I don't know how you got to those numbers, but errr, congratulations, they're all wrong! :D ;)
NV30 is 500x4x8 = 16GFlops for both FP32 and FP16. R300 is 325x8x12 = 31.2GFlops for FP24. Both IIRC. Of course, NV30 is ridiculously faster than that with FX12, assuming no register pressure. But that's arguably not very useful.

Looking at the detailed specs of NV30/NV35 and R300 with some years gone by and some added perspective, I don't think it's fair to claim that the specs made NV3x seem better than R3xx. The clockrate advantage of NV30 against R300 is pretty much the only advantage NVIDIA had spec-wise, for the entire line-up, imo...

Anyway, back to R600 now, please - and sorry for continuing the OT, but it's not like there's much I/we can say on-topic anyway, except perhaps bitch at everyone linking retarded bullshit! :)

Oh, nice math there. I guess AMD should have done a "tech preview" in December to get guys like you to stop the clock? Riiiiight.

How many GPGPU apps that you know of are CPU limited? That's just one example, of course.

That's a good point. I don't know of any GPU isolating test results being discussed in the rumors.

One thing I do have to say: When Nvidia realized that the NV30 was crap, they frantically tried to fix it. AMD/ATI has shown virtually no such sign. Not to say that they aren't frantically trying to fix it behind closed doors, but they're giving the impression that the R600 isn't very good.

What signs would we be looking for that would indicate AMD is frantically trying to fix R600?

There are a lot of signs Nvidia was pushing NV30 hard, such as the high clock and heavy shader replacement, but those weren't really known until after the release, and the extent wasn't clear for quite some time afterwards.

The NV30 was quickly scrapped and replaced by the NV35, with a 256-bit memory bus and somewhat improved shader units. The other things happened relatively shortly after launch, not long after. Not to mention the massive hype machine that nVidia used to try to stem the tide. With the R600, we hear next to nothing. It gives the sense that they don't have anything to show.

Or it could mean that the chip is so fantastic there's no need to hype it to all heaven. ATI didn't hype the R300 pre-launch. They just let it detonate over the marketplace.

Wild speculation on marketing, or lack thereof, is pointless at this juncture. We're a few weeks away, *knock on wood*, from a technology demonstration. And a couple of weeks after that from actual benchies.

Let's see what happens.

Deusp,

What, as in there's no R600? :???:

When did we get the impression that R600 isn't any good, let alone crap? And how does one equate R600 to NV30 with contradictory examples (hype machine vs. subdued PR)?

All this waiting is driving me stir-crazy, too, but let's save the funeral for after the birth. There's no sense in AMD releasing specs or benchmarks well ahead of launch, b/c that just gives NV PR something to shoot at.

No, it means that like a lot of other companies, they realise there is not a lot of point in wasting your marketing resources hyping up a product without actually having it to sell to people at the same time.

You may remember about NV30's quick replacement, but do you remember the preceding six months where Nvidia hyped "Cinematic Computing" to death? CG? The infamous PS 2/3 comparisons using Far Cry? Cheating on 3D Mark and withdrawal from the Futuremark? Nvidia looked pretty silly after all that resulted in the NV30 instantly cancelled product.

AMD would rather keep it's mouth shut than dig itself into a big hole. Right now all they are is late. They don't want to be late and stupid with it.

At this point in time, any leak would help keep me from buying a G80 out of frustration. If they'd say it's atleast similar in speed, then I'd wait for quality comparisons at least.

No. No, no, no! They can't be six months late AND be similar in speed with somewhat better image quality. From a product being six months late, I expect it to be a) considerably faster, b) cooler and consuming less power, c) equal or better image quality! Especially it being on a smaller process. Come on!

They either fulfill those three major points or their product is burried on launch.

Does ATI itself usually hype up its upcoming cards? Lack of direct hype != a bad card.

Well PR is almost always stupid. I rather have a bad product with shit-loads of hype than a mediocre one with no hype. Perhaps it's distasteful, but I think it is the right thing to do from a business standpoint. All this non-noise is only driving people to buy 8800s, which will help dry up demand by the time the R600 comes out unless it's completely awesome. At least with lots of unjustified hype, people would hold out.

I recall that with the R300, while people were indeed impressed, they weren't impressed enough to simply ignore what nVidia was going to show, and only when it was clear the NV30 was crap did most people jump ship. This is not the case with the G80. People are impressed enough that many have already jumped ship.

No in general people were floored by the r300 it was pretty much the first card that had usable AA and AF on new games. The R300 caught pretty much every one off guard with its 8pipes 256bit bus and incredible performance. It completley laid waste to everything in site at the time. The only people who held out were the most delusional Nvidia fans. I can't remember anyone on this site who did not give mad love to ATI for the R300 when it hit.

