NVIDIA AGP8x

[pelly]

http://www.xbitlabs.com/news/story.html?id=1030465153

What AGP8x part will NVIDIA be showing???

:)

[Uttar]

Who said the NV28 would be ready VERY soon? Wasn't it the same person who everyone replied to that "GF4 still sell very well, they won't release the NV28 and NV18 so soon before NV30"

Hmm... Wasn't that me? :P

If they show the NV18/NV28 there, they'll be made official soon before. Would we have the official announcement tommorow? *hopes*

But then again, maybe they'll only announce the NV18 and NV18M...


Uttar

[DIMA]

Well, one thing is for sure:

They will show something! :D

[Fotis]

I couldn't care less about nv28!!:p
Give me nv30!!!:mad:

[TheOneKEA]

I suspect that NV28 = NV25 + AGP8X. I doubt nVidia will do anything earthshaking with the NV25 for the NV28 refresh, IMO.

[DadGT]

Originally posted by TheOneKEA
I suspect that NV28 = NV25 + AGP8X. I doubt nVidia will do anything earthshaking with the NV25 for the NV28 refresh, IMO.

That sounds good except nVidia already annouced that NV28 would have 86 million transistors. I don't think the increase can be attributed to the addition of AGP 8x. I think that NV28 will not be earthshaking, but it could be a very interesting product refresh. I wonder what NV has up their sleeve... Suppose we will get a glimpse tomorrow with the AGP 8x session in Taiwan.

[Philibob]

Originally posted by DIMA
Well, one thing is for sure:

They will show something! :D

This is how a GF4 Ti4600 works on an 8X slot.

:D :D :D

[TheOneKEA]

20mil transistors is a lot, and maybe a 1/4th could be AGP8X. But what about bugfixes? That alone could be anywhere from another 1/4th to a whole 1/2. I guess we'll have to wait and see for the AGP8X session in Taiwan.

[-=DVS=-]

Hmm what difference does it make what new GF4 will be or can do or how much more transistor it will have it is still a DX 8.0 card :rolleyes: and its gonna be prised high again only idiot would buy same old dx 8.0 card for new prise :D dx 9.0 cards just on the corner ;)

[StealthHawk]

yeah, and with all those DX9 games ready and right around the corner right :rolleyes:

[netviper13]

Exactly, we can still count on our fingers the number of DX8 games, I don't foresee DX9 games for quite a while at least. I mean think about it, the majority of people probably still don't even have DX8 hardware.

[Bigus Dickus]

I may do some serious crow eating (with hot sauce), but my money is still on NV28 = 66 million, NV18 = 31 million.

[-=DVS=-]

sure we woun't see DX 9 games for long long time but were is the logic to buy high prised same old product in new colors when you can get for same prise next generation card :rolleyes: with bigger support for future :p


and about the prise i am guessing here that so calld GF4 aka NV28 gonna cost about same as Radeon 9700 :D

I realy wish Nvidia would stop releaseing same card again and again geez lol :rolleyes:

[StealthHawk]

future support just doesn't cut it. if you bought the gf3 over the gf2ultra or gf2pro because the gf3 supported DX8 and gf2 line didn't, then that was more likely a waste of money. neither UT2003 nor Doom 3 is out yet, and heck, neither is anywhere in sight.

that's not to say the gf3 didn't have performance and image quality advantages over the gf2ultra, of course.

there really isn't such a thing as being feature proof with a video card, they just get outdated too fast, although we may see that changing with these DX9 cards, they seem plenty powerful. and without FSAA or AF they are horribly CPU bound(supposedly).

all i'm saying is that if nvidia could *somehow* get NV28 to have awesome AF and FSAA peformance, and have a price at $250 or so, it would be very compelling compared to the R9700. of course that's all but impossible, and not gonna happen.

[Nv40]

Originally posted by TheOneKEA
20mil transistors is a lot, and maybe a 1/4th could be AGP8X. But what about bugfixes? That alone could be anywhere from another 1/4th to a whole 1/2. I guess we'll have to wait and see for the AGP8X session in Taiwan.


what bugs are you talking about ?
The Geforce4 has none harware bugs ,at least
no hardaware reviewers hasnt found something
that deserves a "fix" in the Geforce4 ,can you provide a link why you claim that?

and software bugs
(is there is any in the Geforce4 )can be
fixed through Drivers and patches .. :)
you may give a try to Geforce4 and see how
Smooth this card works ,every old and new game works out of the Box ,graphics aplications too ..
its the Highest quality product ever made by Nvidia today ,congrutalutions to them... ;)


"20mil transistors is a lot, and maybe a 1/4th could be AGP8X."
-----------------------------------------------
is more over 23million transistors ,Geforce4 has 63million .. and TNT2 ultra has 25million
transistors ... and the latest DIrecx8.1 card from Trident (Xabre-Xp4T3) has 30 million transistors

http://www.3dvelocity.com/articles/xp4/xp4.htm


So you can create a new cheap DIrecX8 video card with those extra transistors (without the fancy 2d stuff/conectors) with those extra very welcome transistors that will be avaible in the Nv28..;)

I think Gamers will be very pleased for that kind of performance for just $259 ... ;)

[TheOneKEA]

I wasn't saying it actually had bugs; I was just saying that was what it MIGHT be.

