ATI Reveals Radeon X1900 Details Internally

[nitromullet]

http://www.anandtech.com/video/showdoc.aspx?i=2653

 

[peleejosh]

We all know how reliable the ATI folks are I will believe it when I see one in a store.

 

[secretanchitman]

X1900 will be launched in january/february, while the vendors will get them in stock in june/july, haha...

nah, if what ATI says is true about the X1900, then its sure to be one hell of a card.

 

[nitromullet]

Originally posted by: peleejosh
We all know how reliable the ATI folks are I will believe it when I see one in a store.

Agreed, and it's not exactly as if that article was riddled with details either.

 

[nts]

The R580 at the same clocks as the R520 is theoretically 3 times faster, with higher clocks that thing is going to be a monster.

With the amount of shader power this card will have it looks like they'll get ALU ops pretty much free, just waiting on the texture op's but the ring bus controller should handle those extremely efficiently.

Can't wait for this card to be released.

 

[fbrdphreak]

Well, considering ATI was supposed to launch another product in Jan and that got moved back I wouldn't hold my breath.

 

[KeepItRed]

Holy man...That thing's a monster. I feel bad I own a X1800XT. Maybe I should sell it and get X1900XT PE in Crossfire?

 

[allies]

Originally posted by: nts
The R580 at the same clocks as the R520 is theoretically 3 times faster, with higher clocks that thing is going to be a monster.

With the amount of shader power this card will have it looks like they'll get ALU ops pretty much free, just waiting on the texture op's but the ring bus controller should handle those extremely efficiently.

Can't wait for this card to be released.

This is incredibly false, pixel shader pipelines != pixel pipelines...

EDIT: I'm trying to find a link the explanation of GPU tech...

 

[tvdang7]

48 pixel shaders? does that mean it could be 38 pipes and 10 shaders? or does it mean 48 pipelines and soemthing shaders

 

[DeathReborn]

Originally posted by: tvdang7
48 pixel shaders? does that mean it could be 38 pipes and 10 shaders? or does it mean 48 pipelines and soemthing shaders

I think 16 pipes but all capable of processing 3 shaders at once.

 

[chilled]

Originally posted by: DeathReborn

Originally posted by: tvdang7
48 pixel shaders? does that mean it could be 38 pipes and 10 shaders? or does it mean 48 pipelines and soemthing shaders

I think 16 pipes but all capable of processing 3 shaders at once.

Bingo!

 

[mwmorph]

Originally posted by: tvdang7
48 pixel shaders? does that mean it could be 38 pipes and 10 shaders? or does it mean 48 pipelines and soemthing shaders

it means 48 programmable shaders which means the card can choose anywhere from 47pixel pipes and 1 vertex shader to 47 vertex shaders and 1 pixel shader pipe. the efficency of each pipe isnt as high as say 47 optimized pixel shaders but this way, you can use the gpu more efficently since all the pipes can be in use at the same time(nowadays sometimes, a vertex shader or two may be doing very light duty while pixel pipes struggle to play catchup or vice versa. this way, all 48pipes can operate close to max capacity. it wont be as fast as a 46picel pipe 24 vertex shader card at the same clocks shader for shader if every shader is optimally used, but in les than optimal situations, it can pick up the slack by using all the pipes to therotically improve perfromance. in other words, it's like a xbox360 video processor on steroids.

 

[nts]

Originally posted by: allies

Originally posted by: nts
The R580 at the same clocks as the R520 is theoretically 3 times faster, with higher clocks that thing is going to be a monster.

With the amount of shader power this card will have it looks like they'll get ALU ops pretty much free, just waiting on the texture op's but the ring bus controller should handle those extremely efficiently.

Can't wait for this card to be released.

This is incredibly false, pixel shader pipelines != pixel pipelines...

EDIT: I'm trying to find a link the explanation of GPU tech...

I am pretty confident that I know how it works

Sorry about the confusion, I didn't mean full pipes. Thats why I said that ALU operations should be almost free (48 processed in parallel and over 2 ALUs each) and the only limiting factor will be the TMUs (still at 16) and memory bw.

Anybody know the ratio of ALU:TMU operations in current games, has it hit the 3:1 point?

 

[TSS]

isnt exactly "internally" if we know about it hmm?

but yeah, it looks like its going to be a good card. though i'd rather wait for the benches of the thing. higher clocked looks nice too, 700/1800 stock? wonder how far that'll oc.

maybe with a little luck ati will suprise us all and do an actual hard launch for once (hey, a man has a right to dream!)

 

[nts]

Originally posted by: mwmorph

Originally posted by: tvdang7
48 pixel shaders? does that mean it could be 38 pipes and 10 shaders? or does it mean 48 pipelines and soemthing shaders

it means 48 programmable shaders which means the card can choose anywhere from 47pixel pipes and 1 vertex shader to 47 vertex shaders and 1 pixel shader pipe. the efficency of each pipe isnt as high as say 47 optimized pixel shaders but this way, you can use the gpu more efficently since all the pipes can be in use at the same time(nowadays sometimes, a vertex shader or two may be doing very light duty while pixel pipes struggle to play catchup or vice versa. this way, all 48pipes can operate close to max capacity. it wont be as fast as a 46picel pipe 24 vertex shader card at the same clocks shader for shader if every shader is optimally used, but in les than optimal situations, it can pick up the slack by using all the pipes to therotically improve perfromance. in other words, it's like a xbox360 video processor on steroids.

Its not there yet, the Vertex Shaders and Pixel Shaders aren't unified. Not until the R600.

The 48 shader processors is for the pixel pipes, vertex shaders should stay the same as the R520.

