PDA

View Full Version : What do you expect of the NV35? Here's what I expect.


Pages : [1] 2 3 4 5

Uttar
02-15-03, 03:42 PM
Hey everyone,

After seeing a lot of NV35 rumors and stuff, I was wondering what most think it'll deliver. And I've also wondered what I really think it'll deliver, too. So, I'd like to have some feedback as to what you expect out of it, too :)

nVidia seems to be sleeping. They seem to be focusing on their mainstream parts ( NV31 and NV34 ) - but are they, really? Well, if you'd seriously think that, you'd be dead wrong.
nVidia has three design teams for its GPUs. Each works in parallel, and everything is scheduled in the goal of having a product out every 6 months. If one team is delayed, the others aren't. That's the beauty of the system.
The first team worked on the NV30, and now works on the NV40
The second team works on the NV31&NV34
The third teams works on the NV35&NV36 ( although they won't be released at the same time )
Efficient, eh? That's the power of nVidia: parallism. They put it in their GPUs, in their motherboards, and even in their job methodology! If one of their engineer said he was against parallelism, I'd bet big time he'd be fired on the spot :)
But back to the topic... The NV35.
There are several things we know about the NV35. Those things have been confirmed by numerous sources on forums. They are:
- Low K process
- 256-bit memory bus
- GDDR-2 ( -> 30GB/s+ bandwidth )

That's already tasty stuff. But, is there more? That's a good question. And we've got to think about what nVidia priorities are, if we want to figure it out. Thus, what follows is speculation. But I'll give a pessimist, a conservative and an optimist version, because it's really hard to be objective with speculation.

What was the main problem with the original product? Well, reviews certainly weren't kind. The main problems? Bad IQ with Aggressive AF, forcing you to use balanced AF. And lower AA quality, still giving lower overall quality even with balanced AF.

To fix the AF problem, the obvious solution is to increase the quality and performance hit of the aggressive algorithm ( or evantually keep the algorithm, and add another one calling it something else ) - fundamentally different algorithms are unlikely to be seen. 16x is also possible, but unlikely.
Thus, for AF, the following is conservative: revised algorithm, with better quality ( comparable to ATI's Performance ) but significantly lower performance.
The following is optimist: revised algorithm, with better quality ( comparable to ATI's performance ) but slightly lower performance. Support for 16x Aniso
And the following is pessimist: nothing new

For the AA problem, it's however slightly more complex. There are, really, two problems. The first one are the worse sampling patterns. The second one is that the maximum true MSAA mode is 4x.
To fully fix that, you'd thus need to use rotated patterns for every mode and have a higher native full MSAA mode. But that also means a higher maximum color compression ratio...
Thus, for AA, the following is conservative: Native 6x AA mode, rotated sampling patterns which are roughly comparable ( sometimes worse ) to ATI patterns
The following is optimist: Native 8x AA mode, rotated sampling patterns which are roughly comparable (sometimes better ) to ATI patterns.
The following is pessimist: Either native 6x AA mode, or rotated sampling patterns which are roughly comparable to ATI patterns

And, what other problems was there? Oh, yes, much lower than theorical fillrate. I really don't think it's possible to predict anything about that. It could come from drivers, from minor hardware bugs, or major design problems. It doesn't make sense to speculate on that without ANY info at all...

On the features front, little is to be expected. CineFX is already very flexible, and more would kinda be overkill for now.

