- This topic has 20 replies, 6 voices, and was last updated Oct 17-10:13 pm by Brian Masinick.
-
Posts
-
Wallon
I did 3 tests with Firefox 105 (Webrender enabled) with 3 kernels on the following blackboard site;
https://testdrive-archive.azurewebsites.net/Performance/Chalkboard/The three kernels are;
1) 5.10.142-antix.2-amd-smp
2) 5.19.0-15.2 Preempt_Dynamic liquorix 5.19.-22.1~Bullseye
3) 5.19.0-0.deb11.2-rt-amd64 smp PREEMPT-rt debian 5.19.11-1-bpo11+1Results;
1) 6.56 seconds
2) 6,99 s
3) 7,93 sMaster anticapitalista made us a very good kernel. No need to look elsewhere, we recognize all his expertise!
olsztyn
Master anticapitalista made us a very good kernel. No need to look elsewhere, we recognize all his expertise!
Although it might be expected that later kernels tend to be slower, a surprise to me was that within 5.10.XXX kernel family kernel 5.10.142 in antiX lead to the smallest memory footprint. A difference about 10Mb. As I remember 5.18 and 5.19 entailed much larger memory footprint.
I do not know if it is Debian that introduced such optimization or it is antiX specific optimization.
One way or the other, I completely concur that the 5.10.142 choice is excellent.Live antiX Boot Options (Previously posted by Xecure):
https://antixlinuxfan.miraheze.org/wiki/Table_of_antiX_Boot_Parameterscalciumsodium
I don’t see that much of a difference in boot up and see no difference in ram usage of the different antiX 5.10 kernels.
I compared the antiX kernels 5.10.57, 5.10.104, and 5.10.142.
I started with the 5.10.142 as the only installed kernel and did an autoboot into rox-icewm. The icewm is the latest 3.0.1. Then I used the start-t icewm function in terminal to determine the boot up time to get into icewm. The start-t icewm gives two values, so I averaged them and report them here. I also determined the baseline ram usage after booting into the icewm by reading the value from conky. Then I installed the 5.10.104 kernel and purged the headers and image for the 5.10.142. Then I shut down and booted and performed the similar tests. Then I installed the 5.10.57 kernel and purged the 5.10.104 headers and image. Then shut down and booted and performed the similar tests.
This is what I found:
kernel boot time (s) ram (Mb) ________________________________________________________________________ 5.10.142-antix.2- 25.70 192 amd64-smp_5.10.142- antix.2-amd64-smp-1_amd64 5.10.104-antix.1- 25.80 193 amd64-smp_5.10.104- antix.1-amd64-smp-1_amd64 5.10.57-antix.1- 25.98 193 amd64-smp_5.10.57- antix.1-amd64-smp-1_amd64As I see it, 0.1 or 0.2 of a sec is not that much of a difference. The 5.10.142 kernel is not special compared to the other antiX kernels. Based on this data, it is very consistent with the other antiX kernels.
This is the antiX system that I used to perform these tests:
$ inxi -b System: Host: antix1 Kernel: 5.10.57-antix.1-amd64-smp x86_64 bits: 64 Desktop: IceWM 3.0.1 Distro: antiX-bullseye-b1_x64-full Grup Yorum 11 June 2021 Machine: Type: Laptop System: CLEVO product: S3100 v: N/A serial: <root required> Mobo: CLEVO model: S3100 serial: <root required> BIOS: Phoenix v: CALPELLACRB.86C.0000.X.0000000000 date: 09/13/2010 CPU: Dual Core: Intel Core i3 U 330 type: MT MCP speed: 1074 MHz min/max: 666/1199 MHz Graphics: Device-1: Intel Core Processor Integrated Graphics driver: i915 v: kernel Display: x11 server: X.Org 1.20.11 driver: intel resolution: 1366x768~60Hz OpenGL: renderer: Mesa DRI Intel HD Graphics (ILK) v: 2.1 Mesa 20.3.4 Network: Device-1: Realtek RTL8188CE 802.11b/g/n WiFi Adapter driver: rtl8192ce Device-2: JMicron JMC250 PCI Express Gigabit Ethernet driver: jme Drives: Local Storage: total: 298.09 GiB used: 6.47 GiB (2.2%) Info: Processes: 155 Uptime: 2m Memory: 7.57 GiB used: 206.7 MiB (2.7%) Shell: bash inxi: 3.0.36Wallon
Dear de calcium de sodium,
I took the 5.10.142 kernel because many people (like me) have problems with other kernels on their Intel laptop(s) or Intel dektop. If you followed the forum, anticapitalista asked to test with the new kernel 5.10.142 which has a better compatibility with PCs.
