But throttling is still a major concern in phones. In fact, they almost all throttle quite severely nowadays.
Well, that’s how they’re designed. They’re supposed to throttle.
It’s not a simple comparison as it was in the past - let me elaborate a little bit.
Modern CPUs with the Turbo Boost have (as far as I understand it - bear in mind I’m doing this from memory) multiple boost ‘levels’. A CPU will try to always be at the highest level if the surrounding parameters of time, power and temperature are ok.
The Surface at least 2 levels (I think it might be 3 and I will write it down as 3):
Level 2: ~20W TDP, Lasts ~2 Minutes
Level 1: ~15W TDP, Lasts ~10 Minutes
Level 0: ~7W TDP, Lasts indefinitely.
The V’s CPU also has a similar setting - the default setting of the Core Y CPU actually goes roughly like this:
Level 1: ~15W TDP, Lasts 30 Seconds
Level 0: 7W TDP, Lasts indefinitely
Cinebench actually is a bad benchmark in my opinion to compare that as Cinebench is a rather short benchmark - there are better ones for sustained tests as it runs longer and it comes closer to real-world performance.
Are you missing the point on purpose here?
Thermal design is EVEN MORE of a consideration on phones, as they are always passive, have very little surface area for heat dissipation, and the chips are crammed in tight against heat-producing batteries and displays.
Yet the standard now is to make phones out of glass.
Could it be the same reason why laptops aren’t manufactured from copper? Which is to say, that there is always a cost/use/benefit calculation to be made, and the importance of material the further away you get from the core lessens?
I mean, surely that’s why heatsinks usually have copper blocks followed by aluminium fins - there is an advantage to all-copper designs but it’s too small to outweigh the negatives.
Likewise, I don’t think the thermal characteristics of magnesium matter much at all.
Otherwise, the industrial designers at Micro-
Oh wait, yeah, I forgot, you already know better than everyone Microsoft, sorry.
I’m assuming these timings are very rough estimates?
Either way, it sounds like the Surface Pro chassis - supposedly inferior to the V’s due to the use of magnesium - isn’t impeding heat dissipation
Could it be that the design of the heatsink, number and length of pipes, etc, are a SIGNIFICANTLY MORE IMPORTANT FACTOR?
Not going to mention the various Dells and Lenovos out there that use carbon fibre and plastic in their case designs (I’m using one!). Yes, even the ones that are primarily passive.
My point, by the way, isn’t that there is no thermal difference between aluminium and the specific magnesium alloy you get in Surface. Just that it’s negligible. I don’t know why that’s so controversial.
While I’ve not read thorugh the whole thread I don’t think anyone mentioned that the Surface’s Magnesium Casing was made out of ‘Thermal Toothpaste’.
Heatsinks, their position and the number and length of pipes can contribute greatly to a good gooling solution - I’d not say it’s significantly more important as if you e.g. chose a plastic housing and then put the Surface’s Cooling inside it it’d still impede the heating compared when you just e.g. have a copper plate and a aluminum housing.
There is one actually, however it’s offset by the difference in the cooling solution - the thermal conductivity of magnesium is a good chunk lower than Aluminium however if you can spend a good chunk in developing and manufacturing a custom cooling solution you can offset that.
(On my quick research the thermal conductivity W/K * m of Aluminum is ~200 while the magnesium Alloys I found differ from 60-90)
EDIT: And it seems Carbon Fibre has a pretty high thermal confductivity
Interesting. My google searches yield wildly different results per source!
Of course, knowing the exact alloy make-up of the Surface would help…
Well, in my prototype I noticed that it quickly reaches the full capacity during Cinebench test, and after that it can stay like that indefinitely. I tested it with Prime95 and it kept running at the exact same state for an hour. The same state as ~10s into Cinebench.
So it was at the 15W mark the whole time during Prime95? Or it was at 7W the whole time during Prime95?
Also I think your prototype should’ve boosted for ~30 seconds as that’s the default setting for Core Y CPUs as far as I know
7W, I couldn’t manage to set it above that. 7W was the maximum setting that worked. But in Cinebench, it was also at the same 7W mark. And the same temperature.
Their weighted average or whatever it’s called works in an interesting way, so the 30s timer is not as simple as it seems… It takes into account the last 30s when measuring power consumption, but that doesn’t mean it can go above the TDP forthe whole 30 seconds… It’s more like 10 seconds in real life, from my experience. You can see it for yourself in the Throttlestop logs you attached to your benchmark thread
Oh, and also, I ran Cinebench tests consecutively, without letting the CPU temperature or the weighted average to drop. The results were very similar.