Now, as to what influences performance... The first thing is obviously pipeline number. This will remain identical to the NV30, with some slight optimizations to things like scheduling maybe. Something else is vertex shading per-clock efficiency. Both number of units ( which is still unknown for the NV30 ) and optimizations got to be counted for that. Here's the speculation:
conservative: 5% higher per-clock than NV30
optimist: 20% higher per-clock than NV30
pessimist: same per-clock as NV30

What's left? Clock speeds for the high-end part... Well that's hard to guess. Really hard. The only likely thing is that they aren't trying to get a 1:1 ratio. So... let's get right to the numbers!
Conservative: 550/500
Optimist: 625/600
Pessimist: 500/450

That's great. But it's all a bunch of theorical figures and stuff. What about practical performance figures?
The NV35 true power is with 6x AA and 8x Aniso, with FP32 being used in places needing that kind of precision. Its quality is probably comparable with the R350 using 6x AA, 8x Performance Aniso and FP24 everywhere.
Here is what I expect, in terms of overall performance, for the NV35:
Conservative: 30% faster than the R350
Optimist: 50% faster than the R350
Pessimist: On-par with the R350
My overall guess, right now, is 35% faster. Obviously, when compared to the R400, it should be a quite different matter...


Uttar

-=DVS=-
02-15-03, 04:18 PM
Well.... obviosly that would be great if Nvidia would release such a wonderfull card realy soon :D with Better AA and AF then R300 plus alot more speed then R350 :) and without a dustbuster :drooling: in that case Nvidia would get a new customer.

But seriosly do you realy believe Uttar , that such a card will ship any time soon :rolleyes:

volt
02-15-03, 04:20 PM
I'm having a hard time believing it will not be delayed, based on your specs.

-=DVS=-
02-15-03, 04:25 PM
Even if all teams workeing parallel other teams will have to stop workeing becouse of main delay in core , becouse of the Low K , Nvidia did not confirm that they have managed to get Low K workeing.. and you can't make a card nor test for heat/speed stability if you can't make/produce a card in Low k :rolleyes:

So as long as Real NV30 (low K)is in delay all other projects can't move on...


Quite a Paradox eh :D lol

Uttar
02-15-03, 04:37 PM
Retail avaibility:
Conservative: Early August
Optimist: Early July
Pessimist: Early September
My current guess is mid July.

I'd also guess the price to be between $449 and $499 - hoping for $399 is *way* too optimistic IMO.

DVS: nVidia got Low K working a long time ago AFAIK. The problem is that it was very far from being ready for mass production, or even limited production.
I'd be surprised if it used traditional cooling. I wouldn't be surprised if nVidia provided multiple cooling solutions reference designs to their partners. So, if they wanted, they could use the dustbuster. Or they could use something even more expensive, but less noisy. Stuff like that.
The real question is how much more expensive those alternative cooling solutions will be...

As I said, that's what I expect. I could be wrong. But I'd be surprised if I was wrong on the overall picture...
A slight delay is possible, but it'll ship before the R400, that's for sure.


Uttar

-=DVS=-
02-15-03, 05:31 PM
Offcourse it will ship way before R400 , i expect R400 available in stores not sooner then next year :p but i might be off by alot :D

Whatever comes from Nvidia better come at same time as R350 i am not gonna wait for miracles from Nvidia :rolleyes:

I sure hope that your guess are true Uttar , all i want is more IQ and Better Performance then current R300 (doesn't matter witch company brings it) :p

StealthHawk
02-15-03, 11:16 PM
what is up with nvidia at no support for 16x AF? the gfFX is plenty fast at doing AF. i remember that 3DGPU reported that the original gf3 had hardware support for 16x AF but it was never enabled in drivers because it was too slow. the statement apparently came straight from nvidia, if i remember correctly.

Nv40
02-16-03, 02:56 AM
Originally posted by StealthHawk
what is up with nvidia at no support for 16x AF? the gfFX is plenty fast at doing AF. i remember that 3DGPU reported that the original gf3 had hardware support for 16x AF but it was never enabled in drivers because it was too slow. the statement apparently came straight from nvidia, if i remember correctly.


i dont know about AF ,but i predict a huge increase in antialising modes
for the Nv35 ,like 16x true AA ,the quadroFx boards already supports those settings for profesional market .. and even there are reviewers
showing how to hack your Geforcefx to Quadrofx board.

so its a matter of short time for better AA modes..

it is interesting to see that wihout any special hardware only drivers ,
all geforce3/4 boards supports Geforcefx modes..
same with ATi new catalist that goes up to 16x in AA..

my dream card ,should be able to play DOOM3,quake4,halflife2
at least at 1600x1200 AA4x/AF8x in 60fps ... or 2x times the GeforceFX -ultra perfromance :)

i think/(hope) that an NV35 with an atlonxp 3000+ be more than enough
for my next System.. :thumbsup:


:matrix:

Uttar
02-16-03, 06:02 AM
DVS: Yes, you're off by a lot on your R400 shipping date :) Everything indicate it'll be in stores well before Comdex in a worst case scenario. Probably between September and October, but that part is just speculation.

Stealth: Agreed, it's really strange. But, logically, if the original product ( LOL, note how I'm trying not to say the name :P ) had 16x Aniso, wouldn't the nVidia marketing machine have jumpted on it?

NV40: 16x AA? Unlikely. Even with 256MB of RAM, you still wouldn't be able to use it at high resolutions. AFAIK, if it's enabled in the Quadro FX, it's because sometimes professionals may have uses for having a small thing rendered with amazing quality.

And Doom 3 at 1600x1200 4xAA & 8xAF? Certainly not with a NV35. The amount of fillrate it would need is just way too impressive... Maybe you could get that at 1280x960 when using 4xAA and 8x aggressive AF... Maybe... But probably not more.

The good news with Doom 3 is that it seems very fillrate limited, so AA should be quite cheap.


Uttar

nutball
02-16-03, 07:56 AM
Uttar, whilst what you say about parallelism is true to a degree, there must by definition be at least some degreee of synchronisation between the NV30 and the NV35 teams, surely?

Unless NV35 is a completely independently developed core (which I assume it isn't), the basic design of NV30 has to be pretty well sorted out before the NV35 team can pick it up and run with it, surely?

Now if the NV30 feature freeze, functionali layout (whatever it's called, whatever is important) was on-time (by some sort of "original" schedule), then I might agree NV35 will be "on time". However if the delays in NV30 were due to fundamental flaws in the design (which has been suggested in another place) then I don't see how NV35 can be "on-time".

tazdevl
02-16-03, 04:23 PM
I think your July estimate is a bit far fetched for the NV35 given the specs. Also, those look more like NV40 specs, not the NV35.

I admire your optimism when it comes to nVIDIA, but I think you need to get down to reality. NV40 should be very impressive, but I think you're expecting too much from the NV35.

AngelGraves13
02-18-03, 11:23 PM
I expect it to be twice as fast as the nv30, why do you ask? Because when nvidia was asked about a refresh they said "what's that....we're working on something twice as fast!"

Bigus Dickus
02-19-03, 06:39 PM
Uttar, you forgot to mention gamma correction in your speculation on AA. ;)

borntosoul
02-19-03, 08:29 PM
it would be great to see the nv35 with 30+gigs of b/w . does anyone know if it will come with 256 megs ram (even though i doubt it for obvious reasons) .

borntosoul
02-19-03, 08:41 PM
oh yeh has anyone worked out why the nv30 has crappy minimum fps on unreal torn 2003 ?http://www.anandtech.com/video/showdoc.html?i=1779&p=6 it seems ordinarely low .i hope its just early drivers .

MuFu
02-19-03, 08:43 PM
I think they will fix ZCull with stencil ops - ATi have fixed R350 Hi-Z in such situations and JC has had input in NV35 design - likely he would have requested they try to get that issue sorted (amongst other things).

Perhaps double the number of rendering pipelines... work in progress getting to the bottom of this idea. :p

Originally posted by borntosoul
it would be great to see the nv35 with 30+gigs of b/w . does anyone know if it will come with 256 megs ram (even though i doubt it for obvious reasons) .

