Intermittently slow/stuttery/choppy gameplay. Only sometimes

[tfabris]

Care to give us a quick info on general CPU and GPU temperatures, just to make sure it's not some heat issue you're looking at?

Interesting. I hadn't thought to look at heat issues. Back in the days when I was constructing my own desktop systems for gaming, I did a lot of work with temperature monitoring programs, keeping track of how hot my CPU and GPU got during gaming. I haven't done any of that lately, because it's a laptop and I didn't think I had any control over that, so I didn't think it was worth looking at. But now that you mention it, I could envision a situation where the laptop's built-in management firmware might be doing things like throttling the CPU and/or GPU when they get hot. If I could run a program that watches the temps, and see if the stutter occurs only when the temps hit a certain ceiling, then that might prove that theory. I'll investigate that, see if there's a way I can get temps... Ideally in a realtime graph running in a window alongside.

It was several years ago when I worked with temperature and fan speed monitoring software. Back then, it was all a bit iffy and required a lot of careful configuration and tweaking to get correct data. Is that still the case, or is there an easy-to-use idiot-proof program that can more easily gather that data these days?

[Dr.Disaster]

You can use HWMonitor for CPU and GPU-Z for GPU monitoring. Just run them in the background and switch over when the issue occures. You might want to pay attention to the GPU-Z sensor "Perfcap": green bars indicate a power cut by the system.

[tfabris]

Thanks! I'll try those out.

[tfabris]

Okay , I tried the following:

- Ran GPU-Z, HWMonitor, and Task Manager while playing Grimrock 2.

I just got it to do "The Thing" once. Behaviors observed when "The Thing" happened:

- Memory usage for Grimrock didn't seem to change, although I'm not 100 percent certain about that. Grimrock has been hovering in the 1,200,000 K range for memory usage most of the game. there are times when it dips down to 900,000 K range but "The Thing" didn't seem to coincide with the dip. Though I'm not 100 percent certain of that.

- GPU temperature, though going up and down within the 55-65c range, didn't seem to be the triggering factor. I think that it was around 59c or so when "The Thing" happened, but, later the temp went up much higher. Right now it's at 62c and it's not doing "The Thing".

- CPU temperature similar to GPU temperature. I have max logged values for Core0, Core1, and Package, and their top logged values for this session are 67/64/67, but when "The Thing" happened, the CPU temp was in the 50's.

- GPU Load as shown by GPU-Z is constantly fluctuating in the range of 75-82% percent and high values for GPU load don't seem to be what triggers "The Thing".

- CPU usage as shown by Task Manager for the Grimrock Executable, is constantly hovering in the 20% range. The highest CPU item in the Task Manager is System Idle Process, in the 70-80% range, always, even when I'm wandering in outdoor areas of Grimrock 2. High CPU usage doesn't seem to trigger "The Thing" because it never gets very high to begin with.

- When "The Thing" happens, there's nothing new in the Task Manager, at least not in the top range of CPU usage. The top two items are still System Idle Process and Grimrock2.exe, in that order.

- When "The Thing" happens, after a few moments I can clearly hear the laptop's cooling fan decrease in speed because whatever "The Thing" is doing, it's reducing system load because my frame rate drops to a slideshow when it happens. But I don't think that indicates a *cause*, rather, my impression is that this is an *effect* of the issue, not an indicator of the cause, if you get my drift. I could be wrong.

So I'm stumped. Anything else I can try?

[Dr.Disaster]

Those temperatures seem fine; kinda unlikely they are the cause of your issue. LoG2 memory, cpu and gpu usage also seem normal. With CPU load being unchanged looking at other processes and applications prolly won't yield a lot. Still it would be nice to have a screenshot of the process list.

This "frame rate drops to a slideshow" sounds like the system executes a short power cut to the GPU. Depending on how long those slowdowns are GPU-Z's "PerfCap" sensor should show green lines or even bars. Please have a closer look at this one.

