3 ms·

Doesn't prioritizing efficiency and thermals make sense for any device? Even a desktop has moments when it's doing almost nothing but has tasks that can adequat

by pohl 3y ago

Doesn't prioritizing efficiency and thermals make sense for any device? Even a desktop has moments when it's doing almost nothing but has tasks that can adequately be handled by an E core where powering up a P core is overkill. Doesn't being efficient mean that more of the thermal budget can be held in reserve for when it's actually needed? Don't desktop users desire for their powerful machine to also be quiet?

jeffbee 3y ago
The blue-logo CPU company gives you a choice. You can adopt their most aggressive power profiles and get marginal performance gains if you want them. Or, you can tune them down to levels where they are pretty fast and fairly quiet. Or, you can dial them all the way down to their most efficient operating point, which pretty much nobody wants because that point is around 1W/core and they are pretty slow. Apple is dictating how much power the CPU is allowed to draw and they don't let it scale up. They are also making a static choice about how much die area and power to spend on the GPU, which might not suit every user. I know I personally don't give a rip about the GPU in my M2 mac mini, beyond the fact that it can draw things on the screen.
diffeomorphism 3y ago
Not really, no. Case in point: https://www.notebookcheck.net/AMD-Ryzen-9-7940HS-analysis-Zen4-Phoenix-is-ideally-as-efficient-as-Apple.713395.0.html#toc-6 https://www.notebookcheck.net/AMD-Ryzen-9-7940HS-analysis-Ze... Adjusting the TDP from 80W all the way down to 35W costs you relatively little in performance and gives you about the same efficiency as an M2 Pro. That is not done because it does not sell. > Don't desktop users desire for their powerful machine to also be quiet? No. That is not an "also" but an "instead". People claim they would make that trade-off but that is just not the case. Same for "make it heavier but give me more battery life".
- bgirard 3y ago
  If I can tuck away my desktop under my desk and it acts as a large space heater in the below zero winter I don't mind if it's helping me get my compiles down faster. There's some jobs I want a fast desktop for and other jobs, like browsing cooking recipes, I'll take my M1 macbook air for.
- arcticbull 3y ago
  > That is not done because it does not sell. It's not done because it doesn't sell in the PC market. Apple's kind of your only choice and if they say you're getting perf/W you're getting perf/W :) > People claim they would make that trade-off but that is just not the case. It's not 2 dimensional, tdp vs noise. It's tdp vs size vs noise vs price. You get a large ATX mid-tower and throw some Noctua 140mm fans in there and you won't hear a thing as it dispenses 1KW of heat into your feet area.
- faeriechangling 3y ago
  >Same for "make it heavier but give me more battery life". Consumers purchase large, heavy phones more than one would expect given their pricing and one of the reason they do so is that they have better battery life.
crest 3y ago
> Doesn't prioritising efficiency and thermals make sense for any device? No. If you have the cooling to sustain the higher power mode it can be worth it to pay the power bill to get work done faster. Compared to the productivity gains for compute limited workloads the increased power bill is little more than a rounding error or even a net power saving because you need less supporting infrastructure per worker to accomplish the same assuming that the tasks are even efficiently parallelizable to multiple (human) workers.
GeekyBear 3y ago
> Doesn't prioritizing efficiency and thermals make sense for any device? Intel did just ship 14th Gen Core i9 chips with a 400 watt power draw under load and very little in the way of a performance gain. > In some areas, the extra clock speeds available on the Core i9-14900K show some benefit, but generally speaking, it won't make much difference in most areas. The most significant win performance for the Core i9-14900K came in CineBench R23 MT... the Core i9-14900K sits 6% ahead in this benchmark https://www.anandtech.com/show/21084/intel-core-i9-14900k-core-i7-14700k-and-core-i5-14600k-review-raptor-lake-refreshed/15 https://www.anandtech.com/show/21084/intel-core-i9-14900k-co... Having your best benchmark result only see a 6% gain in return for a more than 400 watt power draw makes it pretty clear that just as with the Pentium IV of old, Intel isn't on a sustainable path. It also makes the M3's 20% performance gain along side efficiency gains look pretty darn good in comparison.

