Transcription
Alrighty guys, well welcome to the video. So today, in this guide, what I wanted to do is sort of a more complex and more comprehensive tutorial on how to actually improve your frame times along with slightly improving your frame rates because the two seem to be kind of mixed up in terms of what people actually try to prioritize in their system. The truth is, is that frame times can sometimes be just as important as frame rates because if you have inconsistent frame times, that is almost just as bad as having inconsistent frame rates. And so, in this guide, I wanted to kind of explain how to actually tweak frame times because frame times seem to be a lot more difficult to actually manage. They seem to be kind of variable even with high-end hardware, so it's not something as simple as just buying faster hardware. Frame times seem to have a few other things that I wanted to reveal to you guys and kind of show you what actually is at play when you're trying to to do those tweaks.
And so, the main things that I did for, I ran three different benchmarks on four different games, and each one of these games has something specific that it's looking for. Overwatch is incredibly RAM and CPU sensitive. If you give that game really good RAM and you tweak the timings in the frequency, that game will just eat it up almost linearly in terms of how well it scales. And then for a game like Escape from Tarkov, it sort of leans around a little bit of the medium between CPU and GPU usage. And then Cyberpunk 2077 definitely seems like it enjoys a really high CPU usage, like CPU frequency, and mostly the GPU if you're running at high settings. And then Call of Duty Modern Warfare 2 again seems to be kind of a nice little blend between the two.
And so, for the first game, what I did is, I'll explain it here. So, there are essentially three tweaks that I did. I changed the priority level along with the thread priority level, and then I changed the CPU clocks, and then I changed the GPU clocks, and I physically locked them to a set frequency. Now, how do I do that for all of these? Well, here's the first way for your GPU. All you have to do is go into this little area right here, and you'll see on this part of this graph, you'll see this little yellow line. Well, if you press L, that is what essentially will lock the actual GPU core clock, so long as our temperature doesn't spike or change too much. And so, if I go right here, and I just go in the same spot that I was at just barely, and I click L, and then I hit the check mark, that now locks that frequency exactly where I need it to be. And I also increased the base clock frequency because at stock, if you actually check it, the graph looks like this. But with the actual application that I did, I not only locked the clocks, but I also raised that base clock in order to also improve the consistency of the GPU. And so, that's how I did it for the GPU.
And then with the CPU, this is sort of a trade-off in terms of what you're doing. The drawback to this is that it draws a lot more heat, but it improves the consistency of the CPU because what it does is, it, it's known as disabling idle states. Now, this one is sort of dependent on how good your cooler is and how good your system is because if you disable it, it can kind of cause more temperatures and it can also be sometimes bad. Or if you enable it, it can do the opposite. So, well, I'll show you what that does actually. So, you'll notice in Task Manager that it says 100% CPU usage. That's what disabling idle states does. It essentially prevents your CPU frequency from fluctuating along with its usage, or essentially its idle state. So, that's different than applying an all-core turbo ratio. It's slightly different than that, but I'll show you what happens here when I apply this command. If I go into here, and then I click that, and I enter it, what it does is it drops the CPU usage, and the CPU usage, the idle state will only change depending on the load that the system is actually experiencing for each CPU. And so, that's essentially the tweak is that it allows a more consistent CPU utilization in a CPU usage. And so, that's how I did it for the actual CPU and GPU side.
Now, for the priorities, I mentioned this software previously, but what this allows me to do is if I go into a specific application or game that I'm playing, I can change basically everything that I want about the game. I can change the priority to real-time, I can change the I/O priority to high, I can change the memory priority or the page priority to normal. And then, if I so choose, I can actually select every single thread within that actual specific service and I can apply what is known as a time critical priority. And so, that essentially makes it to where each individual thread is now set to the highest priority of a system service. So, this can actually be super beneficial if you have a lot of services running in the background and they're all kind of competing for priority or essentially attention in the actual operating system. And so, that's essentially the priority tweaks.
