4.68T Optimization Test
- FlowerChild
- Site Admin
- Posts: 18753
- Joined: Mon Jul 04, 2011 7:24 pm
4.68T Optimization Test
Hey guys, I've put together a little test version so you can try out some of the optimizations I've been working on:
Download Link
Note that this version only includes the client, as these changes shouldn't impact the server side of things. I also don't think there will be any issues playing on a 4.67 server with these changes, despite the warning it may give you upon joining about a version mismatch.
I'd be particularly interested in hearing from people with low end systems as to what kind of impact this has on their overall performance as compared to the last version of the mod. I want to try to establish how much of an impact these changes have before proceeding further to make sure I'm not barking up the wrong tree with all of this.
Please note that I have no idea what MCPatcher will do to the rendering code that I've optimized, so if you're going to post results, please do your tests WITHOUT MCPatcher installed. Just do clean installs with BTW and Modloader, and that's it. Any results with MCPatcher installed don't tell me anything of value given I don't know the actual state of the rendering code after it applies its patches.
I'll keep my fingers crossed that this helps you guys out :)
Download Link
Note that this version only includes the client, as these changes shouldn't impact the server side of things. I also don't think there will be any issues playing on a 4.67 server with these changes, despite the warning it may give you upon joining about a version mismatch.
I'd be particularly interested in hearing from people with low end systems as to what kind of impact this has on their overall performance as compared to the last version of the mod. I want to try to establish how much of an impact these changes have before proceeding further to make sure I'm not barking up the wrong tree with all of this.
Please note that I have no idea what MCPatcher will do to the rendering code that I've optimized, so if you're going to post results, please do your tests WITHOUT MCPatcher installed. Just do clean installs with BTW and Modloader, and that's it. Any results with MCPatcher installed don't tell me anything of value given I don't know the actual state of the rendering code after it applies its patches.
I'll keep my fingers crossed that this helps you guys out :)
- FlowerChild
- Site Admin
- Posts: 18753
- Joined: Mon Jul 04, 2011 7:24 pm
Re: 4.68T Optimization Test
Oh, and BTW, these changes unfortunately do not appear to have solved the nether-stutter problem as I had hoped. The change to fluids was what I was hoping would take care of it, and I suspect it may be better, but I'm still seeing it occasionally happen.
-
- Posts: 61
- Joined: Wed Apr 03, 2013 4:42 pm
Re: 4.68T Optimization Test
My test was fairly inconclusive.
In terms of straightforward FPS, both tests saw around 25-30 FPS average, with spikes to both extremes (4 FPS - 60 FPS). Chunk loading felt smoother in 4.68T in the immediate area around me, but was still mediocre ~100 blocks away.
Not sure if relevant, but I was using Fancy Graphics, Maximum Smooth Lighting and Advanced OpenGL Off.
In terms of straightforward FPS, both tests saw around 25-30 FPS average, with spikes to both extremes (4 FPS - 60 FPS). Chunk loading felt smoother in 4.68T in the immediate area around me, but was still mediocre ~100 blocks away.
Not sure if relevant, but I was using Fancy Graphics, Maximum Smooth Lighting and Advanced OpenGL Off.
- StarsintheSky
- Posts: 41
- Joined: Tue Aug 28, 2012 3:45 pm
Re: 4.68T Optimization Test
4-year-old laptop:
4.67: Generally 40 FPS dipping to 30 when I look out from the top of a high place.
4.68T: Generally 60 FPS dipping to 45 when I look out from the top of a high place.
I spent a little time roaming around my base, digging a little, watching my machines. Chunk loading was smooth. My test was a little limited due to my free time but I can confidently say that there is at least some improvement between 67 and 68T and it may very well be on the order of 10-15 FPS.
For consistency: Fast graphics, Max FPS, Max Smooth Lighting, Advanced OpenGL ON
Edit 2: I'm using MagicLauncher (have been since February) to reduce the memory allocated to Minecraft but I have maintained the same settings for both 67 and 68T. I don't know whether that matters but I could test again without that influence.