 

[Wentelteefje]

Let's just hope the X1700 is something worth looking at... The X1600 should have been great, but was such a letdown...

 

[nts]

Originally posted by: TSS
maybe with a little luck ati will suprise us all and do an actual hard launch for once (hey, a man has a right to dream!)

Considering the R520 delay didn't affect it and it's still on schedule and in production now, I would think there would be plenty available at launch.

 

[mwmorph]

Originally posted by: nts

Originally posted by: mwmorph

Originally posted by: tvdang7
48 pixel shaders? does that mean it could be 38 pipes and 10 shaders? or does it mean 48 pipelines and soemthing shaders

it means 48 programmable shaders which means the card can choose anywhere from 47pixel pipes and 1 vertex shader to 47 vertex shaders and 1 pixel shader pipe. the efficency of each pipe isnt as high as say 47 optimized pixel shaders but this way, you can use the gpu more efficently since all the pipes can be in use at the same time(nowadays sometimes, a vertex shader or two may be doing very light duty while pixel pipes struggle to play catchup or vice versa. this way, all 48pipes can operate close to max capacity. it wont be as fast as a 46picel pipe 24 vertex shader card at the same clocks shader for shader if every shader is optimally used, but in les than optimal situations, it can pick up the slack by using all the pipes to therotically improve perfromance. in other words, it's like a xbox360 video processor on steroids.

Its not there yet, the Vertex Shaders and Pixel Shaders aren't unified. Not until the R600.

The 48 shader processors is for the pixel pipes, vertex shaders should stay the same as the R520.

I couldve swore someone said that there wouild be unified r580.

 

[5150Joker]

I'll get excited when I see the card being produced in mass quantity.

 

[Matthias99]

Originally posted by: mwmorph
I couldve swore someone said that there wouild be unified r580.

There were some random rumors, but that was about it. R600 is supposed to have unified pixel/vertex shaders (since it is basically the PC version of the XBox360's GPU).

The (somewhat) more concrete rumors going around the last couple months match up with this design (16 pipes, 48 pixel shaders, probably similar vertex shader capability to R520). It should put up some VERY impressive numbers in 3DMark05 (which is pretty much just a crushing shader workload) and ShaderMark, but we'll have to see how it shakes out with today's real-world games.

 

[BFG10K]

I'm eager to see what the card can do as I'll probably skip this generation. Of course I hope it doesn't come with a monstrous cooling solution.

Go the Gigabyte passive cooling models!

 

[Ackmed]

Originally posted by: peleejosh
We all know how reliable the ATI folks are I will believe it when I see one in a store.

Because there are 512MB GTX's in any store? We all know how reliable the NV folks are...

 

[dunno99]

Hrm, seems like everyone is arguing over symmentics. I guess I'll take a shot at this, but don't flame me if I get it wrong....

Reading chapter 30 of GPU Gems 2 (available for free on nVidia's developer site...or Google for a direct link), it says that the NV45 (GF 6800) has exactly 16 pixel pipelines (and 6 vertex, but we don't care about that right now). However, the 16 pipelines are divided into units of four (which is also why all the NV4x cards have multiples of 4 pixel pipelines), and fragments from each primitive are processed in adjacent groups of 4 at a time (which would mean from the same primitive). What I think this implies is, at border conditions, the pipeline units are less than fully efficient...as in, each unit may process less than 4 pixels (I'm willing to assume that the hardware is smart enough to align the pixels such that it will result in one less "batch" per horizontal scan, if possible). So this means that it isn't really a fully 16 pipeline GPU (although, the performance penalty is probably minimal, and it can probably make up for it if derivatives are used, compared to other cards).

Furthermore, each pixel pipeline has two fp32 shader units (ALUs, I presume) in series, of which the first shader unit can have its result substituted by a texture fetch instead. Both shaders units process instructions in parallel per clock (assuming no hazards or dependencies). From the looks of it, both shader units are full ALUs. Since they probably both can do vec4, vec3 + scalar, or vec2 + vec2, this would mean, at most, 4 parallel instructions (two coissues per shader unit) are processed at each tick of the clock. This is why latency hiding is especially important, and I suppose which is why each texture fetch should be followed by either more texture fetches or non-dependent instructions. (Note: Because of the deep pipeline structure of these GPUs, branching is basically done via a brute force approach. I believe this is the reason why the NV4x GPUs can only perform 4 nested if/else statements, because by taking all branches, 4 nested branches would equate to 2^4 = 16 different paths...but I'm guessing here.)

On the other hand, I'm guessing that the X1900 will have 16 separate (although they might be able to work together) pipelines, each being able to process 3 separate fragments in parallel each. This seems to me to be like the fragment "units" above (I'm guessing the two companies are using the terminology a little differently). So each pipeline is a unit itself, and each unit is composed of 3 individual fragment processors (each being able to work on one fragment at a time). The 16 texture units would mean that each cycle, only 16 of these fragment processors will get to retrieve data from memory. Given that NV45 has 16 texture units and 16 pixel processors with two full ALUs each, that means a 16*2:16 = 2:1 ratio. On the other hand, ATi has 16 texture units and 48 pixel processors with either 1 or 1.5 ALUs each (I don't know which)...this would translate to a 3:1 or 4.5:1.

If anyone notices anything wrong, feel free to correct, not flame.

 

[ZeGermans]

how many "pixel shaders" does the R520 have in comparison? That's what I wanna know.

 

[dunno99]

16 pixel processors for the R520. (btw, pixel/fragment are used interchangably...since they pretty much only differ in whether they get written to the framebuffer or get rejected)