bandit

sorry my G and H keys were stuck, lol

Its too early for tis, im oin to te beac...

I am purposely inorin te IBM documents not because tey are in accurate but because tey ave an investment in marketin tem. Please verify wit independent test from developers and prorammers. As it stands now branc prediction is NECCESSARY for complex amin application especially wit particularly branc code.

Lets establish a few things.
ALL ALGORITEMS CANNOT BE PARRELELIZED (meaning if you dont want a performance hit on this code you must use branching.) -Also you claim IBM put a theoritical worse case scenario in their white paper on branch prediction?!? Obviously when IBM details its own limitations they cant be right after all they just designed and manufactured the processor... (rolls eyes)

Anyway I will lay out my 3 main points and ask you to flatly refute them.

1) All Algorithms cannot be parrellelized.

2) Cell does not have branch prediction as good as 360s and that branch prediction is neccessary for complex game code due to point number 1.

3) Cell cannot access the RSX vram. (Tech specs will tell you what can read and write to where in the system)

Prove these things which were the points I was making all along and I will not have anything left to say. Feel free to use white papers if you must but obviously im move inclined to believe individuals makin GAMES on the hardware. As IBM as soooo much experience making games and soooo much success in the past selling macs as gaming machines... Besides the fact that IBM says the exact same thing about the 360 in their white papers as far as gaming applications. White papers are marketing brochures for engineers... informative but still the object is to generate sales i suggest looking up independent studies of cell in relation to gaming applications.


For Fun Let detail a few quotes. Good article you should read it...
Also before you repeat yourself, yes the author does acknowlegde the spes and branch hints but suggests they are nearly useless.

http://arstechnica.com/articles/paedia/cpu/xbox360-2.ars/7

"Rumors and some game developer comments (on the record and off the record) have Xenon's performance on branch-intensive game control, AI, and physics code as ranging from mediocre to downright bad. Xenon will be a streaming media monster, but the parts of the game engine that have to do with making the game fun to play (and not just pretty to look at) are probably going to suffer. Even if the PPE's branch prediction is significantly better than I think it is, the relatively meager 1MB L2 cache that the game control, AI, and physics code will have to share with procedural synthesis and other graphics code will ensure that programmers have a hard time getting good performance out of non-graphics parts of the game....
...At any rate, Playstation 3 fanboys shouldn't get all flush over the idea that the Xenon will struggle on non-graphics code. However bad off Xenon will be in that department, the PS3's Cell will probably be worse. The Cell has only one PPE to the Xenon's three, which means that developers will have to cram all their game control, AI, and physics code into at most two threads that are sharing a very narrow execution core with no instruction window. (Don't bother suggesting that the PS3 can use its SPEs for branch-intensive code, because the SPEs lack branch prediction entirely.) Furthermore, the PS3's L2 is only 512K, which is half the size of the Xenon's L2. So the PS3 doesn't get much help with branches in the cache department. In short, the PS3 may fare a bit worse than the Xenon on non-graphics code, but on the upside it will probably fare a bit better on graphics code because of the seven SPEs."


Anyway Im going to the beach...

you must not understand the content you keep posting

Or Dynamic physics versus environment. Im talking about DYNAMIC. examples such as wind generation or wave simulations are not real-time are not true real-time examples because they can and are reproduced in a lab with out user imput, AI rountines, or collosion detection. on dynamic variables. You cannot find an example of this because the results are dismay.

"This chip is DESIGNED to handle physics calculations. You do not seem to understand the SIMD/vector paradigm. It is hard to understand how you can think to argue that the cell is not designed for gaming. Please do READ the IBM's cell research page."

Trust me I understand but I'd love to see how you will use vector processing in conjuction with AI branching and user input on INTERACTIVE physic environment.

"And again, you stated another false statement: "the total lack of OOOe" --> READ.. The PPE has full branch prediction, the SEVEN SPE's have limited branch prediction. They also have limited out of order support."

WOW!!! This is basic Fundamental Computer ARch Info...!!! OOOE DOES NOT EQUAL BRANCH PREDICTION!!! Branch prediction by definition only skips lines of code as it enters the execution window. OOOE actual dynamically reorders the code as it enters the instruction window which need less to same is a much more intensive process. The branch prediction on the SPEs is weak that it is not consider branch prediction at all. but BRANCH HINTS, its a legitimate term look it up as it applies directly to your precious spes. No need to go into detail if you dont have the fundamentals down.

"And again, you do not seem to understand that the need to NOT use branch prediction, as it can be coded around and compiled around at compile time. Done using the 128x128bit register. You vectorize array any branch prediction scheme... "

This comes from a lack of understanding the basic concepts at play here. IN-ORDER processors such as Cell and the 360 can only process rountines as the enter the instruction window. Therefore in order to prevent process unneccessary code it must skip lines of cod or process everything that enters the instruction window. The work around is to structure rountines at compile time such that they execute in order. The problem is in a sandbox game such as GTA4, dynamic AI, Physics, User input, and an open ended world make compiling in a completely parrellelize fashion impossible. The aresome things Particularly with "next gen" games which cannot be parrelized. Therefore True branch prediction in an in-order processor is NECCESSARY.

"Remember this thing has 7 SPE's and 1 PPE..."

Meaning What? The PPE s the only one with decent interger performance or full TRUE branch prediction. And the PPE has to use a large precentage of it cycles to act as manager for the SPEs with managing resources adn assigning tasks. While they can do some task autotonously the still rely heavily of the PPE cores direction. Where as the 360s cores run independently of one another and have full interger and branch prediction on each core as well as OOOe vmx units. Which the PS3 DOES NOT have REMEMBER: Branch Prediction does not equal OOOe (Computer Arch 101)

"The url you posted does nothing to support your argument that the CELL is not designed for games. Sorry..."

