The AMD Execution Thread [2007 - 2017]

It's not just a matter of AMD's direct costs.

What of the matter of the costs associated with essentially giving Nvidia the license to print money in the high end market segment?

The high end was very good for Nvidia, and the positive effect was enhanced by the fact that ATI was a no-show for 6 months.

AMD will have to make something, if only to save Nvidia from the horrible fate of making too much money.

 

Honestly, I think we're just too used to the fact that ATI and NVIDIA have always made a "top to bottom" product line...However, there is nothing saying that either vendor MUST do so..

If we use the car industry as an analogy...

Let's say NVIDIA and ATI have been traditionally much like BMW and Mercedes. They offer cars from ~$28k all the way up to ~$128k or more. Essentially, there is something for every market...

What I'm suggesting is that AMD wants to become a lower-cost, higher-volume player like Toyota. Down the road, they can always offer an "enthusiast" part like they do with their Lexus division...

The Camry (and other Toyotas) sold well before the Lexus brand was even created so I don't think the "halo effect" is a necessity for success (though it certainly helps)

 

I'm not talking about the halo effect.

I'm talking about the fact that Nvidia made money from the 8800, and kept fat margins even fatter for far longer because AMD had no answer in the same segment.

That's extra money Nvidia can use to speed up its R&D, money that would not be there if AMD's products was actually competitive and on-time.

Ceding that market segment means AMD just hands Nvidia free money.

 

And more money taken from enthusiasts/gamers if there is only one company to buy high end cards from.

I could just see it now - the equivalent of 8600 GTS's for $400-550 for the "performance" gamer, and for the high end, nothing but 8800 GTX Ultra's for $899-999. In the mainstream: 10 different choices at $150, but all of them about equal to the 2600XT.

 

Toyota is a great inclusion here even without bringing up lexus at all. In 1993 they produced their flagship sports car, the supra(the earlier years don't necessarily count) for an extremely large cost. After they factored in RnD they ended up losing tons of money on the slow selling car. Even though, all the other companies were developing and touting their flagship sports cars moreso for marketing than for sales (corvette, NSX, viper, etc...) Toyota realized they were simply losing money on these, and pulled the plug for hte us market in 98. Look how good they've done since then. Only today are they deciding to re-enter the dream-car market, with a 80,000 dollar lexus.

Why amd cannot make this decision as well is beyond me. Last i checked, don't these graphics companies make the majority of their money from mid-range OEM parts ? What's the point of developing high end stuff... toyota did just fine without it. I know they are different industries, but similar in some aspects, and i really think it's a worthy idea for AMD to try.

 

The cost of high-end R&D is incremental with that of the mid and low end.
There is no exclusively high-end feature, and R600 has shown that the high end may very well lack features.

In addition, unlike Toyota, there is no sea of competitors fighting for the high end.
There's just one competitor.
The high end ceded by AMD goes immediately and wholly to Nvidia.
There is nobody else keeping Nvidia's margins down or keeping Nvidia from fully capitalizing on owning the high end.

Nvidia's margins have already shown what having the high end to oneself for 6 months can do for the bottom line.
G80 is not a loss leader.

The only argument against the high end apparently is "it's not guaranteed income".

The minute AMD gives up the high end, it does become guaranteed income, for Nvidia.

 

Let's remember that we're talking about AMD and not about ATI.....AMD's "chief" business is CPU production.......Regardless of how well the GPU division performs, AMD is doomed if their CPU business fails to remain competitive...To that end, their primary concern is Intel.

Intel threatens AMD's core business of CPU's. Intel threatens AMD's IGP and high volume mainstream GPU business....etc...(Larabee)

NVIDIA is a large (and growing) threat....but Intel is certainly the 800lb gorilla...

If staying competitive with Intel is primary to surviving, I don't think it would be too crazy an idea to see theme concede the enthusiast segment to NVIDIA and instead focus on CPU's, IGP's, and mainstream GPU's.

 

Except Intel is already working on graphics too. Concede your graphics technology business to Nvidia (which is what you're talking about) then Intel will benefit and crucify you 24 months from now.

AMD's main competitor used to be Intel. Now it's Intel and Nvidia, because AMD now make GPUs and chipsets, as well as just CPUs, and without that business, they haven't moved forward, the are just banging their heads against the Intel brick wall.

This business is about technology, patents, future developments, IP, having the technology lead in every segment. If you don't have stuff in the pipeline, you're nowhere. You seem to think you can just close down a business division for a couple of years, and then switch it back on again a couple of years later as if nothing has happened. You can't because in the meantime your competitors have eaten your lunch and dinner, and breakfast the next day.

 

Toyota is a very bad comparison, the scale is on a completely another level. No R&D costs could ever bring them down, being at least 10x bigger than the likes of AMD, NV etc. So say five billions for them is like 50 millions for AMD, bluntly said.

 

You people are so optimistic about Intel's gfx-development

History always showed otherwise. They are big-mouthed as always and will most probably produce some half-arsed POS gfx-deccelerator as they always did.

 

Last time Intel dropped out of graphics, mainstream GPUs were pretty much just about gaming. Now they are about a lot more than that.

 

They aren't - it's still about gaming, with a casual nod toward video codecs.

The gfx-asic houses has wanted to create another impression to please investors, and have indeed looked at other uses, but these are hypothetical directions. Not real life bulk use. It could be argued that transistors spent on making GPUs more generalized are transistors where it looses efficiency at its main task, but regardless of that, GPUs are still overwhelmingly about games.

We are both making just making officious sounding statements, but really - where is all that GPU heavy lifting that is not related to gaming? The video and odd image filter don't have any need for todays GPU behemoths.

 

That and workstations, which by the way is another reason why R600 is important, unless AMD wants to give up high-end gaming and those super-markup FireGL cards.

Graphics cards and cheap x86 chips were what beat the tar out of the the RISC workstation market.

 

my comparison was from 1993, they were much smaller then... much much much smaller

 

Vista's Aero alone places much, much higher requirements on average GPU spec than anything we've previously seen in the mainstream.

 

I'm not really sure about that... An NV FX5200 64mb card can do Aero Glass; I suppose that's "more" than any previous Windows OS, but it's really nothing awe-inspiring in my opinion.

 

Apple announced a new imac line today at a press conference, all featuring the hd 2400/2600 chips. That's a pretty significant oem win I guess.

Edit: Just a comment that it seems like ATI is getting back some of the low to mid range GPU business they lost during Nvidia's 6600/7600 era. Even the within htpc crowd at avsforum, the 2400/2600 are getting the kind of attention that the x1300/1600 never got.

 

I thought 128MB was a requirement of aero. Or maybe just dx10?

 

Seriously. I'm confused about this. I have an HP Pavilion notebook with a nice notebook CPU/HD/Memory. But a really crappy GM950 that is running aero just fine. Just how demanding is it really?

 

But still huge, with annual revenues of at least $50 Billion, millions of vehicles sold every year and half the Japanese market. A project like the Supra is a drop in the bucket. Certainly far less than what Toyota spends on auto racing, especially after Toyota entered Formula 1.