1. ECE 486/586 Computer Architecture Chapter 12 Turbo Boost Technology
Herbert G. Mayer, PSU
Status 2/28/2017

2. Syllabus Introduction Speedup Parameters Definitions Turbo Boost
Turbo Boost, Actual Performance
Turbo Boost, on DP Configuration
Turbo Boost, for SPEC JBB 2005
Bibliography

3. Introduction
Turbo Boost (turbo) first used on Intel® CoreTM i7 processors and Intel Xeon® processor 5500 Series
Turbo dynamically enables some temporary performance boost
Turbo temporarily increases processor core clock in defined, discreet frequency steps (AKA bins) for the benefit of higher, but safe performance
Permissible, only while conditions on physical chip safely preserve the physical health of the CPU
We discuss, how much a core frequency can be raised as a function of the number of active cores and of other architectural parameters
Not to be confused with overclocking or its opposite thermal throttling

4. Introduction
Key learning for ECE students, who in the future may work in Hi Tech Processor Development:
Lab setup differs significantly from final product
For performance evaluation: Take into account open vs. close chassis –air flow! E.g. see page 26!
Also consider rotating fans vs. passive heat sinks
Measure each performance data point repeatedly, and devise consistent policies, how to use various data measured for supposedly “identical” test case
Have an objective method for: Which cases to throw out, which to keep? Which ones to measure repeatedly, and why? Always the same method!
Whichever EE policy is adopted:
use it consistently!

5. Frequency & Voltage Independent Interface
Introduction
Core i7 introduced turbo: Intel 45 nm High-K Silicon technology, launched as High-End Desktop platform with 1 socket; later used as server with 2 sockets
Frequency & Voltage Independent Interface
DRAMs
Last Level Cache
QPI
Pwr
&
Clk C O R E 1 2 3
IMC
DDR3 U N S
High Level Nehalem Architecture

6. Speedup Parameters
Without changing any other HW design parameters, the following technology attributes increase computational speed of uni-processor CPU:
Faster core clock
Artificially cooling CPU, then accelerate core clock
Wider memory bus
Faster bus
Larger caches
Single-clock caches
Multi cache hierarchy
Faster memory technology (e.g. lower latency)
Specialized instructions, to replace multiple ops with single
Instructions that avoid change in flow of control, e.g. CMOV
Uniform instruction format, e.g. RISC
Typically accompanied by increased cost of HW!

7. Definitions Enhanced Dynamic Acceleration Technology, or EDAT
Prior to Turbo Boost in Core i7, Intel’s previous generation Core 2 Duo introduced EDAT
Allows one core to automatically increase operating frequency, if other cores of same product are in sleep state or idle; is a safe boost!
Depending on the number of active hardware threads and on electrical and thermal parameters, before taking advantage of a clock boost within the product constraints
Like Turbo Boost, EDAT is a Green technology, as it provides performance on demand, while keeping power consumption low, when other processors’ operation is temporarily not needed

8. Definitions Turbo Boost Turbo is distinct from overclocking & EDAT
Turbo boost is part of a silicon product spec
It does not change the durability of part, i.e. the increased clock rate is a defined attribute of the silicon product!
Overclocking, by contrast, increases clock frequency by running outside parts spec.; will overheat the part; or requires special cooling
Turbo technology runs processor within product specification; aims to take advantage of optional thermal headroom available during under- utilized conditions, i.e. when other cores are idle

9. Definitions –Turbo Boost
Hyper-Threading Looks Like Multi-Processor Under Right Conditions

10. Definitions Overclocking
Overclocking is a forced increase of the natural processor clock speed beyond specification!
It is an Unnatural Act  unhealthy for life of CPU
Overclocking results in running processor outside its specified and thus safe limits
Voids Warranty!
Overclocking not a Green technology, as it forces increased power consumption continuously regardless of temperature
Overclocking can break a processor; so done in Las Vegas speed completions, but usually with extraordinary cooling methods

11. Definitions –Overclocking
Overclocking: A Special Tool Meant for Performance Enthusiasts

12. Overclocking Not Meant for Stable Execution!
Definitions
Overclocking Not Meant for Stable Execution!

13. Definitions Thermal Throttling
When thermal parameters change, or when the number of active cores increases, prior clock increase is reversed, saving the chip from possible damage, and saving power
Thermal Throttling results in decreased performance
Assumes μP to be running in steady state of execution, acknowledges that temporary hot spots are possible
Which then results in thermal throttling to save chip from self-destruct!

