Your Ryzen laptop is (probably) throttled quite heavily when on battery - and it's (probably?) not a huge deal

[D13H4RD]

If you've been keeping track of Intel's notoriously petty and face-on-ground marketing, especially for Tiger Lake-based systems (especially when it came to the "Evo" branding), you'd probably see a slide not unlike the one below, where Intel was claiming AMD gets its battery performance claims because it's putting a muzzle on how fast it can go when away from the charger.

This was on a Zen 2-based Renoir APU, and Intel rather cheekily (and quite obviously) left out the fact that its boost behavior does result in higher battery drain during such periods. Now though, with a brand-new lineup of Zen 3-based Cezanne APUs and lots of high-end designs that utilize the best of AMD, one question that some have is whether boosting behavior has been tweaked?

 

Well, it hasn't, not by much anyway.

 

Tests by Gordon Ung noted that while the Ryzen 5800U has shown to smoke the Intel Core i7 1185G7 in many tasks, many of which were while the laptop is connected to AC power. When running on battery, the script does flip, sometimes by a little, sometimes by a lot.

 

 

 

 

The delta isn't massive when in the "Best Performance" preset, with around a 20-25%-ish performance delta between plugged and unplugged, though the delta between plugged and unplugged when on the "Better Battery" preset is quite significant, though perhaps not entirely unexpected.

 

It's also worth noting that the Intel machine doesn't keep that performance for nothing, as it does chug down quite a lot more watts in bursts in order to maintain that level of performance when off the charger.

 

Quote

You can see part of the magic of the Core i7 in the MSI on battery is due simply to using more power—sometimes a lot more. In the first Cinebench R20 run, it’s almost a 40-watt load, spiking up to a massive 70 watts in “Best Performance” mode before completing the run mostly in the 55-watt to 50-watt range.

 

The Ryzen 7 in the ZenBook is far easier on the gas pedal, with the battery discharge rate at roughly 15 watts in “Best Battery” and about 20 watts in “Best Performance.”

 

As you can guess, putting that heavy of a load on a battery means you’ll run it down far faster. In our testing, the dent in the MSI Prestige 14’s battery was visibly different after the Cinebench runs, unlike with the Asus. To be fair, the Asus’s battery is 38 percent larger, so any reduction in capacity appears larger on the MSI. But more power used means less run time, no matter how you cut it.

 

Interestingly, Daniel Rubino of Windows Central noted that the SSD is also throttled when on battery, due to a feature called PCIe Speed Power Policy (PCPP).

Quote

As to why this happens, it is clear AMD would like to prolong battery life, which is an admirable goal. Razer confirmed the SSD performance drop is due to PSPP (PCIe Speed Power Policy) set by AMD and is not something it can control. I assume something similar is happening with the CPU and GPU, too.

 

My thoughts (this is going to be a long one, but you all need to read this!)

The big question that I'm sure many would be asking right now is "Is AMD wrong in doing this?"

 

And the answer is a big straight-up no. What AMD is doing is effectively not that much different from many older laptops, where off-charger performance is significantly hampered in an effort to conserve energy, particularly as processors of that era have not reached the level of performance-per-watt that current-generation designs have managed to achieve. In fact, some of the bench graphs show that Intel does also throttle performance a bit when off the charger, particularly in lightened workloads.

 

And it is definitely worth noting that while these charts do indeed show a drastic drop-off especially in the "Better Battery" mode, it is important to note the context, in which these involve ultra-portable low-wattage laptops. It's more likely that a user would be doing tasks like browsing the web or doing document work on such a machine, where the additional responsiveness of Intel's boost behavior may not be as apparent, but battery saving measures, even at the cost of performance, may be more appreciated. Even Gordon makes that case in his article.

Quote

For those who mostly do short, light boosty work, any disparity between Intel and AMD rival CPUs won’t be as apparent.

Neither AMD and Intel have a straight up "wrong" strategy here, moreso that they are very different. Intel is very clearly focused on responsiveness at the expense of outright efficiency, whilst AMD is perfectly happy to ease off the pedal significantly in order to gain more work-per-watt. Depending on your priorities, you might find one or the other more appealing.

 

With that said, here is where I get a bit more opiniated and personal. While I perfectly understand each company's reasoning for what they're doing, I am disappointed that there is no real way to fully adjust the way they behave through Windows' power settings. Intel's power consumption spikes are significantly calmer when in "Better Battery", but not usually by a big amount, whilst AMD's performance improves in "Best Performance", there's no real way to fully lift off the muzzle if, for whatever reason, you need all the CPU performance off-the-charger to finish off some work, such as editing photos on Lightroom.

