PDA

View Full Version : When will Nvidia regain the "performance throne" ?


Pages : [1] 2 3 4

cy_a253
01-19-04, 09:10 PM
The "King" officially died on July 17th 2002, or so, if you recall. (http://www.tomshardware.com/graphic/20020718/index.html) And my GeForce4 Ti 4600 is starting to show its age... (well not really, HL2 is not out yet afterall.)

I've been out of the loop for a while, so when is Nvidia's next product coming? And will it outperform ATI's next card?

So, basically, what should I buy next?

(I have a Athlon XP 3200+ and 512MB of DDR400, so the CPU won't be a bottleneck...)

The Baron
01-19-04, 09:15 PM
Answer: Nobody knows. To any of your questions.

digitalwanderer
01-19-04, 09:40 PM
From what I know, IMHO, yadda-yadda-yadda-ya; not this year, no way.

The Baron
01-19-04, 11:30 PM
Originally posted by digitalwanderer
From what I know, IMHO, yadda-yadda-yadda-ya; not this year, no way.
And from what I know, you're wrong, but hey, that's the funny thing about rumors. :)

ragejg
01-19-04, 11:31 PM
ugh, this banter starts ALREADY??? :rolleyes: :p

jeez, wait a month, guys... :p

-=DVS=-
01-20-04, 04:00 AM
Originally posted by ragejg
ugh, this banter starts ALREADY??? :rolleyes: :p

jeez, wait a month, guys... :p

We can't wait need all NOW !!!!!!!!!!! :banghead: :fu:

Unit01
01-20-04, 05:28 AM
According to 3dmark03 and the forceware drivers they already have the performance crown :D

cthellis
01-20-04, 05:36 AM
The answer is WITHOUT A DOUBT, definitely, positively "who knows?"

Sotos
01-20-04, 06:29 AM
It has some chances to Nvidia become king again with the Nv40

It seems very powerfull even compared to the newcomer ATI R420

but Core clock numbers somethimes is not enough to beat in performance... as was proven with the R3xx and the Nv3x especially under Directx9.0

We will see in CEBIT 2004 and later of who maybe will be the king these time

Sotos
01-20-04, 06:33 AM
Originally posted by Unit01
According to 3dmark03 and the forceware drivers they already have the performance crown :D

I do not think so

Ati released the Catalysts 4.1 just yesterday and it makes the R3xx cards still superior(especially under DX 9.0) from the current forceware drivers with the current Geforce Fx cards

StealthHawk
01-20-04, 06:41 AM
Originally posted by ragejg
ugh, this banter starts ALREADY??? :rolleyes: :p

jeez, wait a month, guys... :p

More than a month :p

Originally posted by Unit01
According to 3dmark03 and the forceware drivers they already have the performance crown :D

Not if you use Futuremark approved drivers.
http://www.futuremark.com/community/drivers/?approved

I'm sure if you make comparisons using valid drivers NVIDIA will not have the performance crown in 3dmark03.

DMA
01-20-04, 07:48 AM
Originally posted by digitalwanderer
From what I know, IMHO, yadda-yadda-yadda-ya; not this year, no way.

Well, if you really know you don't have to be humble man.. :D

Some say NV40 will kick butt and others say R420 will be so fast your monitor will explode.

Who to believe then? I say don't listen to all the "Ey, i have some secret info, but i won't tell you. Actually i don't know ****, but i can give you a ;) and pretend i know just to look cool and gain some respect in this community"

Bah..gimme a break. :cool:


:afro:

saturnotaku
01-20-04, 08:40 AM
NVIDIA will 0wn j00, as will ATI. Though we pretty much know nothing about the next-gen cards at this point, I have a strong feeling they're both going to kick some series rear end.

I don't think there's going to be a clear cut winner, and that's not necessarily a bad thing. ;)

Unit01
01-20-04, 09:03 AM
Originally posted by DMA
Some say NV40 will kick butt and others say R420 will be so fast your monitor will explode.
We already have that :D
Don't you remember the Bull**** that was spread around before about monitors going toasty? Toasters exploding? people getting divorced?
It was all because of NV3x and R3xx

The Baron
01-20-04, 09:09 AM
Let's see. Not sure about AF quality. NV40 will probably have the advantage there, even if it relies on brilinear because it won't be angle-dependent. As far as I know, R420 is angle-dependent, but we'll see about that.