It is said on the internet that the Real Time or Liquorix kernels are the best for gaming or music and even office use. I had noticed on Youtube that gamers were showing that this was not often the case.
That’s why I chose two other special kernels. A Debian Real Time kernel and a Liquorix kernel of the last generation.
The conclusion is striking. AntiX kernels are something to be proud of.
Best regards,
WallonAs I see it, 0.1 or 0.2 of a sec is not that much of a difference. The 5.10.142 kernel is not special compared to the other antiX kernels.
I remember you previously reported this and I have no idea where the difference is.
My comparison was based on my standard composition of antiX 21 Live Runit, *elogind*-free. I did not see a change in memory footprint moving from kernel 5.10.57 to 5.10.104. But then moving from 5.10.142 was a noticeable decrease of memory footprint about 10Mb. This was measured on Thinkpad 410, i5, 4Gb memory. The same on all my three copies of Thinkpad 410. So my observation is based on one laptop model so far. I should do the same comparison on other my laptops, older and newer to make a more comprehensive determination.
So I am not claiming this decrease in memory footprint is for all across the board. This is just what I see on my typical machine using my standard composition of antiX, the only difference being kernel release.
Another very noticeable change as I see (also on the same model machine) is better memory management. Almost no memory leaks. The best I have ever seen…Live antiX Boot Options (Previously posted by Xecure):
https://antixlinuxfan.miraheze.org/wiki/Table_of_antiX_Boot_Parameters
blur13
It would be interesting to see a test of the legacy 4.9 kernel and how it compares to 5.10.
It would be interesting to see a test of the legacy 4.9 kernel and how it compares to 5.10.
In terms of initial memory footprint with the same boot parameters of course:
Booting kernel 4.9 (on the same test machine as before) initial memory footprint appears to be about 5Mb more than kernel 5.10.142. Booted a few times to be sure of consistent behavior.
Again, my disclaimer is that this is what I see on my Thinkpad T410 machine running Live antiX 21 Runit, *elogind*-free. I have not done the same test on other my machines yet, so I am not making a generalized statement but just my observation in my case.Live antiX Boot Options (Previously posted by Xecure):
https://antixlinuxfan.miraheze.org/wiki/Table_of_antiX_Boot_ParametersIt would be interesting to see a test of the legacy 4.9 kernel and how it compares to 5.10.
I have just done this experiment. Surprise! The 4.9 kernel actually boots up slower than the 5.10 kernel.
@Wallon, can you please provide links of where to download those 5.19 liquorix and debian kernels that you had used? I want to try to install those kernels on this test system and report the results alongside the antix 4.9 and 5.10 kernels. It would be great to know the difference.
