open letter to ATi on gamers-depot

I, for one, find speed interesting...

sometimes i crack myself up

 

Well, all Nvidia would need to do is to lower the price on the 6800 GT to 300$ and they would have a nice competitor to the X800 XL. But unfortunetly, there's something called margins that the IHV's need to worry about.

 

LOL, then can you explian why somebody can not buy SLi for its speed?

Ok we are done here, For me, its a waste of time to talk with someone who has no idea about what he is talking about.

 

Ok, this makes no sense whatsoever....what are you trying to say?

Agreed, but I'm gonna try and understand what you say before I dismiss ya.

 

Well, every company has its own interest, strategies and such, but bashing allways backfires. I am not talking about engineering, production, etc. This is at PR level. PR = Public Relations: PUBLIC and the way it percieves the company. Maybe what the autor rants about is very related to the perception that he and the PUBLIC have about the companies and their PR war. If someone forgot, Nvidia supporters (PUBLIC) swallowed a lot from ATI supporters (PR + PUBLIC) related to the FX series.

So, in the end, it is just an emotional issue of the author. What amuses me is the so violent responses it arises 8)

 

True, but the FX series earned every bit of the demigration it got....I don't see the situation as being the same. :?

 

Yes, Nvidia had a terrible product which could not compete with ATI's one, in part because of their fault (the main part), in part because of the good choice of ATI regarding the 24-bit precission an its acceptance by Microsoft's DX9 definition. Anyway, its product was in clear disadvantge.

However Nvidia had to do its best to sell its products, no matter its technical/positional inferiority and it leaded them to put themselves in a hard spot because all the "intelligence" used to shorten the distance to ATI's products sounded as "cheating" (or it was, I do not mind) when discovered.

What happened then? Bashing and thrasing (deserved or not, it depends of the point of view). Given the audience of the products we are talking about (hardware for gaming) and its natural inmadurity (teenagers, twentiers), the outcome is clear: a lot of egos bruised.

When the tide turned, a lot of these people take advantage of every open flank that ATI seems to have to strike back. Even myself did it once here (I promise not to do it again).

I do not thing it has more importance than that...

 

Isn't it obviously? Since he admits that there is a market for SLi just as there is market for high-end CPUs, hence whats wrong with SLI?

Sure you have to read and understand the entire discussion before you join it lol.

 

In retrospect, the memory controller "improvements" may have been mainly to allow it to operate at higher speeds, not to be more efficient.

I think those were due to driver improvements (read: UT2K3 optimizations) that later trickled down to the 9700 series. I'm pretty sure the current speed differences are within the theoretical ones.

 

And that's fine the GF6800GT-128mb is selling for about that now on NewEgg and PriceWatch anyways. I expect that to be the very response nV will take, unless they do indeed produce NV48s on 110nm to fill the gap.

Yoiu miss the point. People are bitching that ATI doesn't have a part to fit that segment, when in effect they do, the X800SE, and just like was done with the R9500PRO and the FX5900XT, they can pop in a part even at a loss simply to fill that sgement of the market. The question is whether that would cost them more than developing the Rialto in the long run. And right now producing R9800PROs to fill the gap and seeling them for less than the GF6600GT is financially sound? I doubt the R360 is much cheaper to produce than a cast-off R420. Heck even ramping up production for that might get more X800XT and XTPEs to market and give people something to sell from the scraps. Solving more than one shortage problem. Producing R9800s for this job leave NO spin-off for the current line.

If more PCIe mobo don't come before too long, then it'd make alot of sense to develop the bridge; however if people get the chance to begin to transition to PCIe and there is critical mass, perfecting the Rialto (as it's been reported there are difficulties) may be a waste of time and money.

Personally I just think people need to move on. Both nV and ATI said they were going to stop making AGP cards, and if it weren't for the poor turn-out of PCIe MoBos in the last few months. Giving people the AGP option just creates more headaches and even more production costs. The only fly in the ointment is the MoBo issue which is something out of the control of the graphics teams, even if it may be nV and ATI's fault to the same extent as Intel, SIS and Via's.

In the end it's like bitching about PCI cards, I just don't care. Cut the Luddites off, and move forward instead of trying to perfect a crutch to sustain their antiquated ways. ATI and nV have alot of other issues to fix, worry about those more than what small gap do we need to fill. I wouldn't be surprised if they still make more money moving R9600SEs and/or X300SEs than nV makes moving GF6600GT-AGPs (which are still rare). How much is a segment really worth?

Either way I don't think it'll have that great an impact. It's the only segment not covered, so is it worth spending more money supporting AGP outside of the R3xx and R420 lines? I don't think so, but I'm sure many AGP-only owners disagree.

 

There are other reasons to buy it (multi-monitor support), but primarily it will be about speed.