Someone's already declared every GPU discussion devolving into R300/NV30 comparisons a B3D Forum Law, right? Me, I prefer car analogies.

I still think R600 won't be cheaper than G80 simply b/c it should have more RAM per comparable SKU. I'm guessing NVIO is a comparatively negligible cost and GPU cost will come out even thanks to 80 vs 90nm. I'm also guessing G80 will continue to have relatively more TMU than shader power than R600, and that ATI most likely won't push more than 4 AA samples/s. That leaves huge bandwidth as the differentiator, and though it wouldn't be out of character for ATI to boost baseline IQ over max performance, I still think performance will be the differentiator at the high end and ATI will have a slight edge with (to borrow an NV term) HRAA scenarios.

Still, the shader scenario continues to intrigue. 320*800MHz vs. 128*1350(++)MHz. Will it matter enough given current benchmark suites? Not to mention that a G80U pushing 128*1800MHz pulls pretty close upon naive inspection, and may not need to clock the rest of its core much higher if it's got more TMUs, which may mean even a G80U consumes significantly less power than a R600XTX.

There, I got all that clueless speculation off my chest. I'm good for another week of waiting.

Kinda like the G80 right now, right? :lol:

I'm pretty sure most people expected Nvidia to respond in kind. Many I recall were quite shocked when it didn't perform as well as the R300.

ROFL :D

Errr...why only 4xAA or was that just a typo? As for HRAA differences I'd personally prefer to see a few more ROP-interna details before I'll speculate too much about anything. Frankly I would expect to see also here single cycle 4xMSAA and thus at least also 8xMSAA, but a finer description of all the ROP capabilities would help a tad more.

As for the theoretical maximum GFLOP numbers which IMHO will come damn close to the theoretical maximum values (for both) only in singled out synthetic benchmarks, I wouldn't be surprised if NV clocks some sort of "Ultra"-whatever slightly higher and gives it a significant shader domain boost to equal at least those rather senseless numbers.

I'm not saying that I don't expect R600 to not end up any fast than the 8800GTX, rather the contrary; but senseless maximum theoretical numbers don't really mean all that much and that goes for both sides.

I must have missed something. 3dmark-06 uses AA/AF?

Ail, whoops, typo. I meant 4 AA samples/clock, like G80.

IbaneZ, I thought 3DM06 uses no AA and 4x AF by default?

I guess we will see a lot of 8800 cards on ebay after r600 with 1gb ram gets out ;)

For me either ati/amd does not seem worried.
The postpone could been good for several reason,
marketing a whole one for all solution with the spanish baby with a red dragster using a cruisecontrol in all levels of products seems like a nice launch,
its just not a videocard anymore, its a full set of products used in conjunction with eachother.

Spanish baby=barcelona K10 quad genuine cpu.
Red dragster= r600 high end videocard.
cruisecontrol= new chipsets.

Use AMD to bring out the power in your videocards.:grin:

http://www.fudzilla.com/index.php?op...d=354&Itemid=1
:wink:

As the old saying goes, haste makes waste. The thing about "jumping ship" is that you want to avoid doing it when you can, unless you have a magic wallet and you don't care if you have to jump back...;) If R600 turns out to be appreciably better than G80, then the early "ship jumpers" are going to wish they hadn't been so hasty. OTOH, if G80 turns out to be appreciably better than R600, then the only "penalty" I can see that people will "pay" for waiting to buy it is that they'll probably wind up with much better G80 drivers from the start than the ones that have shipped with the G80s thus far.

IE, since it is well known that the R600 launch is imminent, it really makes no sense to "jump ship" right now, does it? Far better to wait until the product launches to decide on what ship you want to sail. Besides, at the moment, there isn't a single game out there, or which has recently gone gold that my current x1950 Pro AGP won't run with ease...;)

BTW, the only reason people waited to "jump ship" post R300 (I didn't--I went straight to R300 from nv25) is because ATi was a darkhorse 3d gpu maker at the time, and people weren't accustomed to ATi leaping out ahead of the former pack of 3dfx and nVidia. That made it much easier to believe some of the hype nVidia kept spewing about nV30 right up until nVidia canceled it. Since then, Ati has been out front for a long time, and ATi has completely revamped its driver reputation. So I think that today it's much easier to wait and see what ATi is bringing to the table.

Do we? Last time it was imminent also, don't you think? Not that i think AMD will delay once more the launch, but i think it would be wise not to be certain knowing the past of the R600 launch dates ;)

Hmmm what about those people that held on to pretty weak GPU's while waiting for R600 when they could have been having a much better gaming experience on a G80 based card. You know, like those people suffering with X1950Pro's for example :razz: