What if x360 used a dual core AMD or Intel proc. instead...
- Thread starter bbot
- Start date
"Nerve-Damage"
Regular
A dual core AMD FX-60 should fair better in some cases compared to the Xenon...IMO.
bbot said:
As always, completely simplifying someone else's statements, what he actually said in that interview was:
JC said:
Which is exactly what he had said in August 2005. If you want to sumarise what he said you can say the four points below:
1) If you use the same code practices a high end PC is faster than Cell/Xenon.
2) If you are willing (and able) to spend time optimising your code for them a Cell/Xenon can be as fast as a high end PC.
3) For some applications Cell/Xenon can still be faster than a high end PC.
4) Even in the first scenario a Cell/Xenon is still better bang for buck than the PC when playing games.
Now, as to whether the xbox 360 would be better suited with a AMD/Intel processor. Firstly I very much doubt AMD has the manufacturing capabilities to be a console CPU supplier. Their latest financial report mentions revenues would be higher if only they had a larger manufacturing channel. Secondly, even if they allowed Microsoft's chosen manufacturers to do the excess for them a Athlon FX is way too generalised (thus expensive) for a console. At around $999 at 1,000 volumes I just don't see them showing up in a console. Even the 4xxx+ models are plenty of expensive, some costing as much as the entire xbox 360 console.
Intel would probably have a better chance but it has worse performance in games than AMD and it also suffers from having a lot of technology a console doesn't need. Intel does have the manufacturing muscle to supply a major console though. But worst of all, I'd really like to see a modern AMD/Intel CPU running at even 3/4 their standard speed on that xbox 360 cooler and not be a fire hazard.
bbot said:
They wouldn't, but the XBOX 360 would have cost a lot more and run a lot hotter.
The x86 processors are undoubtably more powerful on traditional software, but they are significantly larger, Xenon sports 165 million transistors while the FX60 has 233 million.
Maximum thermal power for the FX60 is 110 Watts, though this is a broad ceiling for that line of processors, not just the FX60. I can't find hard numbers for some reason, but it looks like Xenon would be around half that.
Utilization-wise, most x86 cores average around .8 or less instructions per cycle.
The explicit threading in Xenon hopefully would beat this, if the software is optimized for it.
PC developers like Carmack would complain a lot about Xenon, because it does require a lot of extra work by the compiler and developer to get comparable performance to an OoO cpu. These days, few developers in the PC space have much experience with this.
However, they would probably then have to bitch about the other sacrifices, such as a severely constrained system platform that would result if a full blown desktop CPU were used.
Whether time-critical game development can handle extensively multithreaded platforms well is a gamble that won't pay off for years.
Alpha_Spartan
Regular
We all know this. You can't run out-of-order code on Xenon faster than a high-end PC CPU. Hell, you can't even run it faster on a middle of the road PC CPU.bbot said:
The point of this thread is moot since console code is specifically tailored towards ONE hardware specification. Saying that a PC CPU is faster than a console CPU is like saying "A Ferarri is faster than a 2-story Caterpillar dumptruck."
No shit. But then again, a Ferrari doesn't haul rocks. If you want to keep comparing a PC to a console spec for spec, then you can do that with a dunce cap on in the corner. It's comparing apples to celery.
Actually I don't think you could run out-of-order code on anything as it by definition, wouldn't work.
I think it's true to say that XCPU and Cell will have problems with certain data structures and their accompanying algorithms.
Once developers figure out which different or modified data structures to use they'll be able to get better results out of both processors. Ironically, PC processors will also benefit from these changes.
So Carmack's not happy about creating for consoles. Is that such a loss? Is it really worth going with a more expensive and potentially less capable platform just to appease a traditional mindset? As nAo has pointed out, for those who rise to the challenge, being 'limited' means finding better ways to do things. Would Carmack 'waste his time' trying to find an alternative to HDR16 when he can just up the requirements of his game to need a more powerful graphics card?
I say so what to Mr. Carmack! There's many a dev out there who won't shy away from learning new tricks and developing new techniques, who'll blow us away. Hooray the inventors and innovators!
I say so what to Mr. Carmack! There's many a dev out there who won't shy away from learning new tricks and developing new techniques, who'll blow us away. Hooray the inventors and innovators!
Shifty Geezer said:
I dont think he's said that he's not 'happy' in the interviews I've seen. Isnt the 360 his new primary development platform? I think he's only been honest and cut through the hype on the new consoles and been honest about whats easier/more challenging for developing for them. If theres other quotes from him that say something different, i havent seen them. (though i dont deny they could be out there...)
Carmack has admitted the need to move to a more concurrent model, so he's not discarding the need for chips like Xenos for the future of performance growth.
OoO would be nice from a dev perspective, the relaxed scheduling rules are a time saver. Also, the huge pool of experience is still very much based on traditional architectures.
The big question is whether the game industry is going to give devs the expanded times needed for optimization and thread debugging that the next gen is going to need to maintaing the (kinda or really bad already) quality of software in general.
OoO would be nice from a dev perspective, the relaxed scheduling rules are a time saver. Also, the huge pool of experience is still very much based on traditional architectures.
The big question is whether the game industry is going to give devs the expanded times needed for optimization and thread debugging that the next gen is going to need to maintaing the (kinda or really bad already) quality of software in general.
Similar threads