14. Definitions Thermal Throttling, Cont’d
Happens when a typical mix of IO-bound & compute-bound execution is replaced by compute-bound only execution, resulting in more heat generation than is safe
Similar to the safety action taken in Turbo mode, the frequency is throttled, resulting in less current, less heat generation, ultimately in lower performance
Architects must decide, which safe technology of performance boosting should be realized in Silicon, one, the other, or both

15. Turbo Boost, sometimes referred to as Dynamic Overclocking:
Definitions
Turbo Boost, sometimes referred to as Dynamic Overclocking:

16. Detail on Turbo Boost

17. Turbo Boost
A number of dynamic parameters dictate the upper limit of Turbo Boost speedup
E.g. the core’s temperature, momentary frequency, the overall current, momentary power, and the total number of active cores
Each frequency step of turbo boots is MHz. For each SKU, fuse values are set during chip manufacturing, to define upper bound, by how many frequency steps maximally a core can grow safely
Table parameters d-c-b-a mean: If 1 core is active, that core’s frequency may increase by a bins. Else if 2 cores are active, these cores can grow by b frequency steps, etc.

18. Turbo Boost
Applying that encoding principle, starting at opposite end, some table entry means that for 3 or 4 cores being active, the frequency may increase by just 1 frequency step (frequency step AKA bin)
But if only 2 cores are busy, the speed may grow up to 4 steps, or 4 bins
And if only a single core is active, the current one may grow by 8 frequency steps, amounting to 1.06 GHz of additional clock ticks

19. Turbo Boost
However, this boost is to be reversed, if for any reason a predefined envelope of maximally allowable current or temperature is exceeded
Decrease is designed not only to save the microprocessor from thermal stress, but to save power and run more green
Similarly, as bound shows, other cores may become active, forcing a current high boost rate to decrease, again to protect the processor and save power
When Turbo Boost Technology promotes core to a higher frequency, the processor will draw more current than it would while running at nominal frequency

20. Turbo Boost
User incurs incremental cost for electrical power consumed in turbo mode
But such cost is minor compared to the power used by the system as a whole
Users may choose to manually adjust the balance between performance and power consumption through the OS power policies
Performance teams at Intel focused on workloads known to be CPU-centric, and concentrated on single- and multi-threaded workloads for turbo performance data . . .

21. Turbo Boost, Actual Performance
Then proceeded by running three baseline frequencies without enabling turbo
Base frequencies were 2.66 GHz, 2.8 GHz, and 2.93 GHz to simulate the lower and upper bounds of the workload.
Setting affinity manually, and forcing workloads to run on a single CPU allowed maximum benefit from Turbo
Affinity: either associate a thread with a dedicated core or hyper-thread

22. Operating Frequency as a Function of the Number of Running Cores
Turbo Boost
Operating Frequency as a Function of the Number of Running Cores

23. Turbo Boost, Actual Performance
Setting processor affinity: an application manually tells the OS scheduler where to run, i.e. restricts available hardware threads where the workload may be executed
E.g. setting Affinity = p3, tells the OS scheduler to only run on Processor 3
Setting Affinity = p0, p2, p3 allows an app to run on hardware thread 0, 2, or 3
Restricting affinity to few, even to a single core, improves performance for singly threaded SW, as migration and associated saving, restoring are minimized
Can improve performance for sophisticated user!

24. Turbo Boost, on DP Configuration
Table on p. 25 summarizes DP Turbo setup
Used in engineering validation board, called Green City: open bench top configuration
With different thermal conditions compared to a production end-user environment in a standard chassis
Each processor has an individual heatsink with active fans attached, in addition to 4 external fans on side
If workload does not hit Turbo constraints, the Core frequency can increase up to GHz dynamically on Green City, depending on number of active cores

25. Turbo Boost, on DP Configuration
DP Configuration Parameters for Experimental Setup

26. Turbo Boost, on DP Configuration
DP Configuration Physical Setup

27. Turbo Boost, for SPEC JBB 2005
Evaluated SPEC JBB 2005 benchmark
Comparing the impact of the simultaneous multi-threading (SMT) under Turbo mode
Turbo provides the upside with and without SMT while best performance is achievable with SMT and Turbo for this workload
In this case, the benchmark rarely hit TDP, which is indicated by the unconstrained Turbo frequency
Def: TDP is Thermal Design Power, AKA Thermal Design Point is the maximum heat, generated by μP that the cooling system can handle

28. Turbo Boost, for SPEC JBB 2005
Turbo Performance for SPEC JBB 2005

29. Bibliography
2009 Markus Mattwandel et al. for Intel SW College: “Architecture and Early Performance Results of Turbo Boost Technology on Intel® CoreTM i7 Processor and Intel Xeon® Processor 5500 Series”
2008 November 3, Intel website %AE%20Core%E2%84%A2%20i7_Overview.pdf “Intel® Core™ i7 Microprocessors, The Best Processor on The Planet”
General SPEC website
2006 August, SPEC website for integer component of SPEC CPU2006:
2003 October, SPEC website for floating point component of SPEC CPU2000:
Intel® Turbo Boost technology,
Intel Turbo Boost technology 2.0: chttp:// technology/turbo-boost/turbo-boost-technology.html