Yes it can support 256MB RAM (as can NV30) - depends on cost whether we'll see such a solution brought to market. Perhaps only in the pro sector. The 256MB board has 16 memory chips, 8 on each side, horizontally opposed (2 chips per partition w/128MBit RAM). For an 8-chip, 256MB board they'll need Samsung's 256MBit BGA, but that is only available in ES form right now (ramp not until Q3).

MuFu.

borntosoul
02-19-03, 11:04 PM
seems to me that evreyone is sure that its going to be 256 bit bus but has there been any official word on it ,i know the ceo said that it will be twice as fast as the nv30 maybe he means bandwidth ,anyway itll be cool if it was

Lezmaka
02-19-03, 11:10 PM
There won't be any official word about anything until the day it's unveiled.

Ady
02-20-03, 10:09 PM
Originally posted by 99 to Life
I expect it to be twice as fast as the nv30, why do you ask? Because when nvidia was asked about a refresh they said "what's that....we're working on something twice as fast!"

did they say it was the nv35 they are working on. Perhaps they are reffering to the nv40.

Uttar: nVidia, Ati and many other companies have multiple design teams. It's definitely the smart way to go. I wonder how they work the refreshes though. As Nutball said, surely they can't pick up production until the chip they are making the refresh on is completed.

Ady
02-20-03, 10:22 PM
Originally posted by Uttar
The first team worked on the NV30, and now works on the NV40, The second team works on the NV31&NV34
The third teams works on the NV35&NV36

Just noticed this..

If you're right on this Uttar, it seems like a good reason why nvidia is having trouble keeping up. It seems like it would be an awfully tough task for the "first team" to go straight from a nv30 into designing a another "nextgen" chip. That would give them less than a year for design and production. I know that Ati worked it so the team that worked on the R200 went straight on to the R400. The team that designed the R300 went straight on to the R500. this gives them alot more breathing space.

borntosoul
02-20-03, 11:11 PM
so maybe for example it worked something like this
team 1 done gf3............ team 1 start gf4
team 2 done gf3 ti......... team 2 start gf fx
team 1 done gf4............ team 1 start gf fx2
team 2 done gf fx...........team 2 gone missing
team 1 done gf fx2.........team 1 start gf fx3
team 1 done gf fx3.........team 1 start gf fx4

Uttar
02-21-03, 02:17 AM
Originally posted by Ady
Just noticed this..

If you're right on this Uttar, it seems like a good reason why nvidia is having trouble keeping up. It seems like it would be an awfully tough task for the "first team" to go straight from a nv30 into designing a another "nextgen" chip. That would give them less than a year for design and production. I know that Ati worked it so the team that worked on the R200 went straight on to the R400. The team that designed the R300 went straight on to the R500. this gives them alot more breathing space.

Hmm, now that you mention it, I could be wrong.
Or nVidia could be doing the wrong thing.
Anyway, there could be a good way to explain what I said - nVidia doesn't have everyone in those "teams".
So, some people would work on the concept and basic design of the NV40 before any team is even able to work on it.
That's pretty much what someone at nVidia said a while ago when it was asked if we'd see GigaPixel tech in future chips...
"It's not the time yet, but we've got a lot of smart people looking at it".

borntosoul: There are three teams, not two! :)

BigusDickus: LOL, now the GFFX AA would still be inferior, even with my changes! Good catch! :)


Uttar

borntosoul
02-21-03, 02:24 AM
uttar that was a joke :)

Uttar
02-21-03, 11:30 AM
Originally posted by borntosoul
uttar that was a joke :)

LOL, didn't figure out the "Gone Missing" part immediatly! :o


Uttar

GlowStick
02-21-03, 08:21 PM
I feel the NV35 will be released as it was orginaly scheduled. I belive they can get it out on time, worst case they have to hire more people to make it work on time.

The only thing for Nvidia to do when ATI pushes them, is to push back harder. Then again i like to take all things on one last gamble :D