Discussion [Windows Central] Some AMD laptops reduce system performance for better battery life, but is that OK?

[tamz_msc]

Is AMD 'cheating' to get better battery by reducing laptop performance?

AMD laptops bring a lot of power to the table, literally. You only see max AMD performance when plugged into an AC outlet instead of running on battery. Is that practice OK, though, and could Intel get away with it as well? Let’s talk about how laptops are changing.

www.windowscentral.com

Looks like a One Plus 9-esque situation - sacrificing performance in order to prolong battery life. FYI on Intel performance is purely a function of temperature. PL1 and PL2 will change on the fly depending on how hot the CPU is for the given workload and duration, regardless of AC or DC power. That's why on laptops with less than beefy cooling subsequent Cinebench runs show a gradual drop in performance. This will happen despite the performance profile set in OEM software, like Omen Command Center for HP, Armory Crate for ASUS, Alienware Control Panel for Alienware, Lenovo Vantage for Lenovo etc.

Like for example when opening an app, for a fraction of a second clock speeds on Intel will shoot up to the Max turbo, even if it means that the temperature shoots up to near or at throttling limit. This happens irrespective of whether the laptop is on battery or not.

TL;DR: instantaneous performance with Intel is a function of temperature, while with AMD it is a function of AC/DC power.

 

[VirtualLarry]

TL;DR: instantaneous performance with Intel is a function of temperature, while with AMD it is a function of AC/DC power.

My first introduction to this sort of scheme, was with my Asus '510 ultrabook, with Intel Pentium B973 processor. It was set up by Asus, to run at only 800Mhz while on battery, thus crippling what was supposed to be "good Intel performance". On AC power, it ran at full 2.3Ghz performance.

 

[tamz_msc]

My first introduction to this sort of scheme, was with my Asus '510 ultrabook, with Intel Pentium B973 processor. It was set up by Asus, to run at only 800Mhz while on battery, thus crippling what was supposed to be "good Intel performance". On AC power, it ran at full 2.3Ghz performance.

Not exactly relevant because that particular processor doesn't have turbo boost.

 

[WelshBloke]

I'm not seeing a massive difference between the two approaches tbh. They both throttle at various times. I guess it would nice to have some user control over both situations.

Look at it this way the performance on the AMD system is going to be infinitely faster when it's clocked down to save the battery than if the battery dies!

I'm fairly sure my intel Dell XPS clocks harder when it's plugged in as well, as least the fans kick in a lot more often when it's connected to the mains power anyway.

 

[tamz_msc]

I'm not seeing a massive difference between the two approaches tbh. They both throttle at various times. I guess it would nice to have some user control over both situations.

Look at it this way the performance on the AMD system is going to be infinitely faster when it's clocked down to save the battery than if the battery dies!

I'm fairly sure my intel Dell XPS clocks harder when it's plugged in as well, as least the fans kick in a lot more often when it's connected to the mains power anyway.

The difference is that on Intel the throttling depends only on temperature and the nature of the workload. It doesn't matter to a large extent what power source the laptop runs on. With AMD performance is drastically different on battery and mains, enough to negatively affect user experience in my opinion.

 

[WelshBloke]

The difference is that on Intel the throttling depends only on temperature and the nature of the workload. It doesn't matter to a large extent what power source the laptop runs on. With AMD performance is drastically different on battery and mains, enough to negatively affect user experience in my opinion.

I mean if your CPU throttles when it has to do work that's negatively affecting the user experience as well.
If I had time I'd check my laptop because I'm fairly sure it does the same thing as you're saying the AMD one does.

 

[JWade]

my laptop with an amd 5800H doesnt do this. but then again i also went into windows power settings and changed the settings when i got my laptop. The way i see it, with the default settings, with the clockspeeds, it tries to get more run time when on battery. I always figured that, so when i get a new laptop i always change the power settings. Only when i travel will i run my laptop on battery for more than 20 - 30 minutes, which i have a power plan settings saved for better battery performance.

