R200 will have 6 TMUs NOT 4 as wrorngly published by Reactor Critical !!
- Thread starter Sir_Pent
- Start date
That'd definitely be nice for future games... right now I doubt it'll be too beneficial for most games though. I always did think Reactor Critical's leak of 4TMU's on the Radeon 2 to be a bit surprising, I wouldnt be surprised if it was incorrect.
"I always did think Reactor Critical's leak of 4TMU's on the Radeon 2 to be a bit surprising, I wouldnt be surprised if it was incorrect."
I though RC said it would feature two - but be configured like GF3 (goes to 4 in two cycles). However, this makes no sence seeing as SMARTSHADER / DX8.1 spec features 6 texture's per pixel. Now, if R200 had 4TMU's why would it only use two in the next cycle? Or if if featured two TMU's it would mean that it could do 6 in 3 clock cycles - so why not do 8 in 4 clocks or 10 in 5 etc, etc? Kinda doesn't add up. I still think it will be three TMU's per pipe (like Radeon) with a single loop to 6.
However, I guess we will find out reasonably soon.
I though RC said it would feature two - but be configured like GF3 (goes to 4 in two cycles). However, this makes no sence seeing as SMARTSHADER / DX8.1 spec features 6 texture's per pixel. Now, if R200 had 4TMU's why would it only use two in the next cycle? Or if if featured two TMU's it would mean that it could do 6 in 3 clock cycles - so why not do 8 in 4 clocks or 10 in 5 etc, etc? Kinda doesn't add up. I still think it will be three TMU's per pipe (like Radeon) with a single loop to 6.
However, I guess we will find out reasonably soon.
Sorry had to run out and pick up a few socket A coolers...
any how I would like to direct your attention to a few things if I may:
1. (a few illustrations from ATI's released roadmap)
Slide 6 of 22:
<center>
<img src="http://www.radeonfaq.com/rv200/3dengine.gif">
</center>
Here we see that the 3 TMUs remain intact in the RV200, within the Pixel Tapestry block we see two renderers (pipelines) each with 3 texture units.
then onto slide 13 (of 22), paying particular attention to the third entry:
<center>
<img src="http://www.radeonfaq.com/rv200/rendline.gif">
</center>
This simply illustrates that the R200 will have twice the rendering piplines as compared to the Rv200 (4 vs 2).
then onto slide 14 (of 22), again paying attention to the third line:
<center>
<img src=http://www.radeonfaq.com/rv200/6tex.gif">
</center>
This is what truly solidifies the deal, similiar to how the GF3 with its 4 pipes and 2 texture units is able to "combine" each renderers lines and TUs to render up to 4 textures per pixel per pass. If the R200 is to have the same amount of both rendering lines (4) and texture units (4?) then the ability to do 6 textures per pixel per pass becomes nearly impoosible unless ATi has bent the laws of mathematic equations so that (4*2)/2 = 6. The ONLY real possible solution to this is that the R200 has 3 TUs per renderer (4*3)/2 = 6 = True.
2. Now onto ATI's official documents on SmartShader:
(http://www.ati.com/na/pages/technology/hardware/smartshader/smartshader_white_paper.html )
Listed as one of the key improvments in the R200:
"Support for up to six textures in a single rendering pass, " again confirming what was already pointed out in slide 14 of ATI's Roadmap (see above.)
The following two pictures give an visual representation of this process:
<center>
<img src="http://www.radeonfaq.com/rv200/pixshad.jpg">
</center>
<center>
<img src="http://www.radeonfaq.com/rv200/pixshad2.jpg">
</center>
Now, scrolling down the page to "SMARTSHADER Advantages", in a comparison of DX 8.0 Pixels Shaders, PS (as utilized the the current GF3) we see the following:
Taken all these facts into consideration and the very uniform increase of 50% in instructions and texture support. These increases in support over DX 8 (4 texture untis for example) can NOT be simulated directly without the inclusion of addition hardware (TUs), this is why Reactor Critial's assumptions as to the amount of texture units is very flawed ... to go any further into details..well, Tuesday is approaching soon enough...
any how I would like to direct your attention to a few things if I may:
1. (a few illustrations from ATI's released roadmap)
Slide 6 of 22:
<center>
<img src="http://www.radeonfaq.com/rv200/3dengine.gif">
</center>
Here we see that the 3 TMUs remain intact in the RV200, within the Pixel Tapestry block we see two renderers (pipelines) each with 3 texture units.
then onto slide 13 (of 22), paying particular attention to the third entry:
<center>
<img src="http://www.radeonfaq.com/rv200/rendline.gif">
</center>
This simply illustrates that the R200 will have twice the rendering piplines as compared to the Rv200 (4 vs 2).
then onto slide 14 (of 22), again paying attention to the third line:
<center>
<img src=http://www.radeonfaq.com/rv200/6tex.gif">
</center>
This is what truly solidifies the deal, similiar to how the GF3 with its 4 pipes and 2 texture units is able to "combine" each renderers lines and TUs to render up to 4 textures per pixel per pass. If the R200 is to have the same amount of both rendering lines (4) and texture units (4?) then the ability to do 6 textures per pixel per pass becomes nearly impoosible unless ATi has bent the laws of mathematic equations so that (4*2)/2 = 6. The ONLY real possible solution to this is that the R200 has 3 TUs per renderer (4*3)/2 = 6 = True.