 

While neither philosophies are straight-up wrong, I'm disappointed that I can't really have the option of straight up gimping performance significantly when I need to conserve my watts, or going full-ham on battery when I need to finish off that landscape photograph on Lightroom.

 

Sources

Tested: Is Ryzen 5000 battery performance really that bad?

Some AMD laptops reduce system performance for better battery life, but is that OK?

[Arika]

I would say that most people probably expect the Intel method when looking at the "better battery" and "better performance" slider, but without looking at comparisons side by side like the, would assume the AMD method is how it would behave on battery anyway.

[D13H4RD]

Just now, Arika S said:

I would say that most people probably expect the Intel method when looking at the "better battery" and "better performance" slider, but without looking at comparisons side by side like the, would assume the AMD method is how it would behave on battery anyway.

Pretty much. I don't think either philosophy is wrong, necessarily, just a bit perplexed as to why there isn't much information on this until recently. Especially the SSD throttling as that bit is new.

[leadeater]

Wasn't there already a news story about this a little while ago? I'm getting very strong vibes of seen this before.

Edited July 11, 2021 by leadeater

[leadeater]

1 minute ago, valdyrgramr said:

Who do you think you are, Uatu?

That, or crazy. Maybe both?

[D13H4RD]

7 minutes ago, leadeater said:

Wasn't there already a news story about this a little while ago? I'm getting very strong vibes of seen this before.

With Ryzen 4000? I think there was several months ago. This one's about 5000

[Dabombinable]

My old Asus A3H used to drastically lower the BUS speed to save power - due to the Celeron M380 lacking speedstep. Intel really shouldn't be trying to throw shade when they used to be far worse.

[D13H4RD]

Just now, valdyrgramr said:

TBH, I really don't feel like either are news.  If you go into the power options in Windows, on a Ryzen system, there's two Ryzen official plans.  One called balanced and one called high performance.  It's rather obvious that a CPU is going to be throttled on battery, for one, and on balanced/default it's more of a "no shit sherlock".  They're not using the Windows default versions, and it's not like some secret conspiracy.

I'm actually curious about whether these power plans exist on a Ryzen laptop.

 

On my Lenovo Slim 7i 13, the only Power Plan in Windows is Balanced. My best guess is that the power plans are relegated to Vantage, where the power plans are there, albeit in different names ("Extreme Performance" for high-performance, "Intelligent Cooling" for balanced, and "Battery Saving" for power-saver). I'm not exactly sure how much the Ryzen laptop differs in that regard.

[D13H4RD]

2 minutes ago, valdyrgramr said:

I don't have my Ryzen laptop anymore, replaced it with an i7/3080 laptop for rendering on the go and gaming on the go.  This is just what my workstation says which uses a higher end mobo, an Asus ROG Crosshair VIII Hero (WI-FI).   However, I think it shows up after installing the AMD chipset driver.  Been awhile since I did that.

Yeah, same thing on my Ryzen desktop. Came with the chipset driver. IIRC, I think those plans are better tailored for the way Precision Boost works, at least on the desktop.

 

Might differ on a laptop. I could test on 2 identical laptops, one on AMD Cezanne, and one on intel Tiger Lake, if possible.

[leadeater]

24 minutes ago, D13H4RD said:

With Ryzen 4000? I think there was several months ago. This one's about 5000

Thought so, it's not like it should have changed with Ryzen Mobile 5000 anyway. When it comes to the platform and how power is handled nothing realistically changed between Zen 2 and Zen 3. Like I'm sure there were optimizations but fundamentally the behaviors is the same.

 

Also I'm pretty sure when on battery most AMD laptops still have an inbuilt 10 second delay before boost will happen, you can remove that and then a lot of "PC" benchmarks get huge increases in scores.

 

A lot of these benchmarks and tests don't really depict how real world application usage and performance is like though. For AMD anything CPU intensive will kick up in to proper boost after 10 seconds and then drop down to TDP for basically ever or until CPU is deemed idle again then cycle repeats etc. If it's a game then similar story except you'll get individual cores being able to boost higher than others and above "all core".

 

Intel laptops boost immediately then drop down rather quickly then will drop to configured TDP and run at that until CPU is deemed idle again, repeat etc. So any application and usage with high and sustained CPU load Intel will get trounced, battery or AC. For gaming basically same story there too. The only time this behavior Intel is trying to point to matters is very light office and web usage where the CPU is very often entering and exiting boost, and then does CPU performance really matter in this type of use case? I'd say battery run time would be much more useful.

[D13H4RD]

1 minute ago, valdyrgramr said:

Might depend on the bloatware too.  But, even so I'd assume since most laptop default to the balanced plans they're designed to throttle to save battery life.  If you swapped to high performance on either Intel or AMD while on battery then you're going to take a hit to battery life.

