[Testing Branch Update] - 370532

[Ipsquiggle]

Hey friends. The solid conduit changes are being backed out for now so that they don't hold up the rest of the improvements. This will resolve all the various solid/liquid/gas plumbing inconsistencies that cropped up. 

It's possible that save files made on the testing branch will not work correctly with this rollback, so we recommend going back to your last save from the Live branch, or starting a new colony.

Have a good weekend!  

Changes:

• Reverted the change "Conveyors now use the same optimized and multithreaded algorithm that the Liquid and Gas pipes do."
• Fix typo in key binding string
• Fix wire load status becoming invisible if the load is safe

 

View full update

 

[Moah]

Yeah!

Edit:

Tested a bit, savegame works, game seems very stable. Thanks, good work!

Also a good weekend to the whole ONI Team!

Edited September 27, 2019 by Moah

[kapoff]

Thanks for this. I can really test it now !
Good wenk-end the team !

Edited September 27, 2019 by kapoff

[Loraty]

My steam engine setup still behaves very differently: With the public build steam temperature stays above 170°C with doors mostly open while the testing build sees a steady decline down to 140°C with doors closed.

The weird disinfecting has stopped, at least.

[Guest aresd]

21 cycles. Initial biome, minimum buildings. Out of 150 FPS, at best, 115 remain on average already 100 or less. Your optimization brought exactly 0 results

[Steve Raptor]

1 hour ago, aresd said:

21 cycles. Initial biome, minimum buildings. Out of 150 FPS, at best, 115 remain on average already 100 or less. Your optimization brought exactly 0 results

There won't be any difference because the resources used by the game when you have almost nothing built are minimal...

Those optimizations are targeted at reducing performance hits caused by stuff like conveyor rails and pipes and other things that generally make a difference when you build them in large amounts.

Edited September 27, 2019 by Steve Raptor

[Guest aresd]

3 minutes ago, Steve Raptor said:

There won't be any difference because the resources used by the game when you have almost nothing built are minimal...

Those optimizations are targeted at reducing performance hits caused by stuff like conveyor rails and pipes and other things that generally make a difference when you build them in large amounts.

Well, I'm based on a huge previous patch. There was much more information about it.

what's the point of optimizing things that don’t affect the game because the players cannot come to this because of the incredibly slow game at an earlier stage? I tried my base in 3500+ cycles before the patch and after the patch and did not notice 0 changes in performance. I tried to create a new map and on the 40th cycle, I had 90 frames left from 150 frames, and the studied biome area increased only by 15 percent from the screenshot. That is - 0% performance gain.

[woodenstars]

What's the point to be so edgy when you have 115 fps lol? Everyone else feels improvements, well, sucks to be you.

[Guest aresd]

3 minutes ago, woodenstars said:

What's the point to be so edgy when you have 115 fps lol? Everyone else feels improvements, well, sucks to be you.

because such serious losses in performance at the beginning are probably the cause of problems not only for me, but for everyone who is used to playing at 15 fps, while they could play at 60

[woodenstars]

2 minutes ago, aresd said:

because such serious losses in performance at the beginning are probably the cause of problems not only for me, but for everyone who is used to playing at 15 fps, while they could play at 60

Teach me how can I get 60, PLEASE

Edited September 27, 2019 by woodenstars

[Guest aresd]

4 minutes ago, woodenstars said:

Teach me how could I have 60, PLEASE

Swap your transistor binary modulator to Pentium 3, obviously

[nakomaru]

On 9/20/2019 at 8:33 PM, nakomaru said:

Here are my results for my cycle 21298 base, attached. Time is measured until I can move the camera.

• Original 10x speed zoomed in FPS: 23-24
• Aggressive 10x speed zoomed in FPS: 35-37

Above are my live branch FPS with save file included. Additionally, here are my performance patch 370532 stats:

• Patch 370532 aggressive pruning 10x speed zoomed in FPS: 57

In practical camera situations, it's about 40-45 at 10x. Great work. With these latest two changes, a mod and a patch, I've more than doubled the fps with much less hitching to boot. I hadn't seen 57 fps at 10x in 20000 cycles on that world.