2. Now onto ATI's official documents on SmartShader:
(http://www.ati.com/na/pages/technology/hardware/smartshader/smartshader_white_paper.html )
Listed as one of the key improvments in the R200:
"Support for up to six textures in a single rendering pass, " again confirming what was already pointed out in slide 14 of ATI's Roadmap (see above.)
The following two pictures give an visual representation of this process:
<center>
<img src="http://www.radeonfaq.com/rv200/pixshad.jpg">
</center>
<center>
<img src="http://www.radeonfaq.com/rv200/pixshad2.jpg">
</center>
Now, scrolling down the page to "SMARTSHADER Advantages", in a comparison of DX 8.0 Pixels Shaders, PS (as utilized the the current GF3) we see the following:
- DX 8.0 PS supports a Max Texture input of 4 where as PS in DX 8.1 supports upto 6 (an increase of 50%, or 2),
- DX 8.0 PS under Max. Program Length, supports 12 instructions:
- 4 Texture sampling, 8 color blending
- DX 8.1 supports up to 22 instructions in Program length, the same amount of instructions are used in color blending, however again we see a 50% increase in texture sampling (up 2 from 4) to equal 6 and also includes 8 new instructions, being texture blending.
Taken all these facts into consideration and the very uniform increase of 50% in instructions and texture support. These increases in support over DX 8 (4 texture untis for example) can NOT be simulated directly without the inclusion of addition hardware (TUs), this is why Reactor Critial's assumptions as to the amount of texture units is very flawed ... to go any further into details..well, Tuesday is approaching soon enough...
Damn, I thought you had some new information for us 
Most of what you've listed has been known for the past few days though, and is the basis for the suspicion that RC was mistaken.
Either way I think your correct, the Radeon 2 will have 6TMU's.
Most of what you've listed has been known for the past few days though, and is the basis for the suspicion that RC was mistaken.
Either way I think your correct, the Radeon 2 will have 6TMU's.
<< Umm no posting images ???? >>
Nope, you can't post images.
Nor can you use HTML in your posts, though you can simulate some HTML commands by using the symbols.
Under the posting interface where it says the following...
<< HTML code is not permitted. Special symbols permit some HTML functionality. >>
click on "symbols" and it'll show you what symbolic commands FuseTalk will recognize.
(FuseTalk being the software the forum is built upon)
Nope.. no NEW info, (not yet ), have to wait until monday at 11:59pm .. hehe !!
Well, I have reapeatedly eamiled and posted my concerns about this to the staff at RC, so far all requests have gone without response. Even so far that I was accused of "RC Bashing" after presenting facts and the difference between facts and rumors, I noticed that RC appears to be momentarily down... maybe that are theying to remove the cream pie from their face..
also the comment on RC: that the RV200 not being a gaming solution... well its funny that the RV being a Radeon Ultra equivilant is no longer a "gaming solution".... doesn't say much for the entire GF line (exempting GF3 of course)..
), have to wait until monday at 11:59pm .. hehe !!Well, I have reapeatedly eamiled and posted my concerns about this to the staff at RC, so far all requests have gone without response. Even so far that I was accused of "RC Bashing" after presenting facts and the difference between facts and rumors, I noticed that RC appears to be momentarily down... maybe that are theying to remove the cream pie from their face..
also the comment on RC: that the RV200 not being a gaming solution... well its funny that the RV being a Radeon Ultra equivilant is no longer a "gaming solution".... doesn't say much for the entire GF line (exempting GF3 of course)..
<< Nope.. no NEW info, (not yet
), have to wait until monday at 11:59pm .. hehe !!Well, I have reapeatedly eamiled and posted my concerns about this to the staff at RC, so far all requests have gone without response. Even so far that I was accused of "RC Bashing" after presenting facts and the difference between facts and rumors, I noticed that RC appears to be momentarily down... maybe that are theying to remove the cream pie from their face..
also the comment on RC: that the RV200 not being a gaming solution... well its funny that the RV being a Radeon Ultra equivilant is no longer a "gaming solution".... doesn't say much for the entire GF line (exempting GF3 of course).. >>
Well the RV200 isnt intended to be a gaming solution by ATi. it's intended to take the place of the Radeon VE as a workstation dual-head card.
But looking at the specs and the unverified benchmarks given out by Rageunderground it's pretty clear that the RV200 will make a very good gaming card for the budget market. I mean it's likely to be priced only slightly higher then the current Radeon VE and offers performance that should match if not outperform the current Radeon SE/Radeon "Ultra". And may even overclock to perform beyond it.
If it's available next month at a decent price as it's supposed to then it may just be my next video card.
If RC doesnt think the RV200 can hold it's own in gaming situations then by their standards I'd imagine that nothing short of the GF3 and R200 is a viable option.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 21K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter