What if x360 used a dual core AMD or Intel proc. instead...

[j^aws]

...
No man 1MB is sufficiënt but like you said above it takes optimising.
Typical PowerPC can cope with 256KB L-2 per core no problem.
It just takes some effort.

Typical PowerPC CPUs are OOOe designs that hide latency better than in-order CPUs, so they can also get away with less cache...

 

[Titanio]

I know what you're getting at but I've also read the Xenon can do fluid dynamics but it just needs some more effort.
And if it can't the Xenos will jump in.

It's not a binary on/off thing, there are levels of performance. Just a guess, but Xenon and a PhysX chip will likely generally be in a very very different ballparks as far as physics go (if PhysX lives up to its promise).

 

[Blazkowicz]

saying a PC PCU costs $1000 is not fair.
high end CPUs have always been overrated crap, they are about paying 200 or 300% more for 10 or 20% increase in clock speed.
A X2 3800+ is still expensive (more than 300 euros) but you have to factor in that it competes against its single core counterparts, AMD couldn't sell its single cores if dual core were too cheap.

 

[ADEX]

AMDs and Intels are designed to do everything equally well.

XCPU (and Cell) are tuned to do some tasks very, very fast, the downside of this is they will do other tasks slower.

They will run branchy code slower, but on liner code (e.g. Linpack) they'll both get very high scores.

What developers need to do is figure out how to get the problems the branchy stuff solves to use code that acts more like linpack.

--

BTW modern x86s are nothing like 386s.
Comparing the XCPU / PPE cores to VIA C3s or 486s is equally silly.

 

[Guilty Bystander]

AMDs and Intels are designed to do everything equally well.

XCPU (and Cell) are tuned to do some tasks very, very fast, the downside of this is they will do other tasks slower.

They will run branchy code slower, but on liner code (e.g. Linpack) they'll both get very high scores.

What developers need to do is figure out how to get the problems the branchy stuff solves to use code that acts more like linpack.

Cell and Xenon don't have to run Windows or some other inefficiënt OS and are good in things they have to do.
It doesn't matter they're not good in things they don't need to do.
By the way a 2,4GHz 6SPE Cell won from an Intel P4 3,2GHz in Wintel.
Article is floating arround on the forums somewhere.

Xenon is not that bad in branchprediction where in Cell's case it doesn't matter considering the SPE's work with local storage memory.

They need to optimise and I don't mean optimise like what Ati and nVidia do with their PC GPU drivers leaving things out and other driver tricks.
I mean developers need to tailor their code to run at maximum sufficiëncy on both cores meaning Xenon and Cell.

BTW modern x86s are nothing like 386s.
Comparing the XCPU / PPE cores to VIA C3s or 486s is equally silly.

You can't compare a 3,2GHz 386 to a 3,2GHz Amd 64 or P4 obviously but in escense it's architecture is the same.
PowerPC's on the other hand have really nothing in commen with the Via C3 or 486 besides both being processors.

 

[London Geezer]

Well have you even played Quake 4 on the Xbox 360?
Cause it seems like you didn't otherwise you would know what I'm talking about.
He knows more of programming than I do that's true but then again I'm no programmer.
Other programmers know a lot more then he does and he's clearly overated and I've got PGR3, Kameo, Condemned and King Kong to proof my argument.
Games which all look better and run better and I could have compared Quake 4 to GoW, N3, FN3 and BF2: MC but I decided to just compare it to other launchgames.

Can you read?

I even bolded it.

What did he screw up with Quake 4?

As in, what did he personally do to screw up Quake 4? Attacking an individual is ALWAYS easy, especially when you have no knowledge of what you're talking about.

 

[Guilty Bystander]

Can you read?

I even bolded it.

Can you even get your fingers to do some informative typing?

As in, what did he personally do to screw up Quake 4? Attacking an individual is ALWAYS easy, especially when you have no knowledge of what you're talking about.

This is without a doubt the dumbest sentence ever.
And c'mon enlighten us with your wisdom cause I've never seen you post anything informative and from responses like these I think we'll never will.
How you became a B3D role model I'll never know.

But what Carmack screwed up on Quake 4 is easy he had insight on the project and should have postponed the release from taking place.
If the great Carmack said so it would have happened.

 

[Hardknock]

Guilty, the port of Quake 4 to the 360 was not done by Id. And further more was rushed to make launch. There was nothing Carmack could have done about it.

 

[London Geezer]

Thanks Hardy.

How you became a B3D role model I'll never know.

Uhm... I changed the tag myself, dumbo!

 

[ERP]

Spagetti code will run better on OOOe processors than in-order processors. Optimising spagetti code will benefit both types of processors. However, it will benefit in-order processors more. Adding OOOe support costs more transistors and it was traded for more ALUs for it to have legs over 5 years in a closed console design.

I want to make this clear it isn't about "spagetti code" and not "Spagetti code", it's about any code.

