R420 = Ps3.

CleanSlate

Supreme [H]ardness
Joined
Mar 28, 2003
Messages
5,258
Where do ppl get this non-ps3 stuff from? I distinctly remember the leaked presentation saying that it supports ps3, but one part hurts performance and they'd tell dev's to work around it in their code. That's ps3 compliant, is it not.

~Adam
 
can i quote you on that?
or can you atleast make a quote of the presentation?
 
I'll post a sn of the presentation. gimme a minute.

Doh I formatted I need to install ms office :D


~Adam
 
Originally posted by CleanSlate
Where do ppl get this non-ps3 stuff from? I distinctly remember the leaked presentation saying that it supports ps3, but one part hurts performance and they'd tell dev's to work around it in their code. That's ps3 compliant, is it not.

~Adam

Didn't those leaked emails or whatever of ATi execs say otherwise? I thought they distinctly said R420 doesn't have SM3.0, so they are going to try to down play its usefulness.
 
"Steer people away from flow control in ps3.0 because we expect it to hurt badly. [Also it’s the main extra feature on NV40 vs R420 so let’s discourage people from using it until R5xx shows up with decent performance...]"


That is that direct quote from the leaked presentation. It says nothing about ps3 being left out of r420 and infact talks about how they want to get developers to steer clear of their weakness until they can fix the weakness in the r5xx. Thus I'd like to tell everyone to please stop talkinga bout r420 with no ps3 shaders because it's obviously in there.
 
Did you actually read the quote you posted?

You seem to be reading between the lines so much that you forgot to actually read the lines themselves.
 
Maybe you don't udnerstand what he is talking about. The first sentence says that one piece(?) of ps3 hurts performance, thus in the second sentence it says that it will be important and to discourage ppl from using IT as in the piece that is hurtfull to performance (NOT PS 3 as a whole!!) from what I gather (understand about the english language? lol.). Not a hard thing to understand.

Someone tell me I'm wrong and provide proof please.

~Adam
 
Originally posted by CleanSlate
"Steer people away from flow control in ps3.0 because we expect it to hurt badly. [Also it’s the main extra feature on NV40 vs R420 so let’s discourage people from using it until R5xx shows up with decent performance...]"


That is that direct quote from the leaked presentation. It says nothing about ps3 being left out of r420 and infact talks about how they want to get developers to steer clear of their weakness until they can fix the weakness in the r5xx. Thus I'd like to tell everyone to please stop talkinga bout r420 with no ps3 shaders because it's obviously in there.
I do not believe that is a real Slide from ATi at all. Nor is that a real Quote form ATi. This is pure FUD.

when ATi is presenting to people they point out their Strengths not their supposed Weaknesses. You either edited that and added words to it yourself or the whole thing is a sham.

Besides Flwo Control is not the main feture of PS3.0 that hurts performance. Its Branching. And Nv40 hardware is slow as molases at Branching anyway.. :rolleyes:
 
When I read that quote I interpretted it to mean that there was some problem with 'flow control' (whatever that is, im no expert) in PS3.0. My interpretation of that quote in fact shows that PS3.0 will most definitely be in the R420 as they stated they are having a problem with flow control in PS3.0 (how could you have a problem with flow control in PS3.0 if there is no PS3.0 in your card lol)
 
Originally posted by Hellbinder
I do not believe that is a real Slide from ATi at all. Nor is that a real Quote form ATi. This is pure FUD.

when ATi is presenting to people they point out their Strengths not their supposed Weaknesses. You either edited that and added words to it yourself or the whole thing is a sham.

Besides Flwo Control is not the main feture of PS3.0 that hurts performance. Its Branching. And Nv40 hardware is slow as molases at Branching anyway.. :rolleyes:

Yes, when they are presenting to the public, not when they are presenting confidentially to marketing execs
 
Originally posted by Hellbinder
Besides Flwo Control is not the main feture of PS3.0 that hurts performance. Its Branching. And Nv40 hardware is slow as molases at Branching anyway..

Branching instructions are a subset of flow control instructions ;)

Edit: And here are the flow control instructions (and the other ones)

PS2 Extended
New instructions:

Static flow control instructions - if bool, call, callnz bool, else, endif, rep, endrep, label, ret
Dynamic flow control instructions - break, break_comp, breakp, callnz pred, if_comp, if pred, setp_comp - ps
Arithmetic instructions - dsx, dsy
Texture instruction - texldd

PS3
New instructions:

Setup instruction - dcl_usage
Static flow instructions - loop, endloop
Arithmetic instruction - sincos (new syntax)
Texture instruction - texldl
 
Originally posted by Cryect
Branching instructions are a subset of flow control instructions ;)
Specifically *branching* is the problem. Its a problem with the Nv40 as well.

