Xbox overclocking????

JustHalf

These are synchronous circuits, guys. Everything is synchronised to the clock. Data is only accepted as processed on the clock pulse.

So you need to speed up the clock to speed up the chip.

po0k

The funnyt thing is this is basicly an under-specced PC with a streamlined OS( Win2k core afaik)
If you were to load it with the full Win2k setup it would croak.
Tis a piece of pish imho....though maybe that's just me being a PC user

astrofool

Originally posted by DeadBankClerk



if you overclock something, when it runs it gets hotter than it did prior to overclocking. Overclock it enought and it will melt. Slap a big fan on and you can overclock more without it melting.

the bigger the fan the more you can overclock.

the upper limit due to impurities in the silicon is reached well before temperature becomes the main concern.

The kryotech systems cool it down to -40 and keep it there, yet a 50% overclock is about as high as you go with them.

suppafly

Originally posted by Inspector Gadget
Sound familiar yet? Yup, it's a PC in console's clothing.

Yup thats its a smaller versin of a PC made for pure gaming!

po0k

I think I'll be sticking with my rig for the meantime, then build another new one in the summer/autumn and setup a good LAN at home with me WAN connection

cerebus

Originally posted by JustHalf
These are synchronous circuits, guys. Everything is synchronised to the clock. Data is only accepted as processed on the clock pulse.

So you need to speed up the clock to speed up the chip.

I guess it is a question of semantics - speed vs. throughput.

Lower temperatures mean silicon is faster. Faster silicon means faster devices (transistors and gates)

As has been pointed out, these are synchronous systems - so system *throughput* will not change without changing the clock frequency - one of the unfortunate side-effects of being synchronous.

Instead, lowering the junction temperature while keeping clock period constant means you should end up with a slightly better timing margin for each combinational path between sequential elements (which manifests itself as greater stability among other things).

I'd imagine there might be some (maybe not measureable?) performance improvement if you look at the time taken for data to propagate to/from the first/last clocked element from/to an I/O pin on the package - it might be hypothetically feasible to sample outputs slightly faster at a lower temperature.

Another interesting concept (for anyone with an interest in mobile/low-power computing and signal processing) is that lowering the temperature should also allow you play with your core voltage to reduce the dynamic power consumed by the processor - extra timing margin means you can drop your voltage (delay is pretty much inversely proportional to the supply voltage) and save power (power is proportional to the square of the supply voltage) while still meeting timing due to the extra margin lowering the temp has bought you (you're trading speed for power). This brings in all kinds of other nasty issues though, like core noise margins and ensuring correct transistor operation (i.e. don't try this at home kids)

Not really sure how relevant this is to XBox overclocking of course... mea culpa.

Shinji

The GPU is not an NV25 core, particularly not with an extra vertex pipe. It's only got two vertex pipes. If it was an NV25+extra vertex pipe, it would have 3 vertex pipes.

Um, let me see, do I trust Anandtech or do I trust the guys I know working on Xbox games.... Hmm, tough one I'm sure you'll agree

(Officially, because the clock speed on the Xbox NVIDIA GPU is low (233mhz, originally announced at 300mhz... Boy, were the guys doing Malice (which burns fillrate like nobody's business) annoyed when that one happened) and there's the extra vertex pipeline (adding 50% to the speed of vertex transforms, which is nice for most developers but inconsequential for some things that are more fillrate dependent), it's the "NV2A". However it was originally announced as the NV25, shares the same silicon as the PC NV25 chips (GeForce 4 Titanium range) and most of the documentation given to developers calls it "NV25". "NV2A" is there for two reasons - firstly because of the low speed and extra pipe, but much more importantly, so the PC fanboys don't stamp their little feet over getting the Xbox GPU repackaged on a graphics card for the same price as an entire Xbox...)

EDIT:

I just read some more of that Anandtech piece - specifically the bit where the guy starts wittering on about the PS2 GS. He clearly has absolutely no idea whatsoever what he's talking about - I'd never write about hardware in that depth because I acknowledge that I'm no expert, but even I can tell that he doesn't actually understand on any level how the PS2 is programmed. Managing to get through an entire paragraph about the Xbox GPU and the PS2 GS without mentioning the Vector Units on the PS2 is downright astonishing.

logic1

TIS IS MY PECE!! IT RUENS SOE FASTE KAUSE OF AL TEH FANES!!!

.logic.

The_Bullman

nice one

_CreeD_

Fill Rate shouldn't be an issue, (HDTV aside) it's intended for the low resolution of a TV. It should have more than enough bandwidth to run at 32bit with at least the equivalent of 2x FSAA (One major disappointment I had was that fact none of the release games use FSAA, nor that it was an option within the box itself - just like the AA override built into Nvidia's drivers).

JustHalf

Originally posted by Shinji
(Officially, because the clock speed on the Xbox NVIDIA GPU is low (233mhz, originally announced at 300mhz... Boy, were the guys doing Malice (which burns fillrate like nobody's business) annoyed when that one happened) and there's the extra vertex pipeline (adding 50% to the speed of vertex transforms, which is nice for most developers but inconsequential for some things that are more fillrate dependent), it's the "NV2A". However it was originally announced as the NV25, shares the same silicon as the PC NV25 chips (GeForce 4 Titanium range) and most of the documentation given to developers calls it "NV25". "NV2A" is there for two reasons - firstly because of the low speed and extra pipe, but much more importantly, so the PC fanboys don't stamp their little feet over getting the Xbox GPU repackaged on a graphics card for the same price as an entire Xbox...)

The "extra pipe" is relative to the NV20.

po0k

http://www.arstechnica.com/cpu/2q00/ps2/ps2vspc-1.html

I think I stumbled on taht from a link either Yellum or Chernobyl posted.
Goes through the differences and explains them in nice laymans terms using clear analogies.
Gets my

saik

logic:

TIS IS MY PECE!! IT RUENS SOE FASTE KAUSE OF AL TEH FANES!!!

I WUZ JSUT TIUNKIN I'D DO THAT.

I HAEV GOE FASTAR STREPS? U TRIED THEM. IT GOES FASTAR NOEW. MOER PREFORMANCE!!!!111111

po0k

Every pe0n and his duck knows that RED WUNZ GOE FASTA!

astrofool

lol at shinj donning a hardware hat, go back to your game thingies

Gerry

www.ars-technica.com have some good articles on how the ps2 works, here.

Shinji

lol at shinj donning a hardware hat, go back to your game thingies

*grin*

Nah, if I was donning a hardware hat I'd be over at the Tech board. This is really a software hat, it's just that where consoles are concerned it's important to know a hell of a lot of detail about the metal is all...


Gerry - those articles are both very old, but they're very good indeed. Some of the detail is glossed over, and some things are now out of date (like the claim that PS2 coders aren't using both Vector Units), but for the most part the info is correct and the guy really knows what he's talking about - really good reading for anyone genuinely interested in the inner workings of the PS2.