Intel details big changes in Comet Lake-S power limits, Tau values

[HEXUS]

In previous gen PL2 was a maximum 1.25x TDP, now it is 3.5x. Tau refers to the PL2 duration.

Read more.

[DevDrake]

holly! that's explains some power consumption spikes in benchmarks.
Wonder if there is something similar for Ryzen.

[cheesemp]

holly! that's explains some power consumption spikes in benchmarks.
Wonder if there is something similar for Ryzen.

I can get 100W spikes on my 65W 3600 if I turn on PBO (Precision boost overdrive) with a slightly better cooler than stock (The wraith prism). I think it will go even higher with a better cooler but yeah not on default power settings.

[ohmaheid]

LOL. "Energy saving models"
Jesus wept.

[philehidiot]

LOL. "Energy saving models"
Jesus wept.

I saw that and chuckled. I thought Tau could be defined by the mobo and that enthusiast models set Tau to 999 or something to allow boost for as long as she can stand?

[DavidC1]

holly! that's explains some power consumption spikes in benchmarks.
Wonder if there is something similar for Ryzen.

The base 3950X uses 140-150W when loaded and it has 105W TDP rating.

In desktops you can use water cooling or heatsinks that weigh much as an ultrabook so CPU and mobo makers go overboard on power settings.

The PL2 being 1.25x is a guideline but laptops already exceed this. Some Cometlake and Icelake models are set at 50W PL2. Of course unlike in desktops it can't be sustained for more than 30 seconds.

[Gentle Viking]

My CPU head are 900 Grams :-)

[ohmaheid]

I saw that and chuckled. I thought Tau could be defined by the mobo and that enthusiast models set Tau to 999 or something to allow boost for as long as she can stand?

Intel. The only company that sells you a CPU with an option for insurance.

TAU 999 With a legal disclaimer saying "Only to be used in asbestos-lined cases."

[ik9000]

traditionally power-sipping, efficient 'T' suffixed processors with a standard 35W TDP might consume up to 123W in bursts of nearly half a minute in length.

In other news Intel updates the ultrabook specification to include mandatory extension cables to a permanent mains supply, and a LN2 feed.

Call me a cynic but I'd be very worried what that kind of spike would do for both thermals and battery life! I see my xps13 throttle regularly as the cooling can't handle simultaneous graphics demand and CPU workload. This kind of proposal would just not be realised in practice - not in that chassis anyway.

[tygrus]

High burst power used (W) all for the sake of better looking benchmarks and marketing numbers.

The smaller real world benefits comes at the cost of exponentially higher power usage (25% higher MHz = 56% higher power usage) which needs better cooling than stock to maintain. Intel provided only small benefits over time from optimising masks & fab process to get slightly higher MHz from the same voltage. It's still Intel 14nm for the CPU's listed above so it's like AMD back with 32/28nm for longer than expected.

[DavidC1]

In other news Intel updates the ultrabook specification to include mandatory extension cables to a permanent mains supply, and a LN2 feed.

Call me a cynic but I'd be very worried what that kind of spike would do for both thermals and battery life! I see my xps13 throttle regularly as the cooling can't handle simultaneous graphics demand and CPU workload. This kind of proposal would just not be realised in practice - not in that chassis anyway.

This is not for Ultrabooks, even the 35W series.

Ultrabooks already have a PL2 higher than PL1, but unlike in desktops it has to come back down to PL1 level after 28 seconds. Your XPS 13 does the same thing, PL1 of 25W.

[3dcandy]

I think we all knew that this happens - but Intel has pushed it about as far as they can and remain credible now. 2700X here and maximum I've seen power draw wise is around 100 Watts and even then it's not that accurate (wall mounted energy meter and checking from idle to full chat)