4.67: Generally 40 FPS dipping to 30 when I look out from the top of a high place.
4.68T: Generally 60 FPS dipping to 45 when I look out from the top of a high place.
I spent a little time roaming around my base, digging a little, watching my machines. Chunk loading was smooth. My test was a little limited due to my free time but I can confidently say that there is at least some improvement between 67 and 68T and it may very well be on the order of 10-15 FPS.
For consistency: Fast graphics, Max FPS, Max Smooth Lighting, Advanced OpenGL ON
Edit 2: I'm using MagicLauncher (have been since February) to reduce the memory allocated to Minecraft but I have maintained the same settings for both 67 and 68T. I don't know whether that matters but I could test again without that influence.
IGN: Beschaulichkeit
- FlowerChild
- Site Admin
- Posts: 18753
- Joined: Mon Jul 04, 2011 7:24 pm
Re: 4.68T Optimization Test
Ok, there is something fucked up here. I'm currently getting higher frame-rates in my debug version of the mod, than I am with the release version I sent to you guys.
That should just never be the case.
Investigating.
That should just never be the case.
Investigating.
- FlowerChild
- Site Admin
- Posts: 18753
- Joined: Mon Jul 04, 2011 7:24 pm
Re: 4.68T Optimization Test
Ah, nm, turns out I had advanced GL turned on in my release version, which decreases my framerate significantly.
Really though I was losing my shit there for a second :)
Really though I was losing my shit there for a second :)
-
- Posts: 61
- Joined: Wed Apr 03, 2013 4:42 pm
Re: 4.68T Optimization Test
I've played Minecraft on every manner of video card and driver, and I've never once experienced a reduction in lag thanks to Advanced OpenGL, which is odd since occlusion culling is supposed to prevent unnecessary rendering...FlowerChild wrote:Ah, nm, turns out I had advanced GL turned on in my release version, which decreases my framerate significantly.
- DaveYanakov
- Posts: 2090
- Joined: Sat Jul 23, 2011 5:17 am
Re: 4.68T Optimization Test
Nether stutter on chunk generation is still present but has been reduced in duration. It feels like I'm recovering from each freeze somewhere between half again and twice as quickly. Overworld chunk generation has been improved with a complete lack of observable cave-revealing world holes after ten minutes of sprinting around a new region. There is definite improvement here.
Better is the enemy of Good
Re: 4.68T Optimization Test
Overall I think I saw a 3-5 fps gain. Depending on the situation.
Running on a i7 laptop, with Intel HD 3000 graphics and 3Gb of memory given to Minecraft.
My best results were with Open GL on, Graphics on Fast, Smooth lighting on minimum, and Performance set to max fps. With a range of 25-40fps.
But definite improvement in world rendering in general no sign of surface holes that I can see the caves through..
Running on a i7 laptop, with Intel HD 3000 graphics and 3Gb of memory given to Minecraft.
My best results were with Open GL on, Graphics on Fast, Smooth lighting on minimum, and Performance set to max fps. With a range of 25-40fps.
But definite improvement in world rendering in general no sign of surface holes that I can see the caves through..
Re: 4.68T Optimization Test
I got on my 2008 13" MacBook for this test.
CPU - 2.4 GHz Core 2 Duo
GPU - Intel GMA X3100
RAM - 2GB RAM
OS - Snow Leopard 10.6.8
Render distance - short
Fancy graphics - off
I started taking note of my FPS on my laptop when the 1.5.1 update was around the corner to see if it gave me any improvements. My GPU is most likely my bottleneck. I did the tests walking around Gringo Town on PK_420's server. Gringo town has lots of diverse BTW blocks, farm animals, redstone devices, many types of crops, testificates, and buildings.
My FPS would fluctuate from like 6-20, but I think the average was a little higher in 4.68T.
