AMD: R9xx Speculation

Doom and gloom for the red team, quoting from the fudzilla piece http://news.softpedia.com/news/AMD-...t-Faster-Than-GTX-480-Rumors-Say-171298.shtml :runaway:

The problem is that sites like Fudzilla don't care if the slide is fake or not. They are happy to report it and then report the opposite a day later because it garners them hits, and something to post about. When you're a rumour site like that, you just post rumours, they don't need to be true or not, and everyone picks it up and the fake snowballs out of control. Sites like Fudz are not about reporting the truth, they are about getting advertising hits, much as TV stations are not about producing good quality programming, but about selling advertising in the commercial breaks. So to them, an obvious fake is just as useful for a posting an article as is a lie or the truth. I wouldn't be surprised if Fudz knew B3D had immediately revealed the slide as a fake before they even posted their story. Look at Softpedia. They didn't take their article down, just left it up with an update saying the slide is a fake. They even managed to turn the reporting of them being taken in by a fake and not checking their facts as more content so they could get more hits!
 
IMG_3199-R3422.jpg

http://forums.guru3d.com/showpost.php?p=3801534&postcount=59
 
The most obvious of answers would be: „Because AMD wants it so and it is in their power to decide.”
We'll see.
 
link




enough with the propaganda in favor of nvidia...ENOUGH...!!!...blasphemy rumors all the time...

hd6970 will be equal or faster on all aspects in comparison with gtx580...keep that in mind...did you forget that official good working drivers are the 10.12.? f*ck the leaked benchs, are ALL fake with "ghost" drivers...i will be laughing until the morning when i'll see the benchmark results...

the difference between 6950 and 6970 is 20%(2gb version, dont mention the 1gb version)...now are you sure, that gtx 580 and gtx570...are exactly a match to them...???...keep that in mind... and stop burying hd69xx as nvidia and fudzilla want to make people buy sooner than expected more gtx5xx...hahahaha...it's nonsense...!!!


A qoute from someone who actually has the cards.
 
How long it's been since Fuad got something right in regard to GPU?
I removed Fudzilla from my bookmarks a while ago, lot of people should do the same.

More on topic the 15th can't some soon enough. The sooner the cayamns are launched the sooner we will more in regard to Kurt :)
 
I'm glad I will have off work from 21st Dec onwards :smile:.

Cayman for me might or might not be faster than GTX580. It will be different architecture to Cypress and that alone is enough to make me happily test it, analyse data, spent endless hours tweaking things, checking new AA modes, AF settings, etc. In other words - Merry Christmas for me!

More on topic:
Doesn't heard anything about SidePort lately, will it be reintroduced with Cayman or it is a dead project? Also I would like to know more about Off Chip Buffer for Tesselation but that probably will need to wait till launch ...
 
It will be different architecture to Cypress
It might be the biggest change since R600. I think it's at least equivalent in magnitude to RV770's change.

Obviously the ALUs are completely new. I suspect the texturing is quite different too.

I have mixed thoughts about the possibility that the L2<->L1 architecture is revamped: have my fingers crossed but actually think it'll still fall short of Fermi.

ROPs are clearly getting an upgrade and I have a sneaky suspicion that the changes will have a gratifying effect on games...

I still guess that Cayman is ~70% faster than Cypress in games, for what it's worth :p

and that alone is enough to make me happily test it, analyse data, spent endless hours tweaking things, checking new AA modes, AF settings, etc. In other words - Merry Christmas for me!
I'm really hoping that a new version of Stream Kernel Analyzer or GPU Shader Analyzer will appear with Cayman support - making it easy to investigate the performance of the revised ALUs.

More on topic:
Doesn't heard anything about SidePort lately, will it be reintroduced with Cayman or it is a dead project?
It had better return with a vengeance.

Also I would like to know more about Off Chip Buffer for Tesselation but that probably will need to wait till launch ...
Plenty of speculation on that topic if you care to rummage. I wonder how Mintmaster's tessellation experiments are coming along :p
 
I have mixed thoughts about the possibility that the L2<->L1 architecture is revamped: have my fingers crossed but actually think it'll still fall short of Fermi.


I still guess that Cayman is ~70% faster than Cypress in games, for what it's worth :p

I think there's little motivation for AMD to massively invest in the L1/L2 structure because of their (comparatively) massive register files. For conventional workloads this approach seems to work just fine, as long as you have more math to hide memory subsystem accesses. Fermi on the contrary needs it's elaborate memory subsystem partly to compensate it's smaller register file.

WRT to the 70%: Do you mean HD 6970 vs. HD 5830? Or fully blown (and clocked) versions respectively?
 
Back
Top