 

[tamz_msc]

I mean if your CPU throttles when it has to do work that's negatively affecting the user experience as well.
If I had time I'd check my laptop because I'm fairly sure it does the same thing as you're saying the AMD one does.

Sure but what I'm saying that the maximum potential performance on Intel is a function of thermals alone. With AMD it depends on thermals and also AC vs DC power. This is especially true in thin and light laptops like the Surface Book, though it seems that the Razer Blade 14 is not immune to this either.

Moreover, AMD sets it's SoC to halve PCI-e bandwidth on battery which is something that OEMs cannot control, according to Razer. If you're video-editing on the go for example on the Razer Blade 14, which is a valid use case, then not only have you got to deal with the lowered CPU performance but you also get only 50% of your potential disk performance and we all know that storage performance is important in video editing.

The problem is that AMD doesn't give the user choice on whether to opt for this kind of behavior.

 

[TheELF]

Sure but what I'm saying that the maximum potential performance on Intel is a function of thermals alone. With AMD it depends on thermals and also AC vs DC power. This is especially true in thin and light laptops like the Surface Book, though it seems that the Razer Blade 14 is not immune to this either.

Moreover, AMD sets it's SoC to halve PCI-e bandwidth on battery which is something that OEMs cannot control, according to Razer. If you're video-editing on the go for example on the Razer Blade 14, which is a valid use case, then not only have you got to deal with the lowered CPU performance but you also get only 50% of your potential disk performance and we all know that storage performance is important in video editing.

The problem is that AMD doesn't give the user choice on whether to opt for this kind of behavior.

Intel has TDP down so it can happen with intel laptops as well where they use the lower TDP setting to get better battery life.
The only issue, if there is any, is if reviews use one setting to get performance numbers and the other to get battery life numbers...

 

[tamz_msc]

Intel has TDP down so it can happen with intel laptops as well where they use the lower TDP setting to get better battery life.
The only issue, if there is any, is if reviews use one setting to get performance numbers and the other to get battery life numbers...

TDP down doesn't affect short, bursty workloads in my experience. It only affects sustained performance. Evidence:

I just ran the Mathematica built-in benchmark. On AC, power profile was set to optimized, Windows set to 'Better Performance'. PL1 was 25 W, PL2 was 51 W. Score: 2.61

On Battery, same settings as before but PL1 was 15 W and PL2 was again 51 W. Score: 2.5

A difference of a mere ~4%.

 

[TheELF]

TDP down doesn't affect short, bursty workloads in my experience. It only affects sustained performance. Evidence:

So?!
It still changes power draw to prolong battery life, you just provided proof that yours does.

Change your argument to 'AMD reduces performance way too much to prolong battery life'

 

[VirtualLarry]

TDP down doesn't affect short, bursty workloads in my experience. It only affects sustained performance. Evidence:

I just ran the Mathematica built-in benchmark. On AC, power profile was set to optimized, Windows set to 'Better Performance'. PL1 was 25 W, PL2 was 51 W. Score: 2.61

On Battery, same settings as before but PL1 was 15 W and PL2 was again 51 W. Score: 2.5

A difference of a mere ~4%.

It still changes power draw to prolong battery life, you just provided proof that yours does.

Change your argument to 'AMD reduces performance way too much to prolong battery life'

LOL. So much self-ownage.

 

[WelshBloke]

If you're video-editing on the go for example on the Razer Blade 14, which is a valid use case, then not only have you got to deal with the lowered CPU performance but you also get only 50% of your potential disk performance and we all know that storage performance is important in video editing.

The problem is that AMD doesn't give the user choice on whether to opt for this kind of behavior.

I mean if I was video editing on the go I'd be totally worried about battery life as much as performance given how long it takes to get any video editing done!

I'd agree that a bit more control over throttling would be great, but I'd say that was an issue with computers in general!

The only issue, if there is any, is if reviews use one setting to get performance numbers and the other to get battery life numbers...

Yeah. That's definitely a valid point.

 

[tamz_msc]

So?!
It still changes power draw to prolong battery life, you just provided proof that yours does.