I don't think people pay mcuh money for products simply because they are interesting. They'll download a tech demo for something like that. But unless you're paying by the mb, then downloading Ruby or Nalu, or an FB Demo (or something new from Humus) isn't going to cost you anything extra.

Utility gets the money, Interest gets the time, at least in my experience.

 

It will cut profit margins too much to sell the 6800 cores at a hundred bucks less with 256mb gddr 3 ram, thats why it hasn't been done yet. Also maybe they might use the nv43 cores instead?

The AGP market is where the high end cards have to be. A person buying a video card to upgrade in his comp is not looking for a 100 dollar upgrade. Thats why nV is still focusing on the AGP market its still very important. Compare that to the amount of people who buy a top end computer with a top end graphics card (would need to spend close to 3k I think).

ATI never really had the x800se, it was out of stock at Dell. Dell dropped that card BTW, and replaced it with a 6800 non ultra.

producing the r360 core would be alot cheaper. Its built on .13 the fabs for .13 is less in price now plus the chip isn't comparatively no where near as complex as today's chips. Yeilds are excellent. Its not the point of it being cheaper, it can't keep up with the 6600gt.

Making 9800 pros aren't going to magically help the x800 series yeild issues to be resolved. You still arn't see great yeilds of the x800 xt pe, same with the x700xt, which the x800 xl is based on? ATi has been saying thier yeilds are good until last month, when they anounced the x850, its supposed to make yields better for them. But how the chip is basically the same just higher clocks. If they couldn't get the kinks out the r300 core when they moved to x13 and low-k how can one expect that to change now? The core is mature, they should be very well aware of how to get better yields, the process doesn't change much from going to low-k, as didn't the chip.

 

r360 is .15

X800 xl is nothing to do with x700 xt or the x800 xt, this is another chip (16 pipe .11)

They have never “moved r300 to .13 low-kâ€￾, they have made huge changes when they move the high end to low-k. There are always opportunities to further tune the chip, if you go back and pay the money to do it.

 

my mistake .15 thank you for the correction

x800 xl is the same as the x800 xt, xt pe (x700's are similiar to these as well just less pipelines), just on a .11. All these chips are pretty much identical, just a different fab size and pipeline numbers.

the structure of the chips have few changes, but not so much as in makes a huge change in the way the core works and how the load of the data is used through the pipelines, so this leads to over all similiar situation. Efficiency with core clocks have to come by better management of data, as nV has done from their fx line to the 6800 line, and by doing so its an entirely "new" core.

 

That's why I said the 128mb, which is already selling for $318, so those $18 are making the big difference. If need be make a GF6800GP that uses cheaper memory. If it can be done already then a mature process should yield more chips and you once again have shifts in costs/margin.

Who's talking about High-End or $100 cards? It's covered. ATI has the High-End covered and even half of the mid-level and all the entry level covered. The area it lacks is the top half of the mid-end AGP cards, the bottom of the mid-end is well filled by the R9600series and the R9800pro.

Wow, Really? Yet I can buy it from 3 retailers here in Calgary (not the Global Hotbed of Graphics I would say), and I can get it from a bunch of e-tailers too. So what was this about 'never... yada,yada'? NCIX can send you one in a few days if you'd like. The issue at hand is not existence it's whether it's possible to bring the price down enough to have it compete for the top part of the mid-level market.

So Dell dropped them, that doesn't really impact anything, so you'll have to explain how it's relevant other than for an economies of scale argument.

As was already pointed out the R360 is 0.15, so your argument holds no value. That 0.15 chip which obviously cost alot to make when it was an R9800XT that couldn't be sold for $200, is it that much cheaper NOW to produce than the cast off R420 chips that couldn't make it as X800XTs or Pros? Only TSMC and ATI could tell us this, but I doubt the differences are that large. And since the R9800PRO sells for $180, that's about $40 to $70 market difference between the R9800PRO and the GF6600GT. So there's no wiggle room? I doubt that.

Who said making R360s would help the X800XTPE? I was saying that if you produced wafers of R420 chips you would have a wider selection to chose from, and would end up with more X800XTs as a result. Understand the discussion before replying to it.

Do you even know what you're talking about? I doubt it based on the amount of errors (X700/X800XL, etc.). There's been no shift only to Low-Kd except for the R9600 series which is quite successful in it's transition. The shifts for the cards we've been talking about have been from 0.15 to 0.13Low-Kd which involves a process change. During that shift there was also the addition of Pixel and Vertex units, so not only a change in process but design. Now the X850P/XT/XTPE aren't a change in process from the R423 (which itself was a slight change from the R420, but feature, not process), but they did add that new power 'gate' feature IIRC, which is supposed to help, however it's not a giant leap. The X800XL is a shift to 110nm which is NOT Low-Kd, so in no way is there a shift TO Low-K alone, except for a totally different series.

