Built-in standard benchmark / track so hardware reviewers can use DCS

[Vertigo72]

I would really appreciate a built-in performance benchmark; for a variety of reasons, but one very good reason is that it may make hardware reviewers add DCS to their game benchmarking suite. That would help us DCS users a lot when making purchase decisions as every new CPU or GPU would be benched on DCS, and its also free publicity for DCS.

 

It could be as simple as a standardized track that can be played back from the free DCS version, and average (/min/max/99%) FPS being shown, though I am not sure how representative the performance is of playing back a track vs actually playing the game? Does the physics and AI engine work in the background?

 

Also the last time I experimented with playing tracks and benching them, I also found the results could vary wildly between runs for reasons I dont quite understand, this would have to be improved too.

[The LT]

oh yes, we need this so bad...

[drPhibes]

Any kind of benchmark based on a piece of software that is in constant development (with frequent reports of performance changes from update to update) is completely worthless.

[Vertigo72]

I disagree. Its a snapshot that provides relevant data. Blender is also updated every few months, its still used as a benchmark in every CPU review ever, and the results are still relevant to me. The only thing irrelevant is comparing old benchmark results with new ones. Although even that is not irrelevant if you keep the hardware the same, as it may show improvement (or not) in the software. If reviewers want to recycle old results, nothing prevents them from keeping an older copy of DCS and use that. It will still be indicative.

[The LT]

Any kind of benchmark based on a piece of software that is in constant development (with frequent reports of performance changes from update to update) is completely worthless.

 

Benchmarking is done on different hardware configurations but on the same version of the software. Even dedicated benchmarking software gets occasional updates.

 

Constant development requires constant benchmarking to accurately gauge performance increase/decrease from builld to build. It will be much more indicative and objective to use a built-in benchmark so players can compare their performance.

[shaz56]

I would really appreciate a built-in performance benchmark; for a variety of reasons, but one very good reason is that it may make hardware reviewers add DCS to their game benchmarking suite. That would help us DCS users a lot when making purchase decisions as every new CPU or GPU would be benched on DCS, and its also free publicity for DCS.

 

It could be as simple as a standardized track that can be played back from the free DCS version, and average (/min/max/99%) FPS being shown, though I am not sure how representative the performance is of playing back a track vs actually playing the game? Does the physics and AI engine work in the background?

 

Also the last time I experimented with playing tracks and benching them, I also found the results could vary wildly between runs for reasons I dont quite understand, this would have to be improved too.

 

+1

[Worrazen]

LOL I was thinking of either making it or posting about the idea myself just 2 weeks ago.

 

It needs to be generated from script perhaps to cover things 100%, it would end up being too much work for a single person, especially when new things are introduced, new units, manually updating each time would take time and effort if one wants to keep it up to date.

 

I'll see what I can do.

Edited June 9, 2020 by Worrazen

[Buzzles]

Any kind of benchmark based on a piece of software that is in constant development (with frequent reports of performance changes from update to update) is completely worthless.

 

Quite the opposite.

 

The performance changes are the exact reason a benchmarking tool which runs the same test consistently is great, as you can then track performance increase or loss between versions.

[Bobik2002]

Quite the opposite.

 

The performance changes are the exact reason a benchmarking tool which runs the same test consistently is great, as you can then track performance increase or loss between versions.

+1

[Tom Kazansky]

such a benchmark could help finding out quickly whether there is a problem on the own settings or hardware or whether it is something that all/most/a lot users have e.g. after patches or in VR.

 

 

it would help optimizing performance and could reduce complaints

 

 

there should be a benchmark for each map/plane though, maybe with those two options.

 

 

 

 

so +1 for the OP

Edited June 9, 2020 by Tom Kazansky

[twistking]

such a benchmark could help finding out quickly whether there is a problem on the own settings or hardware or whether it is something that all/most/a lot users have e.g. after patches or in VR.

 

it would help optimizing performance and could reduce complaints

 

there should be a benchmark for each map/plane though, maybe with those two options.

 

so +1 for the OP

 

 

+1

even if it was only a simple scene with the free tf-51 on caucassus, it would help a lot with troubleshooting, because right now everybody compares different scenes when talking about performance. some even do their testing in multiplayer only ( :doh: ) without doing singleplayer testing first.

a simple benchmarking tool would help so much!

[currenthill]

I definitely agree, a standard benchmark track/mode would be very appreciated!

[Eldur]

+1

even if it was only a simple scene with the free tf-51 on caucassus, it would help a lot with troubleshooting, because right now everybody compares different scenes when talking about performance. some even do their testing in multiplayer only ( :doh: ) without doing singleplayer testing first.

a simple benchmarking tool would help so much!

 

Exactly... it's hard to compare those... I know DCS runs terribly on my rig even if I get almost constant 45fps in my test missions - as soon as I join one of the usual server online, I'm down in the tens. Comparing that to what I see on the Tube literally makes me sob.

 

Having a solid and standardized benchmark would make all those numbers actually comparable.

[Worrazen]

The big thing is also, it would help find false positives and what is causing what. And also very importantly, CONTROLLED CIRCUMSTANCES, so that you can have a clear list of factors of the circumstances so nothing changes, then it's much easier to find a mistaken or culprit factor, perhaps something may be due to some non-DCS factor, like something running in the background or some DCS plugin.

 

A performance dip may be normal due to excessive use of radar-equipped units for example. I have had this on my queue to show for a long long time and didn't got around doing a video about it, I have it in the plans this summer, probably around my own draft example of a benchmark map (mission with all units with some scripted stuff, but thing is IDK DCS scripting yet)

[wju]

1+ benchmark

 

[SharpeXB]

A benchmark cannot be done for DCS because the AI will do different things every time even in recorded tracks.