R420 could be considered PS3.0 compatible, not PS3.0 Complient as far as i understand it.
 
/me shrugs

branching is the majority of the flow control commands and it should be obivious to any graphics programmer they are a bad idea.

Can normally achieve the same thing with some linear algebra tricks instead of branching assuming you dont want something completely different but all depends on what you want to compute.

And really I don't see what the big difference does it matter if ATI has PS3 or not. It doesn't add anything too useful beyond PS2 Extended. Assuming a HLSL is used and ATI's new cards end up being at least PS2 Extended they should do basically everything that a PS3 card can do without killing the framerate.
 
Originally posted by Hiyruu
Playstation 3?


That's what i thought the thread was about until i started to read about it...

And you had my hopes up :(
 
So R420 will not have PS3 right. I don't think by delaying the release date they can include ps3. they can make a faster card. I heard the original R420 was not suppose to pass 10k mark on 3dmark03.
Actually, INQ's prediction, are somehow accurate, sometimes they tend to over do the story, but all accurate.

I don't mind Nvidia becomes king. As long as consumer wins, that is fine with me.
 
Originally posted by Hitokiri Batohsai
So R420 will not have PS3 right. I don't think by delaying the release date they can include ps3. they can make a faster card.

well i think sm3.0 has to do with hardware, could be the reason why the r420 has less transitors than nv40.

not 100% sure on this.
 
Flow control should be referenced to branched predictions.

Which could mean that the branch prediction on the new card is not as fast as ATI would like it to be. It doesn't necessarily mean it does not have PVS 3.0 either.

Similarily if it ends up being true is the Intel Prescott, there the branch is too long compared to the shorter Northwood. The new ATi sounds like the branch is not accurate enough (more branch mispredictions, causing a miss, requiring another pass through the renderer) and not necessarily too long.

Oddly enough, the 6800 has some insanely high Pixel and vertex scores, and yet still is not all that much faster in games like Farcry. I'm thinking they might have massive branch prediction problems too/ Benchmarks will rarely have a branch prediction miss, as its all laid out - In real games though, there may be massive misses (heck, it could get so bad as to go below 30 percent "hits" on either card... definitely something to look out for.)
 
well, In general the 6800 ultra is cpu limited in a big way pretty much across the board, plus in far cry it is running code specifically optimized for the nv3x, hence the IQ problems and possibly some of the performance ones. once that is fixed I think it will help quite a bit.
 
Originally posted by ZenOps
Flow control should be referenced to branched predictions.

Which could mean that the branch prediction on the new card is not as fast as ATI would like it to be. It doesn't necessarily mean it does not have PVS 3.0 either.
Ummm branch predictions being too slow is prolly not it. Mispredicted branching is likely the issue. Any good shader will avoid having to branch unless its of the utmost necessity and which it shouldn't be normally.

Originally posted by ZenOps
Similarily if it ends up being true is the Intel Prescott, there the branch is too long compared to the shorter Northwood. The new ATi sounds like the branch is not accurate enough (more branch mispredictions, causing a miss, requiring another pass through the renderer) and not necessarily too long.
[/B]

Thats called a pipeline not a branch.
 
Originally posted by Hellbinder
I do not believe that is a real Slide from ATi at all. Nor is that a real Quote form ATi. This is pure FUD.

when ATi is presenting to people they point out their Strengths not their supposed Weaknesses. You either edited that and added words to it yourself or the whole thing is a sham.

Besides Flwo Control is not the main feture of PS3.0 that hurts performance. Its Branching. And Nv40 hardware is slow as molases at Branching anyway.. :rolleyes:

Ask someone the url to get the slide.. search for ps3 and its the first slide that comes up, in the confidential notes that only the displayer of the slides would see. Honestly why would I lie about something like this? Makes no sense.

Besides lots of other ppl saw the slides, if I did lie they could just tell you all so.

~Adam
 
ATI accidentally put the .ppt with the notes for the presentation up, they quickly remedied the situation but not before people had downloaded it. I have it and what they are saying is true. search google for the file "save_the_nanosecond.ppt", that should get you the original file with the notes.
 
Yeah, branching is bad to start with, a good programmer will leave as little to a predict as possible.

And oops, yeah its a pipe.
 
Originally posted by ZenOps
Flow control should be referenced to branched predictions.

Which could mean that the branch prediction on the new card is not as fast as ATI would like it to be. It doesn't necessarily mean it does not have PVS 3.0 either.

Similarily if it ends up being true is the Intel Prescott, there the branch is too long compared to the shorter Northwood. The new ATi sounds like the branch is not accurate enough (more branch mispredictions, causing a miss, requiring another pass through the renderer) and not necessarily too long.

