Valdez850xtpe > 6800ultra
x1950xtx > 7900gtx

Individual cards ...yes at times, though as a whole Nvidia have been outperforming ATI in pretty much every class, and the prices keep dropping.

no-one should wish to compare a GT200 against a 4870x2, thats stupid, single gpu vs single gpu people come on.

surely a 3870X2 beats a G92GT/S, but again, not an even comparison. when we go even stevens (G92 SLi vs 3870X2 or 3870CF) then we have a reasonable comparison.

it should be 9900GTX vs 4870
and 9900GTS vs 4850

maybe 9900GTX/S SLi vs 4870X2, thats acceptable, but it is completely unfair to compare single to double.

the only possible way they could compete together is if they both sell at the same pricepoint, i.e. 9900GTX and 4870X2 both retail at $499

wolf.

reason saying is we dont have any gaming tests to compare them both, its just amazing AMD Managed to keep the 4 series so quiet until recently, I know it was expected but we shall see how everything flows, and to say this your better off skipping a few generations than jumping on every new thing in site, i went from a 9800 AIW to a 1950 Pro and the jump was huge

The surprise is that Nvidia is very secretive too this time. Yeah we have lot of "leaked" info, but chances of those being actual ones are so small:

1. 96 TMU? Seems right. Pair it with clock increase for a two fold increase in performance.
2. 384 SP? - Don't think so, not if 1. is true. G92 has 2 SPs per TMU, this is 4 per TMU, balance is broken IMO. Unless they do have other uses for them in mind. Ageia physics? It's very unlikely anyway. Another possibilitily is that they are doing like Ati and in reality it's 192 SPs with 2 ALUs each (one complex one simple), or 96 with 4 each. Unlikely for the latter, but 192 x 2 is possible. It's also possible that the TMUs unlike in G80/92 are capable of FP16 like Ati ones. VERY UNLIKELY, paired with SP number would make this card 3X faster per clock than G92. That's insane and won't happen, right?
3. 64 ROPS? Definately false. 32 ROPs maximum. 24 would suffice IMO.
4. 512 and 448 bit memory interfaces? Could be but it's too much bandwidth IMO even when still using GDDR3. 384 and 320 would suffice IMO. Unless they have the massive physics thing in mind and need that extra bandwidth, which could make sense.

All in all, the specs are plausible, but some of them seem too high to be true. It would be a monster chip, some leaked specs said 1,8 billion transistors and you do need all of them for that specs to be true and that doesn't seem profitable. Not to mention that such a chip would be almost 3x as fast as G92 or have too many spare shaders.

remember though, back in early 2007 they said the next generation core would do 1TF, so it is plausable.

newconroerIndividual cards ...yes at times, though as a whole Nvidia have been outperforming ATI in pretty much every class, and the prices keep dropping.

Only since 8800 came out. the 7900/1900 era belonged to ATI.

Not jumpin the gun!

I can see it already 9900's come out and a few weeks later 10,000 series will be in the rumor mill and the cycle will continue :banghead:

tkpenalty"According to several sources" I haven't seen any sources say that yet, not even fudzilla who are usually on top of this kind of news. AMD will get to launch their card first anyway...

+1 on that statement . . . I doubt that nVidia will be able to launch these cards prior to ATI unleashing the HD4000 series . . . just going by how things between the two companies usually roll, when nVidia declares a release month, estimate 2-3 months after that. When ATI starts leaking specs for their next series, figure an initial release 2-3 months after that.

So . . . by that rule of thumb - we can figure a release of ATI's new goods around June-July; and nVidia's new release around Sep-Oct . . . which sounds about on par when both companies tend to release a new series of cards.


IMHO, this news is really just to start drumming up the support, and forum banter that will build the consumers expectations and start the anticipation process. There's no concrete info, per se, and if we make out the new green camp hardware to sound bettern than ATI's already acknowledged HD4000 series, there's gonna be a ton of people who'd hold off on purchasing a new ATI card until they see the leaked benchmarks and specs of nVidia's newest demons - and I figure, too, we'll start seeing those leaked specs and benchies the closer we get to ATI's release.


So, let's all dry out panties off until we get some solid info.

Wile EOnly since 8800 came out. the 7900/1900 era belonged to ATI.

Maybe, but how far back are we going to go with this? It will be two years since we saw the 8800s come Q4 of this year.

In that time frame we've seen a plethora of cards launched from both camps, even if some qualify as 'rehash.' Never the less, leadership has primarily belonged to Nvidia.


It's a point that even ATI's stance on controlling the non-high end market has been failing. Which further supports the observations that Nvidia doesn't squirm no matter what ATI does.

Even if ATI pulls off some card that out does the current Nvidia 'top dog,' it won't do it by much, and it definatley won't last long.

Hence these 9900s probably won't be anything to kill for, but could be a plentiful upgrade for some users, whether ATI or Nvidia owners previously.

Finally, bringing it to my last point: Right now, and when the 9900s launch, the 9800 series are and will be at a good price, a very good price. I know there's people considering a 9800 card, but are worried it would be premature. It wouldn't be 'stupid,' to get a 9800 now.

