Reaching for Turbo: Aligning Perception with AMD’s Frequency Metrics
by Dr. Ian Cutress on September 17, 2019 10:00 AM ESTA Short Detour on Mobile CPUs
For our readers that focus purely on the desktop space, I want to dive a bit into what happens with mobile SoCs and how turbo comes into effect there.
Most Arm based SoCs use a mechanism called EAS (Energy-Aware Scheduling) to manage how it implements both turbo but also which cores are active within a mobile CPU. A mobile CPU has one other aspect to deal with: not all cores are the same. A mobile CPU has both low power/low performance cores, and high power/high performance cores. Ideally the cores should have a crossover point where it makes sense to move the workload onto the big cores and spend more power to get them done faster. A workload in this instance will often start on the smaller low performance cores until it hits a utilization threshold and then be moved onto a large core, should one be available.
For example, here's Samsung's Exynos 9820, which has three types of cores: A55, A75, and M4. Each core is configured to a different performance/power window, with some overlap.
Peak Turbo on these CPUs is defined in the same way as Intel does on its desktop processors, but without the Turbo tables. Both the small CPUs and the big CPUs will have defined idle and maximum frequencies, but they will conform to a chip-to-chip defined voltage/frequency curve with points along that curve. When the utilization of a big core is high, the system will react and offer it the highest voltage/frequency up that curve as is possible. This means that the strongest workloads get the strongest frequency.
However, in Energy Aware Scheduling, because the devices that these chips go into are small and often have thermal limitations, the power can be limited by battery or thermals. There is no point for the chip to stay at maximum frequency only to burn in the hand. So the system will apply an Energy Aware algorithm, combined with the thermal probes inside the device, to ensure that the turbo and workload tend towards a peak skin temperature of the device (assuming a consistent, heavy workload). This power is balanced across the CPU, the GPU, and any additional accelerators within the system, and the proportion of that balance can be configured by the device manufacturer to respond to what proportion of CPU/GPU/NPU instructions are being fed to the chip.
As a result, when we see a mobile processor that advertises ‘2.96 GHz’, it will likely hit that frequency but the design of the device (and the binning of the chip) will determine how long before thermal limits kick in.
Facebook
Twitter
RSS
144 Comments
View All Comments
Kangal - Tuesday, September 17, 2019 - link
Yeah a Three Scenario figures sound best. It's easiest for consumers to understand. They should apply it for their GPUs and CPUs, and for all products ranging from Mobile to Server.Example:
Base Clock --- Burst Clock -- Boost Clock
3GHz 8 core - 5GHz 1 core - 4GHz 8 core
90W TDP ----- 120W TDP --- 150W TDP
teamet - Tuesday, September 17, 2019 - link
Whatever.. It's fine no matter what. It's just .. I mean .. why couldn't they just write 4.35 GHz on the box or whatever. The performance is there. No need to try and over-exaggerate clock.It feels like a battle between engineers and marketing, and marketing got to write the clock on the box in the end.
mczak - Tuesday, September 17, 2019 - link
IMHO AMD made a mistake anyway trying to extract every little bit of performance. Leave those last 3% to the overclockers - they aren't really relevant but come at the expense of quite a lot of efficiency.franzeal - Thursday, September 19, 2019 - link
So you think it's better they create an inferior product for the masses, so that a few can feel like they got something extra/free? They did good.mikato - Monday, September 23, 2019 - link
Agree. This is a step forward as long as any risks are minimal. Why leave some performance only for overclockers if you don't have to?Arnulf - Tuesday, September 17, 2019 - link
This is an industry-wide thing (have a look at the mobile SoCs, e. g. Qualcomm, ...). They love rounding their frequency figures up, reality be damned!DroidTomTom - Tuesday, September 17, 2019 - link
Reminds me of the days of the AMD Athlon XP 2500+ Barton 1.833 GHz Socket A CPU's. They were faster IPC than Intel but people only paid attention to frequency which was misleading. AMD was trying to put the focus on overall performance per dollar.Dribble - Tuesday, September 17, 2019 - link
The graphs on pages 2 and 3 tell you all you need to know. AMD just run there cpu's far to close to the limit and give you numbers that are higher then what is understood by boost (it would be like Intel advertising boost as 5.2 not 4.6).It's deceptive really as the number does not mean the same thing as Intel's number. I am sure I'll get 20 replies saying how this is fine, but if it were Intel doing this then I bet the same people would be roasting Intel for this sort of miss-information.
Karmena - Wednesday, September 18, 2019 - link
first, AMD does not call it Turbo, like Intel does. It is Boost. The problem is you assuming that those are the same. Same as in legal speak, you have to go and look up the definition of what is claimed. Or keep every company accountable. As most of the time they add that trademark thingy, and in the end that just means that the company gets to set up the definition of what it is. SMT vs HT.Spunjji - Monday, September 23, 2019 - link
I can't believe anyone would make it all the way through such an informative article like this and then still post such an ignorant comment. The only reasonable conclusion is that you did not take the time to read it properly, because everything you said was discussed at length.