• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

ATI r600

albanian

n00b
Joined
Aug 31, 2006
Messages
13
This was taken from my friends german magazine of ati r600 and just translated it for me, idk if its worth it

Price: $449.99

1. Unified shader architecture.

2. 64 pixel/vertex shaders. Since ATI is incorporating an Unified shader architecture, these 64 shaders can be either all pixel, all vertex or a combination of both. In contrast, the R580 based X1900XTX had 48 pixel shaders.

3. 80/65nm fabrication process, though ATI wouldn’t elaborate on the exact split as well as why they weren’t sticking to just a single fabrication process.

4. There was also some passing mention of GDDR4. However, we could not confirm either way whether the R600 would come with GDDR4 onboard. Later variants of the architecture will definitely the updated DRAM standard.

5. As for the clock speed, ATI said they weren’t really too concerned about raising the clock speed to particularly high levels, so we suspect the core clock may not be much higher than the current X1900s.

6. It will obviously be Vista ready and (ATI claims) has been built from the ground up to support DX10.

7. The R600 will also be the first practical implementation of ATI’s GPU concept. This is something we would be very interested in seeing because if this works as well as ATI claims, then apart from cutting down CPU load, it might put certain PhysX processor manufacturers out of business, simply because ATI cards would not need an additional card to do necessary computations for Physics. The onboard GPU will take care of it.

8. The launch of the R600 is expected to be somewhere around october 2006
 
"oh my god, it's full of pipes"

the R600, a graphic pipeline Odyssey.. ;)

looks pretty off the hook to me. :)
 
64 pipes but how are they going to feed all that bandwidth? GDDR4 isn't enough...

we need 512bit memory controllers...

It's so prevalent in 7900GTX... All that fillrate and only keeps up with 1900XTX
 
Marvelous said:
64 pipes but how are they going to feed all that bandwidth? GDDR4 isn't enough...

we need 512bit memory controllers...

Internally the current ring bus on X1000 series is 512-bit.

In a Unified Shader Architecture we really need to stop thinking of them as "pipes".

They are simply shader units or processors that will be capable of either running pixel shading or vertex shading processing.
 
Brent_Justice said:
Internally the current ring bus on X1000 series is 512-bit.

In a Unified Shader Architecture we really need to stop thinking of them as "pipes".

They are simply shader units or processors that will be capable of either running pixel shading or vertex shading processing.

So it's just going to be shader units not actual pipes? What good will that do in real games aside from effects?

edit:

I just read some article about U-S-A... Something like eliminating bandwidth problems for pixel and vertex shaders..
 
Marvelous said:
So it's just going to be shader units not actual pipes? What good will that do in real games aside from effects?

Because with unified architecture you are not limited by pipelines, if you have 64 shader units they can do whatever is needed and not what they can. EG. if a frame has more pixel shading needed and less vertex then the shader units can do that, and vise versa. If a frame hypothetically had no vertex shading and all pixel shading then all 64 units would be doing pixel shading which is really fast.
 
I really don't see anything new in that list that we haven't known or suspected for the past month or two. How they couldn't confirm GDDR4 is beyond me. If we've already got cards hitting the market with GDDR4 it makes sense that a card down the road would likely use it as well.

Internally the current ring bus on X1000 series is 512-bit.

In a Unified Shader Architecture we really need to stop thinking of them as "pipes".

They are simply shader units or processors that will be capable of either running pixel shading or vertex shading processing.

Don't forget geometry shaders as well. Need to add those in as well now.

Maybe we should just start saying 'x' full-processors(pixel) with 'y' sub-processors(vertex/geometry). It seems like vertex/geometry units are a subset of the capabilities of pixel units so it might make more sense to some.

64 pipes but how are they going to feed all that bandwidth? GDDR4 isn't enough...

we need 512bit memory controllers...

It's so prevalent in 7900GTX... All that fillrate and only keeps up with 1900XTX

Also keep in mind they have to run the vertex shaders as well so it's not like you'll see 64 pipes doing nothing but pixel processing very much. Use 16 of them for vertex/geometry processing and you're looking at a similar card to a 1900xt(x). It's not like the memory requirements are changing that much.
 
So basically you have GPU handling graphics.. Unified shaders handling all shaders...

Instead of the old way... GPU has set # of shaders and handle vertex/pixel shaders/graphics.
 
Endurancevm said:
Because with unified architecture you are not limited by pipelines, if you have 64 shader units they can do whatever is needed and not what they can. EG. if a frame has more pixel shading needed and less vertex then the shader units can do that, and vise versa. If a frame hypothetically had no vertex shading and all pixel shading then all 64 units would be doing pixel shading which is really fast.

You still need power either way to fill those triangles not just shaders.. it's interesting though... I wonder how this will effect games and by how much...
 
why is ati releasing this card soo soon when they are releasing the x1950xtx video card later this month?
 
you dont know exactly when theyre releasing it. MAYBE october, maybe not. maybe it gets pushed back... maybe you get a picture of santa with the vista box. Maybe....
 
insane said:
why is ati releasing this card soo soon when they are releasing the x1950xtx video card later this month?
the price on the x1950xtx would drop if this was released because its newer gen. that's not a bad thing ;)
 
Just look at 7800GTX vs 1800XT...