If these 9900s aren't going to be that spectacular, I'd recommend the 9800

newconroerMaybe, but how far back are we going to go with this? It will be two years since we saw the 8800s come Q4 of this year.

In that time frame we've seen a plethora of cards launched from both camps, even if some qualify as 'rehash.' Never the less, leadership has primarily belonged to Nvidia.


It's a point that even ATI's stance on controlling the non-high end market has been failing. Which further supports the observations that Nvidia doesn't squirm no matter what ATI does.

Even if ATI pulls off some card that out does the current Nvidia 'top dog,' it won't do it by much, and it definatley won't last long.

Hence these 9900s probably won't be anything to kill for, but could be a plentiful upgrade for some users, whether ATI or Nvidia owners previously.

I was only commenting on what was said in this thread, but if you do go back thru history, you'll see that ATI and NV have swapped as the top dog every few years. ATI is due to get back on top.

Wile EOnly since 8800 came out. the 7900/1900 era belonged to ATI.

That's true. But X1800, X850, and X800 were basically behind NVIDIA. 9700/9800 were the first time ATI wasn't quite a bit behind. They were fortunate with R3x0 and NVIDIA messing up so bad with NV3x.

It seems to me that NVIDIA has a generally more efficient architecture right now, and historically until you get back to NV3x.

What's most amazing is how R420 and RV670 are both 16 pixel/clock GPUs. Granted, they've gotten more efficient with those resources, but NVIDIA is way beyond that.

swaayeThat's true. But X1800, X850, and X800 were basically behind NVIDIA. 9700/9800 were the first time ATI wasn't quite a bit behind. They were fortunate with R3x0 and NVIDIA messing up so bad with NV3x.

It seems to me that NVIDIA has a generally more efficient architecture right now, and historically until you get back to NV3x.

What's most amazing is how R420 and RV670 are both 16 pixel/clock GPUs. Granted, they've gotten more efficient with those resources, but NVIDIA is way beyond that.

one can't really bring the X1800 series into the 7900/1900 era argument, as the 1800s were quickly replaced with the 1900s.

Anyhow, I can't say nVidia has a "more efficient" GPU architecture . . . nVidia has a long running reputation for running their cards at insane speeds, and having cards that run at thermonuclear temperatures . . . not ideal for those concerned about power consumption, and in many regards, nVidia hasn't made any drastic changes to their current GPU architecure in quite some time, whereas ATi has gone back to the drawing board on a fe occasions. But, both companies' GPUs accel at certain tasks than their competitors. Big reason why during the Phys-ex Bout Part I, ATI layed waste to both nVidia, Aegia and Intel. ATI's GPUs are 1337 at mathematics, and with graphics engines that revolve around this as well. Sadly, there aren't many games like that, so nVidia takes the cake near about always. But the few games that are coded in a way that puts ATI on top - it tends to be a massive margin over nVidia.

I agree with Wile E's statement, every few years the crown gets swapped, and ATI is due to take the lead again. Based on the leaked specs of the HD4000 series, it appears that ATI intends to address the issues that were holding back the RV670 in performance.


Either way, the competition between the two this year in regards to the 4000/9900 series will be very close. The second half of this year will be very intriguing, and I'm damn-straight looking forward to it :rockout:

imperialreignone can't really bring the X1800 series into the 7900/1900 era argument, as the 1800s were quickly replaced with the 1900s.

True, but that's because X1800 was like 6 months late. X1800 wasn't much of a match for 7800GTX. Well, unless you are more concerned with quality than simple speed. GF5/6/7 had plenty of texture filtering quality issues.

imperialreignAnyhow, I can't say nVidia has a "more efficient" GPU architecture . . . nVidia has a long running reputation for running their cards at insane speeds, and having cards that run at thermonuclear temperatures . . .

Not sure I agree here. You'll find that ATI ran their X800 and X850 chips >100MHz faster than NVIDIA 6/7 just to keep up. And they couldn't always keep up. X1800 and X1900 were again clocked very high and certainly were not cool-running. They also had very noisy coolers. GF 7900 was not well matched to X1900, true, but 7900 definitely was cooler and quieter.

I think that the fact that NVIDIA wins in just about every application out there says volumes about their architectural efficiency. R600 was anything but efficient compared to G80. All that touted shader power that went nowhere, pathetic AA performance, excessive memory bandwidth, and extreme heat combined to make a product less power-efficient, slower, and uglier than the competition. And now we have HD 3850/3870 outclassed by 8800GT, 8800GS and 9600GT. 3870X2 has some serious quirks to it (as does 9800GX2 though). At least RV670 was frugal on the juice.

imperialreignBig reason why during the Phys-ex Bout Part I, ATI layed waste to both nVidia, Aegia and Intel. ATI's GPUs are 1337 at mathematics, and with graphics engines that revolve around this as well.

