NV50 specs?

[Natoma]

Well, considering the NV30 was supposedly plagued with many fabrication problems, was possibly released as a much weaker chip than was originally intended, and was nVidia's worst showing to date as it compared against the competition, I would tend to think that nVidia will have a much easier time massively increasing performance (particularly in FP shaders) than ATI, and will definitely have an easier time massively increasing performance this time than next (provided the NV40 isn't a pile).

TNT vs Voodoo 2/SLI? :?

 

[T2k]

NV50 guesstimate
*512 Bit Bus

BS. There's no reason for that.

*32 pixel pipelines (32x1, 64x0)
*16~32 MB embedded memory
*512 MB external memory

Another funny thing: why do you need half a gig memory if you have 32MB embedded?

*VS/PS 4.0+
*1~2 billion verts/sec peak

Why is that?

 

[KimB]

TNT vs Voodoo 2/SLI?

And, if you remember, the TNT was meant to be originally released at .25 micron. This did not happen, and the original TNT was released at .35 micron at much lower clockspeeds. nVidia's final .25 micron product, the TNT2, did usurp the Voodoo2 SLI.

 

[Natoma]

NV50 guesstimate
*512 Bit Bus

BS. There's no reason for that.

*32 pixel pipelines (32x1, 64x0)
*16~32 MB embedded memory
*512 MB external memory

Another funny thing: why do you need half a gig memory if you have 32MB embedded?

You haven't played Painkiller or UT2K4 have you.

I have a 256MB card and with all the texture details up, it can chug at times. More is always better imo.

 

[Natoma]

TNT vs Voodoo 2/SLI?

And, if you remember, the TNT was meant to be originally released at .25 micron. This did not happen, and the original TNT was released at .35 micron at much lower clockspeeds. nVidia's final .25 micron product, the TNT2, did usurp the Voodoo2 SLI.

You're assuming ATI is going to pull a 3dfx though and fall off a cliff. You say that ATI doesn't have an easy a time as Nvidia in increasing performance. I dunno, I think I disagree with that. ATI pulled the 9800XT out of its arse on a 0.15 micron process. From what we've seen in the overclockability of its 0.13 micron process parts, it seems that even they have a lot of headroom in them.

With Nvidia struggling with a 0.13 micron process part in ramping up speeds without using exorbitant cooling, I don't know if they really have that ability moreso than ATI.

Remember, ATI has years of experience in low power, low heat laptop modules, while Nvidia has always thrived on bigger/faster/better/smaller process in order to make its parts effective. That eventually came back to bite them in the ass when the process wasn't fully mature. Will this turn around in one generation? Doubtful imo. That's how I see the situation atm.

 

[KimB]

Another funny thing: why do you need half a gig memory if you have 32MB embedded?

The question should be:
Why would you need a 512-bit bus if you have 32MB embedded?

Anyway, only time will tell whether or not we go for embedded RAM anytime soon. I, for one, doubt it. Eventually we may go there, but not yet. We just aren't that memory bandwidth limited right now, and more advanced shaders should serve to further reduce the memory bandwidth needs of GPUs (in comparison to the amount of processing power needed).

 

[KimB]

You're assuming ATI is going to pull a 3dfx though and fall off a cliff. You say that ATI doesn't have an easy a time as Nvidia in increasing performance.

No, what I'm saying is that ATI's FP shader performance is already high, and thus it'll be hard to increase it much further. nVidia's is not, and so has much further to go.

 

[Joe DeFuria]

No, what I'm saying is that ATI's FP shader performance is already high, and thus it'll be hard to increase it much further. nVidia's is not, and so has much further to go.

Yes and no.

When you add in driver hacks and cheats nVidia's performance is clost to par with ATI's in terms of "game / benchmark performance." It would be just as hard for nVidia to "beat their own cheats" as it would for ATI to beat their current legitimate performance.

 

[KimB]

In most games nVidia doesn't have to resort to any sort of cheats to achieve performance parity with ATI's cards, since FP shader performance is a relatively small part of performance in these games. It's the FP performance where nVidia suffers.

 

[Joe DeFuria]

In most games nVidia doesn't have to resort to any sort of cheats to achieve performance parity with ATI's cards, since FP shader performance is a relatively small part of performance in these games. It's the FP performance where nVidia suffers.

Um right. And it's the FP performance improvements we're talking about here, right?

 

[KimB]

Well, that's pretty much the only place the NV30 suffers. It's not a bad-performing chip otherwise.

 

[Joe DeFuria]

Well, that's pretty much the only place the NV30 suffers. It's not a bad-performing chip otherwise.

??

We're talking about how difficult it would be, relatively speaking, for ATI and nVidia to "improve" PS performance, right?

Currently, nVidia hides the deficiencies it has, through driver hacks and cheats. So in effect, nVidia has to "improve" its shader performance beyond that which its drivers already compensate for the lack of performance.

 

[T2k]

NV50 guesstimate
*512 Bit Bus

BS. There's no reason for that.

