Blur Busters Forums

@kevindd992002, quick reality-check on my part...

I only have a 1080, but I do have a i7-4770k overclocked to 4.2GHz, and I can easily achieve 600+ framerates with V-SYNC off in an online CS:GO match.

After doing so many isolated tests in CS:GO for my input latency article, I was partially conflating the framerate in my test map with the framerate in normal gameplay.

I just played a couple of online matches with everything at max, 1440p, V-SYNC OFF, Multicore Rendering enabled, and I was getting anything from 300 to 500 FPS uncapped (fps_max 999). So, yes, CS:GO is somewhat of an exception to the bottleneck rule, and is more of a CPU limit to itself; brute force CPU power can only help so much here.

That said, in CPU-limited games with a 1080 Ti, you will see, at the very least, a 15% increase in performance with a 7700k over a 2600k, and on average, 20%+. You'll also have more stable frametimes, which means less CPU-induced stutter (frametime spikes).

I was browsing through some of the posts, what's this about MPRF > 1 plus hitting RTSS frame cap = reduced input lag? Is that really a thing?

So if you're running Overwatch at 300fps cap, but only max out at 260fps under the most load, let's say, you can reduce input lag by setting the in-game fps limiter to 258fps?

This doesn't really mean anything with Fast-Sync though, right? That automatically seems to cap FPS on its own (and if you can hit that frame cap without issue, I assume there is no additional input lag beyond what Fast-Sync typically introduces).

RealNC wrote:Single-core performance is the most important stat for CS:GO. If it maxes out a single core, you've hit the wall.

jorimt wrote:

kevindd992002 wrote:I'm not sure if this is off topic or not but here it goes. I initially thought it was a CPU bottleneck but then I checked the maximum thread usage while playing CSGO @ 1080p with my 1080Ti. It only reaches around 88% which means that it's not a CPU bottleneck. But then my GPU utilization/load is just effing 30%!

RealNC is correct, single core speed is what is going to make the largest difference in CS:GO. Not to stay off-topic, but have you tried "Multicore Rendering" both on and off in the game settings?

Exactly, that's why I specified "maximum thread usage" in my previous reply So that value is the max value among all the 8 threads of my 2600K. It's 88% so it's not reaching the max, hence CPU is not bottlenecked.

I just tried disabling multicore rendering right now and I got even lower framerates, for some reason.

jorimt wrote:@kevindd992002, quick reality-check on my part...

I only have a 1080, but I do have a i7-4770k overclocked to 4.2GHz, and I can easily achieve 600+ framerates with V-SYNC off in an online CS:GO match.

After doing so many isolated tests in CS:GO for my input latency article, I was partially conflating the framerate in my test map with the framerate in normal gameplay.

I just played a couple of online matches with everything at max, 1440p, V-SYNC OFF, Multicore Rendering enabled, and I was getting anything from 300 to 500 FPS uncapped (fps_max 999). So, yes, CS:GO is somewhat of an exception to the bottleneck rule, and is more of a CPU limit to itself; brute force CPU power can only help so much here.

That said, in CPU-limited games with a 1080 Ti, you will see, at the very least, a 15% increase in performance with a 7700k over a 2600k, and on average, 20%+. You'll also have more stable frametimes, which means less CPU-induced stutter (frametime spikes).

Thanks for the testing. But with 88% maximum thread CPU usage, will I still see increase in performance when upgrading to 7700K? I would understand this if one of my threads is pegged at 100% but with my situation I'm kinda confused.

drmcninja wrote:I was browsing through some of the posts, what's this about MPRF > 1 plus hitting RTSS frame cap = reduced input lag? Is that really a thing?

So if you're running Overwatch at 300fps cap, but only max out at 260fps under the most load, let's say, you can reduce input lag by setting the in-game fps limiter to 258fps?

This doesn't really mean anything with Fast-Sync though, right? That automatically seems to cap FPS on its own (and if you can hit that frame cap without issue, I assume there is no additional input lag beyond what Fast-Sync typically introduces).

You'll have to ask RealNC for further clarification on that, and I'd have to ultimately test it to verify. It would be very difficult to replicate/isolate properly, so no ETA/rush on this for me.

kevindd992002 wrote:Exactly, that's why I specified "maximum thread usage" in my previous reply So that value is the max value among all the 8 threads of my 2600K. It's 88% so it's not reaching the max, hence CPU is not bottlenecked.

I just tried disabling multicore rendering right now and I got even lower framerates, for some reason.

kevindd992002 wrote:Thanks for the testing. But with 88% maximum thread CPU usage, will I still see increase in performance when upgrading to 7700K? I would understand this if one of my threads is pegged at 100% but with my situation I'm kinda confused.

As far as I understand it, just because you're not "maxing out" one of the threads 100% does not mean your CPU isn't the bottleneck to your GPU. There is more to it than that.

The majority of games rarely use any or all of the CPU cores/threads 100% as is, so even if they are reading 88%, it can be a false positive on the "not bottlenecked" indication; if you're at very high usage percentages with your CPU, and very low usage percentages with your GPU, even at its max core clocks, then either the game engine or the CPU is bottlenecking max GPU usage.

Faster CPUs may not eliminate this behavior in games that are CPU heavy, but they will allow the GPU to fullfill more of its potential. E.g. where 88% core usage on the 7700k would allow more GPU usage, the 88% on the 2600k would not, as the 7700k would have a higher performance ceiling, and what is 88% to the 7700k, could be 108% to the 2600k; It's all relative.

For instance, I have a 4770k and a 1080, and I get 200+ more frames than you do at 1440p, and the max usage on one of my cores can peak at 92% with V-SYNC OFF/Multicore Rendering enabled (average total CPU usage is 38%).