That was never such a link does not exist the point was to educate you on processor arch so that you can make valid points... The cell is not good at certain types of processing which next gen games are beginning to use more and more of. I suggest you quite trying to comprehend these white paper and think about real-world examples of games that your PS3 is suppose to be running....

"The RSX can access the Cells XDR memory, and use it.

This is not a home computer."

Ummm buddy I never said it couldnt... I said you couldnt use it the opposite way. Cell cannot use RSXs VRAM. Thats funny because it looks like you looked it up and realized that what I said was valid all along. Either way I suggest you work on your reading comprehension

@word

Also no cell cannot access vram it doesnt word like that. When is the last time youve seen a computer use gpu ram for system ram? Id love a source on that please

@word

thank you for parroting IBMS research papers now PLEASE tell me what it means in real-world game development
-oh yea you cant

The spes as I explained have limited branch prediction in the form of BRANCH HINTS which is very different from FULL branc prediction. The 360 by contrast has branch prediction closer to IBMs G5 series processors based on Power core. While not the greatest its no slouch either. In addition the 360 as OOOE execution on its VMX units buddy, which utterly slaughters the total lack of OOOe in cell.

While vector processing is powerful no arguement there, BUT there are aspects which cannot be higly parrellized. IE. Collosion detection in combination with AI branching while using real-time physics and alloting room for user input. Individually each of tis acieves a high theorical performance in non-real-time environments but once you essentially mix it together it becomes a coding nitemare were adequeate branch prediction is NECCESSARY and where OOOE would have excelled. Thank fully the 360 as decent branch prediction and OOOe thanks to its VMX units. IF im wrong please post more than a sentence and a link and while we are throwing around links seres a link to IBMs Power4 arch which both cell and 360 are built around.

http://www-03.ibm.com/servers/eserver/pseries/hardware/whitepapers/power4.html

Ok here we go again...

Cell is a ridiculously parrellel processor. It has almost not branch prediction on its spus and its single ppu general processor acts as a controller for both the spes and ram. example cell excells at physics but not real-time. (depending on the game) for instance in GTA4 if a pedestrian is it by a car rag-doll physics may come into play. Ok but then a second car driven by an NPC usin AI occurs with a second break in the already in progress ragdoll animation. And the AI must respond to this. Cell will $#*% itself it does not have the branch prediction capabilities nor general processin to handle that. Branch prediction basically allows the processor to "predict" a rountine thereby skippin processin cycles where poor branch prediction causes the processor to throw out the current process and reprocess which is dramatically slower and inefficient. The 360 cpu is lite years ahead in this regard and it will show in game play. Cell is excellent for vertex renders. Games like GodOfWar are perfect as they have linear simplistic play and button mashin with canned animations and pre-scripted simplistic AI rountines.

What about storage?
It doesnt mean $#*&. it just means you can ship with more crap on the disc and increase dev costs. IT WILL NOT CREATE LARGER MORE DETAILED WORLDS!!! This is a limitation of the SYSTEM RAM. PS3 only has 256MB system RAM and cannot use vram for general ram. 360 by contrast can allocate it where it needs it. The announcement the vram can use system ram too was for fanboys as no dev would ever dip into that small ram space for pretter graphics. it just doesnt make any sense to unless your game is overly simplistic. What this means is that at anytime the total amount of crap in the world must be less than 256MB. PERIOD. Sure you cache to HDD but anything available on screen in the world still has the same limitation the HDD isnt ram its just a faster storage medium than disc. im not even gonna touch the fact that in order to cache to HDD it must travel thru system ram first as the biggest issue is bandwidth and im not even gonna touch that and its poor excuse for a frame buffer....

If you people would take the time to think for one nanosecond you would realize that between 4 graphs you have collectively posted over 250 comments, making this a successful series. That being said, if you F-ing idiots want it to go away then STOP F-ING POSTING COMMENTS!!! Im not saying anything good or bad about the series itself, but for the idiots who continue to complain there is a simple solution. Stop posting, ignore these posts, and you will never see them again. Dont feed the bears. God, you people are a bunch of F-ing idiots...

wow, MS is truely dry-humping the hell out of Epic's new IP. Let's hope the GOW is as much of a hit as MS hopes or they will surely look like complete and total asses... Imagine if the same amount of hype had surrounded Brute Force... including advance play test in movie theaters. That would have been a total disaster. For MS sake I "prey" GOW lives up to all the hype its causing. Personally, Im sticking to the idea that it will be no where near as good as everyone claims so that I will be pleasantly blown away when it turns out to be in instant classic.

While I'm here on my soap box I have a suggestion for MS. Instead of lowering the price of the 360 add 1 year of xbox live and 2 "platinum" games (a $110 value) and we all know it will cost even less than that to bundle those in. That way you look like a true value package when compared to the competition, PLUS you get to hook all those people still holding out from xbox live. Its win, win baby!

In addition I would like to submit the game Gunz: The Duel to be reviewed by the Joystiq staff. The English version is still in beta but set to go live next month. In the meantime anyone can join for free and the game is 100% free. Awesome game, try it out http://www.gunzonline.com/

this is just my opinion but no one person should know this much information about chocolate...

one more addition, an ad supported version of xbox gold, and ad supported versions of arcade games with the option to buy, what i mean is offer Street fighter for $15 dollars with an ad supported version which shows 10,20, or 30 second commercials between matches. Then use that same model for episodic content, such as the upcoming SIN series, which could use a combination of ingame and out of game ads to offset to price for those WHO CHOOSE not to pay the premium price.