Change your argument to 'AMD reduces performance way too much to prolong battery life'

Setting PL1 to 15 W doesn't change the power draw in any significant manner because it turbos during the benchmark run. The performance numbers is evidence of that. On AMD you lose anywhere from 20-40% performance on battery, with 20% being the case for Renoir in light, bursty workloads like Geekbench 5, and 40% being sustained workloads like Cinebench. On Intel you only lose 4% performance on a bursty workload like Geekbench 5 ST or Mathematica.

 

[Thibsie]

Setting PL1 to 15 W doesn't change the power draw in any significant manner because it turbos during the benchmark run. The performance numbers is evidence of that. On AMD you lose anywhere from 20-40% performance on battery, with 20% being the case for Renoir in light, bursty workloads like Geekbench 5, and 40% being sustained workloads like Cinebench. On Intel you only lose 4% performance on a bursty workload like Geekbench 5 ST or Mathematica.

So Intel has a built-in throttle from power on ?

 

[tamz_msc]

So Intel has a built-in throttle from power on ?

Not sure what you mean by that.

 

[Topweasel]

The other week I watched a video about the perf drop depending on battery charge. Seems like AMD takes a bit off the top at the high end (90-100) to extend life a little extra. But if we are okay with a perf drop at 75%, is it really an issue if it happens at 90% if it extends our battery life. I mean if its a temporary thing (start something as you move your laptop to a different desk or something) then the time to completion isn't really a big thing and if you are working an extended period wouldn't it be more important for it to last longer than get max performance?

This is one of those slippery slope scenario's, we won't have a good idea what our laptops perform like at full charge, half charge, and so on. But most manufacturers are already doing most of this in one way or another. All that we are seeing now is some of it imbedded in the chipset. Is this something that the OEM's could override? Would they want to?

 

[tamz_msc]

The other week I watched a video about the perf drop depending on battery charge. Seems like AMD takes a bit off the top at the high end (90-100) to extend life a little extra. But if we are okay with a perf drop at 75%, is it really an issue if it happens at 90% if it extends our battery life. I mean if its a temporary thing (start something as you move your laptop to a different desk or something) then the time to completion isn't really a big thing and if you are working an extended period wouldn't it be more important for it to last longer than get max performance?

This is one of those slippery slope scenario's, we won't have a good idea what our laptops perform like at full charge, half charge, and so on. But most manufacturers are already doing most of this in one way or another. All that we are seeing now is some of it imbedded in the chipset. Is this something that the OEM's could override? Would they want to?

PCWorld has an article about this.

The scenario is this: if you select "best performance" in the Windows battery settings, and run real-world workloads like the Office suite or Photoshop or browsing on battery, the drop off on Intel compared to running while plugged in is a few single-digit percentage points, while on AMD it is much more (often 20% or more).

The issue as far as I understand boils down to how power limits are handled when on battery. You see on Intel PL2 remains untouched, only PL1 is adjusted to be lower on battery compared to when plugged in. This means that despite lower PL1, the cores are allowed to boost as high as PL2 all the time on workloads which aren't constant all-core loads like rendering, so long as you don't hit thermal throttling of course. This is what I observe on my Dell Inspiron with an i7-11370H.

This is how it should be IMO; the same thing is true for the M1 Macbooks - there is no discernible performance difference whether you are on battery or plugged in as long as you don't thermal throttle. On AMD, since it is not the case, it could only mean that the clock speeds are limited from attaining as high values as possible in an attempt to prolong battery life at the expense of performance. I don't have an AMD laptop to test, nor have I seen reviews which specifically test this aspect of performance. Then there is the whole different issue of delayed boost on battery on AMD laptops, which Ars Technica has an article about.

 

[Topweasel]

PCWorld has an article about this.

The scenario is this: if you select "best performance" in the Windows battery settings, and run real-world workloads like the Office suite or Photoshop or browsing on battery, the drop off on Intel compared to running while plugged in is a few single-digit percentage points, while on AMD it is much more (often 20% or more).

