[SirPauly]

Originally Posted by Quasar

The man catches my drift... One other thing to consider would be the omnious answer (and geos - i believe - signature of the nes 6800 GT wearing red) Vijay gave regarding the gap between G80-GTX and -GTS.

This would makes sense but your point was to basically have this by sacrificing the flag-ship based products. My point is: Why can't you have both?

[CarstenS]

Originally Posted by SirPauly

This would makes sense but your point was to basically have this by sacrificing the flag-ship based products. My point is: Why can't you have both?

I should explain myself a little further. I think the delay in R600 was at least in part due to the performance not exceeding G80 by a healthy enough margin. So they tried and stretched clocks as far as they'd go. Then it was discovered, that they'd get only so much R600 running at a high enough clockrate to beat G80 by more than 5-10 percent - plus needing a healthy surplus of power and cooling.

So, instead of taking the blame for another "NV30" (wrt heat, cooling, delay, not necessarily performance) and having to duck it out with expensive cooling, expensive memory and expensive PCBs against the established G80, they could've decided to cancel this product as it would do them no good and instead pretend to have planned a nice, quiet and inexpensive familiy-bbq all along.

If my speculation in the first paragraph was true, then that's at least what I would've done in their place - plus it fit's better with AMDs new profiling as a total platform provider instead of delivering single-high-end-Parts.

[Natoma]

Or it's simply Occam's Razor at work, and what has been stated regarding R600 from AMD is the truth.

Either way, we'll know in a couple of weeks.

[trinibwoy]

Originally Posted by Natoma

Or it's simply Occam's Razor at work, and what has been stated regarding R600 from AMD is the truth.

Actually applying the razor would lead one to the conclusion that something went wrong and forced AMD's hand. The whole cancellation of editor's day as a move to a strategic delay and "family launch" requires a whole lot more imagination/gullibility

[SirPauly]

Originally Posted by Quasar

I should explain myself a little further. I think the delay in R600 was at least in part due to the performance not exceeding G80 by a healthy enough margin. So they tried and stretched clocks as far as they'd go. Then it was discovered, that they'd get only so much R600 running at a high enough clockrate to beat G80 by more than 5-10 percent - plus needing a healthy surplus of power and cooling.

So, instead of taking the blame for another "NV30" (wrt heat, cooling, delay, not necessarily performance) and having to duck it out with expensive cooling, expensive memory and expensive PCBs against the established G80, they could've decided to cancel this product as it would do them no good and instead pretend to have planned a nice, quiet and inexpensive familiy-bbq all along.

If my speculation in the first paragraph was true, then that's at least what I would've done in their place - plus it fit's better with AMDs new profiling as a total platform provider instead of delivering single-high-end-Parts.

Thank you for expanding and understand your point-of-view.

[w0mbat]

3DMark06

X6800@default, Intel 965er Mobo

8800GTX@630/2060 = 11391
R600@default = 12500 (+10%)

http://www.fx57.net/?p=550

[jimmyjames123]

Originally Posted by Quasar

I should explain myself a little further. I think the delay in R600 was at least in part due to the performance not exceeding G80 by a healthy enough margin. So they tried and stretched clocks as far as they'd go. Then it was discovered, that they'd get only so much R600 running at a high enough clockrate to beat G80 by more than 5-10 percent - plus needing a healthy surplus of power and cooling.

So, instead of taking the blame for another "NV30" (wrt heat, cooling, delay, not necessarily performance) and having to duck it out with expensive cooling, expensive memory and expensive PCBs against the established G80, they could've decided to cancel this product as it would do them no good and instead pretend to have planned a nice, quiet and inexpensive familiy-bbq all along.

I agree in general. Also, since AMD/ATI feels that their R6xx cards have a bigger advantage in DX10 vs DX9 in comparison to G8x, then the more they delay the closer they get to seeing DX10 games on the market. On top of that is the fact that a longer delay gives some extra time to work on drivers, and some extra time to improve yields and to build up more inventory of cards, then it's not far-fetched to see AMD going this route.

The biggest problem with a delay tactic is that it gives the competitor time to take good advantage of the situation.

[Natoma]

Originally Posted by trinibwoy

Actually applying the razor would lead one to the conclusion that something went wrong and forced AMD's hand. The whole cancellation of editor's day as a move to a strategic delay and "family launch" requires a whole lot more imagination/gullibility

That's only because you're thinking like a historical GPU industry follower, not a CPU industry follower.

