Not a member yet? Why not Sign up today
Create an account  

  • 1 Vote(s) - 5 Average
  • 1
  • 2
  • 3
  • 4
  • 5
 
Benchmarks Default Kernel vs pf-Kernel vs Liquorix vs XanMod

#1
Exclamation 
i want to share my bencmarks . someone tell me why not try in opensuse ? soo i try and xanmod its a clear winner

some of the test also against pf kernel (bfs+bfq enabled with 1000 freq)

https://db.tt/ya1VsBQl

next time all my bencmarks will be full and not half of the test
Reply

#2
Habernir, thank you very much for the benchmarking task! Quite good results imho!  8)
"(...) the grandest occasion the past or present has seen, or the future can hope to see." -- Cervantes.
Reply

#3
i can keep update this post against each new kernel 4.X against pf-kernel and liquorix kernel  8)

and if someone want me to add benchmark to it just write me here
Reply

#4
Habernir, it's a great idea by your side! Very appreciated! Upcoming results will be welcome!  8) 8)
"(...) the grandest occasion the past or present has seen, or the future can hope to see." -- Cervantes.
Reply

#5
I must say that I am very surprised about the bad results obtained with the PF-patch. ???
https://www.dropbox.com/s/vo2kpznyw3nv8h...6.png?dl=0
Edit: I know that we are good, but I didn't know that we were so good! ;D


Attached Files Thumbnail(s)
   
"(...) the grandest occasion the past or present has seen, or the future can hope to see." -- Cervantes.
Reply

#6
i'm waiting for the next release 4.6 and do much more benchmarks  in opensuse and ubuntu  8)
Reply

#7
soon the reason it because i'm not in my house until end of this month.

and i believe until than the liquorix version and pf-kernel will be also in 4.6

Reply

#8
Thank you Habernir! Here waiting for the new scores from the benchmarking tasks! 8)
"(...) the grandest occasion the past or present has seen, or the future can hope to see." -- Cervantes.
Reply

#9
soon very soon i also waiting for the liquorix and pf-kernel version 4.6 for the bencmarks

they are sooo slow Smile
Reply

#10
(29-05-2016, 12:16 PM)habernir link Wrote: (..) they are sooo slow Smile

Yes, xanmod doesn't depend on external patches, so it's released as soon as mainline is done. Smile And also them are slower than xanmod according to your benchmarking results. Probably mostly of the succcess of xanmod kernel is due to the powerful development efforts since its first release. Smile

Edit:  I must say that it has been a long road of testing and brainstorming task from near all our members, and the courtesy of Xan to take every idea as a potential good point for us. The best feature of xanmod: it's an 'open-minded' kernel, every good idea will be merged.
"(...) the grandest occasion the past or present has seen, or the future can hope to see." -- Cervantes.
Reply

#11
Hi, my 2 cents...

Quote:And also them are slower than xanmod according to your benchmarking results.
Quote:I must say that I am very surprised about the bad results obtained with the PF-patch. ???

Let's start with a question: what are you trying to measure here - interactivity or throughput?

Benchmarks generally are throughput oriented, except ones that are specifically not, they take couple of tasks ftw. Interactivity sacrifices throughput by definition.

There are no much surprises when BFS looses throughput oriented tasks, coz it's just not designed that way.
The same with BFQ, did you try to load your HDD with loads of background reads/writes and then open, say, libreoffice and measure time it takes from cold start to finish with CFQ and BFQ. If you do so, you'll get surprising results.
Reply

#12
first sorry about the delay of the benchmarks . i will do it soon.

well why not both? and this benchmarks its for the people who want to use "custom" kernel and help them choose the right one Smile.
and also to see how xan kernel is against other custom kernel

and which phornix benchmarks do you recommended?
Reply

#13
Hi,

I was not really suggesting a set of benchmarks in previous post, I'm trying to explain that not including, for example BFQ ( just because it didn't do well in some throughput benchmark), in the kernel might not be a wise choice for desktops where interactivity is the key. Really, on regular desktop, noone cares that DB runs great there, ppl need to open libreoffice right now... not later... and no matter the background load... They want to game w/o lags at the same time as downloading torrents or compile a kernel fast and still be able to watch YouTube 1080p w/o stuttering.
Of course, this is true, when there are no stability problems or bugs in particular patches.

On the benchmark note: gaming, gaming on wine, gaming on wine nine, video playback/encoding, browser benchmarks... In short, that's what ppl do in their pcs most of the time, you name it...
My point is, that most ppl won't really care about Postgre working better in high load on their laptops by 1%, because that's seldom when such performance gain is crucial on desktop, servers are for that... The same with C ray, OpenSSL, etc., yes ppl do that, but very seldom..
But of course, there are exceptions Smile

Please ignore this if this is not about "Optimized to take full advantage of high-performance Workstations, PC Games, Media Centers and others. ". "Others" may include servers as well, but I'm not really sure if this is the case. Maybe Tropic can elaborate Smile
And, yes, sorry for noise Smile

regards
Mjasnik
Reply

#14
well thanks of your reply .
and i think we are  exceptions , no? Smile
most of the ppl even dont come to this site or dont know that it exists (the same for liq and pf-kernel) soo all of us exceptions Smile

and my target its for graphics (rendering,encoding.....) and optimizing but i add more test for ppl who want to see the difference .
Reply