*32 pixel pipelines (32x1, 64x0)
*16~32 MB embedded memory
*512 MB external memory

Another funny thing: why do you need half a gig memory if you have 32MB embedded?

You haven't played Painkiller or UT2K4 have you.

I have a 256MB card and with all the texture details up, it can chug at times. More is always better imo.

FYI: I do have a Special Edition DVD set of UT2004...
And I'm playing UT2004 on daily basis w/ my 9700 Pro based rig (check my sig) - I never had any problem on my machine in 1280x768 and 4xaa+8x q af and everything maxed out in game details.

PS: I';m even playing the demo in my office occasionally: A64 3200+ w/ 1GB, R9700 Pro and same settyings I never had any problem there either...

 

[T2k]

Another funny thing: why do you need half a gig memory if you have 32MB embedded?

The question should be:
Why would you need a 512-bit bus if you have 32MB embedded?

Yeah, actually that's a good point as well.

 

[anaqer]

Natoma :


Not sure if that's a problem with video memory... unchecking the "preload player skins" option in UT2004 should help smoothing out gameplay hitches.

 

[Lezmaka]

TNT vs Voodoo 2/SLI?

And, if you remember, the TNT was meant to be originally released at .25 micron. This did not happen, and the original TNT was released at .35 micron at much lower clockspeeds. nVidia's final .25 micron product, the TNT2, did usurp the Voodoo2 SLI.

You're assuming ATI is going to pull a 3dfx though and fall off a cliff. You say that ATI doesn't have an easy a time as Nvidia in increasing performance. I dunno, I think I disagree with that. ATI pulled the 9800XT out of its arse on a 0.15 micron process. From what we've seen in the overclockability of its 0.13 micron process parts, it seems that even they have a lot of headroom in them.

With Nvidia struggling with a 0.13 micron process part in ramping up speeds without using exorbitant cooling, I don't know if they really have that ability moreso than ATI.

Remember, ATI has years of experience in low power, low heat laptop modules, while Nvidia has always thrived on bigger/faster/better/smaller process in order to make its parts effective. That eventually came back to bite them in the ass when the process wasn't fully mature. Will this turn around in one generation? Doubtful imo. That's how I see the situation atm.

I don't know much about the chip designing process, so if anything here is wrong, feel free to correct me.

It would seem the NV30 has a more complex design than the R300. My thoughts are that nvidia spent the majority of the time just getting NV30 to function and didn't have much time to optimize for heat/clockspeeds. If R300 is easier to put into silicon, ATI could spend more time fine-tuning the layout, getting rid of hotspots, etc.

 

[Arun]

AFAIK, NVIDIA targets the NV60, and not the NV50, to have a 512-bit memory bus.
Oh, and completely OT, talking of huge amounts of external memory: what I'd like to see, personally, is varying speeds of memory for different types of data. For example, you could have 512MB of low-end mainstream DDR for textures, 128MB of super-mega-high-end GDDR3 for the Z buffer maybe, and then another 128MB for the Z-Buffer.

Or, you could just go 3DLabs' way: varying memorybus width (example: http://www.3dlabs.com/product/wildcat4/specs.htm)


Uttar

 

[stepz]

Oh, and completely OT, talking of huge amounts of external memory: what I'd like to see, personally, is varying speeds of memory for different types of data. For example, you could have 512MB of low-end mainstream DDR for textures, 128MB of super-mega-high-end GDDR3 for the Z buffer maybe, and then another 128MB for the Z-Buffer.

I really can't see the point of that. You need different pins for different memory chips/speeds. So you may use 192 bits of the memory channel for the slower memory chips and 64 bits for the superduper fast chips. But that would still be be pointless because, if at one point of time you need lots of Z data the 64bits of superfast memory are still slower than 256bits of regular memory or if Z data isn't needed the 64 bits of superfast memory interface are just sitting there idle.

The best way is how they do it now, interleave data on 4 64bit channels of identical memory and hope that some perverse access pattern doesn't require all the data from one channel.

If it were that simple, they would've done it already.

 

[kenneth9265_3]

IMHO, NV50/R500 won't even see DX10 or 4.0 shaders, that pleasure will belong to the NV60/R600. Everything is pointing to an 2006/7 launch date.(XBox 2/DX10/Longhorn) I predict the XBox 2 will have a GPU will have a performance between that of a R500 and a R600. It will be faster than a R500 but have DX10 features. Plus I bet the R600 will be called Radeon X (don't ask how I would know that, you'll would laugh ).

PS.... All of this is base guesses and a lots of surfing of the internet. None based on actual facts and I am not a expert, but a gamer.

 

[Megadrive1988]

that's right, now we can talk about NV60 / R600 since NV40 / R420 are totally finished (probably have been for months) ...while NV50/R500 development continues, Nvidia and ATI now have 1 free team each, for working on NV60 / R600. 8)

I'm not forgetting the refreshes such as R480 / NV45 and R550 / NV55 ...they fit in somewhere