PSX5Central
Playstation/Gaming Discussions => PS3 Discussion => Topic started by: fastson on October 01, 2002, 09:16:10 AM
-
http://www.gamasutra.com/gdce/2002/jason_doig.zip
You\'ll need PowerPoint!
-
Isn\'t there anyway you could transfer the info from powerpoint.. so that we can simply read it here? Or is there too many?
:)
-
It wont be that interesting without the slides :shy:
Wait.. Its 47 000 something letters.. Ill put them in a .txt file.. Hold on
http://w1.480.telia.com/~u48019550/host/Sony-PA.txt
-
It seems thare were some pics with it showing what he is explaining.
In TXT we cant see them.Is there a way you can post a different file with them?
-
No :(
You must have PowerPoint to be able to view all the slides including the pics with the PA2.
-
Downloaded the powerpoint one too.Works afterall.Thanks for the info fastson.
-
fastson, what\'s the date of this lecture?
I only read the text, which was interesting enough for me without the graphs. While it confirms the PS2 has a great deal of potential left to be exploited, it also re-establishes the extreme difficulty in doing so and suggests getting 80 or 90% efficiency (my estimate) within the next few years may be out of reach for even the most talented developers. A lot depends on when the talk was given.
Not that it matters much, I\'m satisfied with the higher end results to date. Though I would feel better if smaller devs had access to, say a 50% potential, without too much hair pulling/trial and error effort. The aid of C compilers seems very limited at this stage. Again, if this lecture was given a year ago there maybe room to be more optimistic.
-
Heretic: Im not sure.. But I think its pretty new.
As he said "the PA has been avalible since April" so its from this year.
Anyways I hope the PA2 can make the same huge difference on PS2 as it did on the PSX.
Remember GT1? :)
-
If I remember right, a PA for PS2 was available long before J&D was ever released, though Naughty Dog never used it. The speaker does mention the one being used for the slides is an updated version. So I dunno. It was just this spring Vector C for PS2 became available, right?
You know what though? Screens of much improved PS2 graphics have been popping up lately. SH3, the F1 racer... Vice City seems to be a huge leap from GTA3...
Yeah, the PS2 still be cookin\' in my book
-
He says in the article (or whatever you want to call it) that its been avalible since April, but only at Sony HQ.
So if you wanted to use it you would have to go to Sony to test your code.
Im not sure about it being avalible before J&Ds release..
I remember the TimeSplitters 2 developers saying that they had been using the PA2 for a few months to optimize.
Im not sure when VectorC was released but now it seems they have gone belly up :(
http://www.codeplay.com/
Screens of much improved PS2 graphics have been popping up lately. SH3, the F1 racer... Vice City seems to be a huge leap from GTA3...
Exacly, I was blown away when I saw the new SH3 screens and that new ZOE2 video.
I cant wait to see what the code gods at PolyPony can cook up for GT4, or what Naughty Dog can do with J&D2 (prays for a sequel) ..
The guy in the paper said he was very impressed by their skills.. :)
Here he\'s talking about J&D (cuz to my knowledge J&D is the only game crunching this many polygons!)
Ok, so just to finish up I’ll show you again the scan from the very first slide. This is a title which achieves a lot of stuff I’ve talked about today. Notably, for most of the time it’s drawing stuff it’s churning out 10 to 20 million polys. That’s no a random benchmark, that’s physically what the GS starts drawing and like I said during the PA description, it doesn’t count zero-area stuff.
Also, the CPU usage is up at about 50% or more, which is more than double what an average title might see, and better than most good titles by a decent margin. It’s also getting dual-issue for about 80% of the time during the critical processing phases (about half a frames worth).
So good performance is possible, and not only that, but this title was written without the aid of the PA, only using the built in performance counters and a lot of skill. It’s also not perfect and I fully expect to see this developers next title beating this performance on every level. Hopefully you guys can do the same.
-
Quite a long read but interesting all the same.
-
Im not sure when VectorC was released but now it seems they have gone belly up
That\'s a shocker :surprised
Perhaps if this PA paper/lecture is recent it was the kiss of death for VectorC. If it was any good I can\'t see Sony letting Codeplay just go away :confused:
Thanks for passing along all the info fast , as usual :cool:
-
Yeah, it would be pretty cool if Sony could pick this VectorC up.
