IIRC the design rationale behind Cell was stream computing, i.e. processing huge amount of unique data. Then, as Mike Acton puts it, code is data.This sort of makes sense, because AI often has a crapload of code that rarely gets executed (on a per-line basis), whereas SPU optimization makes most sense for small amounts of code run repeatedly with different parameters and inputs. Occasionally AI will fit into the latter, but often not for games of this scope.
Unless you're modelling behaviour for 10,000 stupid fishwink
This seems to be a point that lots of posters in this thread are missing. Multithreading alone isn't the problem with Cell, nor is in-order processing. It's the memory model that is a pain in the ass, and that's what Barbarian, Gabe Newell, and others are pissed off about since it needs investment that won't pay off anywhere else.
*Game Development Issues*
- Thread starter valioso
- Start date
- Status
DuckThor Evil
Legend
But I got two games and one extra controller too with $50
The controller has real value, but I'm not so sure about those two games... I'm sure that in most places you can get them dirt cheap as a second hand games, or can borrow them from a friend. Granted 50 pounds is not much for that package. In Finland the difference is 120€ So the 60GB was never an option for me and I like my 65nm Cell processor also. 20GB extra space is pretty meaningless if you ask me. If 40GB is too little you can upgade it into 160GB or 250Gb.
I think we went a bit off topic...
This sort of makes sense, because AI often has a crapload of code that rarely gets executed (on a per-line basis), whereas SPU optimization makes most sense for small amounts of code run repeatedly with different parameters and inputs. Occasionally AI will fit into the latter, but often not for games of this scope.
Unless you're modelling behaviour for 10,000 stupid fish
This seems to be a point that lots of posters in this thread are missing. Multithreading alone isn't the problem with Cell, nor is in-order processing. It's the memory model that is a pain in the ass, and that's what Barbarian, Gabe Newell, and others are pissed off about since it needs investment that won't pay off anywhere else.
I don't think anybody is arguing against the cost of redesigning existing codebase, nor the "difficulties" of fairly limited local memory model. That doesn't mean it's not beneficial. Memory locality is very important for any high performance computation. Of course if Gabe Newell is satisfied with what they are achieving on other architectures, it's only natural for him to avoid any further investment.
Regarding AI though, I don't know what 100(?) AI programmers come up within AC, but I'm having hard time believing framedrops (or tearing) has anything to do with large chunks of poorly written ITEs.
It looks more of a hear-say response to me.
ps: What's up with BC discussion here? Is there any relation between BC and current-gen development that I'm not aware of?
There aren't 120 devs working on the "360 version" in any team, on any game. There are probably 120 devs - programmers, level designers, artists - with Xbox360 test kits on their desks, working on the general, non-platform-specific tasks which take up 90% of game development. Xbox 360 test kits are cheap, available, and perfectly usable for this kind of work, while "full" devkits for both consoles are much more expensive (and the PS3 ones are huge and noisy, at least before the last revision - but the price problem remains). Then you have a handful of guys doing the platform-specific parts and optimization. If they really had 15 people on the PS3 version, this is a huge PS3 effort, and much, much more than most games will ever devote. (Maintaining a decent PC version doesn't take more than 1-2 programmers.)
Also, if you intend to use it a lot, then depending where you live, you could earn back the money on less electricity used alone probably in one year, and certainly over 5 years
Yes, and be more ecological friendly… but our people really care of the ecological impact of their entertainement system (large HDTV start at 200w, console HD more 150 w, great 5.1 or 7.1 system around 250w, so 600w/h), people don't see this, it's difficult to associated entertainment with some restriction.
archangelmorph
Veteran
15 programmers on a PS3 game in the scope of assasin's creed isn't anywhere near "a huge PS3 effort"..
& 1-2 programmers to maintain a PC version for such a game is a little on the short side..
I'm my team there's 3-4 of us currently & we're only doing a PSP game..
Mintmaster
Veteran
Of course it's important, but is it important enough to warrant a big performance decrease for unoptimized code? For a lot of game devs the answer is no.
Of course. These decisions are made using some form of cost/benefit analysis. SPU perf starts off low for generic code, so the benefit is big enough to warrant the cost. It would be better if that cost wasn't there.
This is an important point, because a lot of people say such an optimization is not only beneficial for Cell, but also other CPU architectures. If it was worth the cost on other architectures, then people would already be doing it. Such a line of reasoning is fallacious because the cost has already been judged to exceed the benefit. Cell isn't doing anyone favours by making them write "smart" code.
Yeah, that's the reason I said it "sort of" makes sense. Who knows if its actually true or not.
They are not "programmers on a PS3 game", they are "PS3 gurus responsible for the PS3 version", and I still insist 15 is waaay too much.
