Intel demos fanless, cool 5GHz chip

Thought I should post this <A HREF="http://www.theinquirer.org/?article=2462" target="_new">http://www.theinquirer.org/?article=2462</A> and see what you folks think of it!

Some images:
Detailed picture of the McKinley die: <A HREF="http://www.theinquirer.net/issc1.jpg" target="_new">http://www.theinquirer.net/issc1.jpg</A>
Intel block layout for the 5GHz processor: <A HREF="http://www.theinquirer.net/issc2.gif" target="_new">http://www.theinquirer.net/issc2.gif</A>
Intel's 4MB test die: <A HREF="http://www.theinquirer.net/issc3.jpg" target="_new">http://www.theinquirer.net/issc3.jpg</A>

Shame it's only for servers though.....roll on 5GHz Celeron CPUs!

====
Basmic

Oops just the ALUs will be operating at 5GHz - was too tired that time of the day to read it properly!

====
Basmic

If by 5GHZ you mean ACTUAL 5GHZ, I am impressed but would have a hard time beleiving a 5GHZ processor out there already. If you meant Double Pumped 2.5GHZ to 5GHZ ALUs, more beleivable but would make no sense. So if indeed Intel demoed a 5GHZ fanless ALU, makes more sense, and quite impressive. But it means jack if the rest of the CPU can't be the same!

--
What made you choose your THG Community username/nickname? <A HREF="http://forumz.tomshardware.com/community/modules.php?name=Forums&file=viewtopic&p=19957#19957" target="_new">Tell here!</A>

5 ghz fan or not that not the point

It 32 bit AGU

Now what to do??

P4 twin pump use a 16 AGU this 1 is a full 32 bit ALU AGU so it will be faster clock for clock.

Now what to do??

Current P4's use 32-bit AGU's, that's for memory address. They use double-pumped, 16-bit simple ALU's and 1 32-bit complex ALU that runs at the processor's clockrate. Since most integer operations are 16-bit anyway, I doubt we'll see <b>that</b> big of a performance boost when moving to 32-bit ALU's but hey, you never know.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.

Just think about your typical application. A game for instance. How big a number do you think you're gonna be using? Your HP is what, 100 points? The distance between objects are measured in FP nowadays to insure precision, the AI is mostly branch operations, in fact, in my recent programming assignments for class, the only time I'd use integers is as array indexes. And arrays are almost never 32,768 large. Applications typically are more instruction-intensive (i.e. it does a lot of things) rather than data-intensive (i.e. it uses huge amounts of data). There are exceptions of course.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.

the complex alu also run most of the legacy code only a part of ALU operation can be run on Fast ALU.Be moving all to 32 bit you simpler implemtation and increase performace also raising you die size.

Now what to do??