Can anyone demonstrate Turbo Boost on a Core i5/7 MBP?

[briancl]

I can't.

So I had a few minutes and decided to play around with fan settings to see if I could influence or measure Turbo Boost. My thinking was that if I ran the chip hot, then there would be a smaller thermal envelope and there would be minimal effects from turbo boost. Likewise, if I ran the chip cool, there would be a larger thermal envelope and greater effects from Turbo Boost. Here is how I conducted the experiment.

Core i7 MBP w/ 8gb RAM, OCZ Vertex 2 SSD

1) Used smcFanControl with max fan (7200 rpm) for ~5 minutes to get my CPU temp to 35-37 C. Ran benchmark. Repeat 3 times
2) Used smcFanControl with minimal fan (2000 rpm) and used "yes > /dev/null" in 4 terminal sessions for ~2 minutes to get my CPU temp to around 95-98 C. (temp hit 95C in about 20 seconds. I let it burn in to generate a lot of heat). Ran benchmark. Repeat 3 times.

I used both geekbench and xbench as my benchmarks. Geekbench even separates out the single-threaded performance, which should theoretically benefit the most from Turbo Boost (single-core load turbos 2 bins, multi-core load turbos 1 bin). Regardless of this, my scores never fluctuated when the CPU was hot or cold. Turbo seems to be unaffected by heat, even though heat is allegedly its primary limiting factor.

I also used the MacCPUID tool from Intel and continuously monitored/refreshed the CPU speed to see if it ever changed from 2.66ghz. It never did regardless of the workload or temp. I would have thought for sure that the CPU speed would increase during a maxed single thread load under low temps (7200rpm). Nope. "sysctl -a hw |grep cpufreq" also never changed under any circumstance. It always displays 2.66ghz.

I also tried to get the CPU to SpeedStep/EIST down, but I couldn't seem to monitor that either. I was using sysctl and MacCPUID, but leaving my system idle for 30 second periods under 90 and 80% battery life had no effect.

I'm going to continue to play around with this, but I'm open to ideas on what to try.

 

[Eddyisgreat]

Don't think heat would affect it much but honestly, I haven't put much thought into it .

I'd try the Turbo Boost Widget (if you have windows available). That's been able to show me the boost in real time.

 

[AdamRock]

Don't think heat would affect it much but honestly, I haven't put much thought into it .

I'd try the Turbo Boost Widget (if you have windows available). That's been able to show me the boost in real time.

theirs no OS X version? wtf.

 

[MacsOnAnabolics]

First figure out which application(s) would induce a turbo boost of the intel processor. Then just run that applicatiion...

However, this is a great question. I would also really like to see an answer to this.

 

[briancl]

I don't have a bootcamp Windows partition, but I do have VMware with Windows. However, the Intel widget doesn't work in VMware. Fortunately, CPU-Z does.

CPU-z shows 2.66ghz at all times under all types of single and multi-thread workloads, regardless of temperature. I've tried single and multi-threaded workloads in both native OSX and in my Windows VM. Nothing can make the CPU frequency change from 2.66ghz.

I also retried all of my tests from the top post with the Nvidia GPU forced via gfxCardStatus. I had a hunch that disabling the on-chip Intel GPU would free up some thermal envelope and help induce Turbo, but I saw no change.

 

[briancl]

I've continued some other testing without any luck. Anyone else have ideas?

 

[orangepeel]

Maybe apple disabled turbo boost. They throttle the cpu's, as well as underclock the gpu's all the time. The i7 chip gets very hot, it goes well over 200 degrees fahrenheit. When I used cinebench to benchmark my 2.8ghz i7, the chip went as high as 209 F.

I also monitored the temps while playing starcraft 2. Temperature would go over 200 F, the fans would kick into high and the temp will stabilize at around 190 F. I used istat to monitor the temps.

 

[Eddyisgreat]

Maybe apple disabled turbo boost, they throttle the cpu, as well as underclock the gpu all the time.

The GPU is undercooked, yes, but they tout turbo boost as a primary feature on their product page. To disable it would be false advertising to the max.

 

[orangepeel]

Apple has been know to use false advertising. Did you know that the 5750 gpu in the iMac is actually a mobility radeon 5850?

 

[briancl]

I feel like turbo boost is active and easily detected under Windows via BootCamp on MBP's. I don't have Windows, so I can't really say, but that would at least imply that the hardware is capable of turbo boost. Sure, OSX could somehow block turbo, but that seems like a bit more of a stretch. The OS isn't what tells the CPU to boost or not. It's supposed to be governed by available thermal headroom.

 

[ciaran00]

No turbo boost is not disabled. Why would it be? There is no better alternative to what it offers, period.

 

[tjb1]

You cant force the fans to 2000 so you prove no point with that because when the processor gets hot enough that it cant run at its max speed it will speed the fans up to cool it down.

 

[briancl]

You cant force the fans to 2000 so you prove no point with that because when the processor gets hot enough that it cant run at its max speed it will speed the fans up to cool it down.

This is difficult to decipher, but my experiments were to show that neither cooler or warmer CPU operating temps made any measurable changes to performance. A cooler operating temp should induce turbo boost more often, especially under single core workloads. It doesn't in OSX.

 

[Eddyisgreat]

Well, your proc is probably boosting. The Turbo Boost white paper speaks specifically towards inaccurate reporting:

Most currently available software that calculates frequency relies on the ACPI P-state structures that are provided by the BIOS or calls an application programming interface (API) that relies on these structures. The ‘frequency’ field in the P0 ACPI _PSS object does not reflect actual Intel® Turbo Boost technology frequencies. Instead, it always shows the P0 frequency as P1 frequency plus 1 MHz. Due to this, software that uses the P- state structures, directly or indirectly, will not properly display Intel® Turbo Boost technology frequencies.
NOTE: Operating Systems may also be impacted and may not report accurate frequency.

 

[briancl]

Well, your proc is probably boosting. The Turbo Boost white paper speaks specifically towards inaccurate reporting:

Yep. I totally agree. That's why I haven't been relying on software to "measure" CPU frequency. I've been trying to measure it through benchmarks to actually witness it's effects.

In Windows, it is very easy to do this. Turbo boost behaves as it should. Also, software exists to accurately report boosting CPU's, so it's even easier to verify.

 

[tjb1]

This is difficult to decipher, but my experiments were to show that neither cooler or warmer CPU operating temps made any measurable changes to performance. A cooler operating temp should induce turbo boost more often, especially under single core workloads. It doesn't in OSX.

Using SMC you cannot keep the fans at 2000, when the core gets hot enough that it needs cooled to continue its current workload the fans will speed up and maintain a working temperature. Considering you dont know the limits of the processor, it could Turbo Boost as high as 200F...you have no idea what that limit is so your test proves nothing. AND I bet the software/bios/os whatever is set to keep the processor cool enough to turbo boost if needed.

 

[Eddyisgreat]

Well..uhh...

...this is what I get (native of course)....