All to help the developers get the sweet juices out of PS2 ;)
-
Wow thanks alot for this great info Fast. I was away for 2 weeks (in Malta), but am back again and look forward to some serious posting. I\'ll make sure to comment on the PowerPoint presentation once I\'ve read it.
Ahh, just saw it\'s downloaded... better check it out then. :)
-
I cant wait to see what the code gods at PolyPony can cook up for GT4
Perhaps they could make the game look as good in 2 player as it does in 1 player. :mad:
Then again, people play GT for the single player, not the multiplayer :)
Here he\'s talking about J&D (cuz to my knowledge J&D is the only game crunching this many polygons!)
J&D is only pushing 10 million pps (with information). GT3 is pushing up to 20 million.
In fact lots of games on PS2 are using over 10+mpps
-
Originally posted by Bobs_Hardware
Perhaps they could make the game look as good in 2 player as it does in 1 player. :mad:
Then again, people play GT for the single player, not the multiplayer :)
Right-o.. I think Ive played splitscreen once or twice.. Not fun at all IMO.
J&D is only pushing 10 million pps (with information). GT3 is pushing up to 20 million.
In fact lots of games on PS2 are using over 10+mpps [/B]
J&D is somewhere around 10-15mpps IIRC. (I think Jason Rubin said it goes as high as 15mpps at times but it stays at around 9-10mpps most of the time)
GT3 is not pushing nowhere around 20mpps. PolyPony (Kaz Yamauchi) has said their engine is capable of pushing 20mpps but that does not mean the game is doing so.
Compare the landscape of J&D to the "tiny" (in comparison) tracks in GT3. (Plus 6 cars at around 5000-6000 polys) ..
J&D and R&C are prolly the games that are pushing the PS2 the hardest now.
Btw my friend e-mailed the dude who made this little slideshow and asked him which games he was using as an example. He said he was not allowed to say because of some confidentiality thingy, so maybe its a new game? :D (Criterions FPS comes to mind.. and mebby Project Z (J&DII?) from ND? )
-
I am still waiting for that game that will feature random generated movement to the full(ICO does it in some extend) and random generated polys and textures(for less repeated objects and patterns).
-
Fastson, yes J&D pushes the PS2 more, but GT3 uses more polygons. Polyphony stated that the engine pushes up to 20mpps, but it would reach that point rather infrequently. Which means it CAN be reached. And if it isn\'t I would expect it to be quite close to that number.
And Burnout 2 looks better than GT3 IMO
:)
-
Originally posted by Bobs_Hardware
Fastson, yes J&D pushes the PS2 more, but GT3 uses more polygons. Polyphony stated that the engine pushes up to 20mpps, but it would reach that point rather infrequently. Which means it CAN be reached. And if it isn\'t I would expect it to be quite close to that number.
And Burnout 2 looks better than GT3 IMO
:)
I dont see how GT3 can use more polygons.. Only 6 cars (5000-6000 polygons each) and a pretty sparse background/landscape (the only really impressive tracks are the Tokyo track and Seattle (polygone wise)).
The engine is capable of pushing 20mpps but the game does not need that many polygons. Thats how I remember it :)
I still need to get/tru Burnout 2, cant wait :)
-
Originally posted by Bobs_Hardware
......but it would reach that point rather infrequently. Which means it CAN be reached. And if it isn\'t I would expect it to be quite close to that number.
:)
It can be reached but it never reached.
-
Originally posted by Bobs_Hardware
Fastson, yes J&D pushes the PS2 more, but GT3 uses more polygons. Polyphony stated that the engine pushes up to 20mpps, but it would reach that point rather infrequently. Which means it CAN be reached.
AFAIK, those 20mpps are just textured polygons, but when effects are applied it\'s more like 6-7mpps at most.
As for the PA2, well it seems to me that the new tools and documentation released from Sony (and others) are doing their job quite nicely. I expect that now developers will be able to spend more time making their games more fun to play or developing some original content for the PS2, rather than fighting with the system.
That being said, it should\'ve been done some time ago.
-
Who is to say should\'ve been done some time ago? It would\'ve been ideal to have clearer documentation and key tools sooner but the negative impact hasn\'t really amounted to much as far as Sony\'s bottom line goes.
-
J&D is only pushing 10 million pps (with information). GT3 is pushing up to 20 million.
