AMD Cayman info (or rumor)

[bill_d]

If it perform's like xfire 6850's. I am pretty sure they would have released it already. Paper launch or not, it would have put a pretty strong damper on the gtx 580 launch in addition to raise the 6xxx namesake, to increase sales of the 68xx cards even further.



They are not going to double everything unless they want the card to consume 300 watts + be 500mm in size.

I think the 3-4 tesselation performance is bull. Synthetics show the greatest difference in tessellation performance. People rationalizing that Cayman could be 2-3 faster in synthetics and even faster in games by the vagueness in the rumors shows how misplaced their rationality is.

Synthetics are coded to specialize in showing maximum prowess of a certain aspects. Drivers are specifically coded for them to maximize performance, hence why Nvidia kept on bragging about unigine heaven.

We won't see faster tessellation in games compared to synthetics. Be rational people.

not if they need a complete driver rewrite to show it at it's best, they get one shot at a launch most sites don't review a second time for new drivers

[generics_user]

2x HD6850 have much higher VRAM bandwidth, though.
And we don't know for sure whether everything gets doubled or not.
It's all speculation...

6850 is extremely inefficient in terms of mem BW usage (mem BW scaling is better than on cypress), i bet that Cayman will have a significantly more efficient memory controller like cypress did and will make up for the rest with a higher mem BW

[tajoh111]

not if they need a complete driver rewrite to show it at it's best, they get one shot at a launch most sites don't review a second time for new drivers

They only need the driver rewrite if it doesn't performs as high as they wanted. That is, beating a gtx 580.

From a hardware perspective, I think the 6970 has the innards of a card that can beat a gtx 580. But it isn't going to be super easy and I think the delayed launch meant that it might have not won that battle and drivers are going through heavy revisions to get it to that level.

From how a few reliable rumors leakers from AMD side were going on, it seemed like AMD could have released the 6970 the same week as NV released the gtx 580. Something tells me AMD delayed it because the card probably beat a gtx 480 but not a gtx 580.

[xBanzai89]

According to latest slides (from Nov. 22nd). 4d is similar to 5d performance. Number of ROP number has not changed. TMU number for HD 6850 is 48. HD 6970 will have more than 80 (according to slide.) which is almost double. Double the vram (2Gb), but less likely not memory bandwidth

I'm confused here. Is a 4D arrangement with 1600 SP (with 400 per section) or one with 1280 SP (with 320 per section same as 5870) perform similar to a 5D like the 5870 that has 1600 SP (with 320 per section)?

[-Boris-]

I'm confused here. Is a 4D arrangement with 1600 SP (with 400 per section) or one with 1280 SP (with 320 per section same as 5870) perform similar to a 5D like the 5870 that has 1600 SP (with 320 per section)?

320x4 would perform similar to 320x5. So, 400x4 would be ~25% faster. Now shader performance isn't everything so unless other parts are beefed up these 25% is in best case.

[xBanzai89]

320x4 would perform similar to 320x5. So, 400x4 would be ~25% faster. Now shader performance isn't everything so unless other parts are beefed up these 25% is in best case.

Ok thanks for clearing that up. I've just been a bit confused, since I have seen people on forums saying that.

[zalbard]

The only official thing I've seen regarding 4D performance is "the same performance as 5D while using 10% smaller sized die".

[zerazax]

They only need the driver rewrite if it doesn't performs as high as they wanted. That is, beating a gtx 580.

From a hardware perspective, I think the 6970 has the innards of a card that can beat a gtx 580. But it isn't going to be super easy and I think the delayed launch meant that it might have not won that battle and drivers are going through heavy revisions to get it to that level.

From how a few reliable rumors leakers from AMD side were going on, it seemed like AMD could have released the 6970 the same week as NV released the gtx 580. Something tells me AMD delayed it because the card probably beat a gtx 480 but not a gtx 580.

Why do people keep insisting its performance? First people think they can juts change BIOS and clocks within 3 weeks, now people think drivers are going to give a significant boost in 3 weeks rather than over the course of months? Those would have to be a hell of a driver change to get the card from 480 to 580 levels

Far and away the most likely issue is components or manufacturing delays for the timeframe we are talking about at this point.

6850 is extremely inefficient in terms of mem BW usage (mem BW scaling is better than on cypress), i bet that Cayman will have a significantly more efficient memory controller like cypress did and will make up for the rest with a higher mem BW

They've said that Barts used Redwood's mem controller. Cypress is significantly more complex

[Chickenfeed]

Those would have to be a hell of a driver change to get the card from 480 to 580 levels

Far and away the most likely issue is components or manufacturing delays for the timeframe we are talking about at this point.

I'm with you here. If driver improvements that magical were possible in mere weeks, i'd think they've been holding out on performance intentionally.

Hopefully we get a clear and honest answer when launch time comes.

[Heinz68]

Why do people keep insisting its performance? First people think they can juts change BIOS and clocks within 3 weeks, now people think drivers are going to give a significant boost in 3 weeks rather than over the course of months? Those would have to be a hell of a driver change to get the card from 480 to 580 levels

Far and away the most likely issue is components or manufacturing delays for the timeframe we are talking about at this point.

Right on. I think beating the GTX 580 is AMD last concern right now, plus they still have the world's fastest graphics card HD 5970

AMD strategy is to make smaller and cheaper chips and get the top performance card with dual GPU, if they beat the GTX 580 with smaller chip it would be just a bonus and something Nvidia would have to worry more about.

