there is no NV47, but G70 is on the way

[DegustatoR]

Spring 06' and then Summer 06' a WGF2.0 GPU? That's what I am wondering about. WGF2.0 GPUs still sound to me like late 2006 or even later products.

Next NVs hi-end GPU should be WGF2-capable imho. Whether this will be in Spring'06 (optimistic), Summer'06 (realistic) or Autumn'06 (pessimistic ) is hard to guess atm.

The instead part would make more sense, only I just don't have the slightest idea with what kind of memory they'd build it with

I think they'll use the same approch as NVIDIA with NV47 oops G70 That is -- more internal shading power with the same external frame buffer output.

 

[Geo]

Prices of high-end graphics cards have been going up pretty steadily. The GeForce DDR started at $300. The GF2 Ultra was really high-priced ($400 or $500, I forget exactly).

Yeah, GF2 Ultra, Aug 2000, $500. That was MSRP; somehow I seem to recall street was even higher for a little bit.

Here's a trip down memory lane --gotta love the historical graphs on bandwidth usage. On the one hand, it seems like we are still in a death struggle with the same issue --on the other, all that kvetching over 7gb/s.

http://graphics.tomshardware.com/graphic/20000814/index.html

 

[JoshMST]

I agree with Rys, this is all about margins. While I think they have good supply of chips onhand, selling it for that price with 256 MB of memory is all about getting a good chunk of change for both NVIDIA and the board partners.

I am guessing the Ultra version will have 512 MB of memory

Those willing to shell out $800 to $1000 for that little number...

 

[PatrickL]

I am a bit dubbious about that because at least in france none was offered to preorder when I checked online yesterday.

 

[JoshMST]

Of course none were offered for pre-order, the part hasn't even been officially released yet! NVIDIA is keeping everyone as clammed up about this one as possible.

 

[Geo]

Of course none were offered for pre-order, the part hasn't even been officially released yet! NVIDIA is keeping everyone as clammed up about this one as possible.

Yeah, the BFG page disappeared pretty quickly once it hit the net it was there (they hadn't had it linked --their webmaster must be snarling at all in sight). Tiger Direct had a page up with pricing, haven't checked this morning if it is still there.

This would be the downside for an IHV with simultaneous announce/availability --enough people are in the loop that the sieve is even more leaky than usual, so you lose to some degree your ability to make the big splash on announce day, at least amongst the hardcore.

 

[Geo]

I'm not convinced they've got a real load of chips to sell (at all, really), but the price smells like "yay, margins!!!!!!!" to me.

Well, it does protect GF6 pricing that much longer as well, so you may be right. It also suggests they don't expect an R520 with comparable/better performance to be putting price/performance pressure on them for a bit, which is consistent with what we're hearing and supports your "make hay while the sun shines" theory.

 

[JoshMST]

Yeah, and they also have to unload all of those 512 MB 6800 Ultras! Wouldn't make much sense to offer a 512 MB 7800 at around the same price as the 6800 U 512MB. This gives those guys a couple extra months to try to pawn off those monstrosities before ATI releases their 512 MB high end card, and NVIDIA has to match that bit of action.

 

[swaaye]

It's pretty interesting to me how they're blurring their production lines though. With 3 variants of 6800s with the same chip they can reduce production costs of the ultra-high end parts, because they aren't really all that special but are just higher clocked mid range parts. It boosts their margins since they aren't REALLY building low volume, high end parts and don't have to retool production lines.

Obviously it doesn't work out ideally usually though because NV has developed things like NV42 and NV41 which are pure 12-pipe chips, not 16-pipe defectives. I have a NV42 in my notebook. Same thing with the 9500Pro debacle where ideally they would be able to make use of expensive, defective high-end chips in mid-range cards, but the yields were better than they projected so it ended up being a financial screw up.... Well maybe not a screw up, but they didn't make as much money as they could've if they'd planned their product range differently.