Now, let's get to those benchmarks and I'll show you guys what they look like. So, starting off with Overwatch, we'll see here that we have three of them. And we'll notice that the main improvements, this will be very consistent going through the next three games that we show. We don't really notice a ton of change in the average or the one percentiles. What you notice is the 0.2 percentiles do actually increase a ton, or actually a fair amount. And so, if we set our mouse right here, we'll notice that the actual improvement is about 12% from stock value. And so, that's a nice bump, especially for a game like Overwatch where you have thousands of tiny little particles flashing on your screen at any one point. So, improving consistency in this game with frame times is much more noticeable because of how, how sort of, uh, microscopic the game is in terms of particle effects. And so, then if we check the line charts here, we can see that there are three different charts, and it's very tight grouping because this game already runs at 600 frames per second. But even here, with such a consistent frame time overall, we'll notice that if we highlight the green box right here, the frame time spikes are much less frequent than with the stock or the priority changes. When we apply both of them, the game seems to benefit a ton from that. And then if we go into the frame time variances, we'll notice that again, we've crunched the graph significantly. We've cut a little bit of the eight millisecond frame time variance, and we've almost completely eliminated most of the actual large 12 millisecond variance. So, Overwatch definitely seems to appreciate just about all of these tweaks, but it's also an incredibly well-optimized game, so it makes sense that it would benefit from these kind of tweaks.
And now let's go to Tarkov. So, with Tarkov, unfortunately, I wasn't able to get the priority tweaks. I was only able to get the priority tweaks and the locked clock tweaks. So, this essentially is because I died on that part. So, we'll even notice that here in the same raid that I did, we'll notice that the frame time variance is significantly tighter. Now, we no longer have that annoying hitch of above 12 milliseconds for a fraction of a second. We'll notice that the line charts also, it looks very similar in terms of the frame times, but when we actually highlight over the tweaks applied, we'll notice that we don't get as, as large spikes as we did before. We'll notice that it's much more consistent to you in terms of how much it's delivering, and it's actually slightly lower than the stock value in terms of the consistency. So, in a game like Tarkov, again, it does seem to benefit a little bit from these tweaks. And then, as we see in the actual frame time charts, or actually the frame rate charts, we'll notice about a 20 to 10% increase in the 0.2 percentile lows. So, this is sort of becoming a pattern as we are going to see going forward.
And then with Cyberpunk 2077, we'll notice that it seems like the frame rate doesn't really change a whole lot, especially the variance. Again, it's only less than 10%, so it's not a ton, right? And then if we look at the line charts, we'll notice that this is actually where this game seems to benefit a lot from doing these tweaks. Is if we highlight over here, it's a very, uh, unoptimized game. So, something like this probably won't be as beneficial, but it will at least help in terms of a game like Cyberpunk, which is a very demanding game. And so, it's a much significantly lower value than what is normally on the blue and the orange. And so, then if we go to the frame time variance again, it's a slight improvement, but not a crazy amount like Overwatch or Tarkov. But that's still a nice thing to know that we can still make a few different improvements in the frame time variances just by doing these simple tweaks.
And so, then with the final game, we have Call of Duty Modern Warfare 2. Ran the in-game benchmark, and we'll even see here that actually applying these tweaks is probably the most beneficial for this game because not only do we cut all of the red and orange, which are the really, really bad frame time spike variances, but if we actually look at the line charts as well, again, we can see that stock, the frame time spikes can get significantly high, almost 16 milliseconds, which is as low as 60 frames per second. But if we highlight over the orange, we'll notice that we only have two or three spikes throughout the entire benchmark run. Whereas if we do just the regular priority tweaks, we'll get a sort of reduction as well, but doing all of them helps the most. And then finally, with the frame rate chart again, it's only really about 10 to 20% on the 0.2 percentile lows. But that's a massive improvement because you don't really feel the averages when you're playing a game. You feel the one percentiles and the 0.1 percentiles because those are the fractions of a second that you'll notice, and that's what actually makes the game feel, um, unresponsive or essentially feels cluttery or clunky.
And so, yeah guys, hopefully you enjoyed this guide. Um, please like, please comment, please leave any suggestions that you guys want to have in the comments. Um, I will link all of the different commands, and I will also link the guides on how to do this because this is all stuff that we've covered in this channel before, except System Informer, but I'll link a guide for that as well. And so, yeah guys, hopefully you enjoyed this video. Please like, subscribe, comment, favorite, share, all of that fun stuff. Anyway guys, have a good one. My name is Savitarix, and I'm out.