New Radeon Series Lacks Full SM3.0 Support

Status
Not open for further replies.
tranCendenZ said:
Even better would be a weekly "face-off" by a forum member on each side of the NV/ATI fence on some heated topic. Of course, each member would be voted on by the general forum populace every week, the debate would be moderated by kyle/brent, and the results would be posted on the front page each week ;)
Click to expand...

I can only hope that some day my site will be popular enough to do that kind of stuff...cuz I have the skillz to make that kind of thing happen...vbulletin is pretty easy to modify...
 
Jbirney said:
Have proof on this? Link?


Topweasel,

Your blowing things way up. ATI meets the techincal specs of SM3 and this is just one of many of SM3 features. So the may not have followed the spirt of the SM3 specs..there are still many other parts of SM3 that they did a better job..
Click to expand...

whoa whoa whoa...it's better according to THEM...you know, the same people that told you the cards would be in stores on launch day?
 
Can't any of you find ANYTHING more important in your lives then bitching about something you can't do anything about, for hours on end?
 
SnakEyez187 said:
Can't any of you find ANYTHING more important in your lives then bitching about something you can't do anything about, for hours on end?
Click to expand...

blah...I can multi-task...:)
 
razor1 said:
You are the one bashing canadians, not me, sorry but Sander's really didn't lie take a look at Hexus's and PC
Click to expand...

Can you please stop say this crap over and over in the different froums..you have seen the B3D thread where they show his number are off by as much as 40% You use two web sites that are close but you ignore ALL the others that show his numbers to be way off.
 
^eMpTy^ said:
whoa whoa whoa...it's better according to THEM...you know, the same people that told you the cards would be in stores on launch day?
Click to expand...

Nice try...here's a cookie...while your reading that go check out some of the tests xbit did on the PS3.0 code path...then look over the dynamic branching (another part of SM3.0) and see how the x1600 is FASTER than a 7800GTX...then you can figure out if theirs was right.
 
Jbirney said:
Nice try...here's a cookie...while your reading that go check out some of the tests xbit did on the PS3.0 code path...then look over the dynamic branching (another part of SM3.0) and see how the x1600 is FASTER than a 7800GTX...then you can figure out if theirs was right.
Click to expand...

how do you read a cookie?

And yeah, dynamic branching...like just about everything else ATi does...it looks wonderful on paper...but I don't see it making a difference anywhere else...
 
Jbirney said:
Can you please stop say this crap over and over in the different froums..you have seen the B3D thread where they show his number are off by as much as 40% You use two web sites that are close but you ignore ALL the others that show his numbers to be way off.
Click to expand...


Well if Sanders cherry picked the benchmarks, that doesn't mean he lied, the only one that lied was.......

My point at B3D is the benchmarks are all over the place, its very hard to say Sanders was lieing since there are benchmarks out there from reputiable sites that have similiar results. Unless Sanders is a damn good fortune teller, and his technical know how somehow gave him insite on performance vartiations as close as 1% without getting the info from someone who tested the cards, is that what you are suggesting?

Anyways that is a different discussion and doesn't belong in this thread.
 
^eMpTy^ said:
how do you read a cookie?
Click to expand...

I was throwing you a cookie to reward you week attemps at trying to be smart...too bad they failed..

Now did you even bother to look for your self and see if some of the PS3.0 "stuff" runs faster on the R520 series yet? Yes? No? Was the R520 faster at some of the PS3.0 test? Yes? No?

Well if Sanders cherry picked the benchmarks, that doesn't mean he lied, the only one that lied was.......
Click to expand...

I have no idea what the hell he thinks he tested and no one really dose. Razor seems to think he did...however looking at more than just a few web site it seems his numbers were that of a 1800xl and not the xt as he claimed...
 
razor1 said:
My point at B3D is the benchmarks are all over the place, its very hard to say Sanders was lieing...
Click to expand...

Well then if its hard to tell if he was lying, its also hard to tell its the truth..you said it your self the numbers at some match but the numbers at other simular places do not..if they are all over the place then why do you constantly say his numbers were right????
 
Jbirney said:
I was throwing you a cookie to reward you week attemps at trying to be smart...too bad they failed..

Now did you even bother to look for your self and see if some of the PS3.0 "stuff" runs faster on the R520 series yet? Yes? No? Was the R520 faster at some of the PS3.0 test? Yes? No?