My 15-20%+ increase in performance estimate across CPU-limit situations with a 7700K vs. a 2600k still applies, and may be conservative in some instances, especially considering your 1080 Ti. The 7700k also has much faster single core speeds when compared to the 2600k, which may help in games such as CS:GO.

jorimt wrote:

drmcninja wrote:I was browsing through some of the posts, what's this about MPRF > 1 plus hitting RTSS frame cap = reduced input lag? Is that really a thing?

So if you're running Overwatch at 300fps cap, but only max out at 260fps under the most load, let's say, you can reduce input lag by setting the in-game fps limiter to 258fps?

This doesn't really mean anything with Fast-Sync though, right? That automatically seems to cap FPS on its own (and if you can hit that frame cap without issue, I assume there is no additional input lag beyond what Fast-Sync typically introduces).

You'll have to ask RealNC for further clarification on that, and I'd have to ultimately test it to verify. It would be very difficult to replicate/isolate properly, so no ETA/rush on this for me.

kevindd992002 wrote:Exactly, that's why I specified "maximum thread usage" in my previous reply So that value is the max value among all the 8 threads of my 2600K. It's 88% so it's not reaching the max, hence CPU is not bottlenecked.

I just tried disabling multicore rendering right now and I got even lower framerates, for some reason.

kevindd992002 wrote:Thanks for the testing. But with 88% maximum thread CPU usage, will I still see increase in performance when upgrading to 7700K? I would understand this if one of my threads is pegged at 100% but with my situation I'm kinda confused.

As far as I understand it, just because you're not "maxing out" one of the threads 100% does not mean your CPU isn't the bottleneck to your GPU. There is more to it than that.

The majority of games rarely use any or all of the CPU cores/threads 100% as is, so even if they are reading 88%, it can be a false positive on the "not bottlenecked" indication; if you're at very high usage percentages with your CPU, and very low usage percentages with your GPU, even at its max core clocks, then either the game engine or the CPU is bottlenecking max GPU usage.

Faster CPUs may not eliminate this behavior in games that are CPU heavy, but they will allow the GPU to fullfill more of its potential. E.g. where 88% core usage on the 7700k would allow more GPU usage, the 88% on the 2600k would not, as the 7700k would have a higher performance ceiling, and what is 88% to the 7700k, could be 108% to the 2600k; It's all relative.

For instance, I have a 4770k and a 1080, and I get 200+ more frames than you do at 1440p, and the max usage on one of my cores can peak at 92% with V-SYNC OFF/Multicore Rendering enabled (average total CPU usage is 38%).

My 15-20%+ increase in performance estimate across CPU-limit situations with a 7700K vs. a 2600k still applies, and may be conservative in some instances, especially considering your 1080 Ti. The 7700k also has much faster single core speeds when compared to the 2600k, which may help in games such as CS:GO.

That makes sense and is actually the answer I was expecting. I just wasn't sure if CPU bottlenecking equates to one of the threads being pegged at 100% so I wanted someone to confirm.

The o ly thing I don't understand now is why in Overwatch I get 99% GPU load at 1080p and I get arpund 250fps which is good.

If you get high GPU usage, it means the game uses the GPU. More complex graphics (or badly optimized graphics) need more GPU. Overwatch has more demanding graphics than CS:GO (by a large margin.) CS:GO has low-end graphics (even the new maps don't really need much GPU). That means that since the GPU isn't used much, it's the CPU that will be the bottleneck in most cases.

In OW, the GPU can be the bottleneck. That means that reducing graphics settings, or playing at a lower resolution, or getting a faster GPU, will increase FPS.

Also, at any given framerate, being GPU limited will have higher latency than being CPU limited. So it can be worth using an in game framerate cap even if you're not hitting your monitor's maximum refresh rate(just below your minimum framerate).

drmcninja wrote:I was browsing through some of the posts, what's this about MPRF > 1 plus hitting RTSS frame cap = reduced input lag? Is that really a thing?

I posted a test you can do yourself. Should work in every game, even OW.

So if you're running Overwatch at 300fps cap, but only max out at 260fps under the most load, let's say, you can reduce input lag by setting the in-game fps limiter to 258fps?

I expect the effect in OW to be minimal, since OW avoids excessive frame buffering and uses MPRF 1. There still should be a slight input lag reduction even in OW, but it's probably very small.

The effect seems very pronounced in games that are not optimized for low input lag.

RealNC wrote:I expect the effect in OW to be minimal, since OW avoids excessive frame buffering and uses MPRF 1. There still should be a slight input lag reduction even in OW, but it's probably very small.

The effect seems very pronounced in games that are not optimized for low input lag.

To get back on-topic, I actually briefly tested something related to this yesterday.

I was curious on whether Overwatch would benefit from MPRF 1 in a very specific but typical G-SYNC scenario, and it appears it does not.

Same test method as my article, but latest driver version and my XB271HU was being used at 144Hz instead. In-game settings maxed, G-SYNC + V-SYNC, 141 FPS in-game limit, Reduced Buffering enabled:

MPRF Default:
MIN: 20
AVG: 23
MAX: 26

MPRF 1:
MIN: 21
AVG: 24
MAX: 28

I only did a single 10 sample run per, but it is enough to know whether there is up to 1 frame of reduction/delay. If I had done four runs per, they would have been within 1 ms of each other (margin of error).

So Overwatch does appear to be using an internal value of MPRF 1, unless of course the setting would only have an effect in another scenario.

Again, the MPRF setting "depends" on a variety of factors; it would be interesting to do some real in-depth tests in the future. Once I recover from my test fatigue, I may.

MPRF should never have an impact when your framerate is limited by an in game cap.