NVIDIA: Beyond G80...

[INKster]

When AMD (i mean ATi) make some good ingame performance incrase with new drivers they not called it as Catalyst 14.8.
NV drivers vista performance still very poor, im not surprised NV is crying like a b**** for dx10 XP version

There is no DX10 in XP, eveyone knows that for a long time.
And i bet ATI would rather have to code for a DX10 title in XP too. Don't think for a minute they are any different...


edit
I've just counted the number of entries on that misterious "G98" in the new FW 165.01 for Vista:

32 instances

 

[Arun]

"G98" (not a typo )

!!! Is there any precedent for chips being in the drivers prior to tape-out? Because if that means G98 taped-out (or is even back from the fab), it'd be the first G9x to tape-out; and this could also imply quite a lot of things about today's launch...

G84 features one TFA per TFU again... The basis for GF8900?

Why would it be?

 

[vertex_shader]

The is no DX10 in XP, eveyone knows that for a long time.
And i bet ATI would rather have to code for a DX10 title in XP too. Don't think for a minute they are any different...

I know this, but NV still want dx10 for XP because they driver team have very big problems understanding vista in driver level. (i not say AMD vista drivers are perfect, but much better).

The number jumps its just funny nothing more, i think some user eat this and think big driver numbers means big performance incrase and stability :smile:

 

[INKster]

I know this, but NV still want dx10 for XP because they driver team have very big problems understanding vista in driver level. (i not say AMD vista drivers are perfect, but much better).

The number jumps its just funny nothing more, i think some user eat this and think big driver numbers means big performance incrase and stability :smile:

Did you test them ?
I did.

- new control panel (as it should be)
- lots of bug fixes
- considerable performance gains across the board
- stability has increased somewhat
- supports lots of other non-GF8 cards

No easy feat delivering this simultaneously for both XP 32bit and Vista 32/64bit.
Sure, there's still a lot of work ahead for them, but for someone like me and lots of others who have had to endure so many hair raising issues in the past few months, it's nice to see so much finally changing for the better, albeit not yet perfect.



edit

!!! Is there any precedent for chips being in the drivers prior to tape-out? Because if that means G98 taped-out (or is even back from the fab), it'd be the first G9x to tape-out; and this could also imply quite a lot of things about today's launch...

Arun:

// 06E0 - NVIDIA G98
// 06E1 - NVIDIA G98
// 06E2 - NVIDIA G98
// 06E3 - NVIDIA G98
// 06E4 - NVIDIA G98
// 06E5 - NVIDIA G98
// 06E6 - NVIDIA G98
// 06E7 - NVIDIA G98
// 06E8 - NVIDIA G98
// 06E9 - NVIDIA G98
// 06EA - NVIDIA G98
// 06EB - NVIDIA G98
// 06EC - NVIDIA G98
// 06ED - NVIDIA G98
// 06EE - NVIDIA G98
// 06EF - NVIDIA G98
// 06F0 - NVIDIA G98
// 06F1 - NVIDIA G98
// 06F2 - NVIDIA G98
// 06F3 - NVIDIA G98
// 06F4 - NVIDIA G98
// 06F5 - NVIDIA G98
// 06F6 - NVIDIA G98
// 06F7 - NVIDIA G98
// 06F8 - NVIDIA G98
// 06F9 - NVIDIA G98
// 06FA - NVIDIA G98
// 06FB - NVIDIA G98
// 06FC - NVIDIA G98
// 06FD - NVIDIA G98
// 06FE - NVIDIA G98
// 06FF - NVIDIA G98

I'm at a loss of words right now, so please shed some light on this...
There are actually more G98 unique references in this driver than for the entire current G8x line (including the new G84 and G86 derivatives).

 

[CJ]

Well there was talk about nV being busy with it a G9x GPU which would be released at the end of the year. But then again, maybe it's a similar situation like we had with Catalyst 5.10a which contained a lot of strings to the R499, which turned out to be ATi's reference to R580.

 

[karlotta]

G98= placecard for mobile chip?

 

[_xxx_]

because they driver team have very big problems understanding vista in driver level.

Oh, you got some insider info?

A-ha, so they have a bunch of monkeys with typewriters writing their drivers instead of experienced programmers? Or do they only employ low-IQ people?

 

[INKster]

G98= placecard for mobile chip?

32 of them ?
And why not G88 ?

 

[Arun]

32 of them ?

I can see two reasons for this. The first is that they are just reserving numbers for no good reason. The second is that they are completely changing something quite deep in how the drivers communicate with the chip because of virtualization. Neither really satisfy me, though...

And why not G88?