I have no idea what the hell he thinks he tested and no one really dose. Razor seems to think he did...however looking at more than just a few web site it seems his numbers were that of a 1800xl and not the xt as he claimed...
Click to expand...


most likely it was the underclocked XT that was sent to developers and board partners along time back, like the ones at 525-550 mhz. I knew his numbers were close, because I too saw these cards and various benchmarks. Oh btw go tell Rys that his benchmarks are off, I think he will be really happy with you :). Doesnt' mean anything if Sander was right or wrong, since the numbers are out, if anything he has more credibility now but he is still a ass of the first level.
 
ATI claimed "SM3.0 done right" and clearly they did not do it right. If you have to use a "work around" then you did something wrong.

If using a vertex texture fetch is part of the spec and they do not support it, then they do not fully support SM3.0.
 
Jbirney said:
I was throwing you a cookie to reward you week attemps at trying to be smart...too bad they failed..
Click to expand...

Actually, if you read your post, you tossed me a cookie, and suggested that I "read" it...unless I'm mistaken, which I could be, but I dont' think I am...

Jbirney said:
Now did you even bother to look for your self and see if some of the PS3.0 "stuff" runs faster on the R520 series yet? Yes? No? Was the R520 faster at some of the PS3.0 test? Yes? No?
Click to expand...

I read that yesterday...I read pretty much every review I can find...

And yeah, according to "xbitmark", whatever that is, the x1000 series does much better with dynamic branching...but it's a synthetic test...so while it's interesting...it doesn't really matter...

Luckily, since nvidia implemented ps3.0 over a year ago, anyone who's designing anything to do dynamic branching is doing it with nvidia cards in mind...;)
 
Jbirney said:
Have proof on this? Link?


Topweasel,

Your blowing things way up. ATI meets the techincal specs of SM3 and this is just one of many of SM3 features. So the may not have followed the spirt of the SM3 specs..there are still many other parts of SM3 that they did a better job..
Click to expand...


the point is when you buy a product that says " SM 3.0 done right" you expect a full feature set. not Including this feature might put strain on devs to include extra code to determine what the user is running in order for the game to run properlly.

Also when a card is missing a feature how can it be " done right". it should be SM 3.0 done right according to ATI.

how exactly can u do a better job at supporting SM 3.0?
 
^eMpTy^ said:
Actually, if you read your post, you tossed me a cookie, and suggested that I "read" it...unless I'm mistaken, which I could be, but I dont' think I am...
Click to expand...

Oh good god I made a typo..sorry did not realize you were prefect in that regaurds I will be more carefull....



And yeah, according to "xbitmark", whatever that is, the x1000 series does much better with dynamic branching...but it's a synthetic test...so while it's interesting...it doesn't really matter...
Click to expand...

There were more test of the PS3.0 pipeling then just dynamic branching and they all ran very will on the new r520 design..
 
swatX said:
the point is when you buy a product that says " SM 3.0 done right" you expect a full feature set. not Including this feature might put strain on devs to include extra code to determine what the user is running in order for the game to run properlly.

Also when a card is missing a feature how can it be " done right". it should be SM 3.0 done right according to ATI.

how exactly can u do a better job at supporting SM 3.0?
Click to expand...

The card supports the specs of PS3.0 there is no doubt on this..yes its a loop hole but the fact remains. Let me ask this..do you actually believe any PR for face value?
 
swatX said:
the point is when you buy a product that says " SM 3.0 done right" you expect a full feature set. not Including this feature might put strain on devs to include extra code to determine what the user is running in order for the game to run properlly.

Also when a card is missing a feature how can it be " done right". it should be SM 3.0 done right according to ATI.

how exactly can u do a better job at supporting SM 3.0?
Click to expand...


oh be quiet please:

Vertex fetch is optional, so it's not needed to be SM3.0 compliant.

You basically have three options. Do like nVidia and implement it, but be slow and very limited. Or spend an assload of die space to put equal texturing capabilities in the vertex shaders, at the expense of general performance or other features you could use that space for. Or defer this feature until it makes more sense and use R2VB instead for today's needs as that's going to outperform vertex fetch and will work on previous generation hardware as well. Personally I think the last option is the right choice.
Click to expand...

^ = truth, you = propaganda.

You're taking a limited feature that ATI dropped out on because its just that, limited, and you're making it seem like it makes one card far superior over the other. Saying its now not SM3.0 capable is total crap as well. This is just not true. A 6600 can support FP16 HDR, does that make it generally superior to an X800? Nope.
 
Jbirney said:
Oh good god I made a typo..sorry did not realize you were prefect in that regaurds I will be more carefull....

There were more test of the PS3.0 pipeling then just dynamic branching and they all ran very will on the new r520 design..
Click to expand...

