AMD: R9xx Speculation

[Harison]

I don't think people are disputing that ATi are in a better position overall. What is being pointed out is that they are in a relatively worse position with Cayman than they were with Cypress. They look like having a similar performance delta but they have a bigger die size.

You are overstating, Cayman is in a slightly worse position than Cypress, and thats the worst case scenario. You know whats a probable better case? Mature 40nm process increases yields for both manufacturers, but AMD yields gain will be exponentially better than massive die GF100b, so AMD even with slightly bigger die they could be in the same position as before.

If we factor supply issue, AMD with all Cypress advantages couldnt benefit fully due to limited supply, its much lesser issue now. UKOC reported they have huge stock already, and it seems its common thing. Therefore AMD could very well profit from Barts/Caymans way more than they could from 5xxx generation.

ATi made the same amount of money over the last year as Nvidia, it was pointed out a few posts back. Which means even with their small die strategy ATi haven't been able to take advantage, which means they have less chance now since they are in a relatively worse position.

NV main profits came from pro market, not from consumer division. Plus compare significant market share shift to AMD favor, I guess they took advantage after all

 

[Richthofen]

You can't really compare nvidia 's profits against amd. They have a hugely sucessful professional busniess where those huge dies don't matter because cards sell for many times what a consumer board sells for .

Well you can because first of all i don't think that their professional business was able to drive Q3 that much. The professional segment is not that volatile to perform such a change like the one shown in the financial results from Q2 to Q3.

And even if it would only be their professional segment - which i doubt - than where is the point for the consumer? I mean if Nvidia can afford huger DIE size due to their professional segment, AMD lost the DIE size advantage in my view.

Everytime when it get's likely that AMD is not able to beat Nvidia you hear the DIE size story. Who cares?

Let's imagine if Nvidia's professional business doubles and they earn millions of money. That would give them the opportunity to subsidize their consumer products (this is not the case at the moment - its just an example). And so? In the end the consumer product would be price competitive. This is all the consumers has to and will care about and not how Nvidia is doing that in such a scenario.

 

[gkar1]

What does any of this armchair stock analyst talk have to do with the architecture of the card? Getting tired of seeing this thread derailed by the same usual suspects.

 

[Richthofen]

What does any of this armchair stock analyst talk have to do with the architecture of the card? Getting tired of seeing this thread derailed by the same usual suspects.

You have to ask this to the people always bringing up the same lame story "yeah but AMD's DIE size is much smaller and so the efficiency is better". once it is getting rough for AMD performance wise.

 

[DavidGraham]

I myself got bored of this die size story , we have been discussing it for what now ? 3 years ? Yes NVIDIA does have bigger dies , but they manage to survive and make a profit out of them EVERY TIME , which suggests their philosophy isn't as flawed as people try to make it seem , if they were losing money because of it , they would have stopped long time ago .

It doesn't matter if they do it in the professional or the consumer segment , we are not business analysts here , they have far more capable people doing that job for them anyway .

As a consumer I never care about die sizes , what matter is performance then power then heat and acoustics , anything else doesn't concern me as a customer .

The devil is truly in the detail , and I never can believe that a regular journalist/user could really get a hold on the full financial situation of these products , there are so many variables , the die size thing is one of them , but it never was the final deciding factor , this is a business , nothing there is black and white ,only shades of gray .

 

[thatdude90210]

Someone is wrong on the internet, I must correct them.

 

[Arty]

I think the die size has some relevance here - esp if you consider that if 28nm is really delayed for another 1.5-1.75 years. Nvidia is quite close to the max reticule size and dont have much wiggle room there while AMD does.

I think that die size as part of price/mm2 was brought up by the usual suspect with the doom & gloom talk for Cayman. Its a few more days .. we'll see if Cayman sucks or bites.

 

[Shtal]

Now the Radeon HD 6970 proves that AMD doesn't have a single-GPU answer to NVIDIA's GeForce GTX 580. http://benchmarkreviews.com/index.php?option=com_content&task=view&id=13177&Itemid=8

How do they know about cayman performance ??

 

[Richthofen]

I think the die size has some relevance here - esp if you consider that if 28nm is really delayed for another 1.5-1.75 years. Nvidia is quite close to the max reticule size and dont have much wiggle room there while AMD does.

