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
Korguz - Thursday, September 19, 2019 - link
so i guess you are ok with intel marketing their chips to only use the watts they state, but under real usage, they can, specially when overclocked, they can use MUCH more ??Cooe - Tuesday, March 23, 2021 - link
It doesn't say "4.7GHz Turbo" it says "Max 4.7GHz Turbo". Aka "UP TO!". It's not AMD's fault if "Average Joe" literally doesn't know how to read...Oliseo - Wednesday, September 18, 2019 - link
Also, if YOU have to write a multi page highly technical article on why AMD is right and all the customers are wrong, then you're just a fanboy who doesn't give a **** about the customer.We get enough of this in the comments with people in the media joining in to make excuses for large multinationals at the expense of the average consumer.
Shame on you!
eva02langley - Wednesday, September 18, 2019 - link
Here, these people will escort you to your new home, the asylum... cheers and enjoy you stray jacket...Calabros - Wednesday, September 18, 2019 - link
This the funniest First World Problem I've seen in 2019.regsEx - Wednesday, September 18, 2019 - link
I can't find any reference in CFL documentation that PL1 is for base clocks. It doesn't seem to have any correlation.edzieba - Wednesday, September 18, 2019 - link
Sounds like the "Intel Performance Maximizer" is exploring and characterising the frequency-space above the all-core turbo clock. That could be done at the factory to provide faster out-of-the-box performance, but would either introduce chip-to-chip variation is stock (rather than OC) performance, or create even more SKU bins. And people already throw a wobbler over he number of SKUs Intel produce (even though 90% of those are not a consideration unless you order direct from them in $multimillion batches).samboy - Wednesday, September 18, 2019 - link
My biggest concern is for the 3950x.This has the highest Turbo specification of all the processors; which was a good selling point when I first saw the specifications
However, it is becoming clear that the all-core base clock speed specification is more important. Particularly given that the 3900x seems to take the least "advantage" of Turbo out of all the current chips.
Comparing the specifications of 3950x to 3900x we see a 300MHZ drop off on base core speed; in exchange for an extra 100MHZ in boost - which seems somewhat questionable now.
A 50% increase in cost for an extra 4 cores and likely slower throughput for tasks that use 12 or less cores (which covers almost everything that is run in practice today) this doesn't look like particularly good value. My untested assumption is that the 300MHZ less in base speed is the more important number and will translate to slower throughput.
I'll wait for the reviews for the 3950x; but I expect that I'll be leaning for the 3900x for the second system I need to build.
oleyska - Wednesday, September 18, 2019 - link
3900X boosts higher than any other ryzen chip out today.I think 3950X will be faster than 3900X, on ryzen amount of cores doesn't matter at all.
Have a ryzen 1700? it isn't at all limited in comparison to a 1600 and on average seems to hit higher frequencies and it's absolutely true for 3000 series, even More so!
ajlueke - Thursday, September 19, 2019 - link
"A 50% increase in cost for an extra 4 cores and likely slower throughput for tasks that use 12 or less cores (which covers almost everything that is run in practice today) this doesn't look like particularly good value."The 3950X will almost certainly be faster than the 3900X in just about every scenario. If you consider the way the two chiplet CPUs are binned, they have one fast chiplet (CCD) and one that is not nearly as fast.
In the 3900X these CCDs are 6 cores each, while they are 8 cores in the 3950X. Which means in an 8 thread workload, the 3950X can do that entire workload on the "better" CCD0, which should have a better frequency/voltage curve. The 3900X will have to dip into 2 cores on CCD1. The 12 core workload is a similar story, the overall frequency/voltage curve will be better on the 3950X because it is boosting more "better" binned cores than the 3900X with the same power envelop.