If you are working on a PSP game and you are 3-4, probably most of you will have direct contact with the platform - but if you grow to 30-ish, again only 3-4 people will work directly with PSP devkits, the rest will probably work only on their PCs; besides, the PSP is much more constrained and there's little place for platform abstraction.
archangelmorph
Veteran
How in the world did you get that?
There doesn't seem to by any inference to there being "15 PS3 Gurus with possible other programmers working on higher level stuff" at all so I'm not sure where you pulled that from..?Jesus2006 said:
It's likely there are only 3-4 expert senior's on the platform version (1 per 5 man team) if that, with the other 11-12 working on general porting/grunt work.. 15 "gurus" as you put it is a rediculous estimate & you'd likely be under-using most of their expertise considering alot of the low level work would be very focused and require only 1-2 individuals per sub-system to design and implement in a well organised/managed team..
???
I like the way you're telling me how our development strategies/team organisations are implemented!!!
(Seriously some people around here are too much!!!)
At present there are only 2 of us that have done any really low-level work since our engine was ported (from PC, PS2, Xbox360, PS3 & DS..), there are also several artists and designers on the project and one producer, ALL of which have devkits.. Also like I said our entire engine is multiplatform and thus, abstract completely all of the platform specific/dependant code from rendering to networking to handling audio/video playback etc..
Besides that, yes the PSP is much more constrained as a platform than say the PS3 but mostly in the focus of what we're doing with it.. For example we could port our physics code over to the PSP engine but such work would be unnecessary since the game we're working on doesn't need it.. Also i've worked on a 30+ team on a PS2 game prior to this & at most we had 12 programmers with an average of around 7 across the entire durartion of the project.. & this wasn't a "ground-up" project either being a sequel which used much of the existing code base..
Don´t you think it will become more important and more worth the cost in the future? I mean when the games on the 360 is starting to stress the CPU harder and cache trashing becomes more of a problem the benefits of better data locality will be visible in the 360 games as well.
Maybe that is one of the reasons behind Midways decision to make the PS3 the lead platform even though they are not saying it explicitly?
I know what you're getting at and you're right that there are some nigh obsolete games which tend to crush and all but, with that in mind, expert PC users can run the oldest software.
Sure, some legacy software might have a hard time running on modern PCs and OS (nothing unfixable, I think), but anything relatively recent should run without problems.
There are people who see some old PC games as cult objects. I've never been much of a PC gamer but had I to name a couple of games, I'd choose Age of Empires series and Total Annihilation, plus Doom.
BC is great, sometimes the nostalgia comes out and you feel like playing old games, perhaps not like in the past but just for some minutes. Also, classic gaming appeals to some people because for whatever reason the prefer 2D games and define 3D games as consolish and *for kids*. That's not an exaggeration, ShootMyMonkey.
I understand you, though, I imagine every Creature's games out there are BC in the PC world and they make you remember a time when you were unhappy, and the ordeal.
Anyways, back on subject, I'm getting the idea that, currently, perhaps the PS3 is not an easy machine to develop for let alone making it fully BC without a carbon copy of the PS2 and PSX chips in the innards, even with the best tools ever.
Cheers m8
I really don't know what to make of that, except everything I've seen the 360 looks on par or edges out the PS3 version by a TAD.
What I make of it is that it took their 'A' team to make the PS3 version comparable to what their 'B' team did on the 360 version. Worded another way, it takes alot more money to make the same game on PS3 compared to 360. It would have been interesting to see what their 'A' team would have accomplished on 360.
Indeed. It's curious that they're pushing very hard in a number of places the view that the PS3 version is the superior version, when that's not the case, the 360 version is better (granted there's pretty much nothing in it you'd notice unless you had a magnifying glass).
Probably just a business thing, however that doesn't really make sense when the 360 had the exclusive beta. Maybe just making a strong effort to ensure their "best programmers" feel valued?
djskribbles
Legend
some textures appear to be a little better on 360 and some on PS3. shadows seem to be of higher res on 360. but the PS3 version has much better texture filtering. i don't really see how you can say either version is better... i'm sure there are many things throughout the game that are better in one version over the other, but i don't see how PARANOiA came to the conclusion that the 360 version is better. all we've had to compare are some short clips from gametrailers and some screenshots by some guy from neogaf... hardly enough to come to any conclusions IMO. on the other hand, some people from IW are saying the PS3 version is slightly better.
I just get the feeling that Activision is using the situation to score some points with Sony. They know the 360 version is going to sell well and they are trying to protect their PS3 sales by going out of their way to point out how the PS3 version is on par if not even better. I think it's a smart move, especially if they end up getting some more love and support from Sony.
- Status
Similar threads