I don\'t hold any information on the numbers, but AFAIK it\'s a known fact that J&D is pushing far more polygons than GT3 by a large margin. If I recall correctly, GT3 is pushing around 7 million polygons/sec maximum (that information goes back to a interview with Yamauchi). Also, Naughty Dog is the one developer pushing the PS2 most at the moment, although Rachet & Clank despite the blury textures might come damn close. What they did with J&D is incredible, as the engine has to handle 100 millions of polygons and convert them to something between 10 to 20 million when sent to the GS for rendering. Something like this hasn\'t even been achieved on PC yet...
As for the question which game the PowerPoint presentation is refering to: I think it\'s a good bet that he\'s refering to J&D, as at an earlier point, he says that people should check out Naughty Dog\'s presentations on their development and work for some really good stuff.
BTW; very good read... thanks again for the info. I especially enjoyed the part on how he gives tips to keep it all "balanced" - and it also gives quite a good view on where the advantages are and the disadvantages of the EE. That can only help improve games significantly. If Silent Hill 3 is any indication (I\'m still a bit sceptical, but if this will look this good, we have a Doom 3 challenger in the graphics department), then I can\'t wait to see what Naughty Dog might present us next, or what GT4 may look like.
:bounce:
-
See, I knew I was wrong but I still kept arguing..
Seven, you\'re so cool *hugs*
-
Originally posted by Falgarok
AFAIK, those 20mpps are just textured polygons, but when effects are applied it\'s more like 6-7mpps at most.
As for the PA2, well it seems to me that the new tools and documentation released from Sony (and others) are doing their job quite nicely. I expect that now developers will be able to spend more time making their games more fun to play or developing some original content for the PS2, rather than fighting with the system.
That being said, it should\'ve been done some time ago.
Huh?? Where u get that information about 6 to 7 million pps? I believe it\'s 20 million texture and lighted according to most sources.
-
What they did with J&D is incredible, as the engine has to handle 100 millions of polygons and convert them to something between 10 to 20 million when sent to the GS for rendering. - seven
Erm.. didn\'t Sony already announce (repeatedly) that the PS2\'s maximum capabilities are around 78 million untextured/unlit polygons per second.. with the number dropping down to anywhere from 20 to 25 million pps when in the context of an actual game enviroment?
Yeah, we\'re talking a 20 million polygon per second difference.. but still.. 20 million polygons per second is an awful lot and a far cry from 100 million pps. How can Jak & Daxter handle over 100mpps (raw) when the PS2\'s preformance doesn\'t even match that number?
-
Erm.. didn\'t Sony already announce (repeatedly) that the PS2\'s maximum capabilities are around 78 million untextured/unlit polygons per second.. with the number dropping down to anywhere from 20 to 25 million pps when in the context of an actual game enviroment?
Yeah, we\'re talking a 20 million polygon per second difference.. but still.. 20 million polygons per second is an awful lot and a far cry from 100 million pps. How can Jak & Daxter handle over 100mpps (raw) when the PS2\'s preformance doesn\'t even match that number?
Well, to be precise, 66 mpps is only the limit of VU1, but that\'s besides the point. What the game does though, is significant:
The whole world (or island) in Jak & Daxter is being rendered at once. You can go off to anyplace and you\'ll see everything unchanged. This is only possible though by simplifying objects that are further away. In other words, imagine yourself in the town looking at the house in all its glory. Go a few miles up to the jungle and you\'ll have quite a nice view on to the town you were just moments before. Upon looking down at the house you just looked at, it seems to be unchanged. What the engine did though is simplifying that object from say a few thousand polygons into something that looks just the same, but using only a few polygons. The clue is, no difference is seen as the house is too far away to make out the small details anyway. Following up on that, you do have an engine performing calculations on 100\'s millions of polygons (I\'m taking a guess here), but in the end, you are only rendering something between 10 to 20 million/sec. So, what I did mean is not that the PS2 is rendering that many triangles per second, but that the engine has to handle and manage that many, while only rendering only what\'s necessary.
I remember an interview with Jason Rubin a while back explaining this in more detail (and about the sheer number of polygons the engine has to handle). I\'ll look for it later on.
Anyway, be carefull when talking about the specs Sony released. Sony mentioned two specs (for EE and GS independently); one is refering to perspective transformation performed on VU1 (66 mpps) and the other for the GS\'s rendering performance (75 million, 32 pixel, untextured, raw polygons/sec). This number though is only under certain circumstances, i.e. you could surpass this number by running the VUs in micromode devoted to perspective transformation and rendering smaller triangles on the GS.