backtop

Hardware XBOX 360 CPU To Go 65nm
Sven Olsen (Blog) - April 21, 2006 3:46 PM
Print E-mail del.icio.us 22 comment(s) - last by dexvx.. on Apr 23 at 2:14 PM

The 90nm graphics core designed for XBOX 360 - Image courtesy Anandtech.com
XBOX thermal issues got you down? 65nm to the rescue!

VNU reports that Microsoft has solidified a deal with Chartered Semiconductor to change the process size of the XBOX 360 CPU from 90nm to 65nm.  As foundries begin to pick up 65nm production, it will be cheaper for Chartered Semiconductor to produce the CPUs on a 65nm process since the company can dedicate resources to new production lines rather than stick to legacy production.

As an added bonus, the thermal envelope associated with the CPU is expected to decrease as the process size decreases.  The clock speed and design will remain the same for this new iteration.  There is no word yet on whether or not the XBOX 360 graphics core designed by ATI will also undergo a die shrink.

With the switch to 65nm, Chartered will also save on wafer costs, as more dice may be cut from a single wafer.  Hopefully, that additional savings will also translate to cheaper XBOXs for the rest of us.  The new processors will not enter into production until Q1'07.

Comments     Threshold


This article is over a month old, voting and posting comments is disabled
Dice??
By osalcido on 4/21/2006 6:08:20 PM , Rating: 2
More "dice" from a single wafer?




RE: Dice??
By Burning Bridges on 4/21/2006 6:26:40 PM , Rating: 1
I think the word meant is "dies" as in the plural of "die" (the computing term) :/


RE: Dice??
By MrSmurf on 4/21/2006 6:29:19 PM , Rating: 2
"dice" is an alternate version of "dies"

lol @ the hope they pass the savings to us.


RE: Dice??
By bunnyfubbles on 4/21/2006 7:10:00 PM , Rating: 2
Joke all you want, by Q1'07 PS3 should be available (although perhaps not in large enough quantities), it'd be the perfect time for a price cut whether or not the 360 is cheaper to produce.


RE: Dice??
By LuxFestinus on 4/21/2006 7:17:00 PM , Rating: 2
I for one welcome our relatively new game making masters. :)


RE: Dice??
By TheWarden on 4/22/2006 1:10:35 AM , Rating: 2
Anandtech had a blurb about this a few months ago. Apparently they asked a bunch of hardware engineers whether the correct plural term was "dice" or "dies" and the former won out. I think it sounds ignorant, myself. For eons now dice have been for playing games while engineering widgets were dies. Not sure what brought about the change.


RE: Dice??
By OvErHeAtInG on 4/22/2006 1:20:18 PM , Rating: 2
The term "die" comes from the tool which is used to cut out something square. So the "die" of a microchip core and the "die" which you use to shoot craps bear the same meaning. So the plural can be "dies" or "dice" in either situation.


RE: Dice??
By Furen on 4/23/2006 12:05:23 AM , Rating: 2
OK, here's a story I picked up from an "old" engineer. He said that when cooking silicon ingots you need to slice and dice them (the cooking terms). First you slice the ingot into wafers, then you dice it into dice, for which the singular is die. While dies sounds to be the proper name remember that dies are used for shaping other things, they're not shaped themselves, while if dicing the wafer is the proper name for the process the parts resulting from it could very well be named after it.


a
By hans007 on 4/21/2006 6:07:51 PM , Rating: 2
well the difference betweeen a 90nm prescott with 2mb cache and a 65nm cedar mill is pretty large.

a 65nm dual core presler is also ALOT cooler than a 90nm dual core smithfield, and the 65nm has more cache.

the only time the shrink didnt help was going to 90nm from 130nm and that was due to a huge change in the architecture and also 1mb vs 512k cache on that generation.




RE: a
By TomZ on 4/22/2006 2:44:56 PM , Rating: 2
Guess you need to define what you mean by large.

For example, Presler is about 13% power consumption less than Smithfield during load, which is nice, since Presler has more cache. But during idle, power consumption between the two is the same. So, if you figure your computer is idle 80% of the time, the net power savings is only 20% * 13% = 2.6%.

In my book, a 2.6% power reduction is "not much."

ref. http://www.xbitlabs.com/articles/cpu/display/presl...



RE: a
By Furen on 4/22/2006 11:50:55 PM , Rating: 2
You're really bad at this power-consumption thing...

Presler has 13% less power consumption compared to Smithfield at a higher clock speed, on a faster frontside bus with double the cache. Also, intial Preslers were missing a lot of the powersaving features that were used in the later 90nm cores. Prescott was a completely redesigned core that almost maintained the same die-size even after a die-shrink that would have given northwood half the die-size.