Physics on GPUs has gone absolutely nowhere aside from forgotten promises from NV & ATI. And there are synthetic tests out there that show ATI's shader design to be less efficient that NV's in multiple ways as well. Check out some of Digit-Life's reviews to see some signs of this. You'll find that while it can perform extremely well in some cases (geometry), it gets battered badly in others (SM4).
www.digit-life.com/articles3/video/rv670-part2.html

How good at math they are in some cases hardly matters when they are dramatically behind in texture fill-rate anyway. The chips are just totally off balance. I'm worried about RV770 after seeing how it may have 16 ROPs still. That means a max of 16 pixels per clock output.

imperialreignEither way, the competition between the two this year in regards to the 4000/9900 series will be very close. The second half of this year will be very intriguing, and I'm damn-straight looking forward to it :rockout:

Indeed. I like the ultra-agressive mid-range products we're getting. But, I think that NVIDIA has a monster GPU in the works while ATI is going to rely on dual RV770s for the top-end. I doubt that ATI can corner the high-end with a dual GPU design if NVIDIA does indeed have a 1.3 billion transistor chip coming. They just don't work out reliably/efficiently in all apps, as shown by current dual GPU cards and SLI/CF. A big single chip board doesn't have these driver issues and is going to be innately more efficient than two GPUs communicating externally.

swaayeI think that the fact that NVIDIA wins in just about every application out there says volumes about their architectural efficiency. R600 was anything but efficient compared to G80. All that touted shader power that went nowhere, pathetic AA performance, excessive memory bandwidth, and extreme heat combined to make a product less power-efficient, slower, and uglier than the competition. And now we have HD 3850/3870 outclassed by 8800GT, 8800GS and 9600GT. 3870X2 has some serious quirks to it (as does 9800GX2 though). At least RV670 was frugal on the juice.

TBH, I think when it comes down to their massive lead in gaming benchmarks, it has a little more to do with their collaboration with game devs. I'm not claiming every title under TWIMTBP campaign, as they've also worked with others who didn't join up - and, as far as nVidia is concerned, this has been their biggest marketing achievment to date. ATI doesn't collaborate with game devs anywhere near the level that nVidia does, and that is blatantly obvious in the massive performance differences between each company. But, of note, games where ATI has spent a lot of time collaborating with the game devs, we see performance levels on par between the companies, and many times ATI will lead in those tests (at least, in regards to newer generation hardware, in lower AA levels). Games like FEAR and Call of Juarez, for example. ATi has the capability, but there's just no follow through with game devs, and no reach around for us consumers - which I think is kinda sad at times.

Physics on GPUs has gone absolutely nowhere aside from forgotten promises from NV & ATI. And there are synthetic tests out there that show ATI's shader design to be less efficient that NV's in multiple ways as well. Check out some of Digit-Life's reviews to see some signs of this. You'll find that while it can perform extremely well in some cases (geometry), it gets battered badly in others (SM4).
www.digit-life.com/articles3/video/rv670-part2.html

How good at math they are in some cases hardly matters when they are dramatically behind in texture fill-rate anyway. The chips are just totally off balance. I'm worried about RV770 after seeing how it may have 16 ROPs still. That means a max of 16 pixels per clock output.

The whole physics thing was brought up quick, and dropped rather fast as well - especially after ATI's announcement of physics in a crossfire setup . . . I agree as well, though that ATI's shader design have been sub-par since after the X1950 series. Those cards were great with SM3, but ATI has been behind the gate with SM4 with newer generations.

True, for the most part a GPU being good at math has little to do with texturing performance - but, again, if ATI were to spend more time collaborating with game devs to the extent that nVidia has, it would make all the difference in the world - unless ATI just decides to revamp the architecture, which is what RV770 so far will be doing. Sure, it might only have 16 ROPs, still, but if everything is clocked independantly, I don't think that will be a limitation at all . . . only time will tell, though.

[/QUOTE]Indeed. I like the ultra-agressive mid-range products we're getting. But, I think that NVIDIA has a monster GPU in the works while ATI is going to rely on dual RV770s for the top-end. I doubt that ATI can corner the high-end with a dual GPU design if NVIDIA does indeed have a 1.3 billion transistor chip coming. They just don't work out reliably/efficiently in all apps, as shown by current dual GPU cards and SLI/CF. A big single chip board doesn't have these driver issues and is going to be innately more efficient than two GPUs communicating externally.[/QUOTE]

We haven't really seen anything "high-end" from ATI since they exclaimed they're staying out of that market. Sure, the 3870x2 is priced for that market, but it's more two 3870s in one package, which still doesn't qualify for high-end, IMO. Although, if ATI get's back on the ball like they were during the X1900 series, watch that statement go out the window - I'm sure we'll see another 1337 card come from the red camp, but only if they feel it can compete with nVidia's 1337 beast.

I think that ATI must re-think on their marketing strategy cause without marketing they will never catch or be on par with nVidia cause marketing its one of the big things that nvidia do better than ATI putting aside the core clock, frequencies and stuff, they should focus on their designs and slogans, cause many ppl prefer nvidia cards because of that. They should think on the image of ATI... its my opinion putting aside the performance quality of both...

I wouldnt believe chit ati or nvidia said about a reliese date. If your saving for a video card and thinking your going to get the new generation in june I got some land by the sea to sale ya also, in arizona.