" DX8 part with just 5 million transistors less than the Super-Ultra GF4 Platinum (when has any of NV's value chips had just 5 million transistors less than it's flagship chip? That makes no sense at all).

Nothing new here but AGP8x and higher clock speeds it would appear.