Edited September 28, 2019 by nakomaru

[dallion]

2 hours ago, nakomaru said:

Above are my live branch FPS with save file included. Additionally, here are my performance patch 370532 stats:

• Patch 370532 aggressive pruning 10x speed zoomed in FPS: 57

In practical camera situations, it's about 40-45. Great work. With these latest two changes, a mod and a patch, I've more than doubled the fps with much less hitching to boot. I hadn't seen 57 fps at 10x in 20000 cycles on that world.

are you serious? at 1300+ circle with 3x speed, I can hardly got 10+ fps

[nakomaru]

Yes. The base is designed with performance as the top priority, and I have a nice computer.

[avc15]

I wanted to do some actual empirical testing, seen a lot of opinions about whether this update improved things or not, but it's important to keep your own bias in view on something like this. I'm using MSI Afterburner.

fast save.sav

(colony is Sticky Committee, Fast Save is just a backup I made before switching to the fast save mod)

I'm running on an i7 5930k hex with 16 GB of RAM.

Test 1: cycle 1660, with build 370532: (this performance update)

Loaded game, 3x speed, closed ALL programs but afterburner & ONI, zoomed to default widest view and centered camera on my power plant room (one of my busier areas)

31 FPS, one big stumble.

Test 2: cycle 1660, build 366134: (live branch)

Loaded game, set to 3x speed, closed ALL programs but afterburner & ONI, zoomed to default widest view and centered camera on my power plant room. Allowed it to record data until sunset, as before.

22 fps, frequent significant stumbles.

You can see in the screenshots that there are about 180 seconds worth of data in each test.

I have to say just really phenomenal work. It comes out to a 43% performance improvement for my late game colony.

Rebuttals, one at a time

"This performance patch is bad" - please do some unbiased measurement, maybe you're gauging your response to what you hoped for just as much as what you're used to.

"There are more stutters" - count them, you might just be noticing stuttering more easily than previously, because it's not happening constantly anymore.

 

Edited September 28, 2019 by avc15

[blash365]

5 hours ago, avc15 said:

I wanted to do some actual empirical testing,

I agree with your approach, i concur with your findings. But making 2 measurements is hardly empirical. Of course the task to perform all the measurements shouldnt be on your shoulders alone (you provided the first step by revealing your methods and hopefully other users with different hardware and colonies will follow), but with just one colony, one computer and 2 measurements (1 day in the colony, if i understood correctly), it is still very prone to artefacts.

Especially the finding "there are not many spikes".

[avc15]

empirical literally just means measurable, evidence based, reproducible. But let's not get stuck on semantics.

If a few other people repeated this simple test, & shared their results as well it'd be a lot more meaningful than me going and doing 10 more of them.

What I'm saying is, instead of throwing shade - go spend the same 20 minutes I did and count stutters. Let us know your results.

Edited September 28, 2019 by avc15

[Senteliks]

So in conclusion, no matter how much you do the optimization, you will get bested by Unity.  Best you can do, is reduce the impact of performance hogs, hope for the best, and also, that new generation CPU's (preferably intel) will leverage multi-threading to the max. 

Edited September 29, 2019 by Senteliks

[Oozinator]

3 hours ago, Senteliks said:

you will get bested by Unity.

Unity is not to blame for it.
There are many well optimized games out there, running on Unity.
 

[OxCD]

31 minutes ago, Oozinator said:

Unity is not to blame for it.
There are many well optimized games out there, running on Unity.
 

With much more GPU requirement for sure, but with as much CPU calculation as ONI, not so easy to find one I think.

[nakomaru]

The performance update shows both that there is room for improvement and that they are working on it with some success.

Edited September 29, 2019 by nakomaru

[Oozinator]

1 hour ago, OxCD said:

With much more GPU requirement for sure, but with as much CPU calculation as ONI, not so easy to find one I think.

Cities skylines?
 

[OxCD]

1 hour ago, Oozinator said:

Cities skylines?
 

I rely on your experience then, I've never touched it

[Oozinator]

1 hour ago, OxCD said:

I rely on your experience then, I've never touched it

When you run a big city, with all that traffic calculations there, i think it could be CPU hungry, like ONI is, but only Klei coders could know for sure.

When you have a "fork", you can not eat "chicken soup" well with it, even when it's the finest chicken soup ever.
Blaming the fork does not stop your hunger. All you can do is eating some cookies, even when you like chicken soup much more.

[OxCD]

1 hour ago, Oozinator said:

When you have a "fork", you can not eat "chicken soup" well with it, even when it's the finest chicken soup ever.
Blaming the fork does not stop your hunger. All you can do is eating some cookies, even when you like chicken soup much more.

That was... deep. ^^

Edited September 29, 2019 by OxCD