BTW 4.65 (Minecraft 1.4.7) - 9 FPS
BTW 4.66 (Minecraft 1.5.1) - 9 FPS
BTW 4.67 (Minecraft 1.5.1) - 9 FPS
BTW 4.68T (Minecraft 1.5.1) - 11 FPS
I did a second test in a flatworld but got 30 FPS in both 4.67 and 4.68T. My FPS fluctuates a lot less in an empty flatworld, but that doesn't produce quite as useful test data.
=======
I did another test on a 2011 13" MacBook Pro:
CPU - 2.3 GHz Intel Core i5
GPU - Intel HD Graphics 3000
RAM - 4GB RAM
OS - Lion 10.7.5
Render distance - short
Fancy graphics - off
I went back to Gringo Town to test.
4.67 - 24 FPS
4.68T - 26 FPS
=======
I've got a few more computers I could test on if you like:
2008 iMac (Core 2 Duo and ATI Radeon HD)
2012 MacBook Pro Retina (Ivy Bridge i5 and Intel HD 4000 graphics)
2004 Dell desktop with Windows XP, Intel Pentium 4, and some crappy on-board graphics
2008 home-assembled desktop with Windows XP/7, Intel Core 2 Duo e8400, and nVidia Geforce 9800 GTX
CPU - 2.4 GHz Core 2 Duo
GPU - Intel GMA X3100
RAM - 2GB RAM
OS - Snow Leopard 10.6.8
Render distance - short
Fancy graphics - off
I started taking note of my FPS on my laptop when the 1.5.1 update was around the corner to see if it gave me any improvements. My GPU is most likely my bottleneck. I did the tests walking around Gringo Town on PK_420's server. Gringo town has lots of diverse BTW blocks, farm animals, redstone devices, many types of crops, testificates, and buildings.
My FPS would fluctuate from like 6-20, but I think the average was a little higher in 4.68T.
BTW 4.65 (Minecraft 1.4.7) - 9 FPS
BTW 4.66 (Minecraft 1.5.1) - 9 FPS
BTW 4.67 (Minecraft 1.5.1) - 9 FPS
BTW 4.68T (Minecraft 1.5.1) - 11 FPS
I did a second test in a flatworld but got 30 FPS in both 4.67 and 4.68T. My FPS fluctuates a lot less in an empty flatworld, but that doesn't produce quite as useful test data.
=======
I did another test on a 2011 13" MacBook Pro:
CPU - 2.3 GHz Intel Core i5
GPU - Intel HD Graphics 3000
RAM - 4GB RAM
OS - Lion 10.7.5
Render distance - short
Fancy graphics - off
I went back to Gringo Town to test.
4.67 - 24 FPS
4.68T - 26 FPS
=======
I've got a few more computers I could test on if you like:
2008 iMac (Core 2 Duo and ATI Radeon HD)
2012 MacBook Pro Retina (Ivy Bridge i5 and Intel HD 4000 graphics)
2004 Dell desktop with Windows XP, Intel Pentium 4, and some crappy on-board graphics
2008 home-assembled desktop with Windows XP/7, Intel Core 2 Duo e8400, and nVidia Geforce 9800 GTX
Re: 4.68T Optimization Test
Pre 4.68T max 40-42 fps, min (outside in rain) 22-24
Post update max 44-46 min 25-29
Stutter on crossing chunk boundaries is still there in overworld. The stutter generating new areas in the nether is much better. No more lava baths :)
*edit* Just been looking on the mojang bug tracker - The overworld problem listed above seems to be a sound issue with vanilla so disregard
Post update max 44-46 min 25-29
Stutter on crossing chunk boundaries is still there in overworld. The stutter generating new areas in the nether is much better. No more lava baths :)
*edit* Just been looking on the mojang bug tracker - The overworld problem listed above seems to be a sound issue with vanilla so disregard
Re: 4.68T Optimization Test
I am experiencing definite improvement. I am on an ancient laptop with graphics set to fast, smooth lighting at maximum, maximum fps in effect, and OpenGL off. Normally I get about 7-9 fps. I am now getting 16-21 fps. Even better, chunk loading in both the Nether and the Overworld appears to be much smoother. Thank you, FlowerChild!
