AMD: R9xx Speculation

[ZerazaX]

Still, you can see that this particular game IS very memory intensive, because on other games, with similar settings, the GTX580 trounces the HD5870. Then, I still think its NOT a good term of comparison, which was my stance after all.

No it wasn't, your original stance was:

This is no better result than an HD5870 with 2 GB RAM, according to the HardOCP review...

Which is not proven AT ALL.

Besides, at 1920x1080, the cards might not be memory limited anymore, and so the advantage of 2GB vs 1.5 GB might not be so pronounced - I bet the 5870 2GB in 1920x1080 falls far lower because it no longer has the 512MB advantage and now falls behind in actual speed. But unless someone has a comparison at 1920x1080, all we know is that:

at 1920x1080, the F1 results favor the 6970

 

[mao5]

crysis: 1920x1200 4x AA 36.8 VS 38.5 OMG any mem limited ?

 

[ECH]

I'm taking those results with a grain of salt for now. Also, interesting is that he also owns a 480 (from several user posts asking him about it) and he hasn't provided any comparisons . That would be the ideal user review but having been asked that question (along with CCC settings) several times it's gone ignored.

 

[fellix]

Look at the texel fillrate -- full speed FP16 filtering?

And some DC5.0 benching - ComputeMark. Quite slow...

 

[LordEC911]

8.790.6.0

Is that out dated?

Yes.

 

[ECH]

So in a nutshell all his results are not valid because he's not using a driver that's designed for the 6970? Then where the heck did he get 8.790.6.0 from? I thought that was from the cd??

 

[AlphaWolf]

So in a nutshell all his results are not valid because he's not using a driver that's designed for the 6970? Then where the heck did he get 8.790.6.0 from? I thought that was from the cd??

Says test in that picture. So maybe it was distributed to OEMs before the production drivers.

 

[ECH]

Says test in that picture. So maybe it was distributed to OEMs before the production drivers.

Yeah your right there is no telling how accurate that pic is either. But I don't recall 8.790.6.0 as a publicly released WHQL or hotfix driver. It will be interesting to note what driver reviewers use.

 

[Rangers]

http://www.techreport.com/articles.x/20088/9
Metro 2033

Very high
5870: 22
GTX570: 27
Difference: 22%

High:
5870: 31
GTX570: 40
Difference: 29%

Medium:
5870: 35
GTX570: 54
Difference: 54%

Now look at the max settings of the other games I mentioned.

http://www.techreport.com/articles.x/20088/5
Lost Planet 2
5870: 21
GTX570: 31
Difference: 47%

http://www.techreport.com/articles.x/20088/11
Dirt2 DX11
5870: 80
GTX570: 100
Diiference: 25%


The difference from 5870 to GTX580 is a lot higher. All I´m trying to say is that 6970 must be much faster than 5870 to be faster than GTX580 across the board, specially at the DX11 games.

I said Anandtech.

Go look at Anandtech numbers for Warhead, Metro 2033, whatever other game I mentioned...5870 is very close to 570.

All 1920X1200 numbers, high settings and AA. These were the games I mentioned in my post

Metro 2033 DX 11 5870=33 FPS 570=36.5 FPS : 10%
Warhead 5870=45.9 570=48.5 : 6%
Civ 5 5870=35.6 570=41.5 : 16%

 

[Xenus]

Yes and your one data point is no better then his one data point. You need to go through many more reviews than that to make a reasonable conclusion either way.

 

[Rangers]

I wasn't claiming to make a big conclusion. If you read the post he responded to.

Her had said something like "in latest games with DX11 Nvidia is farther ahead" and I said that's funny, from just glancing at the AT review I had got the opposite impression. And so then I double checked a few games at AT and threw them out hastily.

 

[Harison]

the problem is:
334mm2 Cypress was fighting with 530mm2 480
385mm2 Cayman will fight with 530mm2 580
See? mm2 advantage is less now.

That depends if Cayman faster or slower than GF100b. Even if its slightly slower, AMD still has a clear advantage, 385mm2 vs 530mm2, who is in better position?

If AMD was unable to make money from Cypress, how are they going to make some with Cayman, now their expenses are higher not only from bigger chip, but from twice the memory too.
Increasing TMUs? They are already more than in the competitor.

Say what? AMD made plenty of money from Cypress, with ~46% margins and how many sold? You have no basis to claim AMD wont make money from Cayman. Or look at this way - if AMD isnt making money with much smaller die, then how much loses NV would have by selling huge die chips in the same price/perf brackets?

 

[mapel110]

new CCC

 

[eastmen]

not sure how i feel on the new ccc

 

[gamervivek]

old cow with new lipstick.

 

[NathansFortune]

That depends if Cayman faster or slower than GF100b. Even if its slightly slower, AMD still has a clear advantage, 385mm2 vs 530mm2, who is in better position?

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.

Say what? AMD made plenty of money from Cypress, with ~46% margins and how many sold? You have no basis to claim AMD wont make money from Cayman. Or look at this way - if AMD isnt making money with much smaller die, then how much loses NV would have by selling huge die chips in the same price/perf brackets?

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.

 

[Richthofen]

That depends if Cayman faster or slower than GF100b. Even if its slightly slower, AMD still has a clear advantage, 385mm2 vs 530mm2, who is in better position?

Well looking at the most recent Q3 results i would say Nvidia as they were already able to make far more money than AMD's GPU division even with their much larger DIEs. As the difference gets less now i would say that their position is getting even better.

Say what? AMD made plenty of money from Cypress, with ~46% margins and how many sold? You have no basis to claim AMD wont make money from Cayman. Or look at this way - if AMD isnt making money with much smaller die, then how much loses NV would have by selling huge die chips in the same price/perf brackets?

I am not going to say that AMD will not make money because we will only be able to see that 2 quaters down the road. But regarding the 46% margin i have to add that Nvidia hit that number to in Q3 and that with their huge dies. And history of 2010 proves that Nvidia is selling larger chips in the same price segment or sometimes even cheaper (GTX460 and 470) and still they make good money. So for the consumer that means, that the DIE size difference is rather irrelevant. You can bet that Nvidia will follow any price cut because of two things:

1) they want to regain market share at any cost at the moment
2) they can easily afford that due to their nice professional segment and due to the fact that they are in no debt.

In the end AMD will get into trouble if Cayman will struggle with the GTX570 performance wise.

 

[eastmen]

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 .

 

[Richthofen]

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.

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.

In the end you go into a store and check pricing. If Nvidia will continue to be price competetive, which i think they will do until they regained their lost market share, the consumer will choose the better product. DIE size does not matter to him.

 

[eastmen]

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.

In the end you go into a store and check pricing. If Nvidia will continue to be price competetive, which i think they will do until they regained their lost market share, the consumer will choose the better product. DIE size does not matter to him.

Once again , you are lumping in nvidia's professional market. Not the consumer market. When they can sell a quadro for 2 thousand dollars it paints a very diffrent picture