First Look At Intel Tejas & Socket 775

Anonymous Indian writes "The snoops at Anandtech have unearthed some details and photos of Intel's rumored Tejas 90nm CPU which draws 150 watts of power, a 50% jump compared to Prescott. It's also got an interesting locking mechanism instead of the traditional metal clip from hell for most processors." There's not much info beyond the photos, but it's still interesting to see what lies ahead for Intel.

Comments

And this is interesting why?

[Pike]

The both the slashdot sum-up and the linked article fail to explain why, exactly, I should be excited about this.

Re:And this is interesting why?

[alexatrit]

My thoughts exactly. The power consumption is the only thing going for it, and it's understated as it is.

AnandTech: Here are pictures of the thing. We don't have a clue what it does or how it works, but we presume that the extra power draw is part of Bush's Mars Mission.

Re:And this is interesting why?

[Ralph Wiggam]

The dominant CPU maker is releasing a new CPU and a new socket. That's news for nerds. It may not be the most exciting news for nerds ever, but it's still news. The 150 Watt consumption is somewhat interesting.

-B

Re:And this is interesting why?

[bhtooefr]

Intel desktop and server processors. Have you looked at the Pentium M lately? By the way, I didn't think the Athlons ran exactly cool (cooler than any Intel desktop CPU nowadays, though).

Re:150 Watts?

[aldoman]

The only trouble is that power usage inside computers is skyrocketing lately.

150 Watts CPU, a 100 Watts graphics card and then you have all the fans to cool it, and you'll be lucky if you are under 300W and that doesn't even factor in storage, optical drives, motherboard, audio, communication devices.

I mean you will need a hell of a power supply unit to provide that sort of power reliably.

Re:150 Watts?

[Anonymous Coward]

You only seem to equate power wastage in terms of your personal $s.

Try and think a little bit bigger. Think in terms of global warming, energy depletion, entropy etc.

Using less power (whatever the application) can only be a good thing.

wtf?

[autopr0n]

Mod parent as troll.

Why should it be modded as a 'troll'? just because it betrayed some lack of knowledge about CPU design? I think your post should be modded as troll.

Re:So much power

[BitchKapoor]

To say that SMT is "IBM's version of HyperThreading" is backwards-speak. Simultaneous MultiThreading (SMT) is actually a generic term for any technique which multiplexes threads on a single CPU core by allowing any thread to use any available resource during any clock cycle. HyperThreading (nee Jackson Technology) is Intel's name for their implementation of SMT -- see http://www.intel.com/technology/hyperthread/.

From my description of SMT, you might conclude that it is able to make the fullest use of CPU core resources of any multithreading approach. Then why are some benchmarks slower with HyperThreading enabled? I have three ideas. First of all, since OSes tend to treat a single CPU with hyperthreading like a dual-CPU system, they may schedule lower-priority processes on one of the threads, resulting in less total CPU time available for the main benchmark process. Secondly, there is some overhead to multiplexing all of these resources amongst two threads, so perhaps Intel had to "tone down" superscalar operation or memory queues to meet clock timing when in HyperThreading mode (I know this is vague). Thirdly, perhaps in ostensibly non-HyperThreaded mode the second thread is actually used for speculative execution, and hence slightly boosts the performance of a benchmark.