[StealthHawk]

GF4 has hardware bugs, anyone who doesn't believe that is deluding themselves. what about anisotropic filtering in D3D? are we to believe that the horrible performance is simply a software bug that can be fixed this late in the date? it's obivous that it can't be, and it has been for some time. nvidia is NEVER going to fix it, because they can't.

it is a FACT that the GF4 has a much bigger performance hit with AF in D3D than the GF3.

[sbp]

Originally posted by StealthHawk
GF4 has hardware bugs, anyone who doesn't believe that is deluding themselves. what about anisotropic filtering in D3D? are we to believe that the horrible performance is simply a software bug that can be fixed this late in the date? it's obivous that it can't be, and it has been for some time. nvidia is NEVER going to fix it, because they can't.

it is a FACT that the GF4 has a much bigger performance hit with AF in D3D than the GF3. But does that {GF4 anisotropic filtering in D3D peformance} prove its a bug?

[StealthHawk]

Originally posted by sbp
But does that {GF4 anisotropic filtering in D3D peformance} prove its a bug?

take your pick, it is either a software(driver) bug or a issue with hardware. all the evidence seems to support the fact that D3D AF with a GF4 is either horribly inefficient, or moderately "broken."

let's review the history of this problem.

1) the GF4 is launched and reviewed, and it is discovered that both OGL and D3D AF is about on par with the GF3(ie that there is an even bigger performance hit for the GF4)

2) Matt Burris at 3dgpu emails nvidia about this and it is stated that there is a bug nvidia is aware of and is working on fixing

3) fast forward a few months later, to driver release 28.90. OGL AF is boosted in terms of performance across the board, including for the GF3 line of cards. people speculate that nvidia will do the same with D3D but is waiting for a "Detonator5" release and/or will first include AF options for D3D in the control panel.

4) a few months after that, the 30.xx drivers are released, which contain new registry settings that allow tweaking of AF. once again, OGL AF can be boosted across the board. there are also new options for adjusting D3D AF, but this process is a game by game basis. additionally, the tweaking is available only for the GF4

conclusion: nvidia has been trying to solve the problem, but has so far been unable to. this points not to a driver problem but more likely, a hardware problem. not to say that nvidia will not be able to find a "solution" but it will not be a quick across the board fix. quite frankly, i doubt many people are still holding their breath when expecting a solution. there is still the possibility of a "det5" release with the NV28 debut though. and maybe nvidia will present some kind of solution then.

of course my guess is that nvidia is much more preoccupied with the launch and preparation of NV30 to care about the GF4. and that really begs the question if they would even bother to fix the problem once NV30 hits. afterall, any weakness in the GF4 can in turn be construed as a strength for NV30, right? i firmly believe that if by the time NV28 arrives and there is no resolution, there never will be one.

[pelly]

Ok......yeah.....getting back to the real subject of the thread.....

Has anyone seen that blurb on DigiTimes talking about NV18? Evidently, NVIDIA showed it off at the VIA-AGP8x show.....

For some reason, nothing about this "showing" or any details about the VIA AGP8x convention has been on the net?

Strange....:confused:

[Gator]

I think the NV28 will probably be advertised as such:

Nvidia GeForce4 TI5000 128DDR at 325/650 and AGP8x

Starting price, $300us. Will people buy it? Probably. Will it compete with R9700? Probably. Will they kick themselves with NV30 is release? Probably. LOL

I'm not sure NV28 is a good idea unless there is some definite architecture changes. However I'm also not sure what Nvidia plans to accomplish by releasing not 1 but 2 new chips around the same time. I guess they're thinking econo-market vs hardcore-gamer market much like ATI is with 9000 vs 9700. Only problem is you know the NV28 will be anything far from cheap. ROFL

ok, i'm done rambling

[SurfMonkey]

Off topic mode:

Originally posted by sbp
But does that {GF4 anisotropic filtering in D3D peformance} prove its a bug?

The Aniso performance isn't due to a hardware bug, it's down to how nVidia decided to implement Aniso in hardware. Enabling Aniso effectively disables one set of TMUs.

If we wanted to talk bugs we could mention *cough*45 degrees*cough*, but in marketing speak I guess that would be a unique feature, not a bug...

Besides if we want to talk hardware bugs, check out JCs comments on why he thinks the 8500 isn't as fast as it should be.

Off topic mode off.

I was under the assumption that the latest cards would be AGP8x and no more. But I may be wrong, also aren't there going to be three flavours N28, N18, nForce2. All DX8.1?

[saturnotaku]

If what I'm reading is correct, the integrated graphics in the nForce2 will be nothing more than a souped up GF4 MX. So that means we won't even get DirectX 8 compatibility. :mad:

[-=DVS=-]

i wonder if there is big advantage in AGP 8x :rolleyes:

[StealthHawk]

SurfMonkey,

then you are saying the efficiency of AF in OGL vs D3D is due to the differences between APIs? because OGL is fine and dandy, and D3D AF is horribly slow with a GF4...