kernel boot time (s) ram (Mb) ________________________________________________________________________ 5.10.142-antix.2- 25.70 192 amd64-smp_5.10.142- antix.2-amd64-smp-1_amd64 5.10.104-antix.1- 25.80 193 amd64-smp_5.10.104- antix.1-amd64-smp-1_amd64 5.10.57-antix.1- 25.98 193 amd64-smp_5.10.57- antix.1-amd64-smp-1_amd64 4.9.0-326-antix.1- 26.14 179 amd64-smp_4.9.0-326- antix.1-amd64-smp-1_amd64Again, this is the test system:
jakersfan@antix1:~ $ start-t icewm 26.11 26.17 jakersfan@antix1:~ $ inxi -b System: Host: antix1 Kernel: 4.9.0-326-antix.1-amd64-smp x86_64 bits: 64 Desktop: IceWM 3.0.1 Distro: antiX-bullseye-b1_x64-full Grup Yorum 11 June 2021 Machine: Type: Laptop System: CLEVO product: S3100 v: N/A serial: <root required> Mobo: CLEVO model: S3100 serial: <root required> BIOS: Phoenix v: CALPELLACRB.86C.0000.X.0000000000 date: 09/13/2010 CPU: Dual Core: Intel Core i3 U 330 type: MT MCP speed: 666 MHz min/max: 666/1199 MHz Graphics: Device-1: Intel Core Processor Integrated Graphics driver: i915 v: kernel Display: x11 server: X.Org 1.20.11 driver: intel resolution: 1366x768~60Hz OpenGL: renderer: Mesa DRI Intel HD Graphics (ILK) v: 2.1 Mesa 20.3.4 Network: Device-1: Realtek RTL8188CE 802.11b/g/n WiFi Adapter driver: rtl8192ce Device-2: JMicron JMC250 PCI Express Gigabit Ethernet driver: jme Drives: Local Storage: total: 298.09 GiB used: 6.40 GiB (2.1%) Info: Processes: 171 Uptime: 2m Memory: 7.60 GiB used: 203.9 MiB (2.6%) Shell: bash inxi: 3.0.3Wallon
Dear Calciumsodium,
For the Debian kernel, it is from Synaptic. You have to search on “linux-image” and on “linux-headers”.
You have to select the 2 packages. It’s surprising, but Synaptic doesn’t install both at the same time.
I chose the “rt” packages. This means Real Time (no latency).For liquorix, I had made a request on the forum. I will do a search to give you the thread. It was our friend Brian who kindly answered me.
Best regards,
WallonWallon
Dear calciumsodium
Here it is the url regarding Liquorix kernel.
https://www.antixforum.com/forums/topic/how-to-install-kernel-liquorix/
Please note that after installing this kernel, antiX will not be able to update i386 packages.
I don’t know why antiX wants to update i386 packages on an antiX 64bit installation. Normally, as I understand it, i386 packages are for 32bit.
Best regards,
WallonIt would be interesting to see a test of the legacy 4.9 kernel and how it compares to 5.10.
Out of curiosity I got around to compare initial memory footprint on several laptops using the same antiX 21 Live Runit *elogind*-free, using Lxtask as measurement tool. Prior to this I checked also Htop and results were pretty much correlated. Also, unlike my previous post I actually wrote down results.
Measurement was each time on fresh boot, after network connected and after about 15 seconds to calm down.
It appears that lower memory footprint could be in favor one kernel vs. another depending on machine:Machine kernel 4.9 (MB) Kernel 5.10.142 (MB)
Thinkpad X61 (Core2Duo), 4Gb: 126 129
Thinkpad T410 (i5), 4Gb: 174 164
Thinkpad T520 (i7), 8Gb: 177 175
Thinkpad Z61P (Core2Duo), 3Gb: 143 146
Thinkpad X220 (i7), 6Gb: 184 171Looks like on three out of five machines I see lower memory footprint using kernel 5.10.142, vs. 4.9
I did not measure speed of booting, just memory footprint.P.S.:
Sorry – tabulation got messed up. Actually I was not successful in tabulating on this site…
The first memory figure (MB) was for kernel 4.9, the second figure was for kernel 5.10.142.
The exact memory footprint savings on Thinkpad T410, which I roughly reported previously is larger than reported before. It is actually 10MB lower using kernel 5.10.142.- This reply was modified 6 months, 3 weeks ago by olsztyn.
- This reply was modified 6 months, 3 weeks ago by olsztyn.
- This reply was modified 6 months, 3 weeks ago by olsztyn.
Live antiX Boot Options (Previously posted by Xecure):
https://antixlinuxfan.miraheze.org/wiki/Table_of_antiX_Boot_ParametersI have now collected all the data and am reporting the data.
This has been a good exercise in installing various linux kernels: antiX, liquorix, debian, jxself libre.
The libre kernels use the least ram compared to their kernel version peers.
The 5.19 kernels uses the most ram and takes more time to boot then their lower version counterparts.
But the difference in boot time between the fastest kernel and the slowest kernel is only about 5 seconds using the antiX21 system!