Understand one simple thing, the X800XL(+ the X800 based off of it) and the X850 series have NOTHING to do with this conversation other than to say that they fill equivalenet gaps in the PCIe market, because the X800SE is not based on either of those two products, so just forget about them for now. The only issue that Bjorn and I were focusing on is the AGP market, where currently ATI have a noticeable hole to file. The questions are how do they do it, and even whether they will bother to fill it with anything more than the R9800pro answer they've been giving for the past while.

Sure, sure whatever; and the Gigabyte 3D1 is just a Volari Duo on a different process. I'm not going to debate your simplification of what's involved since you couldn't get your facts correct in the first place. It also has little bearing on what we're talking about. Regardless of whether the X4xx is based on the R3xx or the Rage Fury it's cost to produce, and it's XT/PRO/not-Pro yield rates are what's important. Performance and price are the issue, not whether you feel it's enough of a technical leap to warrant your/our praise.

 

I don't see how you do not understand his post. He is basically saying that the reason a consumer would buy an SLI setup is no different than why JVD stated someone would buy a P4 EE. Of course JVD is going to find it uninteresting since it isn't ATI's technology. Every person in this forum knows that. What I find especially humorous is JVD's attitude towards anyone who gives kudos to Nvidia. He tries to shame them to death in a forum. Hehehe. Woohoo, anonymous torture! Personally, I would be embarrassed to dedicate my postings on these forums fighting for some illogical cause. They're videocard/gpu companies. Get over it already....

 

I hate to have to do this, but

HOLY CRAP! IT'S THE FX ARGUMENTS ALL OVER AGAIN! THE FANBOYS ARE LINING UP, AND THIS TIME IT'S WAR!

Seriously, folks, can't we get over the FX crap?

Things about the FX vs. R300 debate:

1. NV3x--Crappy chip. All of them.
2. Crappy drivers because they cheated.
3. Yes, they cheated. No, it's not an optimization, even if they were only using FP16 where it "should have" been used ("Valve hates us! Waaaaaah!")
4. ATI had the far superior product. Hell, even its AF was superior because you couldn't get full trilinear on the FX no matter what you did at the time.. Seriously, comparing R3x0 and NV3x is like comparing... uhhhh... Opteron to Itanium? Something like that.
5. FP32 doesn't mean anything if you can't use it on more than one shader at a time or performance will tank.
6. With regards to MS screwing over NVIDIA or whatever rubbish you've heard, compare the FP24 speed of R300 to the FP16 speed of, well, any NV3x chip. They're similar in some cases, but I recall NV3x still losing the majority of the time to the R300 chips. Not surprising, given the horrible register limitations of NV3x, but seriously, NV dug their own grave with shader performance on NV3x.

The end.

Now, as for more on-topic points:

1. Razor, you have no idea what you're talking about. Seriously, do some research before you spout off nonsense like "The R300--they moved it to .13u low-k!" B3D has chip tables with fab information for a reason. oh, and "x800 xl is the same as the x800 xt, xt pe (x700's are similiar to these as well just less pipelines), just on a .11. All these chips are pretty much identical, just a different fab size and pipeline numbers."--that line? you might as well call the NV40 and the R420 basically the same chips because they both have pipelines.

2. jvd, of fucking course there's a market for SLI. IT'S THE SAME MARKET WHO BUYS $500 GRAPHICS CARDS THE DAY THEY'RE AVAILABLE. hint hint wink wink. please, stop being a fanboy for a second and acknowledge that (is that word still banned?). there are always people who will pay out the ass for the Biggest and Newest just because they have money and don't have kids or something silly like that.

3. Chris, I seem to recall that the early disparity in per-clock performance between R300 and R350 was primarily because of driver improvements that were not originally enabled for R300 cards (in drivers available at the time of the R350 launch). a few months later, they were, and they were just about even again. so no, I don't think R350 had any major IPC improvements.

god, this is a silly, silly thread.

 

Quick lock this thread before Baron’s common sense spreads and takes the fun out of it.

 

Perhaps you're a bit to hard on Razor here. I'm thinking that the quads in the X700 are very similar to the quads in the X800 so i don't see the problem with calling them similar with the exception of the number of pipelines. I'm sure that there are differences but leagues away from the differences between the NV40 and R420.

I'm no 3D hardware engineer though so this is from a consumers/enthusiast point of view though.

 

I thought those driver improvements were just with Aa/AF though? Cat 3.2 or something introduced it. I do remember several people "Modding" there 9700 to 9800 ((if you remember the omega drivers which automatically did this. I heard there were some errors with it however)). You could very well be right though. I dont have a 9700 Pro to compare to my 9800 Pro. It really doesnt matter at this point except to crazy people who like wave rattles at the IHV which offends them most