AMD Cayman info (or rumor)

[informal]

Cayman is officially Chuck Norris approved design

[kuroikenshi]

Cayman is officially Chuck Norris approved design

now that's gotta hurt!

[FU1G0R]

hope this is kinda new info XD

http://translate.google.com/translat...zona,22055.htm


Cayman XT - 1920SP, 900Mhz core, 6GHz GDDR5, 3.5 TFLOPS.
Cayman PRO - 1536 SP, 800Mhz core, 5Ghz GDDR5, 2,5 TFLOPs. Cayman Pro is supposed to have - 1536SP, 800Mhz core, 5GHz GDDR5, 2.5 TFLOPS.

who nows...

until dec 13

[informal]

Complete slide deck(photographed):
http://www.computerbase.de/forum/sho...9&postcount=92

[generics_user]

looks like a significantly bigger change than expected, it looks like a real next gen, not just like a minor refresh like 68xx and geforce 5xx

PS: i hope chucknorris smacks informal because his chuck norris posts are annoying

[spursindonesia]

So, from the chio's mArch diagram, no decoupled TMUs means a 120 TMUs chip ??

[-=DVS=-]

So, from the chio's mArch diagram, no decoupled TMUs means a 120 TMUs chip ??

I count 56 only , assuming those yellow squares of fours represent texture units, for comparison 5870 has 80 of those little squares.

[zalbard]

Complete slide deck(photographed):
http://www.computerbase.de/forum/sho...9&postcount=92

It looks so f-ing awesome!
Compare this to the excuse of the chip GTX580 is... We fixed transistors... The power consumption is not as bad anymore... Etc etc...

[geo]

price wars incoming !!!!! i'm assuming it'll also have UVD3.0,EyeSpeed etc not mentioned in the slides

[Lanek]

I count 56 only , assuming those yellow squares of fours represent texture units, for comparison 5870 has 80 of those little squares.

Not all the SIMD are shown on the 6970 diagram ... ( the little black point show there's more, they don't wanted to show the real numbers . )

4 TU/ 20 SIMD = 80 TMU ( HD5870 ).. all will depend now of the number of SIMD, rumored at 30 ... 30x4 = 120TMU.

[Chumbucket843]

Not all the SMID are shown on the 6970 diagram ... ( the little black point show there's more and don't wanted to show the real numbers of SMID. )

4 TU/ 20 SMID = 80 TMU ( HD5870 ).. all will depend now of the number of SMID, rumored at 30 ... 30x4 = 120TMU.

just to end the confusion, that's a diagram of barts.

[spursindonesia]

I count 56 only , assuming those yellow squares of fours represent texture units, for comparison 5870 has 80 of those little squares.

This :

just to end the confusion, that's a diagram of barts.

1920 4-VLIW SPs, 120 TMUs, 30 SIMDs, 2 tesselators & setup engines, 32 beefed up ROPs, better GPGPU functionality, fast memory controller, if the chip is still under 400 mm^2, that would be a marvelous engineering feat from AMD graphic division engineers.

[Chrono Detector]

Needs a higher memory bus and more bandwidth, other than that the specs look good.

[FischOderAal]

Here: ALL DIAS

Thx a lot! Cheers!

Complete slide deck(photographed):
http://www.computerbase.de/forum/sho...9&postcount=92

Taken down by CB.

[Oliverda]

Needs a higher memory bus and more bandwidth, other than that the specs look good.

192GB/sec won't be enough?

[ubuntu83]

[Tim]

Hmmm I don't like what I see with the "Tesselation done right / adaptive tesselation", I'm afraid this is a whole new level of driver optimizations and it's going to be hard to see the different level of IQ but the performance increase or decrease could be very great indeed. I see some more PR fighting going on from nVidia and AMD, telling reviewers their level of tesselation is best.

[FischOderAal]

Hmmm I don't like what I see with the "Tesselation done right / adaptive tesselation", I'm afraid this is a whole new level of driver optimizations and it's going to be hard to see the different level of IQ but the performance increase or decrease could be very great indeed. I see some more PR fighting going on from nVidia and AMD, telling reviewers their level of tesselation is best.

Exactly what I was thinking.

[Manicdan]

well my opinion about tessellation is that its suppose to be built around adapting based on distance. you dont need a billion polygons from every character a mile away, but as they get closer you want to see them (or if your scoped in on them). it should be a stuipdily simple setting to determine how much is needed and it simply scales from there. so i would say i want normal, up close is 10 million, at far is 1 million. or i go with heavy and up close is 20 million, and far is 2 million. the levels between the close and far might not have perfect steppings, but this may be more of an issue with the game designers rather than the drivers. but like AMD is trying to say with their 16 pixels per poly hopefully has this handled. so a huge mountain which always takes half your screen will still have the same polygon intensity as a character which can take up half or near none of the screen.

a hint of slightly less IQ with 2x the framerate sounds like such a horrible trade off, lol. for people who want the very best IQ, get mad at the console ports and not the optimization of the gpus. since the guys who want the best IQ probably also want ray tracing, yet somehow have been ok with rasterization for how many decades now?

[bamtan2]

"Under NDA until November 22"

ARGGGGGGGGGGGGGGGGGHHHHHHHHHHHH

[bamtan2]

Give it to me

[Sn0wm@n]

Hmmm I don't like what I see with the "Tesselation done right / adaptive tesselation", I'm afraid this is a whole new level of driver optimizations and it's going to be hard to see the different level of IQ but the performance increase or decrease could be very great indeed. I see some more PR fighting going on from nVidia and AMD, telling reviewers their level of tesselation is best.

and is there really a best method really ???


its the same in the fast food industry and people dont complain too much

[Tim]

and is there really a best method really ???


its the same in the fast food industry and people dont complain too much

Maybe, maybe not. But performance difference between cards is usually reasonably small, but with some smart adaptive tesselation, the two cards reviewed could have a much bigger difference, hence one will come out on top and could therefore have the highest sales. But like you said, which would be best?

I really don't see how any of this applies to fastfood by the way.

[informal]

looks like a significantly bigger change than expected, it looks like a real next gen, not just like a minor refresh like 68xx and geforce 5xx

PS: i hope chucknorris smacks informal because his chuck norris posts are annoying

Come on ,give me a break,I rarely use Mr Norris in posts and it was a joke

"Under NDA until November 22"

ARGGGGGGGGGGGGGGGGGHHHHHHHHHHHH

I suppose this is why we see the slides in the first place,since it's 22nd today .

[Heinz68]

Complete slide deck(photographed):
http://www.computerbase.de/forum/sho...9&postcount=92

Thanks for the link. I like what I see and it's definitely not HD 5*** series refresh as some of the Nvidia schill and fanboys predicted.

Hope AMD manage to get some Cayman cards on the market before Christmas and for me I hope to see Antilles early in Q1 2011.