That is saying a lot.
kernel boot time (s) ram (Mb) ________________________________________________________________________ 5.10.142-antix.2- 25.70 192 amd64-smp_5.10.142- antix.2-amd64-smp-1_amd64 5.10.104-antix.1- 25.80 193 amd64-smp_5.10.104- antix.1-amd64-smp-1_amd64 5.10.57-antix.1- 25.98 193 amd64-smp_5.10.57- antix.1-amd64-smp-1_amd64 4.9.0-326-antix.1- 26.14 179 amd64-smp_4.9.0-326- antix.1-amd64-smp-1_amd64 4.19.0-256-antix.1- 26.46 196 amd64-smp_4.19.0-256- antix.1-amd64-smp-1_amd64 5.19.0-16.2-liquorix-amd64 28.29 213 5.19.0-0.deb11.2-rt-amd64 30.26 208 (Debian) 4.9.330-gnu1 (libre) 24.86 165 5.10.147-gnu1 (libre) 26.98 178 5.19.15-gnu1 (libre) 28.30 187Machine kernel 4.9 (MB) Kernel 5.10.142 (MB) Thinkpad X61 (Core2Duo), 4Gb: 126 129 Thinkpad T410 (i5), 4Gb: 174 164 Thinkpad T520 (i7), 8Gb: 177 175 Thinkpad Z61P (Core2Duo), 3Gb: 143 146 Thinkpad X220 (i7), 6Gb: 184 171Looks like on three out of five machines I see lower memory footprint using kernel 5.10.142, vs. 4.9
I did not measure speed of booting, just memory footprint.P.S.:
Sorry – tabulation got messed up. Actually I was not successful in tabulating on this site…
The first memory figure (MB) was for kernel 4.9, the second figure was for kernel 5.10.142.
The exact memory footprint savings on Thinkpad T410, which I roughly reported previously is larger than reported before. It is actually 10MB lower using kernel 5.10.142.Hi @olsztyn,
To use tabulation in this forum, simply click the “code” format (at the top) before you create your table and click “code” after your table. It does not have to be a table, it could be anything.
@Wallon, thanks for creating this thread. Because of this exercise in trying out the various kernels, I have actually expanded the use of the libre kernels on more of my computers. Also, I have begun to use the liquorix kernels on my newer computers. Even though it booted up a few seconds slower on my test system than the antiX 5.10 kernels, on the newer computers, that difference is even smaller. Installing the liquorix kernels is SUPER SUPER easy. One command in terminal! Don’t have to worry about setting up repositories.
@Brian, thanks for showcasing the liquorix kernels in your what are you “here” with thread. They work great on my newer computers.
Because of this exercise in trying out the various kernels, I have actually expanded the use of the libre kernels on more of my computers. Also, I have begun to use the liquorix kernels on my newer computers. Even though it booted up a few seconds slower on my test system than the antiX 5.10 kernels, on the newer computers, that difference is even smaller. Installing the liquorix kernels is SUPER SUPER easy. One command in terminal! Don’t have to worry about setting up repositories.
Hi Calciumsodium:
First, thank you for tabulating my results…
Also, could you please expand more on your reasons to use liquorix and libre kernels vs. Debian/antiX kernels, in spite of showing worse results (as it appears to me) in your benchmarks? E.g. is this a matter of being more cutting edge, being more real-time responsive, etc…
I am curious if it makes a difference in average user experience, such as not playing time sensitive games or real-time critical applications…Live antiX Boot Options (Previously posted by Xecure):
https://antixlinuxfan.miraheze.org/wiki/Table_of_antiX_Boot_Parameters
Brian Masinick
Regarding the kernels, though the 5.10 kernels are indeed efficient, I have had enough problems with them overall to make the decision to use other kernels.
Why? #1 because they’re available
#2 for me they match the typical systems and workloads I use.
#3 on my newer hardware I am forced to use newer kernels than the ones provided in order to get networking working properly.
#4 on my oldest hardware the 4 series kernel works best, usually the 4.9 version.The 4.4 kernels also work well on systems prior to 2010.
--
Brian Masinick
- You must be logged in to reply to this topic.