- FlowerChild
- Site Admin
- Posts: 18753
- Joined: Mon Jul 04, 2011 7:24 pm
Re: 4.68T Optimization Test
Aha. Now that's interesting. Everything so far has been rather circumstantial sounding, but that seems quite solid. Sweet.Vexalor wrote:I am experiencing definite improvement. I am on an ancient laptop with graphics set to fast, smooth lighting at maximum, maximum fps in effect, and OpenGL off. Normally I get about 7-9 fps. I am now getting 16-21 fps. Even better, chunk loading in both the Nether and the Overworld appears to be much smoother. Thank you, FlowerChild!
Re: 4.68T Optimization Test
Ok. I have a fairly high end setup but I'm still noticing a modest improvement of about 5-10 fps. Chunks are loading noticeably faster in the overworld. No framerate drop from extended play but that wasn't ever much of a problem for me.
The reason i'm reporting is that for me at least, the nether is smooth as butter. I had a heavy chunk load stutter before trying this test and walked around for about 10 minutes with NO stutter after installing it...
fancy graphics on, performance set to max fps.
The reason i'm reporting is that for me at least, the nether is smooth as butter. I had a heavy chunk load stutter before trying this test and walked around for about 10 minutes with NO stutter after installing it...
fancy graphics on, performance set to max fps.
Time slows down in a high gravitational field. This fact is the pinion of my plan to destroy us all...
Re: 4.68T Optimization Test
There is definitely an increase of about 20-30% on this system, a ThinkPad W510. I'm still getting my main problem, which is a hellish drop to about 7 fps with serious hangups right after sunrise and sunset. But the drop and stuttering that used to happen if I moved too far too fast is all but gone (I can still see a drop in fps if I use the debug tools, but it's mostly unnoticeable ingame).
Lots of planets have a north...!
Re: 4.68T Optimization Test
I have a really low-end ancient system but don't have access to it until Sunday night. Can you please leave this patch available until then?
I'm on vacation.
I'm on vacation.
There's only one V in my name, thanks.
<TaterBoy> I figured out why there's so much lag. We have too much iron.
<TaterBoy> I figured out why there's so much lag. We have too much iron.
-
- Posts: 181
- Joined: Mon Sep 24, 2012 1:49 pm
Re: 4.68T Optimization Test
After playing a few hours watching my FPS without the update I averaged 5 fps after the new update average 9. Yep my system is pretty low end but i did notice an improvement. THanks FC
-
- Posts: 596
- Joined: Fri Sep 30, 2011 2:25 pm
- Location: In the ether...
Re: 4.68T Optimization Test
Yeah, 15-20 before, 17-27 now.
Chris Martin wrote:The morning is for sleeping.
- FlowerChild
- Site Admin
- Posts: 18753
- Joined: Mon Jul 04, 2011 7:24 pm
Re: 4.68T Optimization Test
Perfect man, those were the kind of results I was hoping for, and definitely tell me I'm on the right track.OldMarriedDude wrote:After playing a few hours watching my FPS without the update I averaged 5 fps after the new update average 9. Yep my system is pretty low end but i did notice an improvement. THanks FC
My own system is too high end to reliably perceive a difference. Chunk loading definitely *feels* smoother to me, but that's not the kind of thing that is reflected too well in my frame rate figures.
