A warning about Intel processor power consumption and high CPU usage

[Simon]

Yes faster clock speeds do draw more power. My understanding as I stated earlier is that it's the actual action of opening and closing transistor switches in the processor that causes the power dissipation not the steady states that they may sit in.

[rrinker]

 I think you are misunderstanding how the turbo mode works. It only gets to that peak frequency if executing a single threaded application. To stay within the specified TDP envelop, as more cores go active, the top speed has to come down. Run something like BOINC set to use just one core and compare the top speed vs having it use 4 or all 8. With no modern multicore processor can you directly compare peak speed and what it runs at with a decent workload - great, it can do 3.1GHz on a single core (so you install DOS?) but 4 cores running at 2.5GHz are going to complete an awfully lot more work than 1 core at 3.1GHz.
 The tools mentioned like CPU Temp will show a lot of useful information. There's another similar one called CPU something or other which shows the load on each core as well as temps for each one plus the overall temp of the processor. Also in any of the newer Windows versions, the Resource Monitor (not Performance Monitor) will show a lot of details about each subsystem including the CPU with a chart for each core (real and virtual - and yes, Hyperthreading DOES help a LOT of types of workloads, even if it does little for this one specific case).
 As for how long it can sustain max power, that really depends on how well the hardware is designed. With proper cooling, the maximum can be sustained forever. With a less than adequate cooling system, the CPU will throttle back when it gets too hot.

[Simon]

Well one core only gets me 3.07GHz but like you say it is more efficient to run slower while chewing on more work. That was kind of my point.Yes I know hyperthreading will help lots of stuff baring this as it just loads down the processor and needs resl hardware to run on whioch is why i don't use more than 2 cores for it and lock them.

Taking it further with 2 fully loaded cores (or even trying 4 which is pointless) I get up to 2.85GHz but the extra power demand and stresses are probabvly not worth it and if I am knowingly loading my processor down disabling turbo boost is a good ideas which was also my point in creating this thread......

[Twoflower]

TDP is not the power the chip is using, it is the amount of heat it generates.

First law of thermodynamics.  It's not generating sound or light, it's not radiating a significant amount of power via EMI, it's not charging or discharging an energy storage device, etc., which means power in = heat out.

OK, parts of the energy feed into the Processor are not converted into heat, but I think that's just a minor part. I think Towger means that the TDP doesn't have a 1:1 relationship to the power consumption. The processor is actually allowed to generate more heat than the TDP indicated. But that just for a very short time frame. It takes account the thermal mass of the cooling system. The TDP should be used to calculate the cooling system; if you use this value to perfectly match the power supply you're calling for trouble. To my understanding the accumulated thermal load over a longer period of time will not exceed the stated TDP.

Still this TDP value is nowadays widely ignored by laptop manufacturers. It doesn't matter if you buy a cheap or a expensive one. The cheap ones usually save money on the cooling system, the expensive ones try to minimize the volume/weight of the Laptop. Both solutions have a direct impact to the system performance under long lasting loads. An exception might be the so called 'Gamer Laptops'. The question today is only how well the cooling system it is designed. Probably the best solution is to look for a system that's available also in more powerful versions (higher TDP rated CPUs) and hope the system you buy (with lower TDP) has the same cooling system. That may or may not indicate to a better cooling system.

But it even gets more variable: The TU Dresden recently published a document showing how the CPU performance behave over a large number of identical processors: The Shift from Processor Power Consumption to Performance Variations see figure 1b. The performance varies about +/-5% between individual CPUs. To my understanding similar results were made elsewhere too with same or worse results. If and how much this will translate to a Laptop stays on another page. But my personal assumption: This variations will increase in future as the silicon is squeezed out to the last bit and all kinds of variations (Process speed, minor changes in the cooling system and many more) gets visible.

[SpaceCow]

I'm happy that  I don't have to worry about the temp or clockspeed of my laptop. It handles that all by itself.

[rrinker]

 Good point. I gave up on even bothering to overclock anything years ago. The newer processors are so much lower in power draw than what was out 10 years ago that you (well, at least I) don't even bother with aftermarket cooling systems in my desktops, the supplied fan in the Intel boxed processors is sufficient for the machine to run hours on end at max CPU load without throttling. My only laptop is an older I5 (dual core, no hyperthreading) provided by work, despite being only 2 cores, I've had it running labs with 3x Server 2012 R2 VMs on top of the Win10 host with no problems. The fan does run at full speed most of the time when doing that, but it kept going with no slowdowns while I build and configured the 3 servers to simulate a system I was designing for a client. During a typical day, when I am just doing some web browsing and have email open, the fan is mostly silent. This is not a gaming laptop, there's no discrete GPU, but it's also not an ultralite design - it's fairly hefty compared to many new laptops.