 

[Farid]

Latest L'Inq tentative for some pages views before the NDAs are considered history
http://www.theinquirer.net/?article=24068

Dell's favorite 3D benchmark, 3Dmark05 will score 7600 on a non-overclocked machine, while two cards in SLI can get 11200. Although we all kind of expected more, it’s a nice score. And SLI could kind of provide even more.

Two 6800 Ultras from Gainward clocked at 425MHz can score 10200 - just a thousand less than two super-duper 7800 GTX cards. So whoever launched the story claiming a 7800 score for Geforce 7800 GTX was not too off at all, even if the name comes from Nvidia's proven naming nomenclature.

 

[wireframe]

Yeah, and they also have to unload all of those 512 MB 6800 Ultras!

I'm sure that won't be a problem. They will sell...both of them!

 

[ClyssaN]

Ail - when are your expectations for R580 and G70 refresh?

so you are saying that the card competing with the g70 is the r580 ?

 

[N00b]

Ail - when are your expectations for R580 and G70 refresh?

so you are saying that the card competing with the g70 is the r580 ?

I'm pretty sure he meant it this way:

Ail - when are your expectations for R520 refresh (=R580) and G70 refresh?

 

[ClyssaN]

Ail - when are your expectations for R580 and G70 refresh?

so you are saying that the card competing with the g70 is the r580 ?

I'm pretty sure he meant it this way:

Ail - when are your expectations for R520 refresh (=R580) and G70 refresh?

i guess ...

 

[Geo]

I certainly hope that the 3dm scores we've seen aren't really indicative of what G70 can do, otherwise NV will be guilty of what they used to mock ATI of "back in the day" --coming out with a card to beat the competitions current generation instead of their next.

Looks like roughly 30% stock vs stock against 850xtpe.

But then this might be another implied argument for that presumptive Ultra they still have in their pocket.

 

[digitalwanderer]

I certainly hope that the 3dm scores we've seen aren't really indicative of what G70 can do

I don't think they are, I think 3dm2k5 scores will be totally out of whack with actual performance this generation.

 

[martrox]

I don't think they are, I think 3dm2k5 scores will be totally out of whack with actual performance this generation.

Can't do anything but hope so........it's about time to lay Futuremark to rest.....

 

[KimB]

Obviously it doesn't work out ideally usually though because NV has developed things like NV42 and NV41 which are pure 12-pipe chips, not 16-pipe defectives. I have a NV42 in my notebook. Same thing with the 9500Pro debacle where ideally they would be able to make use of expensive, defective high-end chips in mid-range cards, but the yields were better than they projected so it ended up being a financial screw up.... Well maybe not a screw up, but they didn't make as much money as they could've if they'd planned their product range differently.

Well, I doubt that. More likely it's the natural way things work: as times goes on, yields improve, and nVidia has a product sitting on the back burner ready to launch once yields improve past a certain point. Once they do, the second product comes forward and takes the place of the ones with disabled parts.

 

[digitalwanderer]

Can't do anything but hope so........it's about time to lay Futuremark to rest.....

Actually 3dm2k5 pretty much did lay 'em to rest for me, it's definately one I put almost no stock whatsoever by.

I still like 3dm2k1, and I never really did like 3dm2k3...it's definately looking that FM's useful life-span is rapidly reaching it's end.

 

[Geo]

I still like 3dm2k1, and I never really did like 3dm2k3...it's definately looking that FM's useful life-span is rapidly reaching it's end.

The problem tho, is that niche is a useful one, going back to ZD WinBench (which got "gamed" way worse than 3DM from my memory) thru the various iterations of FM 3DM.

The community needs a ubiquitious easy/free to use/run benchmark to swap l/i/e/s/ scores with and get a easy thumbnail sense of what kind of improvement they'll see if they invest in an upgrade, and whether maybe their system has some issues somewhere when their box doesn't score within reasonable % of likewise boxes. Not everyone has D3/HL2/FC [fill in your favorite here] to compare with.

If FM is going down in flames, that will leave a big hole in my estimation.