[flippin_waffles]

Shouldn't AMD's upcoming parts be designed to compete against what nv would have had on the market if they wouldn't have screwed the pooch? 580 is basically what AMD planned on competing with last generation. so what will nv have to compete with AMD's next gen, ie Cayman?

[cegras]

Why do people keep insisting its performance? First people think they can juts change BIOS and clocks within 3 weeks, now people think drivers are going to give a significant boost in 3 weeks rather than over the course of months? Those would have to be a hell of a driver change to get the card from 480 to 580 levels

Far and away the most likely issue is components or manufacturing delays for the timeframe we are talking about at this point.



They've said that Barts used Redwood's mem controller. Cypress is significantly more complex

Not really one hell of a driver, but more like rewriting the driver to use the new architecture.

[tajoh111]

Not really one hell of a driver, but more like rewriting the driver to use the new architecture.

I actually agree with you on this.

[cegras]

There's been no tech news of note recently, the forums have grown so stagnant ... even Engadget is quite the zzz.

[Heinz68]

Not really one hell of a driver, but more like rewriting the driver to use the new architecture.

The software and hardware teams works together from day one when designimg new architecture, there is no such thing like rewriting to driver in last minute and getting some huge improvements like from faster than GTX 480, after rewrite faster than GTX 580 as somebody said, that is complete nonsense.

I believe more the zerazax argument in post #1633 this thread, after all there were few rumor post about it.
AMD has nothing to gain to hold back the cards if the product was ready.

[slaveondope]

I blame Dr.Mos

[Heinz68]

Shouldn't AMD's upcoming parts be designed to compete against what nv would have had on the market if they wouldn't have screwed the pooch? 580 is basically what AMD planned on competing with last generation. so what will nv have to compete with AMD's next gen, ie Cayman?

Well if the GTX 580 was ready last year (actualy it was supposed to be GTX 380) AMD had the HD 5970 to compete and they still do.

[zerazax]

Not really one hell of a driver, but more like rewriting the driver to use the new architecture.

That's not how the software side of things gets developed, unless AMD is really that bad with their development. Again, these things don't get rewritten with just 3 weeks to go, when the product has been in development for over a year. This isn't some phone application...

[Maxforces]

Well if the GTX 580 was ready last year (actualy it was supposed to be GTX 380) AMD had the HD 5970 to compete and they still do.

And It`s multi-GPU card . In one place of game you've got 80% scaling that in the second place fell to 30% scaling + .
The gameplay is smoother on the GTX480 than on the HD5970, there is no FPS drops

[Heinz68]

If it perform's like xfire 6850's. I am pretty sure they would have released it already. Paper launch or not, it would have put a pretty strong damper on the gtx 580 launch in addition to raise the 6xxx namesake, to increase sales of the 68xx cards even further.



They are not going to double everything unless they want the card to consume 300 watts + be 500mm in size.

I think the 3-4 tesselation performance is bull. Synthetics show the greatest difference in tessellation performance. People rationalizing that Cayman could be 2-3 faster in synthetics and even faster in games by the vagueness in the rumors shows how misplaced their rationality is.

Synthetics are coded to specialize in showing maximum prowess of a certain aspects. Drivers are specifically coded for them to maximize performance, hence why Nvidia kept on bragging about unigine heaven.

We won't see faster tessellation in games compared to synthetics. Be rational people.

I hear for about half year how great the Nvidia cards are in tessellation and how the AMD cards sucks and now when the AMD improved the tessellation performance it's all about synthetics, what a waste AMD.

The fact is AMD had to improve the performance because Nvidia is getting more benchmark and now even game benchmarks. LINK

Anyway 3-4 faster tessellation was somebody post on some Russian forum and everybody knows that is baloney since the AMD slides show 2-3 faster.

I'm not going to comment on the rest of the post since it was already discussed in this thread, no need to repeat.

[Heinz68]

And It`s multi-GPU card . In one place of game you've got 80% scaling that in the second place fell to 30% scaling + .
The gameplay is smoother on the GTX480 than on the HD5970, there is no FPS drops

Sure it is much smoother at least untill Nvidia gets the dual GPU card, maybe you should tell all the people how Crossfire and SLI sucks, I bet most of the real users want agree with you.

[safan80]

And It`s multi-GPU card . In one place of game you've got 80% scaling that in the second place fell to 30% scaling + .
The gameplay is smoother on the GTX480 than on the HD5970, there is no FPS drops

The 5970 minimum frame rate is still on par with the 580GTX. The 5970 is still the best single slot card out there. The 580GTX is last year's news. For those of that have had the 5970 for a full year the 580GTX is too little too late. You can make a 5970 run at 5870 speeds no problem with a clock and voltage bump.

[railer]

Few more weeks and we will know if Cayman has a good bite
I could not wait to play all the new console ports that are out so I opt out for 460.

[cegras]

That's not how the software side of things gets developed, unless AMD is really that bad with their development. Again, these things don't get rewritten with just 3 weeks to go, when the product has been in development for over a year. This isn't some phone application...

Considering nvidia's supposed 20-30% gains over their driver releases, I'd say things aren't as perfect as you think, physical silicon being different from simulation being a key factor.

[zerazax]

Considering nvidia's supposed 20-30% gains over their driver releases, I'd say things aren't as perfect as you think, physical silicon being different from simulation being a key factor.

Yes.... OVER their driver releases. Meaning, this is often over many many months + working with developers to optimize their code

Not a three week project