nv40 ....nv45 for next gen games

[borntosoul]

now is not right time to invest for the next generation games with the nv30 or nv35 ! spending $400/$500 just to have that thought of future proof is a waste IMO, speeding up todays games is a different matter all together:). it will be interesting to see how taxing (on the gpu) the other games based on the doom3 engine will be! the net step up obviously will be nv40 and that wont be a huge jump in performance over the nv35, that will happen when .09 micron and gddr3 become available sometime mid next year (around the time for the nv45). everyone wanting much better dx9 speed will hopefully get it by then. next windows will be dx9 based (not dx10). its obvious that we can never keep up with the fast moving pc industry but the timing of our purchases can make a difference. instead of buying now for doom3 atleast wait till the game is out and tested and by that time we will have faster gpu's anyway.

[mikechai]

The cycle will never end...

[pat777]

I'm excited about the NV40 because I think nVIDIA is actually going to work hard on it instead of slack off like they did with the NV3X series. It'll probably stomp ATI's R400 due to the fact nVIDIA probably learned a lesson from developing the NV3X series.

[Philibob]

I've decided to buy the 5900ULTRA this time, I've waited for the next product since the GF2MX.

[borntosoul]

hehe youre in the same boat as me :) i might just get a cheaper 3d card just to fill in between now and mid next year , see how cheap the nv35 value is and how it performs. ill have to get a loan card from my bro just to keep me going.

[Uttar]

Originally posted by pat777
I'm excited about the NV40 because I think nVIDIA is actually going to work hard on it instead of slack off like they did with the NV3X series. It'll probably stomp ATI's R400 due to the fact nVIDIA probably learned a lesson from developing the NV3X series.

No idea it would have stomped the R400, considering the real R400 is RIP.
If you're refering to Loci, well, it'll be an interesting battle IMO.


Uttar

[Paul]

borntosoul: The NV40 should represent the same jump as normal refresh -> next gen chips almost always give. Think NV25 to R300 :) (If everything goes as planned, of course).

Loci vs. NV40 will be a pretty much straight heat in current games, with Loci *perhaps* coming out on top. But the NV40 is definitely the more "advanced" of the two, and should pull ahead comfortably when the 2004 games come out.

I guess the same things could have been said before the NV30 release, but there you go.

[Uttar]

Well, I believe the jump likely won't be as big as NV25->R300

My guestimate is more like NV15->NV20 - both in features and performance, actually :)

Some things run a LOT faster ( AA, AF in the NV20 case, shaders in the NV40 case I'd bet ) , most things run faster, but not by a lot and improved by drivers.

In the NV20, the 2x AA quality was improved. Let's hope AA quality will also be improved in the NV40, hopefully more than that :)

Also, the overall IQ was improved in the NV20 by the way of making higher IQ features, mostly 32BPP, cheaper.

Finally, in the NV20, you had a TON of new features, mostly shaders. In the NV40, you'll get the PPP, VS3.0., PS3.0. and a lot more. It's great stuff, really.


Uttar

[GlowStick]

WEll on the 'dont upgrade now, wait' part, there is never a good time to upgrade. Almost with in a half year your card will be outdated, but the way you can get the 'most' time out of your card is to buy a card the day its released!

EG.

9800pro's started shipping monday (preorders got thursday/friday/saturday) i got mine saturday. so it worked out.

[suburbanguy]

obviously you can say NV35 isnt good enough for today's or tomorrows games, that NV40 and NV45 will be what you want, but then NV50 and NV55 will be coming next when NV45 is here and its the same thing all over again.

[StealthHawk]

Originally posted by Paul
borntosoul: The NV40 should represent the same jump as normal refresh -> next gen chips almost always give. Think NV25 to R300 :) (If everything goes as planned, of course).

In other words, heed this disclaimer.

NV40 will only bring a significant increase in performance at certain times, mainly with FSAA. One reason is that nvidia will switch to MSAA 6x from 6xS, and they will claim a 600% peformance increase :p

[mikechai]

It takes years to gain trust and NVIDIA certainly has lost it.
They will have to work 2x as hard to slowly gain it back.

[borntosoul]

Originally posted by suburbanguy
obviously you can say NV35 isnt good enough for today's or tomorrows games, that NV40 and NV45 will be what you want, but then NV50 and NV55 will be coming next when NV45 is here and its the same thing all over again.
yes thats true. like i said before- we can never keep up, but i still think timing makes a bit of difference.

added this bit.....
well im just saying that doom3 and all the games based on that engine will be coming out after new years. so buying early/mid next year would be better if your thinking getting the most out of those games. buying extreme high end now to still be lacking next year is a waste of cash considering the premium users have to pay for those cards.

[borntosoul]

Originally posted by Paul
borntosoul: The NV40 should represent the same jump as normal refresh -> next gen chips almost always give. Think NV25 to R300 :) (If everything goes as planned, of course).

paul, that jump needs some big horse power, i believe we will get that type of jump from nv40 to nv45. nv 40 should fix most of the weak spots on the nv3x line. nv45 should be brute power to push dx9 shaders to usable levels.

[Paul]

Originally posted by StealthHawk
In other words, heed this disclaimer.

NV40 will only bring a significant increase in performance at certain times, mainly with FSAA. One reason is that nvidia will switch to MSAA 6x from 6xS, and they will claim a 600% peformance increase :p

Well yeah - Of course the huge jumps won't come in every single area. But there will be definite gains across the board, no doubt.

