Core 2 Quad - bring on the heat.

[BillyBuerger]

Tom's Hardware - Intel's Core 2 Quadro Kentsfield: Four Cores on a Rampage

It's an early sample and all. But wow, sounds like it's going to be "HOT". The total system drew 100W more than a similarly clocked Core 2 Duo at full load. TDPs are going to be back up over 100W with these. Although I guess that's not much different than AMD's FX series.

Just thought I'd mention it.

[MikeC]

TDPs are going to be back up over 100W with these. Although I guess that's not much different than AMD's FX series.

Let's not forget that the vast majority of Intel desktop processors being sold today are Prescotts and Pentium Ds -- ie, over 100~130W TDP. Intel's own forward-looking roadmaps indicate that Core 2 might represent 30% of their total desktop processor sales by the end of the 1st quarter next year.

Hot Intel processors are not behind us yet. The mainstream is still sucking up those >100W babies, and Intel is happy to keep production lines active and move stocks while their factories move to core 2.

[smilingcrow]

I’ve read in a number of places that the mainstream Quad cores due in Q107 will have a TDP of 80 to 85W and I think that’s good for up to ~2.4GHz.
If Intel cherry pick Core 2 Duo chips they should be able to bin them for low power/voltage and use the best for the Quad CPUs. I imagine that they might well need to use lower voltages than standard to achieve 40W per chip at 2.4 GHz. In comparison, undervolting an E6600 should enable you to hit ~40W at present.

These seem a lot more attractive to me if I wanted a very high performing system with decent power consumption. Running a Core 2 Duo at ~3.3 GHz as in the recent SPCR article is very power inefficient at idle and load. Hopefully, Intel will implement Speedstep in an efficient way on these dual chip CPUs. The ES versions doing the rounds don’t seem to support Speedstep at all!

[BillyBuerger]

I guess I just got fooled by Intel's marketing of performance/watt stuff that they were going to try to stay away from the hot, power hungry CPUs. I kinda forgot that the Pentium Ds are their current budget-mid range CPUs now. Which is where more people are opposed to the high-end performance freaks. And Kentsfield will make the high-end performance people happy. Once their native Quad-cores come out, they'll probably be better on power consumption.

And yes, I assume the final versions will have SpeedStep enabled.

[vitaminc]

There's not a lot of reason to release low power C2Q for Intel because power consumption isn't on the mind of those who want quad core on desktops.

What's interesting is that Intel has a 50W quad core server processor coming in 1Q of 2007. God damn that's fine.

[smilingcrow]

I guess I just got fooled by Intel's marketing of performance/watt stuff that they were going to try to stay away from the hot, power hungry CPUs.

Aren’t you overreacting a bit here? These aren’t mainstream CPUs but top end parts at top end prices and have an insignificant impact on the overall market. It would be like berating Fiat for making fuel hungry Ferraris, whilst they simultaneously produced fuel efficient mainstream family cars.
80-85W for the mainstream parts if this is a true figure, will lead to a very healthy performance per watt for those applications that can utilise all 4 cores efficiently.

Also, if you look at the current power consumption of a workstation class dual dual-core Xeon or Opteron system, you’ll see that even the 120W Quad cores gives a better performance per watt, especially when you look at the overall system power consumption. At a much better price to if they cost ‘only’ $1000.
The real power hogs are the first AMD 4x4 chips at 90nm which have a TDP of 125/130W per CPU; ouch.

And yes, I assume the final versions will have SpeedStep enabled.

I’m wondering how they will implement it seeing that the chip is two cores on one package. It may be a compromised solution.

[gb115b]

Running a Core 2 Duo at ~3.3 GHz as in the recent SPCR article is very power inefficient at idle and load.

what article is that?

i was under the impression that core2duo automatically underclocks to 1.6ghz at idle.... ??

[smilingcrow]

Running a Core 2 Duo at ~3.3 GHz as in the recent SPCR article is very power inefficient at idle and load.

what article is that?
i was under the impression that core2duo automatically underclocks to 1.6ghz at idle.... ??

This article
Intel’s Core Duo series of chips run at idle with a multiplier of 6; this leads to the mobile chips idling at 6*166 = 1GHz, the desktop chips at 6*266 = 1.6GHz and the Server chips up to 6*333 = 2GHz.
As you can guess from this, if you overclock a Core Duo you not only increase its full clock speed but also its idle clock speed.
e.g. An E6700 run at 333 FSB will idle at 2 GHz.

[jojo4u]

Let's not forget that the vast majority of Intel desktop processors being sold today are Prescotts and Pentium Ds -- ie, over 100~130W TDP.

Intel is also making steps forward in this area. Source is german, though.
http://www.heise.de/newsticker/meldung/77721

There is a new D0 stepping due for Netburst:
new Celeron D 360 (3,46 GHz, FSB533, 256 KByte) 65 W TDP (VRM spec 06)
Pentium 4 631, 641, 651, 661 soon with same characteristics
Pentium 960 soon with 95 W TDP (vrm spec 05A)

All are said to have working C1E and Speedstep.

[pony-tail]

With the aftermarket coolers around at present cooling a 100watt cpu is not that difficult even in a quiet machine and with the brute power these things should have would not be a bad trade off if you have a use for it .
Out of my league from a financial standpoint - but it would halve my compile times . other than that it would most likely not be of sufficient use to even consider one . I am running a dual core at present and even it has not been of any great advantage - Even if there is some benifit for Vista (of which I have no fore knowledge ) I will never be installing Vista on any machine I own , or for that matter any of my immediate family own .