http://www.nordichardware.com/news/71-graphics/41433-amd-radeon-hd-6000-roadmap-leaked.html
3rd slide which coincides with the other slide.
AMD: R9xx Speculation
- Thread starter Lukfi
- Start date
http://www.nordichardware.com/news/71-graphics/41433-amd-radeon-hd-6000-roadmap-leaked.html
3rd slide which coincides with the other slide.
DarthShader
Regular
http://www.fudzilla.com/graphics/item/20797-hd-5970-drops-to-€389
Just emptying stocks? Or a sign Cayman will take longer than expected?
Just emptying stocks? Or a sign Cayman will take longer than expected?
Xt > Pro in AMD terminology.
Where do they coincide? On the Performance slide, Cayman is located a little bit up GTX480/470...
It's hardly the same chart. But as you can see AMD positioned Cayman above GTX480.
caveman-jim
Regular
Hmm, must have missed where the Bart's cards have 5gbps memory...
look at where they put Barts, seems like that road map seems pretty much whats going to come up.
No it doesnt... Barts is faster than Cypress Pro and the chart doesnt show that?
But you might be right, judging by other I had arguments today xD
Last edited by a moderator:
Slide 23 (5th on that page) says that Cayman XT will have only 1 GB of memory. I thought we were expecting 2 GB?
leoneazzurro
Regular
Incorrect - the 6850 is quite faster than the 5770 and it has lower clock. The clock reason could be true only for the 6870 performing better than the 5850 in the average.
They have that kind of memory indeed:
http://ht4u.net/reviews/2010/amd_radeon_hd_6850_hd_6870_test/index11.php
They operate it at a lower frequency though.
That is a market segment chart...
Antilles/Hemlock = Enthusiast
Cayman/Cypress = Performance
Barts = Sweet Spot
Juniper = Mainstream
They are binned 5Gbps, clocks are lower due to TDP/performance.
Unknown Soldier
Veteran
I have a feeling that AMD were taken a back(again since it's happened again and again from NV40 and G80) with the GTX580 results and now they have to decide on clock speeds to try and be competitive.
*please note that this is my own opinion*
I think they are going to have to try and lower the prices vs the 580 to get 6950/70 to sell.
That or TSMC screwed them over again.
*please note that this is my own opinion*
I think they are going to have to try and lower the prices vs the 580 to get 6950/70 to sell.
That or TSMC screwed them over again.
Last edited by a moderator:
You surely mean 6950/6970, don't you? I don't think they were planning to go way above 400 Euros with 6970 anyway. So probably not even a need to adjust pre-launch?
Eh, what? It's not the doubled ROPs? It's not the doubled memory interface? It's not the 40% more ALUs/TMUs? It's not the improved tesselation unit?Razor1 said:
It's 384 4D-units vs. 320 5D-units. I doubt AMD would have bothered to make such a change if there wasn't any noticable increase in efficiency, so those 1536 surely won't perform like 1536 Cypress ALUs would.
Also, it should be mentioned that raw Shader-/Compute-power was/is not the weak spot of Cypress. 16xHQAF, Z Culling, Geometry/Tesselation and some other things are somewhat bottlenecking the shader units, not the other way round.
Anyway, I think AMD's switch to smaller 4D-units has the mid-term goal of doubling the ALUs per SIMD, my guess is that at 32nm Cayman would have been a 2560 ALU/80 TMU (20 SIMDs) chip, and Caymans 28nm successor could very well be Cayman with doubled ALU count (3072) + the usual minor architectural enhancements.
That's what I think, too.
iirc anandtech and others said it's actually because AMD removed some stuff from the memory controller that allowed higher clocks.
True, using the smaller/simpler MC saves PHY but I would also assume they are more power efficienct, though that is also due to running at slower speeds.
Also, according to anandtech, the 6850's have no problem hitting 4.6ghz effective, so the 4ghz reference speed must be for another reason.
Similar threads
