ATI Radeon HD 4000 Series discussion

[Seraphiel]

Thing is though, I think this is designed to be an mcm, not just two gpus on a single board, I don't think there will need to be an external bridge chip. Think about it, if they have a shared ringbus, then that means this is more than just two rv770s in a dual die mcm, there has to be some modifications to basically meld the chips to be detected as one thus eliminating the need for a bridge chip. That's probably the whole point of doing this rather than the 3870x2 design.

Man I just wish we'd finally get some legit info already, or if we have, that ati would confirm it

I totally agree

It really sounds promising and I rather be an optimist than a pessimist. This could be the advantage NVIDIA can't break until after its nextgen (GT-200(?)). I hope so, because NVIDIA needs a little humility... This said as a user that has spend >4500$ on NVIDIA product in the last 18 months.

Humility benefits everyone.

[w0mbat]

http://www.gpureview.com/show_cards....=566&card2=564

[JamesBong420]

http://www.gpureview.com/show_cards....=566&card2=564

not even a month and we do not have a single pic of these cards. i just want some pr0n thats all....

[generics_user]

http://www.gpureview.com/show_cards....=566&card2=564

doubled texture performance, but according to these specs they didnt add any stream processors

edit: RV770

card specs and chip specs dont lline up...

[AliG]

http://www.gpureview.com/show_cards....=566&card2=564

that doesn't have to correct though, that just seems to be exactly what Theo posted, which Andreas confirmed was wrong. the +200 mhz over the core clock for shader gives it away, not to mention if the card will launch on the 31st we would have at least heard something by ati by now, but that doesn't have to hold true with the way both companies have been

[Kai Robinson]

doubled texture performance, but according to these specs they didnt add any stream processors

edit: RV770

card specs and chip specs dont lline up...

Says 480SP's here....maybe they changed it?

[AliG]

those specs are wrong, they quoted Theo, which Andreas (among others) have confirmed was wrong

[Seraphiel]

So what are the certain specs (if any)

To hell with clock speeds for now, what is the specs of the architecture? I have read everything from doubled clocked SP, asyncronious (?) SP clocks, more SP (480), many more SP (800), simpler SP (like NVIDIA's 2+1, and not current 4+1) and X, and Y, and the most certainly Z.... and a combination of some of them.

If this is about to be released in a few weeks, then I would expect more certainty about specs and improvements.

I really hope ATI has a 8800 Ultra killer on their hand that is single GPU socket, unless they change most standards such as we know them.

[NH|Delph1]

that doesn't have to correct though, that just seems to be exactly what Theo posted, which Andreas confirmed was wrong. the +200 mhz over the core clock for shader gives it away, not to mention if the card will launch on the 31st we would have at least heard something by ati by now, but that doesn't have to hold true with the way both companies have been

Actually I said the +1GHz GPU frequency was wrong. It says 850MHz core here, and 1050MHz shader clock. That's not unreasonable.

//Andreas

[AliG]

I thought you said the clock differential was theo's wishful thinking?



Eitherway, I'd personally like to see those shaders running at 2x core speed instead of just 200mhz higher, that's what they need to do to spank g92 imo

[Mungri]

I dont think we will see 1000+ Mhz on GPU's until they shrink to 45 nm tbh.

My birthday is in 9 days and I wanted a 4870

[AliG]

as for an IOU?

[madcho]

I dont think we will see 1000+ Mhz on GPU's until they shrink to 45 nm tbh.

My birthday is in 9 days and I wanted a 4870

do your birsthday when you can buy one then

[NH|Delph1]

I thought you said the clock differential was theo's wishful thinking?

Nono, just the +1GHz GPU frequency, but nothing is set in stone.

//Andreas

[AliG]

Nono, just the +1GHz GPU frequency, but nothing is set in stone.

//Andreas

hmm, I wonder what ati will be doing about the shader clock issue the r600. To be honest I'd rather see the old 320 shaders with something like a 1:2 alu:text ratio and the shaders running at least 2x core speed rather than 480 shaders. Of course many can argue that's copying g80's design, and of course it is, you implement what works into your own design, why else did intel bring 64 bit computing to their cpus?


That, and don't forget, no matter how shaders the rv770 has, unless if the game can make use of them, they're just about useless, the 9600gt proved this by being very competitive with the 3870 with only 1/5th the shaders

[LordEC911]

hmm, I wonder what ati will be doing about the shader clock issue the r600. To be honest I'd rather see the old 320 shaders with something like a 1:2 alu:text ratio and the shaders running at least 2x core speed rather than 480 shaders. Of course many can argue that's copying g80's design, and of course it is, you implement what works into your own design, why else did intel bring 64 bit computing to their cpus?


That, and don't forget, no matter how shaders the rv770 has, unless if the game can make use of them, they're just about useless, the 9600gt proved this by being very competitive with the 3870 with only 1/5th the shaders

Ok... so I am confused by your post...

First you say that ATi needs an independent shader domain.
The you proceed to talk about how "inefficient" ATi's shaders are by comparing across a different architecture?

My main question is why does everyone think they need to implement an independent shader domain when R600/RV670 has shown that the "bottleneck" is elsewhere. An independent shader domain would not only hinder the architecture IMO, implementing it now would basically null the point of RV770.

[whocaresbg]

Computex 2008 AMD won't display RV770 at Computex :

http://my.ocworkbench.com/bbs/showth...981#post430981

.

[roadie]

Don't the shaders do AA? Therefore, the more shaders the better!

I'm not a huge fan of the poor AA performance on my 3870.

[Origin_Unknown]

hype hype hype...

[Scimitar]

Computex 2008 AMD won't display RV770 at Computex :

http://my.ocworkbench.com/bbs/showth...981#post430981

.

That's weird, I thought the cards were going to be released at the end of May. Now they won't even have them ready to show the first week of June? Maybe this is going to be more like a July launch.

[azza21]

R600 launch all over again, uh ohhh

[Calmatory]

Are they hiding something (bad) like with R600?

Another flop?

[AliG]

Don't the shaders do AA? Therefore, the more shaders the better!

I'm not a huge fan of the poor AA performance on my 3870.

Turn off AF, you should see a nice boost. I believe its been agreed on that AA wasn't the big issue, AF was due to the very small texture power

Ok... so I am confused by your post...

First you say that ATi needs an independent shader domain.
The you proceed to talk about how "inefficient" ATi's shaders are by comparing across a different architecture?

My main question is why does everyone think they need to implement an independent shader domain when R600/RV670 has shown that the "bottleneck" is elsewhere. An independent shader domain would not only hinder the architecture IMO, implementing it now would basically null the point of RV770.

Why would it reduce performance? I mean look at how well the 96gt performs, and it only has 64 shaders. I pretty sure unless games are coded to make use of all those shaders, no matter how good the drivers are we probably won't see as much as a boost as we'd like.

Are they hiding something (bad) like with R600?

Another flop?

Nah I highly doubt it, more likely than not gddr5 is the reason for the delay, besides fuad has posted I believe 3 times now that the rv770 will see the return of the r600's ringbus so that means there's a possibility of a gddr3 4870 with a 512 bit memoy bus. I generally don't trust the really shaky things fuad says, but when he posts the samething multiple times it generally turns out to be right

[madcho]

4870 will be a boost about 30 to 50% over 3870

[Calmatory]

4870 will be a boost about 30 to 50% over 3870

Quite likely. Even though the rumours float around and the speculations vary from theoretical ~6+% boost to +100% boost, in practise this is going to be somewhere near 30-50 % average.