Oddly enough, the 6800 has some insanely high Pixel and vertex scores, and yet still is not all that much faster in games like Farcry. I'm thinking they might have massive branch prediction problems too/ Benchmarks will rarely have a branch prediction miss, as its all laid out - In real games though, there may be massive misses (heck, it could get so bad as to go below 30 percent "hits" on either card... definitely something to look out for.)

umm correct me if im wrong, but your talking about 2 diferent things. the 30 PIPELINE prescot is 30 piplines in serial of each other. The 6800 ultra has 16 branches that process in PARALLEL
 
Originally posted by evilchris
Did you prove your THREAD SUBJECT, dickweed? Sit down, or give proof.

Atleast he gives an explanation, and one that actually makes sense, if you were to say R420! = full PS3 support then you might have made sense
 
Why is it that a bunch of assholes come out of the woodworks when new cards launch?

Things never change... people spouting off about stuff they know nothing about and getting their panties in a bind.

Bottom line, nobody knows except ATI and folks under NDA. Everyone cool their jets and wait 1 week.
 
^^ you sir are an intelligent man.... I honestly never quite thought of it that way, but you have it right. And the card will be unveiled in a week, and there is no sense in going nuts over what it will or won't be yet.... though a thread like this has a logical reason to be made (since we don't know if it really has ps3 or not)... they are also bait for fanboys and fuel for flame wars... I say just sit tight and wait.
 
The pipeline is referenced to the branch (at least i think so)

the pipeline defines the path, where it starts to where it ends (hence the pipe) the acutal route it takes in the pipe either 1 or 0, defines the branch (I think) if it makes a bad prediction in the branch, it has to go through the pipe again anf take a different branch.

Not up on my pipe/branch stuff, so feel free to correct me here.
 
This branch instruction you speak of is it the branch they have in assembly? In assembly there is a limit on how many lines you can branch.
 
Originally posted by Hellbinder
I do not believe that is a real Slide from ATi at all. Nor is that a real Quote form ATi. This is pure FUD.

when ATi is presenting to people they point out their Strengths not their supposed Weaknesses. You either edited that and added words to it yourself or the whole thing is a sham.

Besides Flwo Control is not the main feture of PS3.0 that hurts performance. Its Branching. And Nv40 hardware is slow as molases at Branching anyway.. :rolleyes:

Great, glad to see you are here too Mr. FuckingPositive. You know, between all the other forums you shoot your mouth of on, you think you would have seen the posts that showed the slides that were infact from ATI. And I believe someone at your precious Rage3d posted on it along with Futuremark.

And from someone who actually knows business, they do infact show the weaknesses so they can show what is good and needs to be presented before the PR Starts. Hello? R8500 Problems when it first came out? T&L Work arounds for drivers?

Slide was leaked and is real. ATI will do a cover up on it and move on. They are pushing FP24 for a reason and will continue with this product till R500. Thank you. ;)

(You know, I don't mean to jump down your throat, but thats how you come across. Here and Rage3d.)
 
I used common knowledge and little bit of speculation... all you used was a couple cuss words and a not equal symbol. "dickweed".

~Adam
 
Originally posted by SnowBeast
Great, glad to see you are here too Mr. FuckingPositive. You know, between all the other forums you shoot your mouth of on, you think you would have seen the posts that showed the slides that were infact from ATI. And I believe someone at your precious Rage3d posted on it along with Futuremark.

And from someone who actually knows business, they do infact show the weaknesses so they can show what is good and needs to be presented before the PR Starts. Hello? R8500 Problems when it first came out? T&L Work arounds for drivers?

Slide was leaked and is real. ATI will do a cover up on it and move on. They are pushing FP24 for a reason and will continue with this product till R500. Thank you. ;)

(You know, I don't mean to jump down your throat, but thats how you come across. Here and Rage3d.)

Hellbinder's great isn't he? He has the Rage3d members all wrapped around his finger, they believe he is some "in the know" guy who has all the industry execs calling him giving him infoZ! Good info too, like when he SWORE the NV40 wasn't a true 16x1 !
Fact is he's just an ATI fan with no real knowledge of the R420 at all other than Inquirer articles. Example of his FUD spreading:

I just learned something that has me twice as excited as before. Launch day cannot come fast enough.

This is going to be great

The rage3d readers lap it all up, it's pretty amusing. Read this whole thread actually:

http://www.rage3d.com/board/showthread.php?threadid=33753558
 
Wow....just wow...and I thought the Nvnews forums were bad. That thread was a celebration of human stupidity. I hope the cake was poisoned.
 
Back
Top