[Geeforcer]

We also need to consider the whole distribution ecosystem (read: AMD/ATI exclusive partners). If anyone would like to explain to me how not having a competitive high-end product for 6 months helped THEM, go for it. In fact there are rumors that Connect3D in on the verge of bankruptcy.

I think economic equating is fairly simple. People who were looking for a high-end card over last 6 months bought G80s. Unless R600 is amazing enough to convince them to upgrade from G80 to R600, each G80 sale is for all intends and purposes a loss of potential R600 sale.

Furthermore could anyone explain in detail the mechanism by which family lunch will improve sales of either high-end or midrange product as opposed to piecemeal rollout? I was not aware that "halo" effect fades so quickly.

Bob: "Wow, did you see those RV630 cards released today?"
Tom: "Not interested."
Bob: "Why not?"
Tom: "Isn't it obvious? The high-end part, R600 came out 2 months ago! Why would I want RV630-card?"
Bob: "You are right! The fact that R600 came out 2 months ago totally destroys the viability of RV630 as a competitive product!"
Tom: "If only R600 came out today. That would have made RV630, like, totally kickass extreme! To the MAXXXX!!!"
Bob: “Do the Dew!”

[trinibwoy]

Originally Posted by Natoma

That's only because you're thinking like a historical GPU industry follower, not a CPU industry follower.

I don't get it There isn't a history of family launches for CPU's either. It's quite common for a new architecture to be rolled out over a very long period of time.

[trinibwoy]

Originally Posted by jimmyjames123

Also, since AMD/ATI feels that their R6xx cards have a bigger advantage in DX10 vs DX9 in comparison to G8x, then the more they delay the closer they get to seeing DX10 games on the market.

If that's their tactic they should delay till september because that's when DX10 games will be available. Also, if those benches are accurate it's doing quite fine in DX9 as well.

[caffeinated]

Originally Posted by Natoma

That's only because you're thinking like a historical GPU industry follower, not a CPU industry follower.

Forgive me for saying so, but does basic business sense change when assessing *any* product? You can't sell out of an empty wagon - that's pretty basic.

[dnavas]

Originally Posted by trinibwoy

Actually applying the razor would lead one to the conclusion that something went wrong and forced AMD's hand. The whole cancellation of editor's day as a move to a strategic delay and "family launch" requires a whole lot more imagination/gullibility

I've been trying to find a nice way to say this...
It would not be the first time that I've run into execs that believe that packaging trumps content -- Product Suites, "Solutions" vs. "Products", etc. are merely pitstops to believing "it doesn't matter what we ship, only that we make it look and sound good." I think I would prefer to believe that 'something went wrong'. The alternatives seem worse....

http://www.fx57.net/?p=550

I'm glad to see R600 getting better numbers than the 8800. Hopefully we'll get some interesting architectural twists to talk about! "12500" is an awfully round number though

[jimmyjames123]

Originally Posted by trinibwoy

If that's their tactic they should delay till september because that's when DX10 games will be available. Also, if those benches are accurate it's doing quite fine in DX9 as well.

I believe that AMD had to strike some sort of "balance" among the factors I listed above. If they launch in September, then that is almost a full year of no competition at the high end for NV, and it gives NV even more time to work up an awesome refresh. If they launch in March, then it's only about three or four months of no competition at the high end for NV, but then AMD is farther away from showing DX10 advantage, and also there is much room for improvement on driver side, on yields, on inventory, etc.

I found it interesting that when the G80 came out, NV stressed that one of the first and foremost rule of the new forward-looking DX10 architecture was to excel on DX9 games. As R600 is getting closer, AMD is only talking about DX10 advantage. Most likely NV realized that AMD would not be able to gain a significant performance advantage on DX9, and most likely AMD realized that as well and move on to promoting DX10 performance.

[Natoma]

Originally Posted by trinibwoy

I don't get it There isn't a history of family launches for CPU's either. It's quite common for a new architecture to be rolled out over a very long period of time.

When AMD first launched the Athlon, Athlon 64, X2, did they have ~3-5 clock speed variants for each product at launch?

How about Intel wrt the Pentium IV and Core 2 Duo?