The issue as far as I understand boils down to how power limits are handled when on battery. You see on Intel PL2 remains untouched, only PL1 is adjusted to be lower on battery compared to when plugged in. This means that despite lower PL1, the cores are allowed to boost as high as PL2 all the time on workloads which aren't constant all-core loads like rendering, so long as you don't hit thermal throttling of course. This is what I observe on my Dell Inspiron with an i7-11370H.

This is how it should be IMO; the same thing is true for the M1 Macbooks - there is no discernible performance difference whether you are on battery or plugged in as long as you don't thermal throttle. On AMD, since it is not the case, it could only mean that the clock speeds are limited from attaining as high values as possible in an attempt to prolong battery life at the expense of performance. I don't have an AMD laptop to test, nor have I seen reviews which specifically test this aspect of performance. Then there is the whole different issue of delayed boost on battery on AMD laptops, which Ars Technica has an article about.

Doesn't really answer any of my questions. I am not questioning what it is doing, or how that compares to Intel. But performance on a laptop is rarely linear. This is a choice AMD is making with OEM's. But also a choice on Laptops that OEM's can and have set themselves. My question is basically is this something we should care about. The lack of transparency but a lot of AMD decisions are made with OEM not just support but usually at their behest. Considering laptop CPU/Chipset. AMD at that feels its just part of a laptop ecosystem approach and just an aspect of having a laptop.

But that's what I was asking about in regards to actual affect (the in use aspect), its difference compared to the drop in perf when talking 50% and so on. Has anyone measured work done per charge vs. using the full CPU perf? Potentially by lets say knocking of max power usuage or limiting turbo's do we get enough extra longevity to get us more work done compared to a CPU maxed out. I know we like to have more control over the decisions of our systems, but on the other hand Laptops are all about compromise, and one of those things that AMD found was that by limiting on battery perf they maximized total compute performance per whr of battery.

 

[Tams80]

It is bad of AMD to force you to reduce performance on battery. As long as it doesn't present a safety risk (exploding batteries, etc.), then you really should be able to choose.

I do greatly appreciate the ability to reduce performance for increased battery life, and most users will probably want that, but it should be an option. Hell, it can be the default. If someone wants to have terrible battery life while rendering their videos for YuuuTooobeee or whatever, then let them though.

 

[Thibsie]

Force who?
Just change the power options.
And stop spreading fud.

 

[Topweasel]

Force who?
Just change the power options.
And stop spreading fud.

I think you are missing something and that is that all 4th and 5th gen AMD solutions have a full system wide power reduction policy when running on battery.

But here are things we already know.
1. Most OEM's do this to one degree or another on a significant portion of their laptops. Just not generally the "Enthusiast" laptops.
2. All Laptops experience drops in allowed performance no matter the windows settings are as available battery power drops. Most as soon as the battery hits 90%.
3. AMD's Mobile is more efficient and even more that plateaus at a pretty low power setting compared to Intel. Meaning even significant power limitations means minimal perf loss.
4. I am struggling to see a scenario where this actually effects people other than running a R20 bench. I believe this causes the system to run at its more efficient settings which will equal more work accomplished per battery charge. Only really differs from all other systems by that first 10%. If you were doing work completely mobile, a slight loss in instantaneous productivity would most of the time welcomed if it means longer on battery time to do said work (within reason). If its short term, start something, then move to an area you are plugged in the difference in compute performance is going to be relatively small and not something you would measure for when buying a laptop.

I am certainly for more transparency and it's obvious to me that Renoir and Cezanne AMD went all in pushing perf for 15-25w usage. This is an ULV CPU first and formost. By designing it as Low wattage 8c CPU, they put themselves into a market that Intel just couldn't compete on their level. Even now their 10nm 8 cores just take such a hit going below 35w, that AMD would walk all over them. You want a thin and light that can stand with the big boys in productivity you need a 4800U or 5800U. My guess that a lot of this is built into their mobile chipset to help themselves and the OEM's get even better longevity accomplishments in their thin and lights and what this means to benchmark focused enthusiasts took a back seat.