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
yannigr2 - Thursday, September 19, 2019 - link
"(3) because it just wasn’t worth drawing attention to it."Oh, no no, please do an article about this. A comparison with Intel's marketing slides from, lets say, 2013, when Intel had clearly the upper hand would be interesting.
eva02langley - Thursday, September 19, 2019 - link
You don't get it, Intel goal was to draw attention and sow confusion by spreading FUD to damage AMD momentum.Anandtech acted professionally in this matter and I salute them for it. I get your point that writting an article that could backfire on them would have its advantages, however it would just continue feeding this FUD thrown by Intel. It would just help Intel in the end to maintain the doubt that should not even exist in the beginning.
casperes1996 - Thursday, September 19, 2019 - link
Ah, yes. The Heisenberg Turbo-certainty principle at play.And of course not to mention the deeply philosophical quantum of a processor bursting but nobody noticing, did it even burst?
Jest aside, great, balanced article. Good to get misconceptions dealt with. I like AMD's approach to turbo. Only marketting could be argued it a negative I'd say. In general I think it's almost like we're going back to the frequency wars. Marketting should instead emphasis how little importance frequency has. And really it should almost be obvious to people, since pretty much everyone has tried upgrading their computer with a newer chip, experiencing a much greater difference than if frequency was the determining factor. Don't know about you lot, but I'd rather have a modern chip at 1GHz than (if it were possible) a Motorola 68K at 8GHz
AntonErtl - Saturday, September 21, 2019 - link
When I bought my first Intel CPU with Turbo (a Core i7-6700K), my expectation was that base was guaranteed, and Turbo is an unpredictable bonus in speed. So not everyone has the expectations that you claim everybody has because of experience with Intel. Later, my experience was that the base frequency means very little: With little load, the CPU clocks at some low frequency; with load, it clocks beyond the base frequency. And then there are things like the AVX offset, which seems to affect even my laptop's Core i3-3227U (which clocks at 1.7GHz instead of 1.9GHz when running matrix multiplication code), although AVX offset is not documented for this CPU.Back to AMD, in hindsight it might have been a better marketing strategy for AMD to announce the Ryzen 3000 CPUs with 100MHz less boost.
Concerning the article, I found the way strange in which the different approaches of Intel and AMD to deal with the limits of the hardware were portrayed. If I pay the same amount, and get the same stock performance, then I prefer a CPU with more headroom (for either overclocking or future degradation). If I pay the same amount and get the same overclocking performance (i.e., headroom), then yes, a CPU that uses more of that headroom is probably preferable (at least if it does not affect durability); but does the pricing of the CPUs actually reflect the headroom? In any case, for non-overclockers price/stock-performance and durability are relevant, and headroom is merely interesting.
lakedude - Saturday, September 21, 2019 - link
This per core binning is basically giving you a chip with ZERO headroom, none. Intel chips like the famous A300 cel have typically had great headroom with a few exceptions over the years. I applaud AMD's recent efforts but they are pushing their chips to the max and overstating their ability. This not guaranteed crap is for the birds. I'd be pissed if I had purchased a chip and it didn't get the full turbo that was advertised. If the chip can't hit a certain performance level with certainty then use more weasel words like "up to". As in "up to 4GHz turbo possible with some luck"Korguz - Saturday, September 21, 2019 - link
with that, intel cpus should also have on their boxes : at base clocks, our cpus use xx watts, but, in real world usage, and if you over clock, our chips can use 100 to 150 watts more.lakedude - Sunday, September 22, 2019 - link
Agreed, same type of thing with Intel's TDP. They blamed the motherboards, roll eyes.John_M - Saturday, September 21, 2019 - link
Thing is, AMD doesn't call it "Turbo". If tech journalists and reviewers called "Precision Boost" by its proper name there might be less confusion amongst the masses.lakedude - Sunday, September 22, 2019 - link
Precision Boost is more descriptive but the problem isn't so much with the name as the rating.Before reading this article I would have expected even the slowest core to be able to reach the advertised Boost speed. For example a AMD RYZEN 7 3700X 8-Core 3.6 GHz (4.4 GHz Max Boost) should be able to hit 4.4 GHz with each and every core and for longer than a few nano-seconds. Of course being unlocked OC frequencies beyond 4.4 GHz would not be assured.
Now both Intel (TDP) and AMD (Boost) are pushing the limits of their technology and their own terminology.
Good thing we have AnandTech!
Badelhas - Saturday, September 21, 2019 - link
This is an awesome article, this is why I love AnandTech. Thank you.