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
GeoffreyA - Wednesday, September 18, 2019 - link
Excellent article and detective work, Ian! Thank you for it. Also reminds me of observation in QM, where experiment affects the results. Anyway, have a great day.eva02langley - Wednesday, September 18, 2019 - link
"However, given recent articles by some press, as well as some excellent write-ups by Paul Alcorn over at Tom’s Hardware* "Please, I know you are parent sites, but HELL with that. Paul literally test the hardware with the SAME motherboard, the MSI GODLIKE x570 and never... ever mentioned anything close to a BIOS issue. He did an half-ass job that I could call as amateurish at best.
In the meantime, Steve from HardwareUnboxed tested the same CPU on DIFFERENT board and concluded into BIOS immaturity, what I called on the first instance of Toms series of bashing article.
ajlueke - Thursday, September 19, 2019 - link
I wouldn't necessarily agree, but the spirit of the statement is on track. In Paul Alcorn's write-up he attempted to associate the missing boost MHz, to a statement Shamino made about reliability, and then changes in thermal thresholds observed by "The Stilt".He never bothers to explain, why single threaded boosting (the thing everyone is complaining about) would be related to a threshold change from 80C to 75C when those temperatures are never observed during a lightly threaded workload. He then heats the boards up to those temps and looks at boosting, and sure enough, something changed just like the Stilt said. But what, if anything, does that have to do with the missing single thread boost MHz, when temps are well below 75C for most end users?
eva02langley - Wednesday, September 18, 2019 - link
" 1. Popular YouTuber der8aur performed a public poll of frequency reporting that had AMD in a very bad light, with some users over 200 MHz down on turbo frequency,2. The company settled for $12.1m in a lawsuit about marketing Bulldozer CPUs,
3. Intel made some seriously scathing remarks about AMD performance at a trade show,
4. AMD’s Enterprise marketing being comically unaware of how its materials would be interpreted."
And in the meantime in the same week...
1. https://www.extremetech.com/computing/297627-amd-o...
2. https://www.extremetech.com/computing/297785-amd-s...
Like I told AdoredTV... we have a very different definition of BAD WEEK. Honestly, those issue are hiccup of any new platform launch.
eva02langley - Wednesday, September 18, 2019 - link
"Others we ignored, such as (4) for a failure to see anything other than an honest mistake given how we know the individuals behind the issues, or the fact that we didn’t report on (3) because it just wasn’t worth drawing attention to it."The reason why you guys are pros. You didn`t do Intel dirty work for propagating their propaganda... unlike TomsHardware...
quadibloc - Wednesday, September 18, 2019 - link
Both Intel and AMD should start marketing their chips as "an X GHz chip", where X is the base frequency, if the turbo frequency isn't a part of the basic specirication of the chip that it must meet. Since even at the base frequency, apparently AMD chips don't last forever, it looks like I'm going to be underclocking mine a little.ballsystemlord - Wednesday, September 18, 2019 - link
Spelling and grammar corrections:"Certain parts of how the increased performance were understood,..."
Should be "was" not "were":
"Certain parts of how the increased performance was understood,..."
"...(standard is defined be Intel and AMD here, usually with a stock cooler, new paste, a clean chassis with active airflow of a minimum rate, and a given ambient temperature)..."
"by" not "be":
"...(standard is defined by Intel and AMD here, usually with a stock cooler, new paste, a clean chassis with active airflow of a minimum rate, and a given ambient temperature)..."
"This ultimately would lead some believe that this relates to a thermal capacity issue within the motherboard, CPU, or power delivery."
Missing "to":
"This ultimately would lead some to believe that this relates to a thermal capacity issue within the motherboard, CPU, or power delivery."
Uroshima - Thursday, September 19, 2019 - link
Very nice article.From what I understood, AMD has done tried to get as close to the limit of the silicon as possible regarding clocks. This allowed them to "survive" the transition to 7nm. Intel has kept a wide margin to the actual limits of the silicon and at 10nm (which is more or less 7nm of AMD) they struggle as the chips simply can't clock high enough.
Could be, this is the reason Intel will stick with 14nm for high performance until new silicon comes out that is similar to the AMD "to the limits" approach? This would be roughly 3 years from when they decided this (Jim Keller's arrival?).
I have a hunch that this is the future we are going towards, new nodes with diminishing returns (or even reductions) on clocks but advantages in power and number of transistors. Keeping close to the limit of the silicon will be the key for performance, right next to IPC.
On the other hand I would even consider that for some applications, having a refined 14* nm process could be an advantage (up to a frankenmonster of a hybrid 7/14 with UV). Intel, with its vast resources, should definitely explore this option to not only follow the competition but maintain the low thread performance crown.
But then, looks like AMD did their homework this time. :)
eva02langley - Thursday, September 19, 2019 - link
You are bang on. Intel 10nm process cost more, is having low yield and the frequency drop over 14nm++ is not bringing meaningful performances for making the transition.This is why Intel is releasing new server, laptop and desktop CPUs on 14nm++. It cost less, having better yield and perform better.
eva02langley - Thursday, September 19, 2019 - link
However the power consumption just cannot match TSMC 7nm.