Maybe because it's not a G8x/GeForce 8?

 

[dnavas]

I'm at a loss of words right now, so please shed some light on this...
There are actually more G98 unique references in this driver than for the entire current G8x line (including the new G84 and G86 derivatives).

Well, I would summarize my thoughts paraphrasing one of my earlier posts: the release of an Ultra says something important regarding the lack of a refresh part. In my opinion, NV's refresh is late, and the G84/86/Ultra are backup plans.

An alternative theory (which is essentially similar, but with a different spin ;^/) is that NV is rushing towards the 9-series (65nm?) and essentially skipping as much of 8 as possible, trying to take advantage of the R6x release schedule slippage. If NV is willing to give up supremacy from May-July, to hit hard in August, that'd be a worthy goal and a reasonable tradeoff... (?)

Unless NVidia has taken over some large notebook company, I don't think they'd need that many entries for a mobile part ;^/

Other thoughts?

 

[trinibwoy]

Other thoughts?

The monkeys writing Nvidia's drivers are fascinated with the number 98 and thought it would be cool to paste it 32 times? :smile:

 

[icecold1983]

anyone have any links to testing of these drivers for performance? the readme file on nvidias website is down.

 

[INKster]

Well, I would summarize my thoughts paraphrasing one of my earlier posts: the release of an Ultra says something important regarding the lack of a refresh part. In my opinion, NV's refresh is late, and the G84/86/Ultra are backup plans.

Maybe it's not the design of said refresh per se, but actually issues with the process at TSMC.
65nm could possibly be, in part, responsible for lower yields, hence, shortening of the time-to-market window.
And let us not forget that RV610 and RV630 (not exactly high-end/enthusiast parts) are also 65nm designs and have in fact been delayed, where the "simultaneous launch" excuse from AMD is still shady, at best.
So, G84 and G86 would be brought to market at performance mid-range prices, taking advantage of a temporary lack of DX10 competition.

An alternative theory (which is essentially similar, but with a different spin ;^/) is that NV is rushing towards the 9-series (65nm?) and essentially skipping as much of 8 as possible, trying to take advantage of the R6x release schedule slippage. If NV is willing to give up supremacy from May-July, to hit hard in August, that'd be a worthy goal and a reasonable tradeoff... (?)

Unless NVidia has taken over some large notebook company, I don't think they'd need that many entries for a mobile part ;^/

Other thoughts?

In my opinion, August is already late for the back-to-school season, especially if they want to fulfill large OEM orders.
Perhaps early June to late July.

 

[[Brick_top]]

The monkeys writing Nvidia's drivers are fascinated with the number 98 and thought it would be cool to paste it 32 times? :smile:

I have nothing to add to this thread other than saying I thought that was funny, thanks

 

[jimmyjames123]

So, G84 and G86 would be brought to market at performance mid-range prices, taking advantage of a temporary lack of DX10 competition.

That is precisely right. It's the same reason why the 8800 GTX debuted with an MSRP of $649, and why the price has only gone down a little bit since then. NVIDIA (and/or board partners) is taking advantage of the current absence of competition to boost their profit margins.

 

[Geeforcer]

That is precisely right. It's the same reason why the 8800 GTX debuted with an MSRP of $649, and why the price has only gone down a little bit since then. NVIDIA (and/or board partners) is taking advantage of the current absence of competition to boost their profit margins.

AMD has nothing to compete with G80, but they (and ironically enough, Nvidia themselves) have plenty to compete with G84. The reason why G80 was a success had a lot less to do with DX10 then with utterly overwhelming performance advantage. It was a no-compromise product, if you could afford it. G84, not so much.

 

[CarstenS]

Yeah but wouldn't that border on the ridiculous in terms of texturing? That would be like a 45GT/s bilinear fillrate at 700Mhz ! I can't believe that G80's shaders are that texture bound right now.

Why would it be?

I thought along the same lines, but then someone who has a far greater understanding on all these matters than me pointed me towards CUDA.

 

[vertex_shader]

Oh, you got some insider info?

A-ha, so they have a bunch of monkeys with typewriters writing their drivers instead of experienced programmers? Or do they only employ low-IQ people?

Link

 

[_xxx_]

Link

So planing with less manpower or time now equals "having very big problems understanding vista at the driver level"? I don't really think so

 

[Arun]

I thought along the same lines, but then someone who has a far greater understanding on all these matters than me pointed me towards CUDA.

I fail to see how that has anything to do with CUDA... What matters there is the ALU-TEX ratio more than anything, and since that'd increase the number of transistors allocated to TEX, it'd be more of a negative than a positive...