Overclocking the Pi3b+ GPU (Results)

dankcushions

@Brunnis said in Overclocking the Pi3b+ GPU (Results):

@quicksilver said in Overclocking the Pi3b+ GPU (Results):

@Parabolaralus something you could try is to set the CPU governor to performance mode. Can be found in the runcommand options in the retropie setup

Yep, that should be the preferred way of forcing the CPU to max clocks, since it will prevent it from running all out when you're in the menu. Does anyone know why this is not the default in RetroPie? The on demand governor causes issues with some SNES games as well, at least if you use other settings that are demanding.

it's a fair point. i assume it's not set in the default image because ondemand is what raspbian defaults to. depending on what your pi is used for, you may still want it to be ondemand - eg, portable builds. however i would have thought that for rpi3 and 2 builds that 'performance' would be a good default. maybe @BuZz has a view?

pjft

Wow.

@Parabolaralus and @Brunnis setting core_freq to 600 pretty much makes Crazy Taxi on the Dreamcast run a lot smoother without the audio skipping issues that plagued it in the default settings. Only skipping in minor occasions now, whereas previously after a 20-30 seconds it'd start skipping every couple of seconds when there were a lot of polygons on screen.

Thank you!

BuZz

@dankcushions I prefer to stick with OS defaults.

quicksilver

@Brunnis here it is:

https://www.raspberrypi.org/forums/viewtopic.php?f=29&t=6201&start=975

Man that took some digging 😂

There is a post on that page by Millhouse about it, and on the next page it looks like Dom (RPI engineer) confirms it. However firmware is constantly changing and sometimes there are changes which aren't well documented so who knows what's going on now.

mitu

@BuZz I think @dankcushions meant why isn't performance set for the Runcommand setting by default, so performance would be switched on only just during gameplay, but the default (ondemand) is preferred outside of gameplay.
@dankcushions or am I wrong ?

dankcushions

@mitu i didn't but that is a better idea! :)

hhromic

@dankcushions @mitu @BuZz while setting the scheduler to performanceby default in runcommand is a tempting idea, I would advocate against it. Setting this could potentially make RPIs to overheat without users being aware of it. At least in my experience, I have never had the need to set the scheduler to performance and things runs nice for me.
I think the best approach here is to educate the users about the CPU scheduler more than forcing a potentially troublesome option without their knowledge.

BuZz

@mitu I understood. I don't want to default to switching the governor on launch.

quicksilver

Setting to performance is a small change that really shouldnt make a significant difference when it comes to wear and tear on a pi but I agree with buzz that it should be up to the user to make that decision. I think there is some documentation in regards to the CPU governor modes in the retropie docs but maybe we can clarify that performance mode does help a few games/systems run smoother (perhaps it already says this, need to review). I would be happy to make any edits to the wiki if people feel that it is pertinent info.

mitu

@quicksilver There is a note in the Wiki on https://retropie.org.uk/docs/Speed-Issues/ and also on the N64 page, but maybe wen can add it also to the Overclock or Advanced configuration.

dankcushions

it would be useful to see some specific examples (games, benchmarks, etc) as i don't really get why ondemand (which i think is the default) would be slower than performance.

since ondemand ramps up the speed with load. i would have thought there should be no difference between the runtime cpu frequency between governor in cpu-heavy applications. they both should be running the cpu at full speed in a cpu-limited emulator, right?

shavecat

I always deal with the overclock that's what i get.. (crazytaxi 2 runs really nice )
just now the core_freq=600
so give in it afew days see if its stable.

hhromic

Something I don't see mentioned very often and I think is important to keep in mind is that the RPIs have a "warranty bit" that is burned when you overvoltage too agressively. This way, RMA or support can know if users broke the RPI by misuse or the device was faulty from factory.

over_voltage
(...) Values above 6 are only allowed when force_turbo is specified: this sets the warranty bit if over_voltage_* is also set.

force_turbo
(...) Enabling this may set the warranty bit if over_voltage_* is also set.

never_over_voltage
Sets a bit in the OTP memory (one time programmable) that prevents the device from being overvoltaged. This is intended to lock the device down so the warranty bit cannot be set either inadvertently or maliciously by using an invalid overvoltage.

Ref: https://www.raspberrypi.org/documentation/configuration/config-txt/overclocking.md
Ref: https://www.raspberrypi.org/forums/viewtopic.php?p=176865#p176865
Ref: https://www.raspberrypi.org/blog/introducing-turbo-mode-up-to-50-more-performance-for-free/

quicksilver

@hhromic the interesting thing is the current model pis cant overvolt any higher then a value of 4. At over_voltage=4 the core voltage equals 1.394v and it will not increase any higher than that. Values of 6-8 still only equal 1.394v (as @Rascas noted earlier). I think the stock core voltage is set higher on current model pis. I am sure that force_turbo would set the warranty bit but does over_voltage=4 now also set it? Official rpi documents are vague about this.

Parabolaralus

@Brunnis I really dont know if it actually would apply that setting and found no tangible benefit to clocking it that high, but i do know that if i set it to 735 even it would eventually freeze on me...usually within 20-30 minutes of gameplay so 735 is not stable.

My testing method with 733 as well as other settings involved starting a PS1 game (hence moderate load mentioned earlier) and leaving it for a few days. I normally do not play video games during the week so i could leave it running without it being a pain in the butt. Three days later id find the game still running and call it stable.

