Trying The Turbo Boost Max 3.0 Patches On Linux 4.9 With A Core i7 6800K Broadwell-E

[bug77]

Intel is well aware of the capabilities of their products. When you consider they have some CPUs with turbo speeds above 4GHz, you know there's plenty of room to spare on cheaper models.

And good for you (not being sarcastic) - but I can assure you that not everyone can sustain those turbo speeds on the stock heatsink.

To clarify, I have no problem whatsoever of a CPU being pushed hard. My gripe is the reason behind TB 2.0. There is no reason why your 6600k couldn't just be called a 3.9GHz part and just slow itself down when it gets too hot. ...

Oh, but the reason is staring you in the face: 3.9GHz is only sustainable on one core only, when the others are idle.
What you say should happen, is already happening in the mobile space: Qualcomm advertising Snapdragon 8xx at 2GHz+, when they can only sustain that frequency for a minute or so. This, to me, look way more dishonest (yet seems to fly with the public nonetheless). Arstechnica wrote about this. I can provide links if you're not familiar with the subject.

[DJViking]

TBM 3.0 is complementary to TBM 2.0.

Does this mean support for TBM 2.0 already exists? If so, can it be used without the patches for 3.0?

[schmidtbag]

Oh, but the reason is staring you in the face: 3.9GHz is only sustainable on one core only, when the others are idle.
What you say should happen, is already happening in the mobile space: Qualcomm advertising Snapdragon 8xx at 2GHz+, when they can only sustain that frequency for a minute or so. This, to me, look way more dishonest (yet seems to fly with the public nonetheless). Arstechnica wrote about this. I can provide links if you're not familiar with the subject.

To my understanding, TB 2.0 allows all cores to achieve 3.9GHz, or at least more than one. TB 1.0 would only do 1 core at a time (which is why I'm a fan of that generation).

[bug77]

To my understanding, TB 2.0 allows all cores to achieve 3.9GHz, or at least more than one. TB 1.0 would only do 1 core at a time (which is why I'm a fan of that generation).

http://ark.intel.com/products/88191/...up-to-3_90-GHz click on "Max Turbo Boost Frequency" for additional info

https://en.wikipedia.org/wiki/Skylake_(microarchitecture)#Desktop_processors

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

[schmidtbag]

Based on what you posted and further research I've done, we're both right and wrong.

For a CPU like yours, it is possible for all 4 cores to operate beyond their "normal" speed, however, it isn't possible for all 4 to reach the max turbo speed. For every one core that isn't in use, the other remaining cores get a little bit faster. Only one core by itself can reach the max speed.

So: it's less stupid than I thought it was, but the fact that all cores can be boosted beyond their default speed is still stupid.

[schmidtbag]

EDIT:
When I mean "all cores" I mean all simultaneously.

[bug77]

Based on what you posted and further research I've done, we're both right and wrong.

For a CPU like yours, it is possible for all 4 cores to operate beyond their "normal" speed, however, it isn't possible for all 4 to reach the max turbo speed. For every one core that isn't in use, the other remaining cores get a little bit faster. Only one core by itself can reach the max speed.

So: it's less stupid than I thought it was, but the fact that all cores can be boosted beyond their default speed is still stupid.

Maybe I've misworded my comments, I never implied all core go to max turbo. But during video benchmark that only stresses one core, I do see it going up and staying there.
As for why you think this is stupid, I still don't know. It looks pretty straightforward to me. For example, my 6600k is a 91W part. It can stay within that envelope running all cores at 3.5GHz. But if not all cores are busy, some of them will need less power, thus produce less heat. This in turn creates room for the remaining 1 or 2 cores to run faster. For short amount of time, if the CPU has been idle for a while, thus it's cool, all 4 cores can go up to 3.6GHz until a certain temperature is reached. It's just like carbon credits for CPU cores

[schmidtbag]

Maybe I've misworded my comments, I never implied all core go to max turbo. But during video benchmark that only stresses one core, I do see it going up and staying there.

Understood. I don't have a problem with that. Again, I think TB for one core (or no more than half of the cores) is great.

As for why you think this is stupid, I still don't know. It looks pretty straightforward to me. For example, my 6600k is a 91W part. It can stay within that envelope running all cores at 3.5GHz.

For argument's sake, let's say 91W is the actual power consumption of the CPU when it is under full load at it's default speeds. If you push all 4 cores to the max and it turbos to 3.6, it is no longer a 91W part. The fact that the only thing restricting the CPU from reaching 3.6GHz is thermals suggests that there is no reason why Intel couldn't just make it a 3.6GHz part and just underclock where necessary (as stated before, it has to do this anyway even if it operates at its "normal" speed). I don't really know how to clarify that any better.

But if not all cores are busy, some of them will need less power, thus produce less heat. This in turn creates room for the remaining 1 or 2 cores to run faster.

I know - I agree with this. I don't have a problem with this. I am 100% in favor of cores getting boosted, but the exceptions are:
1. Not all 4 should be boosted simultaneously (for reasons I just stated)
2. Whichever cores do get boosted should not exceed the CPU's rated wattage.

[bug77]

schmidtbag Well yes, 4 cores boosting is a bit counterintuitive, I'll give you that. As I've tried to explain, if your CPU has been idle for a while, it's a safe bet it can go a bit overboard with TDP without incurring any damage. After all, 3.6GHz is less than 3% over the standard 3.5.
I think we both understand each other better now

Fwiw, this 91W include the IGP. If you're not using that, you're not going anywhere near the rated TDP. But this isn't linked to how TB should work, this is more me telling you you shouldn't fret so easily for going a bit outside the limits. It's not sustainable, it's only done for seconds, I think.

[bug77]

Ok, it seems intel's own website doesn't do very good job of exposing TurboBoost info, but I was able to dig up these:
http://www.intel.com/content/www/us/...ody_turboboost - see: "Note: Intel® Turbo Boost Technology 2.0 allows the processor to operate at a power level that is higher than its TDP configuration and data sheet specified power for short durations to maximize performance."
http://www.intel.com/content/www/us/...000005641.html - What is the difference between Intel® Turbo Boost Technology 2.0 and Intel® Turbo Boost Max Technology 3.0? Intel® Turbo Boost Technology 3.0 allows you to direct workloads to the fastest core.

I believe I'm going to stop at this point.