 

[KhronX]

For further details:

https://en.wikipedia.org/wiki/Intel_Turbo_Boost

I think you are misunderstanding how the turbo mode works. It only gets to that peak frequency if executing a single threaded application. To stay within the specified TDP envelop, as more cores go active, the top speed has to come down. Run something like BOINC set to use just one core and compare the top speed vs having it use 4 or all 8. With no modern multicore processor can you directly compare peak speed and what it runs at with a decent workload - great, it can do 3.1GHz on a single core (so you install DOS?) but 4 cores running at 2.5GHz are going to complete an awfully lot more work than 1 core at 3.1GHz.
 The tools mentioned like CPU Temp will show a lot of useful information. There's another similar one called CPU something or other which shows the load on each core as well as temps for each one plus the overall temp of the processor. Also in any of the newer Windows versions, the Resource Monitor (not Performance Monitor) will show a lot of details about each subsystem including the CPU with a chart for each core (real and virtual - and yes, Hyperthreading DOES help a LOT of types of workloads, even if it does little for this one specific case).
 As for how long it can sustain max power, that really depends on how well the hardware is designed. With proper cooling, the maximum can be sustained forever. With a less than adequate cooling system, the CPU will throttle back when it gets too hot.

PS: HP laptops, especially those lowly consumer ones are designed and built down to a budget. As in, WAY down. Even the Pavilions from a couple years ago were (and they may well still be). Especially when it comes down to build quality and cooling (and most likely display as well) - those are the few main areas costs CAN be cut.

[senso]

If you think that a 600€/&/£ laptop as a production cost of around 200€ and half of that goes to Intel you start to see why they are so crappy...

No, it will only be allowed to go up to 25Watts in TDP-UP mode, that requires that the BIOS allows the cpu to do that, and you limited cooling system doesn't support 25Watts of TDP, so the TDP-UP option is turned off and removed from the visible options in the user-land BIOS...

It can be turned on, but it will be under temp throottle all the time..

And remember, there is also vregs and ram controller and most important there is the iGPU in all sharing that 15Watts budged, run the laptop in 16 colours mode at 240x480 and you might gain another watt to the cpu cores.

[Simon]

It actually limits at 18W or lower depending on ambient temperature.

Sent from my phone so mind the autocorrect.

[slicendice]

I think there are a lot of misunderstanding regarding the relationship of TDP, Power consumption, Temperature, Thermal dissipation and Turbo Boost.

I start with Turbo Boost. Intel processors can up it's frequency above specs while using n-cores (not just 1-core). This means that a 4-core CPU can TB on all 4 cores as long as the designed max power consumption and temperatures are within defined specs. The TB is defined separately for 1-, 2-, 3- and 4-core operation. The settings could be for instance 2/2/7/9, which means the amount of frequency increment steps the CPU can take above specs. The numbers starting from left to right are defined for 4-, 3- 2- and 1-core boosting.

Example:

We have a 2GHz CPU and it's base frequency is 100MHz, this means that on 100% CPU usage with big enough heat sink the multiplier will be 2000MHz / 100MHz = x20.

So now we know the max frequencies we can get during TB.

Calculations:

n-cores --> ( MULTIPLIER_STEPS + TB_MULTIPLIER_STEPS ) x BASEFREQUENCY = MAX_TURBOBOOST_FREQUENCY
4-cores --> (20 + 2) x 100 = 2200

Resulting table based on above formula and values

Cores	TB Steps MAX	Freq. MAX
4	2	2200
3	2	2200
2	7	2700
1	9	2900

All this info can be found in the CPU specifications (if someone dare to read hundreds if not thousands of pages of CPU specs) :-D

Max power consumption is not equal to TDP. You have to consult the CPU specifications documents in order to know how much power is required to run the processor. TDP is the minimum thermal dissapation capacity/power required for a cooling solution under normal CPU operation/load. You have to consult the specs sheet again for more details.

Finally we have temperature, the CPU speed is increased and decreased according to temp in relationship to Tjunction, again consult the specs sheets for more details, but basically this means the CPU will power off completely at 100C. And for example MAX turbo boost at a temperature of below 45C. (consult docs for exact values).

Simon, if your computer is something cheap, then it is very likely that the cooling solution is not good enough. The heat pipes, the thermal paste could be bad quality and the cooling solution max capacity was designed too close to TDP.

Hope I did clear up more confusions than I created by posting this. :-)

[Simon]

I think ultimately it boils down to the processor having a maximum working temperature. If you can push it faster in a colder temperature with a higher voltage if necessary then so be it. I think a lot of what Turbo boost relies on is the thermal inertia of the cooling system. If you need a few seconds of maximum possible speed that could not be sustained by the cooling system for long periods you can do that relying on the fact that the actual heat sink material can absorb an amount of heat before its temperature hits the maximum that the processor can withstand. Once you hit that temperature limit then the processor will start to slow down again to maintain that temperature without crashing. This means that you will get more performance in a colder environment and in a warmer environment the process that will limit more readily. It's really quite simple. Obviously when you have multiple cores you can start looking at things like how much more speed can you get for one core alone if you practically turn the other core off which is excellent for single thread programs which most programs are.

For example if I am doing work in my 3D CAD program the biggest processor demand is when I moved the model around on the screen. So if I'm going to spend a few seconds at a time moving the model around the processor can afford to put on a spurt of speed to make that movement more responsive and fluid for me.

[slicendice]

Attached is an image of how Turbo Boost should work. My CPU is by default 2.3GHz but as you can see, the frequency is way beyond that though all 4 cores/threads are almost at 100% usage. :-)