To get good performance out of the X360/PS3 cores the compilers simply will not get you there without significant help.

Sony and MS traded the complexity of OOOe for more ALU's, in trivially parallel problems where a programmer can write code to work on more than one element at once you can exploit them to some extent. In the other 90% of code it's a lot harder.

Someone with the Cell simulator should run some simple benchmarks between the PPE and a conventional intel/AMD platform, you might get some idea of why developers like Carmack mention significant work.

 

[Edge]

You're right, and I think this type of comment should be banned. It contributes nothing to the technical discussion and can't be substantiated anyway (unless they keep a record of everything everyone's said!). As per the FAQ, if what you're writing doesn't contribute to the discussion, don't say it.

Pointing out a lack of consistency in the logical arguments around here, is good motivation for people to explain themselves better and helps contribute to the discussion. Anyway, I was more or less just teasing a bit, and hope we can have a bit of harmless fun here also. Maybe we get too worked up on this whole Product A is better than Product B discussions.

 

[Edge]

london-boy I don't agree with your numerious cut-downs of Guilty Bystander. He presented some numbers, and they are just fine, and open to discussion. If you disagree with the numbers, then say why, instead of saying he does not know what he is talking about.

The problems with Quake 4's frame-rate on the Xbox 360 has been mentioned in quite a few places, so Guilty Bystander's criticism of the game, and the person behind it, John Carmark is quite valid. Quite a number of people have been disappointed in this port.

 

[scooby_dooby]

Why don't we wait until Carmacks real console game comes out before insulting the guy, it was a crappy port by raven.

Lets see what they do with Castle Wolfenstein, that will be a much better indication of Id's skills on consoles.

 

[London Geezer]

The problems with Quake 4's frame-rate on the Xbox 360 has been mentioned in quite a few places, so Guilty Bystander's criticism of the game, and the person behind it, John Carmark is quite valid. Quite a number of people have been disappointed in this port.

The whole point is that John Carmack was not behind it! It was developed by Raven Software. John Carmack has a lot better things to do that porting Q4 to X360.
I could be wrong obviously, and if i am i apologise, but i'm quite sure that John Carmack was not behind anything in an active way, therefore attacking him is out of line.

 

[Shifty Geezer]

Indeed, for all GuiltyBystander knows Carmack asked for the port to be postponed and improved but the publisher financing the development said no. If Q4 on XB360 was a private development, headed by Carmack and financed by Carmack and answerable solely to Carmack, GuiltyBystander has a point. If other people are involved in the decision making process (which they were), it's ridiculous to take the one console conversion as evidence Carmack doesn't know what he's talking about.

 

[pegisys]

I don't see the point in putting down Carmack for what he says, if he is programming for the 360 now and going to the metal, he may come out with one of the better looking games nextgen.

I think the point is that you have to put in lots of extra time with the new cpu's that the consoles use to get (better, comparable, ?) performance, which takes more time, which end up costing alot more

Id is a small company and it seems he wants to keep it that way, so I think his biggest concern is cost, and dev time

 

[London Geezer]

Indeed, for all GuiltyBystander knows Carmack asked for the port to be postponed and improved but the publisher financing the development said no. If Q4 on XB360 was a private development, headed by Carmack and financed by Carmack and answerable solely to Carmack, GuiltyBystander has a point. If other people are involved in the decision making process (which they were), it's ridiculous to take the one console conversion as evidence Carmack doesn't know what he's talking about.

Thank you.

 

[j^aws]

I want to make this clear it isn't about "spagetti code" and not "Spagetti code", it's about any code.

To get good performance out of the X360/PS3 cores the compilers simply will not get you there without significant help.

Sony and MS traded the complexity of OOOe for more ALU's, in trivially parallel problems where a programmer can write code to work on more than one element at once you can exploit them to some extent. In the other 90% of code it's a lot harder.

Someone with the Cell simulator should run some simple benchmarks between the PPE and a conventional intel/AMD platform, you might get some idea of why developers like Carmack mention significant work.

The tradeoff has obviously brought it's own challenges of multithreaded coding and early compilers. But the whole industry is moving towards multi-threading and you'll eventually have the same issues in the PC arena. You guys will be pioneers in many respects but in 5 years will you think the trade-off was worth the effort...?

 

[Edge]

I know the game is by Raven, but iD's logo is on the PC version, and wondering if it is on the X360 version? What is iD's relationship to Raven? Any quality assurance going on by iD?

 

[ERP]

The tradeoff has obviously brought it's own challenges of multithreaded coding and early compilers. But the whole industry is moving towards multi-threading and you'll eventually have the same issues in the PC arena. You guys will be pioneers in many respects but in 5 years will you think the trade-off was worth the effort...?

I actually disagree with this, the issue isn't "multithreading" , the issue is in order execution vs out of order. I suspect PC's won't ever forgo the OOOe on their path towards parallelism, in that arena it would be totally none sensical decision. In the console space it's justifyable to some extent.