Look at this: http://www.lostcircuits.com/cpu/amd_fx57/13.shtml

Notice how Winchester (AMD's first 90nm core) basically halved the power consuption of a clawhammer (yes, newcastle would be the better one to compare it to but the one in this example is higher-clocked, and the venice has dual stress liners). Also notice how Yonah pretty much halved the power consumption of Dothan (or rather, Intel crammed twice as many cores into the same power envelope), this is the kind of power consumption reduction that you can expect from a properly executed die-shrink.

Also, whenever you do a die-shrink you will experience across-the-board power consumption reduction and, in fact, this will be higher when CPU load is lower, since most current chips are able to enter power saving states. Another thing: if your maximum power disipation decreases then your cooling solution can be relaxed a bit (or even downgraded, which makes the system cheaper).


RE: a
By dexvx on 4/23/2006 2:14:14 PM , Rating: 2
You ignore the power saving features that were missing in the initial Preslers.

http://techreport.com/reviews/2006q2/core-duo/inde...

The 965XE, with EIST and some other improvements, actually has lower power consumption at idle than a X2-3800+ with C/Q disabled. At full load, the 965XE uses the same consumption as the the FX60, and about 50% less than the 90nm 840XE.


This doesn't mean more cores or any cache
By hstewarth on 4/21/2006 8:58:35 PM , Rating: 1
Just going to 65mm doesn't mean that 360 will have more cores or more cache on the process. It means the die(s) are smaller and likely use heat.

Not even sure if the software is design to handle more cores if even more cores are provide which I don't think they are being provide.

It could be one way that Microsoft in 2007, could increase the cpu power of the 360 with more cores to compete with the PS/3. New version of 360 however.




RE: This doesn't mean more cores or any cache
By vhato on 4/21/2006 10:49:11 PM , Rating: 2
Bil Gates already stated the new HD-DVD won't be for games so he doesn't infuriate the current install base. The PS/3 has a 2 extra, highly efficient, non-scalable cores as well as a an Nvidia GPU based on todays architecture. The 360 has 2 extra, highly scalable, in-efficient cores with a next gen ATI GPU based almost on DirectX 10 features. IMO, they will be WELL matched as game consoles, giving up nothing to each other in the long run.

So knowing this, why would MS add more processing power in 2007 and alienate their existing Xbox360 install base?


By DigitalDivine on 4/22/2006 2:02:19 AM , Rating: 2
ms doesn't necessarily have to increase clock speed or anything for that matter with the die shrink. they will probably keep it the same in favor of a more slim and compact design for the 360 and cut cost. similar to the the PSTwo and the PS2.


By hstewarth on 4/22/2006 10:42:54 AM , Rating: 2
The HD-DVD drive is likely be external at extra cost. This means for HD media support you need the the 360 and external drive or you could have a PS/3. I really hope that the PS/3 is not marked as a game machine. That would make it more attracted to other people - especially people that don't buy games.

PS/3 cell processor actually has 7 cores total plus a PowerPC core that is used primary as controller of 7 SPE cores - however one of those are use for OS, maybe for DVR features.


Thermals Not A Done Deal!
By TomZ on 4/21/2006 4:55:50 PM , Rating: 1
It's not a done deal that the process shrink will significantly decrease power consumption of that processor. Consider the case with Intel's Netburst processors - the die shrink only gave a small decrease in power consumption.

The problem is that while dynamic power consumption decreases with the process change, the static leakage current actually increases. So, the net decrease is the difference between the decreasing dynamic power and increasing static power.




RE: Thermals Not A Done Deal!
By phatboye on 4/21/2006 5:57:14 PM , Rating: 2
that was because intel made other changes to the cpu alongside the die shrink. If all else stays the same there should be a decrease in the amount to thermal energy lost to heat with a die shrinkage.


RE: Thermals Not A Done Deal!
By TomZ on 4/22/2006 2:49:24 PM , Rating: 2
Please see my post below.


hopefully
By Homerboy on 4/21/2006 4:50:19 PM , Rating: 2
quieter too




This is why I don't have a XBox360
By AlexWade on 4/21/2006 8:22:01 PM , Rating: 2
You wait until a price cut, plenty of games, and all the bugs being worked out. Then you buy one.




Xbox360 Mk.II in Q1 or Q2 2007
By kilkennycat on 4/23/2006 12:32:31 AM , Rating: 2
Faster, 65nm CPU and GPU - hence fewer failures due to
overheating of the internal inaccessible-for-cleaning heatsinks, lower incidence of overheated DVD-drives due to the same heat-sinks, and with the DVD drive finally replaced by HD-DVD or Blu-ray, plus far higher-quality all-around DVD-decoding. Only another $500 or so.

Distinctly not good being an "early-adopter".




"Young lady, in this house we obey the laws of thermodynamics!" -- Homer Simpson









botimage
Copyright 2009 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki