What do you expect for R650

[nutball]

how it will stack against G81 65nm

What's G81?

 

[trinibwoy]

D.) Depends by how much R650 will be killer for G80Ultra and how it will stack against G81 65nm

Is Y even eligible to be an X killer if Y is released 9-12 months after X? That's like saying G80 is an R520 killer. I'd be surprised if Nvidia's Q4 part isn't on the horizon shortly after R650 breaks cover.

 

[Shtal]

What's G81?

My guess refresh for G80, unless they skip and go strait for G90 which will not happen until sometime in 2008.

Is Y even eligible to be an X killer if Y is released 9-12 months after X? That's like saying G80 is an R520 killer. I'd be surprised if Nvidia's Q4 part isn't on the horizon shortly after R650 breaks cover.

I understand and I agree , but fudzilla says it is going to be killer even it is released 9-12 months after.

http://www.fudzilla.com/index.php?option=com_content&task=view&id=1042&Itemid=1
http://www.fudzilla.com/index.php?option=com_content&task=view&id=472&Itemid=34


[Edit: based from Fudzilla information - my thoughts for R650]
Just my guess for R650
R600 64 shader units to -----> R650 96 shader units (5x) in 5 blocks
R600 16 TMU's to -----> R650 64 TMU's
But for ROP's, I got a feeling ATI will still keep it at 16.
GPU clock probably be 852MHz-912MHz

 

[fellix]

All depends on the manufacturing technology limits, so again no escape from trade-offs this time:

1. Double the #TMU and quad-fold the texture sampling rate (eg. make it 1:1 as G80), and this will eat more cache, so wont come cheapish.
2. Double the ROPs and bump the shader core with another two 16xALU clusters (requires extending the thread dispatcher), as the increased # of colour sampling per cycle will put sizable pressure on the shaders, where the sample resolving is done. Again -- plenty of bandwidth available to burn, for this.
. . . . . . .
Leave as it is and skyrocket the clock-rate.

 

[Shtal]

All depends on the manufacturing technology limits, so again no escape from trade-offs this time:

1. Double the #TMU and quad-fold the texture sampling rate (eg. make it 1:1 as G80), and this will eat more cache, so wont come cheapish.
2. Double the ROPs and bump the shader core with another two 16xALU clusters (requires extending the thread dispatcher), as the increased # of colour sampling per cycle will put sizable pressure on the shaders, where the sample resolving is done. Again -- plenty of bandwidth available to burn, for this.
. . . . . . .
Leave as it is and skyrocket the clock-rate.

Is it hard to accept that history could repeats itself again; just like R520 had short life before it was replace with R580 3:1 ratio shader power.

 

[Shtal]

Following on from the HD2900 XT, AMD are working to get the 65nm R650 ready for a September launch.

Clock speeds could be around 1 GHZ on the core, and possibly 1GB of GDDR4. There is also a rumour that the amount of texture units will be increased.

Because of the die shrink, power consumption should be considerably lower. I think everyone is hoping R650 will be competitive, but apparently NVIDIA will having something new for us by then, too.

This won't just be a simple die shrink of an existing marchitecture as ATI plans to redesign some parts of the chip. The major architectural difference is that R650 will have more texture power as this was recognizes as one of the weakest points of R600. http://www.pureoverclock.com/story.php?id=1172.

September huh! - That is 4 months from now and about 10 months since G80GTX launch.

 

[NocturnDragon]

All depends by how much improvement Nvidia could make based on 65nm tech,

A.) If their will be no delays from Nvidia
B.) Depends if it is just a clock bump from G80Ultra
C.) Also well possibly ATI could response with extreme high clock R650+ or R680 GPU
D.) Depends by how much R650 will be killer for G80Ultra and how it will stack against G81 65nm

What is an R680?

 

[compres]

ROFL...

Yeah people pulling code names from their...

I specially like the random MHz...lol?

After reading the review here I think, as many other forum members here have stated, that AMD is aiming at something else along with graphics, be it gpgpu or fusion. I know they said they were not working together on this part but it just seems that way, just look at the amount of cache for a video chip.

There is a good chance they will spank nVidia in 6 months or so with their next gen, indeed. But I doubt they will with r6xx.

 

[Shtal]

What is an R680?

I got from trinibwoy:

Hmmmm just in time to start speculating about R680....http://forum.beyond3d.com/showthread.php?p=915263#post915263

You could check the link!

 

[trinibwoy]

I got from trinibwoy

Yaay I get credit for inventing R680!

 

[NocturnDragon]

Yaay I get credit for inventing R680!

If you read a list of the ATI codenames you'll realize why some were ending with 00 and 50 and some others with 20 and 80. So no I don't think a 680 will ever exist!

 

[Shtal]

Yaay I get credit for inventing R680!

Sorry....

No hard feelings.

 

[ninelven]

unless they skip and go strait for G90 which will not happen until sometime in 2008.

I thought G90 was supposed to be available around November

 

[Shtal]

I thought G90 was supposed to be available around November

You are correct, the reason I said that if Nvidia would do release G81 core, then G90 would be push back. Otherwise if Nvidia goes strait for G90 - then yes, it would be this year.


-OR-

In the other corner, we have NVIDIA probably working on a R650 competitor. For now, there are some rumors that the 65nm G90 may be the counterpart of R650, but NIVIDA did not care to reveal details on this aspect. http://news.softpedia.com/news/More-Details-on-ATI-039-s-Upcoming-R650-and-R700-GPUs-54971.shtml

Since Nvidia is working on R650 competitor, and if for some reason R650 be overkill or to powerful for G90 "I never said yes", then Nvidia has to push back into 2008 time frame to improve core and compete with R650.

 

[Geeforcer]

There will never be a "core improvement" in response to competitor's release - the most you can do is try pushing frequencies. Since the design takes years and the chips tape out at least 6-12 months before release, any such changes are impossible.

 

[Shtal]

There will never be a "core improvement" in response to competitor's release - the most you can do is try pushing frequencies. Since the design takes years and the chips tape out at least 6-12 months before release, any such changes are impossible.

Are you saying example like NV30 - clocked was pushed to 500MHz?

 

[Geeforcer]

Yes, if there ever was one, NV30 would be it.

 

[Shtal]

Yes, if there ever was one, NV30 would be it.

Just another quick question, how would you explain R520 ----> R580 situation?

And could Nvidia do same thing with G90 ---->Like R520-R580: if R650 turns out faster then G90.

 

[Geeforcer]

Just another quick question, how would you explain R520 ----> R580 situation?

There is not much to explain. It appears that ATI was committed to 3:1 ratio, and that R520 was meant to serve as the stepping stone to that. It got delayed, R580, which at that point must have been in development for a while did not and thus it was a very shot-lived stepping stone.

And could Nvidia do same thing with G90 ---->Like R520-R580: if R650 turns out faster then G90.

You mean delay it and then quickly replace it so it becomes just a footnote in the history of the industry? They could, but why in the world would they want to?

 

[Shtal]

You mean delay it and then quickly replace it so it becomes just a footnote in the history of the industry? They could, but why in the world would they want to?

I understand!

-------------------------------------

But why did ATI released R520 after then quickly replaced with R580 within (3 months), since R520 would have a short life anyway (what's the point). Same thing with R600 since it will be replaced with R650 in 4 months.