Read about the design, verification, and making of the 165-million-transistor Xbox 360: a first-pass-good processor with advanced debug and test features, designed to the exacting standards of a gaming platform, with high throughput, low latency, low cost and low power – and very tight deadlines.
The Microsoft Xbox 360 CPU Story
About The Author
Follow me on Twitter @thomholwerda
2005-12-08 8:07 pmAnonymous
There’s design, then there’s implementation of both the individual part, and the system that surrounds it. If the manufacturing process isn’t sufficient, the design (and the exactness thereof) doesn’t amount to a hill of beans. Ever hear of Charles Babbage? He actually designed a computer (primitive by today’s standards, granted) in the 1800’s that was mechanical in nature. It failed to work not because of an incorrect or inexact design, but due to the inability of manufacturing to make the parts to a tight enough tolerance; when built sometime in the 1990’s as an experiment somewhere (can’t remember where) they verified that it did, indeed, do exactly as it was designed to do.
So, you can hold onto your assertions about how bad the processor is in terms of exacting design, etc. but reality is that a perfect design for a single object for a given purpose in a larger system can be blown out of the water by all the surrounding components, including user error (ie. not placing something in an environment with sufficient air flow for cooling). If you want perfection, go to academia. If you want something that will actually be used and released at a cost that’s bearable in a time that’s reasonable, remember that engineers must make tradeoffs somewhere, optimizing for certain variables: engineers are feature accountants, when all is said and done. When you throw something into the mix they couldn’t anticipate (a vendor not quite living up to their claims) then even the best laid plans of mice and men are laid waste by Mr. Murphy again.
2005-12-08 8:23 pmAndrew Youll
hmm… I thought Babbage’s “Computer” wasnt built due to lack of funds and an inability to get the funds toi build it… i maybe wrong but thats what i thought the reasoning behind it not being built.
2005-12-08 10:45 pmAnonymous
That depends. Are the xBox problems being caused by the CPU?
Go back to school and learn to read you dumbass.
I’m not sure about the problems on the x360 (I’ve heard of them), but I know for sure it can’t be just the CPU. When you have so many parts working together, you can have any of them fail (same thing with a reg computer). It’s true that many x360s have been affected, but I wouldn’t pinpoint the CPU as the only culprit. I’ll read up on the story though
One thing that I don’t like about the CPU is it being PowerPC. Not that I have anything against PowerPC…. but I’d expect to see something like Intel coming from Microsoft. I thought that MS and IBM weren’t at good terms… who knows lol
Correct me if I’m wrong though.
2005-12-09 6:02 amAnonymous
I am probably pulling this out of my backside, but I do have some experience with such things, so, here we go:
Since the XBOX 360 crash problem can be solved by hanging the power supply with string it means the problem’s source is at the power supply. I have no clue what the supply is doing when it overheats, but there are really only three things it can do: overvoltage, undervoltage or ripple.
Overvoltage of the supply, if it’s not by too much should simply make the system more stable and perhaps overclockable. If it’s too much, things die, but since the systems aren’t permanently damaged by these crashes, obviously overvoltage too much is not the problem.
Undervoltage and ripple are far more likely as these can cause system failures pretty easily. Undervolting a chip can cause slower transitions (or no transitions) and that can lead to data corruption.
My theory is that the problem is not with the CPU. Since the system does detect a failure and show a special crash (not-quite blue screen) screen, it doesn’t follow that the CPU died. If the CPU had died, the system merely would have halted wherever it was.
From looking at pictures of the crashes, it is pretty clear that there is at least a video memory problem. The vertical corruption lines are common for memory failure, since they’re regular, but not constant failures (since you can see some of the picture still). Since the ATI graphics chip has an on die framebuffer using embedded DRAM, it is my conclusion that this is what is failing. It actually makes a fair bit of sense that this would fail since embedded DRAM is most out of its element. It is DRAM in a logic process, which means that the DRAM is non-optimal. This could very well mean that it is less fault tollerant.
So overall my theory goes like this. Overheating of the power supply causes a droop in voltage (either momentary or sustained, it doesn’t matter). This droop caused embedded DRAM corruption in the graphics. When this corruption happened, a checksum failed, which caused this system to go into error handling mode (the BSOD).
.. is the overheating. But *not* the CPU, but the power adapter. It’s *huge*. It’s actually same size as gamecube and heavier too 😀 altough different shape. Place your Xbox360 power adapter on the cool floor, with some free space for heat to go away. However, console itself is fine on the carpet or so, altough it does heat too *if* you block airholes. Other than that, it’s a bit too much FUD overrating the problem.
” While the term VMX is familiar to PowerPC users, the implementation on the Xbox 360 processor is a new design called VMX128 which was specially enhanced to accelerate 3D graphics and game physics.
The number of vector registers was increased from 32 to 128″
Be interesting to see what the performance increase is like, no doubt it would have been called Altivec 2.
2005-12-09 1:34 pmAnonymous
The performance of the VMX registers is still dependant on the core clock. The advantage of having 128 VMX registers is that more data can stay resident in the CPU instead of having to be written out to memory. As fast as L1 caches are, having the data right there in the registers is waaaay faster.
This is nonsense. All electronic devices have a % of failure. The usual average is 3% and the 360 is said to be well below that. This is a very complex device folks. More complex than any PC IMHO. Furthermore, the biggest problem appears to be the power brick overheating. In some cases user error, and not in others. Regardless, this talk of lawsuits etc. is pure rubbish.
Many of you are right, lots of the people who reported the systems freezing on them, after getting the power supply/brick in a better location, or hanging it up for better cooling etc, said that their systems started working perfectly for hours.
Every first batch of consoles has some problems. But with the x360’s case, it was the US market that got the first batch and thus had to live through this minor problem.
In contrast Sony and Nintendo test bed their systems on their own Japanese market first and then ship their systems months later after all the problems are worked out and shippments are higher.
I could be wrong but I beleave the PS2 also had it’s share of problems, but those systems were only sold in Japan.
I agree with Anonymous completely, every kind of complex electronics or machinery will have problems on their 1st batch of shipments. Even though the labs do numerous testing, there sure will be some glitches overlooked.
As the saying goes: “Nobody is perfect” So i believe that minor problems overlooked are somewhat acceptable as long as the problem can be fixed by the users themselves without huge costs.
If the x360’s problem can be solved by hanging its psu in the air, it is probably its air-flow problem.
Use the console in an air-conditioned room or have a fan face the psu directly should solve the problem.
Just an opinion of mine, please correct me if i am wrong.
“designed” to “exacting” standards sounds nice, but I bet all of the people piling on the lawsuits do not agree, I think there are even a couple class actions out thier for the 360.