topic I>>>t is possible that the OS in Software Tuning, Performance Optimization & Platform Monitoring

Disabling Turbo Boost

Samuel_M_1 — Mon, 20 May 2013 23:04:58 GMT

As per the recommendation in this topic (http://software.intel.com/en-us/forums/topic/385319), I have used this sequence to disable turbo boost:

"section 14.3.2.2 OS Control of Opportunistic Processor Performance Operation (Vol. 3B page 14-5 of the May 2012 version)

System software can temporarily disengage opportunistic processor performance operation by setting bit 32 of the IA32_PERF_CTL MSR (0199H), using a readmodify-write sequence on the MSR. The opportunistic processor performance operation can be re-engaged by clearing bit 32 in IA32_PERF_CTL MSR, using a readmodify-write sequence. The DISENAGE bit in IA32_PERF_CTL is not reflected in bit 32 of the IA32_PERF_STATUS MSR (0198H), and it is not shared between logical processors in a physical package."

Source: Patrick Fay, Intel

However, my Intel Xeon X5650 (http://ark.intel.com/products/47922/Intel-Xeon-Processor-X5650-12M-Cache-2_66-GHz-6_40-GTs-Intel-QPI) is still operating at its maximum turbo frequency of 3.06 GHz. Is there something else that I need to do to disable turboboost?

Hello Samuel,

Patrick_F_Intel1 — Tue, 21 May 2013 02:31:33 GMT

Hello Samuel,

Is bit 32 of MSR 0x199 staying set? That is, if you read the MSR after you set the bit, do you see that the bit is still set? And does it stay set?

It is possible that the OS is overwriting the MSR (and re-enabling turbo-mode).

Is this windows or linux? On linux you can control the frequency with the /sys/devices/system/cpu/cpu*/cpufreq/ interface. Let me know if you interested in this.

Pat

Hi Pat,

Bernard — Tue, 21 May 2013 05:08:49 GMT

Hi Pat,

what exactly does "opportunistic processor performance" mean?

Thanks in advance.

Hello Illyapolak,

Patrick_F_Intel1 — Tue, 21 May 2013 11:45:21 GMT

Hello Illyapolak,

"opportunistic processor performance" refers to the strategy of trying to find opportunities to improve the performance while staying within a power, current and temperature envelope. For example, a multi-core processor is designed to not exceed some maximum power usage when all the cores are busy. Much of the time however the user might be running just 1 or a few threads so there is the opportunity to increase the performance on 1 core and still stay within the power and thermal restrictions. So the processor boosts the frequency of 1 core, boosts the performance of the 1 thread and stays within the power/thermal limits.

There is a brief discussion at http://www.intel.com/content/www/us/en/architecture-and-technology/turbo-boost/turbo-boost-technology.html

Pat

Thank you Pat for detailed

Bernard — Tue, 21 May 2013 14:05:16 GMT

Thank you Pat for detailed answer.It always puzzled me what "oportunistic" really means.

>>However, my Intel Xeon

SergeyKostrov — Wed, 22 May 2013 13:54:00 GMT

>>However, my Intel Xeon X5650 (http://ark.intel.com/products/47922/Intel-Xeon-Processor-X5650-12M-Cache...) is still operating >>at its maximum turbo frequency of 3.06 GHz. Is there something else that I need to do to disable turboboost? Could you repeat your tests when you computer is running on a battery? Update: It is Not clear if you're using a Desktop or a Notebook. Please disregard the question if it is the Desktop computer.

@Sergey: It is a Desktop. I

Samuel_M_1 — Wed, 22 May 2013 16:49:00 GMT

@Sergey: It is a Desktop, running Mac OS X 10.8. I am going to try disabling the apple power management kernel extensions, which may be overwriting the value. I will post solution if this works.

I>>>t is possible that the OS

Bernard — Wed, 22 May 2013 17:51:12 GMT

I>>>t is possible that the OS is overwriting the MSR (and re-enabling turbo-mode).

Is this windows or linux? On linux you can control the frequency with the /sys/devices/system/cpu/cpu*/cpufreq/ interface. Let me know if you interested in this.>>>

On Windows there is an option to put a breakpoint on MSR 0x199 access and inspect the code which is modifying(writing) to this register.

Thanks for the update.

SergeyKostrov — Thu, 23 May 2013 00:04:24 GMT

Thanks for the update. >>...It is a Desktop, running Mac OS X 10.8. I am going to try disabling the apple power management kernel extensions, which >>may be overwriting the value. I will post solution if this works... I was thinking that some Power Management software subsystem could affect your tests. It would be very interesting to see your results. Did you check BIOS settings?