AA is more iffy at this point. NV will probably dump all OGMS in favor of RGMS. The fate of supersampling is unknown. Also, whether or not the multisampling is gamma-corrected is unknown (although I would be somewhat surprised if it wasn't). A stochastic AA mode might be implemented, although we're more likely to see that on NV50 (at the earliest). ATI will use the same (or very similar, maybe slightly improved at best) AA methods, still using GCRGMS. 8x AA support will probably be in there this time because of the standard 256 megs of RAM.

Pipelines? No one knows. All sorts of theories, everything from true 16x1 to 8x2 to whatever. It doesn't matter, at least not really.

Floating point support? NV is sticking with FP16/32. ATI is most likely sticking with FP24, although it might move to FP32 because some conjecture that PS3.0 is going to increase the minimum required precision to s21e9 (I think that's right... FP32). ATI will almost certainly be considerably faster in PS2.0 and 3.0. I doubt that they will have the insane advantage that they have now (double the performance of comparable NV cards), but they will probably will still have a sizable lead in shader performance (maybe ~33%?). Either way, both cards will be fast enough in shader-heavy apps that the difference won't be too important, at least while these cards are still the high-end.

Core speeds? R420 will be 475-500Mhz and NV40 will be 500-550Mhz. RAM will likely be above 600Mhz raw for each. NV40 will probably have the memory speed advantage, perhaps 650-700Mhz.

So which will win? Nobody knows. They're going to be much closer in terms of theoretical speed than they are now, and it will probably come down to the quirks and unforseeable drawbacks of each architecture.

cthellis
01-20-04, 10:19 AM
Originally posted by Unit01
Toasters exploding? people getting divorced? It was all because of NV3x and R3xx
Dogs and cats living together... Mass hysteria!

991060
01-20-04, 10:47 AM
s21e9? is it a new foramt? what's the point to have 2bits for the sign???

I think IEEE defines FP32 as s23e8

The Baron
01-20-04, 11:10 AM
Originally posted by 991060
s21e9? is it a new foramt? what's the point to have 2bits for the sign???

I think IEEE defines FP32 as s23e8
*******it, I had s23e8 but I thought about s15e7 and then s10e5 and got confused. bah.

but wait, that would make it s20e9. whatever. s23e8 is what I meant.

digitalwanderer
01-20-04, 11:22 AM
R420 will be FP24, I hate to disapoint ya. :)

The Baron
01-20-04, 11:28 AM
Originally posted by digitalwanderer
R420 will be FP24, I hate to disapoint ya. :)
I know it will, you know it will, everybody who realizes that there is no minimum precision change in PS3.0 knows that R420 will use FP24. But hey, have to keep the kiddies happy, you know :)

991060
01-20-04, 12:11 PM
Originally posted by The Baron
*******it, I had s23e8 but I thought about s15e7 and then s10e5 and got confused. bah.

but wait, that would make it s20e9. whatever. s23e8 is what I meant.

your mind goes wild. :bash:

nForceMan
01-20-04, 12:17 PM
Originally posted by The Baron
And from what I know, you're wrong, but hey, that's the funny thing about rumors. :)

We know him very well. He's always wrong. ;)

digitalwanderer
01-20-04, 12:20 PM
Originally posted by The Baron
I know it will, you know it will, everybody who realizes that there is no minimum precision change in PS3.0 knows that R420 will use FP24. But hey, have to keep the kiddies happy, you know :)
Yeah, I know....case in point:

Originally posted by nForceMan
We know him very well. He's always wrong. ;)

;)

jimbob0i0
01-20-04, 01:44 PM
Baron you *know* how this thread will end up. The answer of "when it happens" has already been given... lock it now for the sanity of us all in this big bit of flamebait ;)

digitalwanderer
01-20-04, 01:49 PM
Originally posted by jimbob0i0
Baron you *know* how this thread will end up. The answer of "when it happens" has already been given... lock it now for the sanity of us all in this big bit of flamebait ;)
Personally I think him and a few other mods keep 'em open for a bit 'cause they bet on who and when will be the first to cross the flibble-line. ;)