I just thought the cookie reading thing was funny...:)

btw...it's "perfect", not "prefect" and "regards", not "regaurds"...;)

Yeah and I can point out a bunch of things nvidia cards can smoke ATi at all day long...like...*gasp*...OpenGL, which actually makes a difference to gamers right now...and how about z stencil operations?

Both architectures are sure to have thier strong points, in the end it's the overall gaming experience that matters...don't you read anything Brent writes?
 
Shifra said:
oh be quiet please:



^ = truth, you = propaganda.

You're taking a limited feature that ATI dropped out on because its just that, limited, and you're making it seem like it makes one card far superior over the other. Saying its now not SM3.0 capable is total crap as well. This is just not true. A 6600 can support FP16 HDR, does that make it generally superior to an X800? Nope.
Click to expand...


How is it limited Shifra? Someone put this in quotes since I'm on his ignore list. Would really like to see his answer.
 
^eMpTy^ said:
Yeah, Humus, the most unbiased guy ever...
Click to expand...

Should he lie about this one?

"Vertex fetch is optional, so it's not needed to be SM3.0 compliant."

If true this thread can go in the trash can, at least the topic title is plain wrong.
 
ATi's method isn't wrong, Dx doesn't give strict guidence on how the silicon should handle the vertex texture fetch. If anything might just cause more work for developers, which kinda sux, but the turn around is it works, I don't think it will be faster then the vertex texture fetch though. Specially since bandwidth is going to play an important roll in this.
 
^eMpTy^ said:
I just thought the cookie reading thing was funny...:)

btw...it's "perfect", not "prefect" and "regards", not "regaurds"...;)
Click to expand...

I thought its common knowledge when you start pointing out grammar/spelling mistakes your loosing your point...anyways..

Yeah and I can point out a bunch of things nvidia cards can smoke ATi at all day long...like...*gasp*...OpenGL, which actually makes a difference to gamers right now...and how about z stencil operations?
Click to expand...

And how on earth do those relate to SM3? I mean last I check thats what we were trying to discuss is SM3 and ATIs "done right" claim...nice dodge of the issue...

Both architectures are sure to have thier strong points, in the end it's the overall gaming experience that matters.
Click to expand...

Then why on earth are you constantly posting in an about ATI's SM3 done right thread if its the gaming experince that matters????


don't you read anything Brent writes?
Click to expand...

Sadly not as much as I use to. While current gaming reviews tell us about how the card plays its limited to how those game play (for example reading this his reviews tells me alot about those game...but it no use for me on other games) and usally misses out on the strenghts and weakness of the card. For example we found out from B3D x800xl 512 mb review that the extra memory could make some difference which did not come out in Brent's review...now we are seing the same thing with COD2 and soon to be other games...
 
^eMpTy^ said:
Said the kettle to the other kettle...
Click to expand...

Well no offense and while he may be biased..he still knows a hell of a lot more about 3D hardware than you or I will...so I would listen to his comments and see if they make sense for myself vrs saying anything about his inputs just because of who signs his paychecks...
 
Jbirney said:
I thought its common knowledge when you start pointing out grammar/spelling mistakes your loosing your point...anyways..
Click to expand...

I honestly just thought it was funny...making two spelling mistakes in a sentence about spelling mistakes...

Jbirney said:
And how on earth do those relate to SM3? I mean last I check thats what we were trying to discuss is SM3 and ATIs "done right" claim...nice dodge of the issue...
Click to expand...

yeah I see your point...I was headed off on a tangent there...I concur that their increased dynamic branching performance among other things gives them the right to say "done right"...early on I was just laughing at this one feature they left out of their "done right" implementation...

Jbirney said:
Then why on earth are you constantly posting in an about ATI's SM3 done right thread if its the gaming experince that matters????
Click to expand...

Because I like talking about this stuff...whether or not I have a point...whether or not it's important...I just like sitting around talking about video cards...that's why I made a website about it...

Jbirney said:
Sadly not as much as I use to. While current gaming reviews tell us about how the card plays its limited to how those game play (for example reading this his reviews tells me alot about those game...but it no use for me on other games) and usally misses out on the strenghts and weakness of the card. For example we found out from B3D x800xl 512 mb review that the extra memory could make some difference which did not come out in Brent's review...now we are seing the same thing with COD2 and soon to be other games...
Click to expand...

Yeah...I was mostly just being obnoxious with that "gaming experience" comment...I hate it when Brent goes off about that stuff...I'm curious about how the card works, not just how well it runs games...
 