[Natoma]

Originally Posted by caffeinated

Forgive me for saying so, but does basic business sense change when assessing *any* product? You can't sell out of an empty wagon - that's pretty basic.

Right. But lo and behold, that's how ATI and Nvidia did business for a very long time. Particularly ATI in recent years.

[Geeforcer]

Originally Posted by Natoma

When AMD first launched the Athlon, Athlon 64, X2, did they have ~3-5 clock speed variants for each product at launch?

How about Intel wrt the Pentium IV and Core 2 Duo?

I am sorry, but the analogy is faulty. Lunching Athlon 64 at several frequencies is the equivalent of lunching (card name) XTX PE, XTX, and XT at the same time - same chip, several price points- which is what original R600 lunch was supposed to be. The CPU equivalent of the family lunch currently planned would be AMD lunching new revisions of Operton, Athlon and Sempreton one the same day. Care to point out when that happed last time?

[Natoma]

Originally Posted by Geeforcer

I am sorry, but the analogy is faulty. Lunching Athlon 64 at several frequencies is the equivalent of lunching (card name) XTX PE, XTX, and XT at the same time - same chip, several price points- which is what original R600 lunch was supposed to be. The CPU equivalent of the family lunch currently planned would be AMD lunching new revisions of Operton, Athlon and Sempreton one the same day. Care to point out when that happed last time?

AMD released the Opteron 100, 200, and 800 series at the same time. And there were significant differences other than just core clock between those lines iirc.

That's just off the top of my head.

[caffeinated]

Originally Posted by Natoma

Right. But lo and behold, that's how ATI and Nvidia did business for a very long time. Particularly ATI in recent years.

Selling out of an empty wagon means you haven't got anything to sell. Since technology has the shelf-life of a fresh fish, technology companies have to consistently produce new things to sell. The article that Trinibwoy linked to showed how 3dfx pretty much died on their own sword by not being able to do this. When technology companies have no new technology to sell...they are selling out of an empty wagon.

Nvidia had this same problem; for details, review how the FX 5800 went . Intel also did this; review the NetBurst architecture for details .

Sony might be an example to the opposite; the PS3 is new and different, but it isn't selling very well. Then again, it had delays, too. I don't know of too many folks that can argue against the idea that Sony is either delusional or grasping at straws when they claim awesome PS3 sales.

The point is, delays in what technology companies have in their wagons to sell cause issues because they aren't selling the most important thing that they have to sell: innovation. That is why I am saying ATI has an empty wagon at present. I love AMD processors, and their platforms; but right now they are also beginning to take a beating because they have no answer to Core 2.

To be honest though, I'm very likely to shut down my computers and move to the Sahara desert if I read on the [H]OCP forum one more post of "I have an AMD X2, am I CPU bound?" I actually saw one of "I have a 5000+ am I CPU bound?" Uh, no. Unless you plan on upgrading to an LN2 cooled Quad core, you are probably okay
The point of that statement is that somehow, Intel has managed to press home the idea that X2's just don't cut it...blah blah blah. I guess it was because they have something new, right?

EDIT: I'll also point out that my X2 PC consistently outperforms Core 2 systems that aren't overclocked and some that are. X2's are not 'CPU bound' they just don't overclock as well. Hopefully with Brisbane some of that changes.

And one last edit : I picked up a 1900GT for $129 dollars that was a "holy crap!" for me. I threw it in my Folding box today. I bought it because the fan died on my ancient 6600 non-gt..LOL, but also because for $129 that's pretty doggone cheap! ...And it folds

[trinibwoy]

Originally Posted by Geeforcer

The CPU equivalent of the family lunch currently planned would be AMD lunching new revisions of Operton, Athlon and Sempron on the same day. Care to point out when that happed last time?

Agreed, it's already a stretch to think that the R600 situation has any relation to CPU launch practices but even if you get to that point there still is no evidence of CPU "family launches". The launch of 3-5 clockspeeds of the same chip is not a family launch since CPU families typically consist of much larger number of SKU's.

jimmy, I think Nvidia stressed DX9 performance because that's what IS important right now. It's not an opinion or some great insight - anybody can look at their current games library and tell you that

[Geeforcer]

Originally Posted by Natoma

AMD released the Opteron 100, 200, and 800 series at the same time. And there were significant differences other than just core clock between those lines iirc.

