i7-6700 upgrade options, and power efficiency.

[daninmanchester]

I'm considering upgrading from a i7-6700 but wanted to save on power.
It's running my XCP-Ng box and ticks along at well under 50% most of the time.
Most servers just sip processing, pfSense, home Assistant, Plex, piHole, unifi controller, resilio, OMV, etc.
Occasionally they go up when watching a movie for example or transferring files
The main thing that takes up processing day to day 24/7 is BlueIris running on windows 10.

Beyond the basics of C and P states I'm not sure how different chips would benefit me e.g. given I have a constant load would a "T" series chip offer me any benefits?

I'm thinking a i7-8700T would give me similar peak performance but with lower power options.
But could I get power savings from say a 9th or 10th gen 10700 and just underutilize it?

Or because I have a constant load I can't benefit from a T chip?
 

[Jumballi]

A T series chip, while power efficient, is not actually close in terms of peak performance to the mainline chip itself. 
 

For what you want to do, the power savings won’t equate to the cost savings, as you still have to pay for the new cpu and possibly other components.

 

It’s better to keep using the older CPU until it dies if it’s meant to be used in a server. And in terms of carbon footprint, buying something to reduce waste will produce a huge ton more waste compared to continuing to use the system you have.

 

 

[daninmanchester]

Thanks, but what I didn't mention is that I will keep the older server bits and build a lower grade backup/lab server.
I need a decent backup but also want to separate out my "mission critical".
the backup will run on a schedule to sync then shut down unless I'm doing some lab stuff.

If I look at a more recent  T series chip it offers similar performance to the 6th gen.
But I dont know how the "T functionality" offers power saving. Does it rely on having no load and down tuning so I would never benefit because of my blue Iris constant load?
Therefore I may as well just get a regular chip which will give me a bit more headroom. 

[unclewebb]

The T series is a marketing gimmick. All Intel Core i CPUs built during the last 10 years use two turbo power limits to control power consumption. If you buy a K series CPU and you are overclocking, the BIOS sets these power limits sky high. If you buy a T series CPU, Intel sets the default turbo power limits low. Other than that, the CPU is identical. Often times K series CPUs are higher quality so can actually run the same speed as a T series with less power consumption compared to the T series. 

 

You can use software like ThrottleStop to adjust the turbo power limits to whatever you like.

 

5 hours ago, daninmanchester said:

I may as well just get a regular chip

That is true. Here is an example of a fully loaded 10 core CPU when set to 65W.

 

 

The CPU slows down to whatever speed is necessary so it does not exceed 65.0W. 

 

Intel's T series does the exact same thing. Might as well create your own T series.

 

Everyone frowns on 10th Gen because full load power consumption is obscene. When properly setup with the C states enabled, light load power consumption is better than previous generations.

 

 

[daninmanchester]

thanks @unclewebb.
Does the lower clock speed help in a T series?
The chip is cruising along at minimal load and its base frequency.
in a T chip it can idle lower and therefore save more power?
can I do more to save power like underclock/under volt and is there any way to do this dynamically?
 

[unclewebb]

1 hour ago, daninmanchester said:

in a T chip it can idle lower

You can run an Intel CPU at whatever speed and voltage you like. The T CPUs might use a lower default multiplier but you can run any CPU at a reduced speed if you want to. Here is an example of a 5000 MHz CPU being limited to 2000 MHz while 16 threads are fully loaded with a benchmark test.

 

 

With Intel Core i CPUs, there might not be any reason to do this. The best way to save power is to enable the C states. Idle cores with nothing to do will automatically save maximum power by entering C7 where they are disconnected from the internal clock and disconnected from the voltage rail. These idle cores will be sitting at 0 MHz and 0 volts. All of these other tricks do not accomplish very much if the C states are enabled and properly working. 

 

Older CPUs used SpeedStep to adjust the CPU speed. 6th Gen and newer Intel CPUs use Speed Shift Technology. Either CPU can have its speed fully controlled if you think that is necessary.

 

Remember, the T series is just marketing. They save power by running slower. If you slow a fast CPU down, it will save power too. 23W at 2000 MHz in my screenshot is way better than the typical 150W when allowed to run that same load at full speed.

[daninmanchester]

Thanks for clarifying. I will need to work out how to do this on XCP-NG. I'm pretty sure the bios I have doesn't provide this level of control.
I do need a new motherboard anyway though so I'll look in to this.

[unclewebb]

5 hours ago, daninmanchester said:

the bios I have doesn't provide this level of control

I have never seen a BIOS that provides this level of control. That is why you use need to use software in Windows if you are interested in full control.