Jbirney said:
Well no offense and while he may be biased..he still knows a hell of a lot more about 3D hardware than you or I will...so I would listen to his comments and see if they make sense for myself vrs saying anything about his inputs just because of who signs his paychecks...
Click to expand...

I've seen that dude be blatantly wrong on multiple occasions...I really don't trust his opinion whatsoever...it's not just that he works for ATi...it's that he's the really overzealous kind that runs his mouth without thinking...a lot...
 
Since you all like to quote b3d forums:
Well, it would certainly work, but they need a way to bypass ROPs and "feedback" raw PS output into triangle setup. Otherwise, you waste power and memory bandwidth writing back intermediate buffers to video memory. Now, they might argue it's a small percentage of bandwidth, but it adds latency to the roundtrip as well.
Click to expand...
http://www.beyond3d.com/forum/showpost.php?p=588813&postcount=25

So by not fully supporting the spec they will take a performane hit.

I guess they could put a gig of ram on the card and have a 3 slot cooler to make up the difference. :rolleyes:
 
Jbirney said:
Oh good god I made a typo..sorry did not realize you were prefect in that regaurds I will be more carefull....





There were more test of the PS3.0 pipeling then just dynamic branching and they all ran very will on the new r520 design..
Click to expand...


http://techreport.com/reviews/2005q4/radeon-x1000/shadermark.gif

not many HDR (More like the version that was used in rthdribl), Envirmental bump mapping (which isn't used much anymore), and dynamic flow control.

So what do you get 3 areas where the r520 is superior. And 22 shader types that it losses.
 
PRIME1 said:
Since you all like to quote b3d forums:

http://www.beyond3d.com/forum/showpost.php?p=588813&postcount=25

So by not fully supporting the spec they will take a performane hit.

I guess they could put a gig of ram on the card and have a 3 slot cooler to make up the difference. :rolleyes:
Click to expand...


Well this is what I said earilier, the hit won't be that bad but I'm guessing right around the same hit has what the vertex texture fetch is.
 
Apple740 said:
Should he lie about this one?

"Vertex fetch is optional, so it's not needed to be SM3.0 compliant."

If true this thread can go in the trash can, at least the topic title is plain wrong.
Click to expand...

Do you think Humus, the ATI employee you quoted, would instead say "Yes, Vertex texture fetch is required for true SM3.0 compliance, and therefore the upcoming X1800XT is not compliant with SM3.0." He'd have his pink slip by the morning :)
 
razor1 said:
http://techreport.com/reviews/2005q4/radeon-x1000/shadermark.gif

not many HDR (More like the version that was used in rthdribl), Envirmental bump mapping (which isn't used much anymore), and dynamic flow control.

So what do you get 3 areas where the r520 is superior. And 22 shader types that it losses.
Click to expand...


Yea I saw that and not sure what to make of it. The 800xt scores faster in those test than the 1800xl (even thought they have the same number of pipes, simular fill rate) so I am thinking its a drive bug or something..as the 1800xl should be faster..but its not. . For example the 1800xl falls way behind all cards (its slower than an 6800u) in the Adv PS2.0 test of 3dmark so something there is just not "right". Keep in mind the extra 8 pixel shader that the GTX has. Also shadermark was a PS2.0 test recomplied to us the PS3.0 pathIn the xbit review they show some results that made heavier use of the PS3.0 shaders where the R520 did much better...

Edit:
I guess there is just not enough data one way or another yet....
 
On another note, I wonder if this VS3.0 displacement mapping demo will work with full IQ/visual effect on X1800 cards (i.e. not ps2.0 emulation of the effect)?
http://www.realityflux.com/abba/C++/Displacement Mapping/DisplacementMapping.zip

If it is doing a PS2.0 emulation you will see that the "wave" is really flat close up. If it is doing true VS3.0 displacement mapping, it will clearly jut out far from the water.

Curious as one mentioned the feature here is often used by displacement mapping in VS3.0... No ATI hardware to test, though :)
 
tranCendenZ said:
Do you think Humus, the ATI employee you quoted, would instead say "Yes, Vertex texture fetch is required for true SM3.0 compliance, and therefore the upcoming X1800XT is not compliant with SM3.0." He'd have his pink slip by the morning :)
Click to expand...

no i really think you're wrong

a curious loophole of the VS3.0 specification is that although the capability bit for Vertex Texture capabilities must be enabled for compliance, there are no actual texture formats dictated for support, so if the capability bit is enabled but no texture formats exposed VS3.0 compliance can still be met
Click to expand...
 
Status
Not open for further replies.
Back
Top