In such a case they have plenty of time to adjust their future products. For now it has no relevance as they make money with it. So in the end the GTX 580 and 570 have set the bar to beat. Not beating it will result in less sales and less market share going forward as the time advantage AMD had with the 5870 will be completely gone.

 

[zshadow]

How do they know about cayman performance ??

They don't. Baseless article conjured up from forum posts.

 

[eastmen]

In such a case they have plenty of time to adjust their future products. For now it has no relevance as they make money with it. So in the end the GTX 580 and 570 have set the bar to beat. Not beating it will result in less sales and less market share going forward as the time advantage AMD had with the 5870 will be completely gone.

Unless of course the 69x0 is cheaper and almost as fast. It would cause either nvidia to drop prices or for ati to have that market segment to themselves.

If nvidia is making ap rofit with a bigger gpu at $400-$550 then ati can make a profit with a smaller gpu at $300-$400

 

[jimbo75]

How do they know about cayman performance ??

Coles reckons the 6850 competes with the 460 768mb, so if he says the 6970 competes with the 570, it's probably a lot closer to the 580 in reality.

You can safely ignore BR, it's getting more ridiculous every time I read any of Coles articles.

 

[Arty]

In such a case they have plenty of time to adjust their future products.

Adjust down you mean?

So in the end the GTX 580 and 570 have set the bar to beat. Not beating it will result in less sales and less market share going forward as the time advantage AMD had with the 5870 will be completely gone.

Oh please, GTX580 was already beaten. Joe consumer doesnt care if his epeen is 1 mm longer than the next guy with his mGPU-on-a-stick SKU. The minute Nvidia has a SKU that can topple 5970 is when we can start the doom and gloom talk, for now please give it a rest here.

 

[ZerazaX]

You just have to look at the latest Q3 2010 results from both

AMD's GPU division operating profit was 1Mio$ which translates into a net loss due to write offs, tax charges etc.

Nvidia in the same quater made 85 Mio $ net profits at a gross margin of 46%. Q3 was the first quater where all 4xx products have been in the market and retail stores for a while. So even with the far bigger DIEs and the very price attractive GTX460 and 470 they were able to make far more money than AMD.
Q3 was from my point of view the turning point and a good proof why DIE size comparison is not relevant for the consumer.

$1 million is a nice number to look at, but you aren't factoring in that in Q3, Nvidia won a ton of OEM contracts in the mobile sector for the 4xx GPUs and that average selling price of cards dropped with the GTX 460's release.

Plus, costs for R&D and what not for Barts and Cayman were probably wracked up during this time as well.

Income/profit is only part of it - need to see where revenues were and what was shipped also.

 

[hoom]

The minute Nvidia has a SKU that can topple 5970 is when we can start the doom and gloom talk, for now please give it a rest here.

Don't forget Antilles/6990/X2 card...

old cow with new lipstick.

Just as long as she has a navigation tree & lipstick is the right shade of red I'm keen

There is something not being told/addressed about those results.

Its already been pointed out that the rumored SP count & clock rate give 6970 near same theoretical max math power to 5870.
Where Cayman should get interesting vs 5870 is cases that the XYZWT setup can't get good utilisation like serial dependencies & code that is high in trancendentals.
If those cases are quite common (which seems likely given the disparity between ATI & NV in raw maths power) then real-life|theoretical max performance should be quite a lot better.

 

[Xenus]

Honestly I expect most of the gains to be under tessellation.

 

[Shtal]

http://www.hardwareluxx.de/community/15863940-post170.html

 

[rpg.314]

Meh, gimme benches ...


PS: FWIW, I expect 6970 to lose to 580.

 

[Lightman]

Meh, gimme benches ...


PS: FWIW, I expect 6970 to lose to 580.

Me too, but not in all cases and by smaller margin than HD5870 is loosing to GTX480.

I think AMD addressed biggest bottlenecks of Cypress compared to GF100 with Cayman.

3 more days to go and we should be able to draw much better picture about product positioning. Hopefully price/performance/power ratio will be at least as good as Cypresses :smile:.

 

[ZerazaX]