With your VAIO featuring an nvidia GPU you can also try this: inside your nVidia driver's 3D settings is an option called "Power Management". It's default setting is "adaptive". Test if the issue still occures when you change this setting to "maximum".

You can also keep an eye on disk activity with the windows resource monitor (Start→All Programs→Accessories→System Tools→Resource Monitor). Perhaps some SSD garbage collection thing (TRIM) is set off.

[tfabris]

I didn't notice any green bars in the GPU-Z graphs, but I will double check.

I'll look at adaptive power management settings in the Nvidia driver. I remember going into the Windows power management settings and making sure that everything was set to maximum performance when the laptop was connected to AC power. However if the Nvidia driver has squirreled away its own setting for that, one which is not reflected in the Windows power management details panel, then that could very well be the culprit. I will check for that.

(Edit: Found the Adaptive power management in the Nvidia settings. As I suspected, it's kinda buried in there, kinda hard to find, and it's definitely hidden in the Nvidia panel, it doesn't exist in the Windows power management settings. I'm at work right now so I can't play the game to see if it solves the problem, but I'll try that this evening.)

I'll also see what the Resource Monitor can show me about CPU and disk usage.

Thanks!

[tfabris]

Changing the power management from "Adaptive" to "Prefer Maximum Performance" did not fix the issue.

Watching Resource Monitor, the disk usage constantly spikes up and down, with Grimrock giving a spike whenever it needs to read data, or save a savegame, or load a savegame. I did not see any particular spikes or anything special when "The Thing" happened.

But here is something interesting. Though the GPU *temp* didn't seem to do anything odd when it happened, the GPU *clocks* did in fact have a major change. Attached is a screen shot of when it happened. You can see both when the problem started (I drew a crappy arrow) and also when the problem went away. You can also see a slight voltage drop, but that's probably a secondary side effect of the GPU core clock drop.

[Dr.Disaster]

It looks like a massive power drop, similar to switching from fullscreen with active 3D stuff to windows desktop.
Too bad your gpu driver version does not support the GPU-Z "Perfcap" sensor to make sure it is.
Which driver version is this? Worth an update eventually?

As example here some sensor readings showing system-triggered GPU power cuts: viewtopic.php?f=21&t=7088&start=20#p73910

IIRC your VAIO has nVidia Optimus features like automatic GPU selection depending on i.e. running on battery. I'd check those option too (fix assignment of LoG2 to nvidia GPU and so on).

[tfabris]

It is true that the system was set to automatic GPU selection. It was always autoselecting the nvidia GPU, and the indicator icon still said that Grimrock2 was using the nvidia GPU even while the problem was occurring, but it was certainly worth trying to set it to force the nvidia GPU. So I went into the nvidia control panel, selected grimrock2.exe, and set it to force the nvidia GPU.

No change: the problem still happens occasionally.

The problem happens maybe once every half hour or so on average. It only happens for a short time, like I dunno, maybe a minute? However long that graph dips in the screenshot I posted. But while it's happening its really hard to make mouse selections in the game because it doesn't register them properly.

I will see about upgrading the nvidia driver next.

[Dr.Disaster]

The problem happens maybe once every half hour or so on average. It only happens for a short time, like I dunno, maybe a minute? However long that graph dips in the screenshot I posted. But while it's happening its really hard to make mouse selections in the game because it doesn't register them properly.

Can't help but this "maybe once every half hour" for "maybe a minute" sounds much like something scheduled being executed again and again. Windows Defender scan plan? SSD TRIM? Forced energy saving? Software update checks? Could be anything that either requires some time to complete or is set to run that long. A process list would really help.

I recalled this:

I remember going into the Windows power management settings and making sure that everything was set to maximum performance when the laptop was connected to AC power.

Give this a try and create a new plan that never tries to save power. Together with nvidia power management set to "maximum" it should eliminate any power cuts (if it's a power cut).