My other pi (3b+) as mentioned before would flat out freeze on me so much as pushing the RAM 10mhz higher. I guess silicon lottery and one is a slight score while the other is a dud for OCing.

On the governer thing. I never really thought about changing that TBH...I think performance would work in place, but i noticed absolutely no slowdowns using force_turbo and kind of left it at that.
Does anyone have anyway to see if it is actually downclocking while running a core/rom with ondemand, or is this something thats more user experience?

mitu

@Parabolaralus said in Overclocking the Pi3b+ GPU (Results):

On the governer thing. I never really thought about changing that TBH...I think performance would work in place, but i noticed absolutely no slowdowns using force_turbo and kind of left it at that.

Using force_turbo overrides the CPU governor and runs the cores at max frequency, so you're already using a 'performance' profile

By default (force_turbo=0) the "On Demand" CPU frequency driver will raise clocks to their maximum frequencies when the ARM cores are busy and will lower them to the minimum frequencies when the ARM cores are idle.

force_turbo=1 overrides this behaviour and forces maximum frequencies even when the ARM cores are not busy.

Brunnis

@dankcushions said in Overclocking the Pi3b+ GPU (Results):

it would be useful to see some specific examples (games, benchmarks, etc) as i don't really get why ondemand (which i think is the default) would be slower than performance.

since ondemand ramps up the speed with load. i would have thought there should be no difference between the runtime cpu frequency between governor in cpu-heavy applications. they both should be running the cpu at full speed in a cpu-limited emulator, right?

It might be mainly a problem if you decrease buffering, for example by setting max_swapchain_images=2. I believe the issue is caused by the on demand CPU governor not being able to handle the spiky CPU load. The CPU will emulate one frame and then push it to the GPU. While the GPU waits for a frame flip, the CPU will more or less idle, before kicking off emulation of the next frame. My guess is that the governor spins down the CPU and loses too much time when spinning it back up again during the next frame.

I would say using the performance governor as default for the run command would be safe. The user should expect (and want) the CPU to be in the high performance state anyway when running an emulator (and have the necessary cooling in place). The fact that the CPU may not always hit or stay at max frequency is the actual unexpected part here.

EDIT: On second thought, I guess the reduced buffering just makes the issue more likely to crop up. The unwanted CPU frequency reduction probably happens all the time at default settings as well, it’s just that there’s an additional frame buffered that will mostly cover the performance drop and prevent frame rate hitches.

@quicksilver said in Overclocking the Pi3b+ GPU (Results):

@hhromic the interesting thing is the current model pis cant overvolt any higher then a value of 4. At over_voltage=4 the core voltage equals 1.394v and it will not increase any higher than that. Values of 6-8 still only equal 1.394v (as @Rascas noted earlier). I think the stock core voltage is set higher on current model pis. I am sure that force_turbo would set the warranty bit but does over_voltage=4 now also set it? Official rpi documents are vague about this.

My Pi 3 B+ actually hits 1.39V already at over_voltage=1. That’s actually pretty high on 40nm, so I wouldn’t want to push it more. I guess the A53 really isn’t made to cope with high frequencies... 1.5GHz on 40nm and 1.4V is pretty abysmal.

BuZz

Most likely safe, but I prefer end users to make decisions like this. Also not everything would benefit - maybe some things that are launched you want to reduce clock if not in use. Frotz? Kodi? No doubt things will run hotter and consume more power. And in many cases it would be a waste. What about handhelds?

It's not going to be changed :-)

hhromic

@Brunnis the ondemand governor is not so primitive for switching speed.
While what you say is true that the CPU idles more with these emulators that use the GPU, that idling time is very short and the governor won't be micro-switching the frequency so fast. In particular, there is a setting for how often the governor will monitor the load to do adjustments:

* sampling_rate:

  Measured in uS (10^-6 seconds), this is how often you want the kernel
  to look at the CPU usage and to make decisions on what to do about the
  frequency.  Typically this is set to values of around '10000' or more.
  It's default value is (cmp. with users-guide.txt): transition_latency
  * 1000.  Be aware that transition latency is in ns and sampling_rate
  is in us, so you get the same sysfs value by default.  Sampling rate
  should always get adjusted considering the transition latency to set
  the sampling rate 750 times as high as the transition latency in the
  bash (as said, 1000 is default), do:

  $ echo `$(($(cat cpuinfo_transition_latency) * 750 / 1000)) > ondemand/sampling_rate

And there are many other settings to control the rather advanced ondemand governor :)

Ref: https://www.kernel.org/doc/Documentation/cpu-freq/governors.txt

Brunnis

@BuZz said in Overclocking the Pi3b+ GPU (Results):

Most likely safe, but I prefer end users to make decisions like this. Also not everything would benefit - maybe some things that are launched you want to reduce clock if not in use. Frotz? Kodi? No doubt things will run hotter and consume more power. And in many cases it would be a waste. What about handhelds?

It's not going to be changed :-)

Yep, for Kodi and the likes it’s definitely not a good idea to use the performance governor. For handhelds, performance governor would still be the way to go for emulation, since it’s still the predictable and stable mode. Any battery life issues should be handled by adjusting frequencies instead.

I definitely understand your stance, though. Development is full of compromises and I’m happy with just changing the setting via the run command menu.