That's just off the top of my head.

1) I thought there was availability delta between 1xx, 2xx and 8xx families.
2) From engineering point of view, what IS the difference between them?
3) Athlon FX/64 are based on the same architecture as Opteron, as was 64-bit version of Sampron. If there were perusing the "family launch" philosophy, should they not have lunched together? Instead, Athlon 64 was released 6 month after Opteron and Sampron 64-bit followed 2(!) years later... and I dare say there is much greater difference between RV610 and R600 there is between Athlon FX and Opterons, design-wise.

[Natoma]

Originally Posted by Geeforcer

1) I thought there was availability delta between 1xx, 2xx and 8xx families.
2) From engineering point of view, what IS the difference between them?
3) Athlon FX/64 are based on the same architecture as Opteron, as was 64-bit version of Sampron. If there were perusing the "family launch" philosophy, should they not have lunched together? Instead, Athlon 64 was released 6 month after Opteron and Sampron 64-bit followed 2(!) years later... and I dare say there is much greater difference between RV610 and R600 there is between Athlon FX and Opterons, design-wise.

1) I believe those processors were released at the same time. Actually, let me amend that. The 800 series did come out later. Maybe a few weeks? But the 100 and 200 series came out at the same time.

2) http://www.amd.com/us-en/Processors/...6_9240,00.html

Highlights:
100 = Single CPU per system, Socket 939, 0 coherent HT Links
200 = Dual CPU systems, Socket 940, 1 coherent HT Link
800 = Up to 8 CPU system, Socket 940, 3 coherent HT Links

I thought there were cache differences as well however. Guess I was wrong.

3) RV variants are basically 1/2 to 1/4 the R no? But functionally they're the same per quad. Oh, and they're typically on smaller processes than the Rs. Anyone want to chime in with other differences?

[Natoma]

Originally Posted by caffeinated

Selling out of an empty wagon means you haven't got anything to sell. Since technology has the shelf-life of a fresh fish, technology companies have to consistently produce new things to sell. The article that Trinibwoy linked to showed how 3dfx pretty much died on their own sword by not being able to do this. When technology companies have no new technology to sell...they are selling out of an empty wagon.

Forgive me, but isn't this called a paper launch? Unless I'm just misunderstanding what you're getting at?

[Sound_Card]

http://rage3d.com/board/showpost.php...43&postcount=1


cool little interview in there.

The reason we decided to delay the launch was that we wanted to have a complete DX10-enabled solutions top-to-bottom. A lot of people wrote that the reason it is delayed is because of a problem with the silicon, but there is no problem with the silicon. We are demonstrating it. We can ship it today. But if you think about it, looking at where the market is at, the volumes are going to be in the RV610 and RV630, so it makes sense for us to do a one time launch of the entire family of DX10 enabled products."

Also increasingly in particular with Vista, as we've seen with the competition. It doesn't matter if you're shipping the silicon if the drivers are not stable. There is nothing more frustrating than having bought a new graphics card and having your system crash repeatedly because the drivers are not ready. Although we today, even by Microsoft standards, have the best and most stable drivers in the entire industry. The few weeks will give us even more time to continue improving the drivers. Again, the decision lies in the fact that we will have a top to bottom DX10 offering with drivers that will have a very very high level of stability and the only difference is a few weeks. So it seems to make a lot of sense to do it that way.

[Geeforcer]

Originally Posted by Natoma

1) I believe those processors were released at the same time. Actually, let me amend that. The 800 series did come out later. Maybe a few weeks? But the 100 and 200 series came out at the same time.

2) http://www.amd.com/us-en/Processors/...6_9240,00.html

Highlights:
100 = Single CPU per system, Socket 939, 0 coherent HT Links
200 = Dual CPU systems, Socket 940, 1 coherent HT Link
800 = Up to 8 CPU system, Socket 940, 3 coherent HT Links

I thought there were cache differences as well however. Guess I was wrong.

3) RV variants are basically 1/2 to 1/4 the R no? But functionally they're the same per quad. Oh, and they're typically on smaller processes than the Rs. Anyone want to chime in with other differences?

So would it be fair to say in recent memory there have been no CPU "family" launches, at least if we were to define a "family" as "fairly different chips based on the same architecture aimed at different market segments"? Yes, Core 2 lunched at several frequencies, but so did NV25.