The Mobile CPU Core-Count Debate: Analyzing The Real World
by Andrei Frumusanu on September 1, 2015 8:00 AM EST- Posted in
- Smartphones
- CPUs
- Mobile
- SoCs
Hangouts Launch
The duration of the test this time is only 3.6 seconds. During the initial application launch, we don't see much activity on the little cores. Cores 1-3 are mostly power-gated and we see that there's little to no threads placed onto the cluster during that period. Once the app opened, we see the threads migrate back onto the little cluster. Here we see full use of all 4 CPU cores as each core has threads placed on it doing activity.
This is the perfect burst-scenario for the big cores. The application launch kicks in the cores into high gear as they reach the full 2.1GHz of the SoC. We see that all 4 cores are doing work and have thread placed on them. Because of the fine granularity of the load, we see the CPUs rarely enter the power-gating state in this burst period as the CPU Idle governor prefers the shallower WFI clock-gating state. As a reminder, on the Exynos 7420 this state is setup for target residency times of 500µS.
In general, the workload is optimized towards 4-core CPUs. Because 4x4 big.LITTLE SoCs in a sense can be seen as 4-core designs, we don't see an issue here. On the other hand, symmetric 8-core designs here would see very little benefit from the additional cores.
Facebook
Twitter
RSS
157 Comments
View All Comments
Gigaplex - Thursday, September 3, 2015 - link
An interesting and thorough analysis, although I'm concerned at some of the assumptions made in some of the conclusions. Just because a queue of 4 threads makes all the 8 big.LITTLE cores active doesn't mean that the architecture is effective. For all we know, the threads are thrashing back and forth, draining precious performance per watt.darkich - Friday, September 4, 2015 - link
Andrei, your articles are in a league of their own. Thanks for the great workmelgross - Thursday, September 10, 2015 - link
I'm still not convinced. The fact that it's doing what it does on these chips doesn't mean that their performance is as good as it could be, or that power efficiency is as good. We really need to see two to four core designs, with cores that are really more powerful, to make a proper comparison. We don't have that with the chips tested.blackcrayon - Thursday, October 8, 2015 - link
Exactly. It should at least show a design with a small number of powerful cores. Obviously with Apple's A series chips you have the issue of dealing with a different operating system underneath, but can't they use a Tegra K1 or something?Hydrargyrum - Friday, September 25, 2015 - link
The stacked frequency distribution graphs would be a *lot* easier to read if you used a consistent range of different saturations/intensities of a single colour (e.g. go from bright=fast to dark=slow), or a single pass from red to blue through the ROYGBIV colour spectrum (e.g. red=fast, blue=slow), to represent the range of frequencies.By going around the colour wheel multiple times in the colour coding it's *really* hard to tell whether a given area of the graph is high or low frequency. The difference in colour between 1400/800, 1296/700, and 1200/600 are very subtle to say the least.
Ethos Evoss - Thursday, November 12, 2015 - link
anandtech always uses weird non-popular words on its own site type ''heterogeneous '' never heard in my life and even usa or uk ppl have to search in cambridge/oxford dictionary :DDDImmediately u can say it is DEFO NOT USA or UK website.. They do not use such difficult words AT ALL :)
Ethos Evoss - Thursday, November 12, 2015 - link
ANd mainly they use when it comes to china products .. like mediatek or kirin or big.little topic etc..This site is DEVOURED or we could say powered by apple.inc :)