by pohl 3y ago

jeffbee 3y ago

The blue-logo CPU company gives you a choice. You can adopt their most aggressive power profiles and get marginal performance gains if you want them. Or, you can tune them down to levels where they are pretty fast and fairly quiet. Or, you can dial them all the way down to their most efficient operating point, which pretty much nobody wants because that point is around 1W/core and they are pretty slow. Apple is dictating how much power the CPU is allowed to draw and they don't let it scale up. They are also making a static choice about how much die area and power to spend on the GPU, which might not suit every user. I know I personally don't give a rip about the GPU in my M2 mac mini, beyond the fact that it can draw things on the screen.

diffeomorphism 3y ago

Not really, no. Case in point: https://www.notebookcheck.net/AMD-Ryzen-9-7940HS-analysis-Zen4-Phoenix-is-ideally-as-efficient-as-Apple.713395.0.html#toc-6 https://www.notebookcheck.net/AMD-Ryzen-9-7940HS-analysis-Ze... Adjusting the TDP from 80W all the way down to 35W costs you relatively little in performance and gives you about the same efficiency as an M2 Pro. That is not done because it does not sell. > Don't desktop users desire for their powerful machine to also be quiet? No. That is not an "also" but an "instead". People claim they would make that trade-off but that is just not the case. Same for "make it heavier but give me more battery life".

bgirard 3y ago
If I can tuck away my desktop under my desk and it acts as a large space heater in the below zero winter I don't mind if it's helping me get my compiles down faster. There's some jobs I want a fast desktop for and other jobs, like browsing cooking recipes, I'll take my M1 macbook air for.
arcticbull 3y ago
> That is not done because it does not sell. It's not done because it doesn't sell in the PC market. Apple's kind of your only choice and if they say you're getting perf/W you're getting perf/W :) > People claim they would make that trade-off but that is just not the case. It's not 2 dimensional, tdp vs noise. It's tdp vs size vs noise vs price. You get a large ATX mid-tower and throw some Noctua 140mm fans in there and you won't hear a thing as it dispenses 1KW of heat into your feet area.
faeriechangling 3y ago
>Same for "make it heavier but give me more battery life". Consumers purchase large, heavy phones more than one would expect given their pricing and one of the reason they do so is that they have better battery life.

crest 3y ago

> Doesn't prioritising efficiency and thermals make sense for any device? No. If you have the cooling to sustain the higher power mode it can be worth it to pay the power bill to get work done faster. Compared to the productivity gains for compute limited workloads the increased power bill is little more than a rounding error or even a net power saving because you need less supporting infrastructure per worker to accomplish the same assuming that the tasks are even efficiently parallelizable to multiple (human) workers.

GeekyBear 3y ago

> Doesn't prioritizing efficiency and thermals make sense for any device? Intel did just ship 14th Gen Core i9 chips with a 400 watt power draw under load and very little in the way of a performance gain. > In some areas, the extra clock speeds available on the Core i9-14900K show some benefit, but generally speaking, it won't make much difference in most areas. The most significant win performance for the Core i9-14900K came in CineBench R23 MT... the Core i9-14900K sits 6% ahead in this benchmark https://www.anandtech.com/show/21084/intel-core-i9-14900k-core-i7-14700k-and-core-i5-14600k-review-raptor-lake-refreshed/15 https://www.anandtech.com/show/21084/intel-core-i9-14900k-co... Having your best benchmark result only see a 6% gain in return for a more than 400 watt power draw makes it pretty clear that just as with the Pentium IV of old, Intel isn't on a sustainable path. It also makes the M3's 20% performance gain along side efficiency gains look pretty darn good in comparison.