But, if the changes are turning things from "unplayable" to "barely playable" on low end systems, then I know I'm heading in the right direction. Nobody really gives a fuck if your framerate goes from 60 to 90 due to an optimization. If it goes from 5 to 9 however, then that's hela-cool ;)
- FlowerChild
- Site Admin
- Posts: 18753
- Joined: Mon Jul 04, 2011 7:24 pm
Re: 4.68T Optimization Test
Yup, will do. I'm likely going to be putting out a release soon with these optimizations rolled into it as is though, so you'll be able to get them there. I think I'd like to broaden my sample size here before proceeding further with them in a future release.Benanov wrote:I have a really low-end ancient system but don't have access to it until Sunday night. Can you please leave this patch available until then?
I'm on vacation.
Re: 4.68T Optimization Test
Reasonably high end system here, but made a new world, seed TEST and ran towards the same mountain, then across some plains before spelunking. All settings as per normal play. Spent 10 mins in each setting.
4.6.7 - Pine forest: 45-50fps
4.6.8T-Pine Forest: 65-70fps
4.6.7 - Mountain: 40-45fps
4.6.8T-Mountain: 55-60fps
4.6.7 - Plains: 60-65fps
4.6.8T-Plains: 65-70fps
4.6.7 - Underground: 40-55fps
4.6.8T-Underground: 40-55fps - went a different route, so not that useful a comparison.
Hope this helps!
4.6.7 - Pine forest: 45-50fps
4.6.8T-Pine Forest: 65-70fps
4.6.7 - Mountain: 40-45fps
4.6.8T-Mountain: 55-60fps
4.6.7 - Plains: 60-65fps
4.6.8T-Plains: 65-70fps
4.6.7 - Underground: 40-55fps
4.6.8T-Underground: 40-55fps - went a different route, so not that useful a comparison.
Hope this helps!
7 months, 37 different border checks and counting.
-
- Posts: 181
- Joined: Mon Sep 24, 2012 1:49 pm
Re: 4.68T Optimization Test
FlowerChild wrote: But, if the changes are turning things from "unplayable" to "barely playable" on low end systems, then I know I'm heading in the right direction. Nobody really gives a fuck if your framerate goes from 60 to 90 due to an optimization. If it goes from 5 to 9 however, then that's hela-cool ;)
Ya it is hella Cool. Wish I had cash to make my system run a bit better but it works so im not gonna bitch. I havnt gone to the nether yet to check on chunk load stutter but i will at some point today and let you know how it goes.
Re: 4.68T Optimization Test
all settings on minimum, short distance, 1024 mb ram given to minecraft, on my main world I had this results:
BTW 4.67 average of 14-15 with spikes of 6-7 and 27-30
BTW 4.68T fluctuating between 17 and 30.
So not a big increase but still I enjoy the improvements.
Thank you FC for your interest in lower end systems, I can't really thank you enough.
BTW 4.67 average of 14-15 with spikes of 6-7 and 27-30
BTW 4.68T fluctuating between 17 and 30.
So not a big increase but still I enjoy the improvements.
Thank you FC for your interest in lower end systems, I can't really thank you enough.
- FlowerChild
- Site Admin
- Posts: 18753
- Joined: Mon Jul 04, 2011 7:24 pm
Re: 4.68T Optimization Test
Well, results like this will definitely motivate me to push it further man. Honestly, the rendering code has so much potential for optimization that I could probably keep going at this for a very long time indeed. I don't want to devote *too* much time to it, but if I could get you to the playable range of say 18-20 fps, then I'd be a very happy camper.OldMarriedDude wrote: Ya it is hella Cool. Wish I had cash to make my system run a bit better but it works so im not gonna bitch. I havnt gone to the nether yet to check on chunk load stutter but i will at some point today and let you know how it goes.
Hmmm...I might take a look at optimizing pathfinding slightly before the next release too as that seemed to have significant results with the last release. I think I'll do that with my morning coffee :)
-
- Posts: 262
- Joined: Sun Mar 04, 2012 7:58 pm
Re: 4.68T Optimization Test
The system I'm on is somewhat high end, but I've got virtual machine software installed for school. Would testing in those be of any benefit to you, or just be theoretical wankery?