Standardised VR benchmark

With so many coming in to VR thanks to the Reverb G2, as well as the 3000 series, we're going to see a lot of people coming in with performance queries. Unfortunately DCS still doesn't have a built-in benchmark, even though it's as simple as running a track file and recording the min, avg, and max fps. So how about we come up with a test ourselves? I would suggest we use assets that are available to all - that being the SU-25 and Caucasus. I'd also suggest we use one of the instant action missions. As for the graphics settings, the presets would be most repeatable between users, as well as come up with one or two of our own, as the VR preset is quite frankly ugly.

What do y'all think?

Someone attempted it already. Doesn't seem like it got much support seeing it only has about 130 downloads. You can search for benchmark on the User Files section.

Pretty sure I've seen such a thread before. But considering the newly released hardware, a new "benchmark" thread could be helpful. Also, and please correct me if I'm wrong, I think there hasn't been a VR only benchmark thread. Limiting to VR would at least filter the results a bit further.

In addition to what you mentioned, you should also set a standard report format (just like the previous benchmark threads).

Something like:

Performance result

Max fps readout during test: ...

Hardware specs

Hmd used: ...

Cpu: ... @ ... (clockspeed)

Gpu: ... @ ...

Ram: ... @ ...

Storage: ssd/hdd

Cooling: liquid/air

To prevent thread cluttering by all kind of graphs and other stuff that require additional software.

As you said, it should be easy to use. So indeed maybe a track file would suffice. (I never actually use track files myself, so I don't know if that's a good option)

The problem with using track files is that they don’t repeat the action correctly ie they’re broken. You’ll get different results every time you run it. Maybe without any AI the tracks work but then that’s not really a valid benchmark.

ED could create a simple utility to test a PC by subjecting it to the things that DCS would normally do. Think of it like Passmark, Linpack or Furmark, inter alia.

That way everyone would download the proggy, run it and post their "number" along with system specs. Though I doubt they will or would care to do it.

ED could create a simple utility to test a PC by subjecting it to the things that DCS would normally do. Think of it like Passmark, Linpack or Furmark, inter alia.

 

 

 

That way everyone would download the proggy, run it and post their "number" along with system specs. Though I doubt they will or would care to do it.

It would be a trivial piece of work for ED to build a benchmark. And considering how often game performance is discussed, along with tweaking suggestions that swing from helpful to utterly absurd, it would help out a huge portion of the players.

I'll check out the other guy's benchmark - no point in repeating a piece of work that's already been done.

The problem with a typical benchmark that shows Min Max and Average is that DCS performance is governed by two distinct things, CPU and GPU which are dependent on particular situations. You’re better off testing those scenarios deliberately instead of looking at just a wide variety over a whole track. Test your CPU-limited performance on the ground or above a large city, a situation with a high object count, aircraft or AI. Test your GPU in free flights. Knowing the limiting factor helps you decide on settings.

I use Passmark to benchtest my PC's. You can test single thread CPU performance and GPU. For reference, my DCS system gets 3044 for single thread and ~19000 for GPU....though DCS will take everything and then some no matter what.

I don't understand why more peeps don't refer to Passmark for performance results. See here for reference:

https://www.cpubenchmark.net/singleThread.html

https://www.videocardbenchmark.net/high_end_gpus.html

I have another system running the i5-6600k that gets 25454 Passmarks for single thread CPU and ~13500 for GPU with the RTX2070 super. I can't run DCS in VR on that one for want of frame rates at 45.

I use Passmark to benchtest my PC's. You can test single thread CPU performance and GPU. For reference, my DCS system gets 3044 for single thread and ~19000 for GPU....though DCS will take everything and then some no matter what.

 

 

I don't understand why more peeps don't refer to Passmark for performance results. See here for reference:

 

 

https://www.cpubenchmark.net/singleThread.html

https://www.videocardbenchmark.net/high_end_gpus.html

 

 

I have another system running the i5-6600k that gets 25454 Passmarks for single thread CPU and ~13500 for GPU with the RTX2070 super. I can't run DCS in VR on that one for want of frame rates at 45.

Somehow your links don't open here.

So you can't benchmark in VR with this software? Did I understand that correctly?

Isn't the monitor/resolution a high factor when it comes to benchmark results? (I mean, I use a very old low res monitor next to my Reverb, so would a benchmark with Passmark give a representative result?)

So you can't benchmark in VR with this software? Did I understand that correctly?

Isn't the monitor/resolution a high factor when it comes to benchmark results?

Passmark is just testing your hardware performance. It’s good to do in order to know whether you’re getting a certain degree of performance from each component. It’s not related to your monitor resolution etc.

No one cares about artificial benchmarks, those don't tell you if you change from a i7 7700k to a i9 10900k your performance in dcs multiplayer vr on a specific test mission will increase by X % which is what people are asking for.

Ed needs to build this into DCS so that in the test the AI behaves the same way every time. Maybe that's not possible with the current engine, but it would be better than what we have now.

It could just be an instant action mission called benchmark where the player doesn't have control over the aircraft nor the ability to change views, and it spits out your fps at the end.

It’s possible that the reason DCS doesn’t have a benchmark track is that

-The AI will do something different in every replay

-Running physics and AI live in real gameplay is more demanding than in a recording.

I did tests like this for myself long ago and really can’t remember the results. It’s easy to do, just run a live mission and then replay it with Fraps running both times. In other sims I would see a big difference between both results. The track function has been broken in DCS for years now though, making this sort of test impossible.

Here I use a low level basic bombing mission filled with a bunch of SAMs to benchmark my system. It's fun, challenging and I get practical results.

I rather see EDs focus towards removing/optimizing legacy code in stead of investing time into a benchmark most people won't use or just use once or twice. :)

In my humble opinion benchmarking is utterly useless.

If hardware reviwers use a DCS benchmark it will popularice the game more than crysis.

It will be good.

If hardware reviwers use a DCS benchmark it will popularice the game more than crysis.

It will be good.

:megalol:

If hardware reviwers use a DCS benchmark it will popularice the game more than crysis.

It will be good.

Nah MSFS2020 is the new Crysis, it is much worse than DCS when it comes to framerate trashing. It will probably get even worse in VR mode.

On a serious note. We should come up with a systematic approach that would allow people to compare the strength of their setup relative to others; and also to see if their setup if really optimized compared to people with similar PC specs. To do this while minimizing bias and confusion, the settings for DCS and mission selection as well as flying altitudes should be the same. And the methods should be easily reproducible by everyone, so we have something in common (since we don't have the exact same hardware setups).

What I have in mind.

1. PD 1.0, Steam VR SS 100%.

2. Fixed presets only (VR vs medium vs high), no individualized settings.

3. Use of instant or built-in or training missions for commonly available planes. For example, low CPU taxing mission such as T51 free flight, as well as high CPU taxing mission such as ground training or flight in highly congested area.

4. Immediately fly to a fixed spot after start (such as WP1) at default ALT for say 10-15 sec then fly low 1000-1500ft.

5. Use of fpsVR to record CPU as well as GPU frametime (to see what is actually bottlenecking the framerate in certain situations), overall framerate, at high and low ALT.

It would be a trivial piece of work for ED to build a benchmark. And considering how often game performance is discussed, along with tweaking suggestions that swing from helpful to utterly absurd, it would help out a huge portion of the players.

 

I'll check out the other guy's benchmark - no point in repeating a piece of work that's already been done.

But you're forgetting the number one reason why vendors don't have performance testing. The results can be used against them by the competition.