benchmark new smartphones:

[mclane]

The throttle is dependent on the temperature of the CPU.
If your CPU is not cool, the results are weaker. If processes in the background are still running, it reduces results. Make sure task list has all processes deleted. Use droidfish. Make sure telephone is not hot or in the sun.

[NoSQL]

Hi mclane,

it's no temperature throttling. Throttling Test app shows the normal behavior for Poco F1 as described in many reviews (it throttles down to around 80 percent within a few minutes. And no, I have no remaining background tasks.

With Droidfish there are peaks of around 5000 kn/s. But when I let Stockfish calculate for a longer time, it goes down and down and ... until it reaches 2300 kn/s. When I stop the engine and restart it again, it again begins witih a higher value and afterwards goes down again until 2300 ... And there are some bigger steps, e.g. it jumps directly from 4000 kn/s to 3200 kn/s which looks like the operating system reduces the number of cores or does some other battery optimizing things ...

For me the peak numbers and also the numbers after 2 minutes have no meaning at all. The only number that counts is the stable one which can be obtained for long running analyses. This afternoon I tested another smartphone with bit.LITTLE architecture, the Xiaomi Mi A2 which is the successor of my Mi A1. But the effects with this snapdragon 660 are the same as with all the snapdragon 845: It starts with a high peak and slows down and down and ...

As mentioned in my last post, my Xiaomi Mi A1 with snapdragon 625 (the only corrent snapdragon with 8 identical cores) starts with 2700 to 3000 and keeps this value even after half an hour. I've used it during the whole hot German summer with temperatures like in the Sahara, and there was no difference.

A pity, my hope from the reports here was that you can keep the great values of near 5000 kn/s for a long time. But this hope is over now. I'll keep my Mi A1 with snapdragon 625 and a constant kn/s value. Better a constant performance than always in the dark, whichever power you get.

[stavros]

NoSQL i completely agree with you, there is a throttling test at youtube for 625 vs 636 vs 660 with results 96% vs 84% vs 68% of the initial speed (15 mins test)
i prefer android but i must admit that apple cpu has the best throttling results (90% of the initial speed).
the 845 (and 835) is for full load burst speeds (secs to 1 min) and its a marketing hype
do you know why lately the new galaxy note 9 (sp 845) has a special cooling system?

[flither]

I'm also not fully happy with my SD835 phone Xiaomi Mi 6 as for chess, my test 2 pages ago shows -40% kNps after 15 minutes.
Simple reason - mobiles are not made for constant max load, that would kill their tiny batteries too fast. But there should be a choice - a simple switch turning off aggressive throttling for those who wanna max CPU for a cost of much faster battery killing.
Generally all this Snapdragon throttling makes a huge difference to battery life - I don't remember a single day with dead battery, and I'm a really heavy user. The battery life is just astounding.
But gimme this little switch for 15 minutes

[mclane]

Some mobile phones have a special Heatpipe, water cooled. Connected with the backside. Different dimensions depending on the model and company .
Nokia 8 good results refer to that. Also we now have special gaming phones like the Razer Phone. And others where you can switch a fan on the outside of the mobile that blows colder air into the mobile phone.
Google gaming phones to read more about these phones. Many companies now have these gaming phones.

[NoSQL]

...
Simple reason - mobiles are not made for constant max load, that would kill their tiny batteries too fast. But there should be a choice - a simple switch turning off aggressive throttling for those who wanna max CPU for a cost of much faster battery killing.
...

Yes, unfortunately that is also my experience. My sd625 seems to be the last snapdragon chip that does respect a users choice of demanding all cores for a long time period. I am well aware that the battery discharges extremely quickly in this use. But I can do this ...

With the much more modern sd845 this seems to be impossible. At least in those smartphones with sd845 that have I tested so far (Xiaomi Mi 8 and Pocophone F1)

[mclane]

Asus Rog phone, gaming phone with (optional) fan.

https://m.gearbest.com/cell-phones/pp_0 ... J4QAvD_BwE

[mclane]

Razer phone 2

Take a look at the big cooling chamber in this video

https://youtu.be/ydGwcq7WRPs

[yurikvelo]

NoSQL i completely agree with you, there is a throttling test at youtube for 625 vs 636 vs 660 with results 96% vs 84% vs 68% of the initial speed (15 mins test)
i prefer android but i must admit that apple cpu has the best throttling results (90% of the initial speed).

it's not "android vs iOs", it is ARM Big.Little architecture, for short period it allows both Big+Little, but for practical need only Big perfomance is relevant

[Thomas Lagershausen]

Is the new Kirin 980 CPU in the Huawei Mate 20 Pro now the fastest for chess in the world of smartphones?