Will the AMD RV790 Architecture be Something Different?

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
The crew at PC Perspective have an interesting AMD RV790 architecture editorial posted today that is definitely worth checking out.

Given AMD’s renewed focus after the R600 debacle, this speculation does not seem that far out of place of what could actually happen. With my luck AMD will release a 1200 steam unit chip without any of the GPGPU or efficiency improvements, and it will simply be clocked faster as well. While it would still make for a fast card, it would be far less interesting than what I have described above.
 
Well.. it certainly will be nice if ATI brings another card out that stomps its previous one.

I went from a 9600XT to a 50% overclocked 7900GS( I heavily modified it) to a 3870, to a 4850. Huge, huge differences every time.

Course I won't be buying one for a while.. no need to with my current card.

Then again.. if GPGPU improves a lot.. I just might have to pick one up.
 
There is as much confusion on the R790 as there was on the R770 chips, we have no idea what it will really be.
 
There was an interview with the guys that made the RV770 and they said it was a huge change and they wont be able to do that every time.
 
If you want to see amd's design philosophy all you have to do is look at the 6 core server chip instead of going back to the drawing board for Istanbul amd used what it had already created as the base essentially it is nothing more than a Deneb core that has been heavily modified. Now take this already established design philosophy of use what we have and modify it now apply that to ati gpu's
 
If you want to see amd's design philosophy all you have to do is look at the 6 core server chip instead of going back to the drawing board for Istanbul amd used what it had already created as the base essentially it is nothing more than a Deneb core that has been heavily modified. Now take this already established design philosophy of use what we have and modify it now apply that to ati gpu's

I thought we had established that the Deneb core on a 45nm process was relatively ok? I mean, it loses to Core i7 but only narrowly to Core 2, and Core 2 is still going strong.

Basically...the Phenom II could be the 3870 to Intel's 8800GT...get me?
 
If a company can fix a broken unit to get twice the performance out of each transistor, they should be able to do the same every year? A better question is if nVidia will be able to copy this trick (of course, this leaves the question as to why PhenonII couldn't copy c2d's magic).

A more interesting question (to AMD directly) would be what plans are available for the x2 cards. Multiprocessor (multiple sockets) opterons share memory and pass the data between themselves. It would be interesting (through possibly too expensive) to stick this type of thing on each processor (GPU memory interfaces tend to be widely more expensive/state of the art than server memory buses).
 
I believe this will be more or less Ati's tick as the 4870/4850's were their tock.
 
I thought we had established that the Deneb core on a 45nm process was relatively ok? I mean, it loses to Core i7 but only narrowly to Core 2, and Core 2 is still going strong.

Basically...the Phenom II could be the 3870 to Intel's 8800GT...get me?

you got completely lost i am talking about the design philosophy that is in use currently with amd's server cpu chips. if what josh says is true amd is taking that philosophy and applying it to its gpu business.
 
Well everyone thought the r700 series was going to have like 500 sp's compared to the 320 in the r600 And BAM!! they released 800 sp. So anything could happen.
 
If a company can fix a broken unit to get twice the performance out of each transistor, they should be able to do the same every year? A better question is if nVidia will be able to copy this trick (of course, this leaves the question as to why PhenonII couldn't copy c2d's magic).

A more interesting question (to AMD directly) would be what plans are available for the x2 cards. Multiprocessor (multiple sockets) opterons share memory and pass the data between themselves. It would be interesting (through possibly too expensive) to stick this type of thing on each processor (GPU memory interfaces tend to be widely more expensive/state of the art than server memory buses).

i disagree with your statements completely but i have prepared some questions so you can clarify

what was broken?

What transistor trick?

were do you base your statement on the pII not being able to copy the c2d magic?

what is this c2d magic?

what is magic?

How does multiprocessor systems come into this debate

Are you saying that amd toss a r790 gpu on a opteron?
 
Those speculated specification with an increase to SPs seem false imo. It would be nice if they would get the die shrink done, but that seems also wrong because of the high leakage currently. A small bump in performance could be maybe seen, because of small optimizations and binning. This is also speculation, of course.
 
Back
Top