I actually did a test on my Lenovo after I read about it. it's not very detailed, though, as it just involves a single GB5 run, but I did find some interesting, yet expected results.

 

 

This was on "Intelligent Cooling" (Balanced) with the Power Mode set to "Better Battery'. There is definitely a small drop off in performance when unplugged, although interestingly, the delta doesn't change that much when the Power Mode is set to "Best Performance".

 

[D13H4RD]

3 minutes ago, leadeater said:

Intel laptops boost immediately then drop down rather quickly then will drop to configured TDP and run at that until CPU is deemed idle again, repeat etc. So any application and usage with high and sustained CPU load Intel will get trounced, battery or AC. For gaming basically same story there too. The only time this behavior Intel is trying to point to is very light office and web usage where the CPU is very often entering and exiting boost, and then does CPU performance really matter in this type of use case? I'd say battery run time would be much more useful.

It's definitely very situational. The only time where I can see this being beneficial is for my specific use-case, which is on-the-go photo editing on Lightroom and Photoshop.

 

For most situations, I think the performance drop isn't that significant when it comes to perceived performance, although some have complained that the Razer Blade 14 drops performance off to the point where Windows felt sluggish, but I doubt this is the case for the majority of laptops out there. Could be specific to Razer.

[D13H4RD]

Just now, valdyrgramr said:

Most likely due to their common heat and bloatware issues.

Wouldn't surprise me. I find Razer laptops overhyped and riddled with issues.

[D13H4RD]

2 minutes ago, valdyrgramr said:

But, their computers are pure garbo.  I'm pretty picky with laptops for mobile rendering due to the amount of heat blender and maya can put on a component.  Not a fan of MSI's bloatware, but the laptop I bought is pretty decent for what I need it to do.

Which is why I find it perplexing how people like Linus continue to vouch for them even though the community seems to have agreed that they aren't very good laptops on the long run.

 

Funnily enough, I picked the Slim 7i 13 specifically for a few metrics, notably the performance in Photoshop and the QHD screen plus the small size. It's a nice laptop, all things considered, if a little plain.

[D13H4RD]

1 minute ago, valdyrgramr said:

But, to relate to the OP my Ryzen CPU did better in my workloads than my i7 did while on battery.  However, I honestly prefer using optix and just the 3080 for rendering when using Blender.  I went back to my old Radeon VII for now in the desktop.   Sadly, I might have to go Quadro if I want Nvidia/Optix/HBM3.    Might stick with team red depending on price/performance.

I'm actually quite interested to know how my system properly fares in real-world photo editing. I've mostly been using my desktop because I never had an opportunity to use my laptop for photo-editing much outside of some real-estate stuff, where it fared pretty well.

 

The benchmarks seem to indicate that it should fare pretty well even on battery due to its nature of being quite bursty, but I'd like to test it out for myself.

[D13H4RD]

To answer one of the questions that one of the source posed, in what should be done moving forward...I think the answer is obvious?

 

Test laptop performance both plugged and unplugged. Who knows? Might be useful to some people.

[porina]

To whoever asked about the AMD power plans, I just looked on my 5800H laptop and it doesn't have any. It actually has a custom plan by Lenovo. I'd also add, I don't think I was brave enough to find and install the AMD chipset driver package which comes with the AMD power plans, given the ability for it to break Windows for no apparent reason. Laptop is running Win10 as supplied so the support in it should be sufficient even if not latest AMD.

 

Back on the thread topic, performance vs power is always a balance and there are many different strategies to obtain that. Perceived responsiveness is helped a lot by CPU speed, but as others said on a longer sustained load it is more interesting to balance power efficiency especially when running on battery. Sustained high power applications aren't the sort of thing you want to run a lot on battery. Then we have the mid ground of applications that use non-trivial power but aren't a sustained load, and this is probably more representative of most battery mode usage. 

[D13H4RD]

6 minutes ago, porina said:

To whoever asked about the AMD power plans, I just looked on my 5800H laptop and it doesn't have any. It actually has a custom plan by Lenovo. I'd also add, I don't think I was brave enough to find and install the AMD chipset driver package which comes with the AMD power plans, given the ability for it to break Windows for no apparent reason. Laptop is running Win10 as supplied so the support in it should be sufficient even if not latest AMD.

So it's the same as my Slim 7i 13, especially if they're all in Vantage. Makes sense considering that every Windows laptop I've seen so far from the major manufacturers have some sort of OEM-provided utility integrated into the system.

