The R7xx Architecture Rumours & Speculation Thread

[Arun]

This thread is exclusively for the discussion of the R7xx architecture, and as such it will have higher quality standards and should be more technical than SKU-specific threads.

Rumoured Data Points
- Evolutionary step of the R6xx architecture.
- Not the first Radeon family with DX10.1 support (that's RV670).
- First chip will be on 55nm according to AMD, same process as RV670?
- Seems designed around multi-chip solutions and scalability.
- Will support both HDMI and DisplayPort natively.
- Slated for Q2 2008 according to Fudzilla.
- Or "mid-2008" according to others...

Noteworthy Internet Rumours
"R700 will support both HDMI and Displayport" [The Inquirer]
"R700 codename is Wekiva" [Fudzilla]
"Big GPUs are set to die" [The Inquirer]

Thread Discussion Starting Points
- Does all of this rumoured information seem reliable to you? What, if anything, sounds fishy?
- Do you believe the ratios of the different units have been changed? How so? How would it affect things?
- What do you think will be the differentiation factor since the process and API are similar? Higher efficiency? If so, how?

Last Updated on August 8th

 

[Sound_Card]

I think what ever ratio RV670 is, that is what R700 will take after. I'm thinking a more texture friendly ratio too.


I also would imagine ATi taking a more elegant approch to the multi chip solution. The chips would be on the same package, rather than on seperate packages.

 

[psurge]

OK this thread is off to a slow start and clearly needs some nonsense injection .

2 chips per package with a specialized and extremely fast in-package interconnect between them - the ring bus now spans across 2 dies. On top of this, each die will have 2 HT3.0 links, allowing 4 dual-die packages to be fully connected and operate in SMP fashion.

Marketing name? OctaFire.
Source of rumour? Currently in very close contact with my chair

 

[_xxx_]

Q2 2008? My own personal opinion: LMAO

I really don't even believe we'll ever see R700 as a high-end chip, but that's just me. I would certainly like to be proven wrong.

 

[Shtal]

Please no delays and please be more competitive!

I still think R700 will be high-end part.

 

[vertex_shader]

Please no delays and please be more competitive!

I still think R700 will be high-end part.

Already visible 2008Q2 or MID release months behind NV schedule,this GPU need almost make the impossible possible, i not see how this can be done without NV made a mistake, but when fail than AMD can sell for 400$.

I'm not optimistic anymore about AMD GPU products, so when they make something great than i can be surprised

Hopefully we not see here in B3D the "last last last R7xx Architecture Rumours & Speculation Thread" , and i hope that too noone quote what Henri Richard say about R700 :smile:

 

[The_Wolf_Who_Cried_Boy]

Should we assume software antialaising resolve will remain, I'm sure NVidia hopes so...

 

[Arnold Beckenbauer]

This thread is exclusively for the discussion of the R7xx architecture, and as such it will have higher quality standards and should be more technical than SKU-specific threads.

Rumoured Data Points
- Evolutionary step of the R6xx architecture.

Stil R400 based?

 

[Arun]

Should we assume software antialaising resolve will remain, I'm sure NVidia hopes so...

There's nothing magical about software downsampling that makes it horribly slow. In fact, I'm not 100% sure that it's slow at all on R600 (although presumably it's still slower than dedicated hardware!) - but if it is, which it might be, then it's more of an implementation problem than an inherent impossibility with the approach.

Stil R400 based?

It's probably more fair to say 'still R500' based than 'still R400 based' IMO - it's very likely that R400 was similar, but it was hardly identical to R500.

 

[Jawed]

There's nothing magical about software downsampling that makes it horribly slow. In fact, I'm not 100% sure that it's slow at all on R600 (although presumably it's still slower than dedicated hardware!)

It's not remotely slow as far as I can tell - having discussed this at length and been wildly pessmistic about this - unless framerates well over 100Hz are important. Prey scales with core clock, about 14%, over X1950XTX (seems reasonable to presume this game has been blessed with a good dose of pixie-dust by the driver team):

http://www.xbitlabs.com/articles/video/display/xfx-gf8800ultra-extreme_10.html#sect1

All the performance blame can be pointed at R600's AA sample generation rate.

Does G80 still resolve AA on each monitor refresh like earlier NVidia GPUs do - the frontbuffer being presented on the monitor contains the raw sample data, doesn't it?

Jawed

 

[Geo]

There's a pretty big semi-official hint in Rys recent Develop day piece about R700. . .

 

[Geo]

I really don't even believe we'll ever see R700 as a high-end chip, but that's just me. I would certainly like to be proven wrong.

What do you mean by this? You mean in the sense of one chip, or you mean in the sense of however it's configured it won't be performance competitive with the best the other fellows can offer as a board-level offering?

 

[vertex_shader]

There's a pretty big semi-official hint in Rys recent Develop day piece about R700. . .

- Not the first Radeon family with DX10.1 support (that's RV670).
I think this was a hint, but this not a hint about r700, so its not a hint about rv670?

 

[Geo]

I'm not going to point it out to you, but if you read it it's there and clear. Really, Arun should have included it in the first post.

 

[nAo]

shader AA resolve is here to stay, it's certainly the way to go for the foreasable future.
I don't know if G8x has a performance problem with AA resolves but if this is the case I hope they will fix it on their DX10.1 GPUs

 

[Sound_Card]

ATi plans on releasing a new set of mainstream cards based on the R7xx arc. by Q4 08. RV730 and RV710.

http://www.fudzilla.com/index.php?option=com_content&task=view&id=2330&Itemid=34

So we have RV635 and RV620 refreshes comming in Feburary. This should be around the same time as RV670 and R670. A nice little winter line up here for ATi. When R700 comes out in Q2, I wonder if they would just simply push R670 and RV670 down and use those as their performance/high end while R700 will be the Ultra high end.

RV635/RV620 on a 55nm node, equiped with DX10.1 and displayport seems likely.

Now back on R700.

I don't think it's impossible to say thet R700 could be 45nm. With 45nm production expected this spetember for TSMC, and R700 to come in say... May or June next year. Possibly with RV730 and RV710 on a 40nm process?

http://www.tsmc.com/tsmcdotcom/PRLi...il&language=E&newsid=2120&newsdate=2007/04/09

 

[Malo]

I'm not going to point it out to you, but if you read it it's there and clear. Really, Arun should have included it in the first post.

You mean this?

At this point it's one of those quietly whispered secrets that AMD's next generation graphics architecture is going to centre around a dual-chip offering for the high-end, with R700 likely comprising two discrete GPU dice.

 

[Arun]

I don't think it's impossible to say thet R700 could be 45nm. With 45nm production expected this spetember for TSMC, and R700 to come in say... May or June next year. Possibly with RV730 and RV710 on a 40nm process?

Take the past into consideration, not just the future...
http://www.notforidiots.com/TSMC.gif

 

[Arnold Beckenbauer]

...

It's probably more fair to say 'still R500' based than 'still R400 based' IMO - it's very likely that R400 was similar, but it was hardly identical to R500.

I thought, Xenos/C1 is the reincarnated R400 with some changes. The R600 is R400's evolution with some R520's ideas. I don't think, it's a coincidence, that all three chips have a tesselation unit.

I expect two thing: smaller threads and wider architecture with smaller SIMDs.

I guess, it's time for the famous ArtX legend: R200 sucked, R400 sucked, R600 sucks. ArtX' R300 was wow, R520/R580 were wow, R700 will be wow!

 

[Geo]

Geek theory of Star Trek movies as applied to graphics?