1800XT 16 pixel shaders - 48GB of memory bandwidth while 7800GTX has 24 pixel shaders - 38.4GB of memory bandwidth...

1800XT beats GTX easily... Unified shaders are good that it doesn't boggle the system because of shaders but I think the graphics still rely on fillrate and memory bandwidth than anything else...
 
I'm not worried about the card so much as the thing it's sitting on. ;) what we're going to need is some serious cpu horse power and mobo memory bw to keep up with the card... but I guess the conroe + new rd600 chipset will provide that? hopefully.. man, it's going to be a fun holiday season for gaming hardware! :)
 
I think the release-date is outdated: TheInquirer stated that the release will be postponed since the X1950XTX was also postponed. I also think so.
 
Can't wait to get one.

Does anyone remember how long it took for the Price Gouging to end when the x1900xt first came out? Was it about a month after, when retailers started selling at the MSRP?
 
i wouldnt worry about it, by the time you wait for the drivers to mature the prices will come down :)
 
first off there is no such thing as a 64 pipeline r600, its either a 64 shader processor chip or something similiar to this with x number of arrays.

These rumored specs are old.

It won't be able to do high level graphics and high level physics processing at the same time, its going to end up being one or the other. I doubt the g80 will either or any GPU for that matter will have enough processing power to do both conciderable well at the same time, until global illumination, tesselation and what not become main stream and GPU's are fast enough to render such things at a constant 60+ frames at very high resolutions.

GDDR4, its confirmed that the r600 will be using GDDR4, don't see ATi going back to GDDR3 the bandwith needed for a 64 + shader chip will require GDDR4 anyways.

Price sounds low if the r580+ is $450 this card will end up $500+
 
In other words, no new news. There is no way we will see R600 before late Decemeber either.
 
The standard 7800GTX was slower than the X1800XT at higher resolutions(look at Tom's Hwardware VGA chart) but the vaporware 7800GTX 512 was faster(but then again, it was insanely clocked).
 
albanian is making stuff up guys. This is the same guy who made a thread where he claimed he is downloading a leaked demo of crysis from a "german website". It was locked and he hasn't sent pm's to anyone with the link. He's getting a kick out of this hoaxing or something. Albanian, all new technology info isnt from Germany first.. just to let you know
 
Big Fat Duck said:
albanian is making stuff up guys. This is the same guy who made a thread where he claimed he is downloading a leaked demo of crysis from a "german website". It was locked and he hasn't sent pm's to anyone with the link. He's getting a kick out of this hoaxing or something. Albanian, all new technology info isnt from Germany first.. just to let you know

if u read the thread was locked so im not risking a ban by creating a new thread claiming it didnt work, obviously if u see no screen shots the shit didnt work.

the demo did not work why the hell would i send u all a link of crysis demo when im risking a ban so im not talking about the demo anymore and it didnt work keeps giving me a file missing error on install. and it is from my friend magazine since hes german he gave me all this stuff

so how about u talk about something that u know cause u dont know much my friend

oh and if ur mad i didnt send u a pm of the link its cause i just figured it doesnt work, and it didnt if it did i would put screen shots up its not that hard i have no need to lie, dont be mad i found a fake link of crysis and u didnt.
 
How so? If G80 has 32 typical pixel pipelines, we're probably talking about 64 ALUs and mini-ALUs. R600 may use the same sort of ALU arrangement as R580: one multipurpose (MADD+MUL+ADD) and one "fairly" useless ADD times 64 (in the case of R600).

From my perspective, nVidia's shader processing is a bit more efficient. Fewer transistors performing, for the most part, the same work at the same clock speeds. The issue with nVidia hardware right now is that there's nothing great handling all the data like ATi's Ring Bus controller, so overall performance will suffer. On the flip side, ATi needs a controller (a load balancer) to handle some duties in R600, which nVidia does not, so that may be the one major negative aspect of the unified design. I'm certain they've figured things out with Xenos, though.

For those concerned about R600 not being able to draw polygons, that's not going to be an issue. We're still in the days of hardware T&L here.
 
Not to sound like a complete idiot here, but is it safe to assume that if these DX10 cards come out in October, and since vista doesn't exist till january, that these cards will run on xp running DX9 fine? I only ask because some people have told me that these cards won't run on xp.
 
They will run fine using XP only

i doubt ATi or Nvidia would release a card that would only work halfway good and only optimized it for an OS with no real set release date
 
All DX10 cards will remain compatible with SM3.0/2.0/1.1 and so on.
 
Unified design would be better than any current card if ATI is going to use 64 no matter what. The point of unified is for the card to be at 100% usage all the time.
 
pArTy said:
Unified design would be better than any current card if ATI is going to use 64 no matter what. The point of unified is for the card to be at 100% usage all the time.


Unification will only give a maximum of 30% performance advantage if the schedular is very effecient. Unificiation isn't about speed, its about flexibility. And if ATi has 64 ALU's and no change to thier pipelines, it will end up around the same performance as the x1900xtx possibly at little bit higher, if you look at xenos it has 48 unified shaders and performs close to a 1800xt :)
 
Back
Top