 

6 minutes ago, porina said:

Back on the thread topic, performance vs power is always a balance and there are many different strategies to obtain that. Perceived responsiveness is helped a lot by CPU speed, but as others said on a longer sustained load it is more interesting to balance power efficiency especially when running on battery. Sustained high power applications aren't the sort of thing you want to run a lot on battery. Then we have the mid ground of applications that use non-trivial power but aren't a sustained load, and this is probably more representative of most battery mode usage. 

I think the heaviest load one can reasonably expect to run on battery is probably photo-editing via Photoshop or Lightroom. I think their nature of being rather bursty just so happens to fit with Intel's boost philosophy. Video-editing, rendering and whatnot, I feel like either would probably be done when it is hooked up to external power, regardless of platform due to their nature of being sustained, and the battery probably wouldn't last for longer than an hour regardless.

 

I thought I'd mention this as it's interesting. I'm using my S7i 13 right now to type and read all of this, and the power plan is set to conserve energy as much as possible without adversely affecting perceived performance. The 1135G7 hovers around 1.1-1.6GHz and average power consumption seems to be about 10W or lower. I honestly didn't notice any perceived dip in performance to a degree where I may go "This is unusable".

[porina]

6 minutes ago, valdyrgramr said:

I've had performance issues without having the AMD chipset drivers installed.  That's why I always install them.

If I'm doing a clean install of Windows then installing the AMD package is one of the first things I do.

 

In this case it is a laptop from Lenovo. You think they might know what they're doing and there is a sufficient driver support package included. Doesn't mean latest, but good enough.

 

Why I'm hesitant to install the latest AMD package on my laptop is the experience from owning three generations of various Ryzen desktop systems. The AMD chipset driver package has a significant non-zero chance of bricking the OS when updating down the line, even if it is ok on 1st ever install. As such I'm treating it as a case of "if it ain't broke, don't fix it".

[D13H4RD]

2 minutes ago, porina said:

Why I'm hesitant to install the latest AMD package on my laptop is the experience from owning three generations of various Ryzen desktop systems. The AMD chipset driver package has a significant non-zero chance of bricking the OS when updating down the line, even if it is ok on 1st ever install. As such I'm treating it as a case of "if it ain't broke, don't fix it".

Same for me with my desktop.

 

I once updated my chipset driver and I've been getting BSODs about 1-2 times a month out of the blue for seemingly no reason. After failing to fix with known culprits, I went on ASUS' website and downloaded the latest certified chipset driver for the TUF X570, which was older. Low and behold, it fixed it. So I'm staying on this chipset driver probably forever.

[porina]

3 minutes ago, valdyrgramr said:

Well, Lenovo typically includes the chipset driver for laptops.  However, with Intel you don't really need to install the chipset driver.  With AMD you have to more than you do with Intel's.   But again, laptop makers include the AMD one for that.  And, AMD's sw tends to update it.

It's not something I'm interested enough to investigate further but Win10 out of the box hardware support has improved over time and for most users I wonder if it is even needed to manually install drivers. Win10 will often download workable drivers if none are present. They may only be minimally adequate but enough to get functioning.

 

Anyway, until such time AMD software proves it doesn't give random problems I'd still exercise caution on unnecessary updating just because something newer might exist. GPU drivers are less of a problem in that when they have a problem it is usually non-fatal, but chipset driver problems are reinstall level pain.

[D13H4RD]

The power plans look like this on my end

 

 

Meanwhile, on Vantage...

[porina]

12 minutes ago, valdyrgramr said:

AMD's sw by default self updates, so you probably never had the need to install them as I said before.

Is there any substantial evidence that AMD chipset drivers update themselves? The last time I manually ran that was around the time of Zen 3 launch. No software was visible on system to suggest any sort of updating going on outside of the initial install. I'd be incredibly surprised if they did that as it is a high risk path with significant stability and security implications. MS might do similar through WindowsUpdate though.

 

AMD's GPU drivers do give the notification there's an update but by default it does not install it without the user's explicit permission.

[porina]

7 minutes ago, valdyrgramr said:

Unless they changed it they typically do, from what I remember.  This is default for me.  The Radeon SW updates the chipset too, iirc.

That looks about as expected, although I haven't run AMD software on a CPU in a while. It actually looks a bit better than I recall, as I don't recall the GPU driver offering other driver updates when I last seriously used it.

 

Maybe I misunderstood your previous comment to mean it would update silently. That setting I presume would give you a notification an update was available and you could choose to install it if you wish, which is the reasonable approach used by most software that isn't Microsoft.

[D13H4RD]

10 minutes ago, valdyrgramr said:

Unless they changed it they typically do, from what I remember.  This is default for me.  The Radeon SW updates the chipset too, iirc.

 

 

I think that just downloads the driver on its own, but doesn't install them without user consent.