Speaking of MS/FSAA however, I'd really like nV to simplify the whole thing for NV40, and cut a lot of the modes out. I think it would be a refreshing change to just scale it back to 2x, 4x, and 6x - But make genuine performance and graphical differences between the three, rather than having no end of different configurations (as now), each offering very little difference in terms of quality.

Originally posted by borntosoul
paul, that jump needs some big horse power, i believe we will get that type of jump from nv40 to nv45. nv 40 should fix most of the weak spots on the nv3x line. nv45 should be brute power to push dx9 shaders to usable levels.

Actually, i'd be very suprised if the refresh jump was bigger than the generation jump. Refreshes almost never give those kind of performance benefits, except in certain circumstances (ala FSAA on the NV35).

Shaders will be pushed very hard in the NV40 from launch, i'd imagine. nVidia would really be shooting themselves in the foot if they didn't support it properly.

[Uttar]

Well, IF nVidia manages to have dynamic resource allocation working ( I didn't hear anything stating they had problems thus far, so I'd imagine they did it right on the first try. Contrary to *cough* ATI *cough* - also their problem might be related to something else, who knows ) - then it's gonna be VERY impressive.

Imagine a complex pipeline with bottlenecks.
And imagine a simple line.

I think you've got my point ;)
Okay, so that's exagerated, because many things still obviously can be bottlenecks: memory bandwidth, for one.

But really, always having the computational units used at the maximum would be a great thing IMO :)
Heck, old games would benefit a LOT! You could simply use the fragment units to do vertex lighting calculations too, for example. Imagine the performance boost! :eek:

Of course, there are few direct indications that the NV40 will use dynamic resource allocation, so let's not put our hopes TOO high yet :)

My guess is that nVidia will keep the NV4x architecture for a full 24 months, however. They've said that their believe product cycles will lengthen soon, and I think this will be when they will.


Uttar

[borntosoul]

Actually, i'd be very suprised if the refresh jump was bigger than the generation jump. Refreshes almost never give those kind of performance benefits, except in certain circumstances (ala FSAA on the NV35).
i dont think its about the refreshes i think its about what technology is available to them and what they make use of at the time. ill say this again that .09 micron and gdd3 will make a huge difference IMO

[Paul]

Originally posted by Uttar
My guess is that nVidia will keep the NV4x architecture for a full 24 months, however. They've said that their believe product cycles will lengthen soon, and I think this will be when they will.

I think what is most likely to happen is something along these lines:

Nov/Dec 2003: NV40
July 2004: NV40 Speed Bump
Nov/Dec 2004: NV45
July 2005: NV45 Speed Bump
Nov/Dec 2005: NV50

The speed bumps are really only there to appease the OEMs, and I doubt they'd make a big deal out of marketting them. So for the general public, they will only see something "new" once a year.

Of course, if they wanted to really be agressive, they could push the NV50 up to Easter 2005, cutting out the speed bumps all together, and bringing the NV45 forward a little. But that would likely mean some cutbacks in the architecture somewhere. Having said that, some early NV50 work was being done quite a few months back, so perhaps it's not all that unrealistic.

It amazes me that, given nVidia missed a whole product cycle, they aren't far behind at all, and could soon push ahead. Some serious luck for them with the R&D problems at ATi.

[The Analog Kid]

Originally posted by pat777
I'm excited about the NV40 because I think nVIDIA is actually going to work hard on it instead of slack off like they did with the NV3X series. It'll probably stomp ATI's R400 due to the fact nVIDIA probably learned a lesson from developing the NV3X series.

No, they don't have to work any hard, they just need to get rid of the 3dFx developers. Think about it, they didn't even put out 32-bit rendering on their cards.

[ReDeeMeR]

Well I'll definately wait till Doom3, seeing as how it's still not clear whether Nvidia really did "specialy" optimized that Doom3 benchmark and ATi was left out with bugs.

In my theory if *****Box is capable of rendering Doom3 then my rig should be twice if not more faster, lets say my CPU and RAM are being ****ed up by windows, but video card isnt and it's more powerfull then Xbox NV3A.

ATi's drivers are still holding me back from purchasing one, so I've decided to wait :)

[StealthHawk]

Originally posted by The Analog Kid
No, they don't have to work any hard, they just need to get rid of the 3dFx developers. Think about it, they didn't even put out 32-bit rendering on their cards.

Voodoo5 supported 32bit color, so did Voodoo4 :rolleyes:

[reever2]

Will the speculation ever end?

[rwolf]

Originally posted by GlowStick
WEll on the 'dont upgrade now, wait' part, there is never a good time to upgrade. Almost with in a half year your card will be outdated, but the way you can get the 'most' time out of your card is to buy a card the day its released!

EG.

9800pro's started shipping monday (preorders got thursday/friday/saturday) i got mine saturday. so it worked out.


That isn't always true the 9700 has been out for a while and it wouldn't be worth upgrading to a 9800 or 5900.

If you bought a 5200, 5600, or 9000 you might want to upgrade again in a couple of months. ;)

[GlowStick]

Originally posted by rwolf
That isn't always true the 9700 has been out for a while and it wouldn't be worth upgrading to a 9800 or 5900.

If you bought a 5200, 5600, or 9000 you might want to upgrade again in a couple of months. ;)

Yeah, i hear that alot from people who dont have a 9800Pro and a 5900 ultra ; D

[rwolf]

Oh sure, rub it in....