#15
habernir, thanks for good reply!
We most likely are exceptions, I'm at least for sure, but not to extremes, I just want my junk hw to perform to max on desktop/laptop workloads Big Grin
Reply

#16
Thank you both Mjasnik and Habernir for your feedback and thoughts. Big Grin

I could write some words to explain my extense point of view, but I've felt a little alone here sometimes. :-[  I'm very happy when members like you both collaborate and comment, so thanks again! Well, let's go for work... Sincerely, I can't understand how an amazing compiled kernel with more than 4500 downloaded times by now (this year) is making so petite noise in other sites. I have been reading for years about BFQ, BFS, 1000Hz and so forth and the single truth is this: a 500hz plus tuned CFQ (yep, tuned CFQ, come on, so easy?) and tuned CFS (yep again, tuned CFS, come on, so easy?) is having such amazing benchmark results against BFQ+BFS, two external patches with thousands of entries on the net.  :o

Do you realize that Xanmod doesn't depend on external patches anymore? Xanmod doesn't wait for nobody to be built and it's released faster than any other kernel! Do you realize that Xanmod is using a "tuned CFQ" that now is currently faster than the overbloated BFQ? Please, guys, tuning the CFQ is just what we really needed at the "end" of a long road of development! And what about the old CFS, now tuned to be efficient like a big turbo machine?  8) Mainline should be released with both Xan's tuned CFQ and tuned CFS by default IMHO -- with its own sysctl.conf file! Things are better when them work! Wink

Edit: "You guys have somehow achieved that using CFQ. Kudos for that!" -- Sarfaraz.
http://xanmod.org/forum/index.php/topic,3.msg1846.html#msg1846

Edit2: also Xan's tuned Ondemand governor is really appreciated here, of course!
Edit3: probably we are making together History here, surrounded by silence...
"(...) the grandest occasion the past or present has seen, or the future can hope to see." -- Cervantes.
Reply

#17
Certainly, I am very interested on how the new real-time lowlatency xanmod kernel (currently under development and already available for testing at Xan's Mega site) will work on Habernir's Opensuse environment while benchmarking its performance. Tongue
Let's bump this topic! 8)
"(...) the grandest occasion the past or present has seen, or the future can hope to see." -- Cervantes.
Reply

#18
well this is the benchmarks for opensuse kernerl against xanmod and pf-kernel
pf-kernel was very disappointed such a poor result with the pf-kernel ( may be it have some regressions in bfs+bfq in kernel 4.6?) . i will check the  next stable pf-kernel and publish here.
with liq kernel there was a few problems with the kernel i will check and ask there.

in the bottom line xanmod its  a clear winner . i wonder how kernel 4.4.13-xanmod will do against 4.6.2  Smile
i just want to ask what is the best way to test the stability of the kernel?



Attached Files Thumbnail(s)
       
Reply

#19
Hello again Habernir! Thank you very much for the benchmarking tasks and also for publish them in that nice way! You are completely right, xanmod is the clear winner! Smile

My way to test the stability is push the kernel to its limits, running some heavy software or some programs at the same time, any virtual machine or some powerful games and just look for freezings (the most common sight of unstability), hangs and kernel panics (uncommonly but mostly the clearest sight). As I said somewhere in this forum, my notebook is bad for notice the best but incredible to discover the worst (i.e. if the kernel is not good my notebook has keystroke delays instantly). Smile

4.4.13-xanmod18 against 4.6.2-xanmod4? In theory the LTS branch should be a little slower but more stable, but that's the theory considering that both come from mainline -- and mostly important, with no external unstable patches like BFQ and BFS. Wink
"(...) the grandest occasion the past or present has seen, or the future can hope to see." -- Cervantes.
Reply

#20
Hi!

I would like to share my benchmarks of plain simple Unigine Valley Valley run on my system.
HW: Phenom II 975 BE, Radeon HD7850, 8GB RAM, Samsung Spinpoint F1 (750GB)
Test method (for each run): boot up, wait for system to settle down, run the benchmark, assess the results.
Unigine Valley settings: OpenGL, 1680x1050, Medium

Please see tested kernels and results in attachments.

Conclusions.
For gaming, when there is no concurrency involved, Ubuntu standard kernel with 250Hz performs the same as tuned kernels.
For gaming, when there is some scheduling involved, BFS performs the best by a quite large margin.
For file activity , when there is scheduling and disk activity involved BFS + BFQ performs the best by a really large margin.

P.S. I didn't have 4.4 my compilation of 4.4 kernel available at hand, but I bet difference could be like couple of percent... Even GCC versions might gain or loose couple of %... So at least in my book this has to be +/ fair from this perspective.
P.P.S. At the end of results there is a "Gain (%)" column which actually shows added value in percent, not mathematical "this is x times better than y" Smile


Have a nice day,
Mjasnik


Attached Files Thumbnail(s)
       
Reply



[-]
Quick Reply
Message
Type your reply to this message here.

Image Verification
Please enter the text contained within the image into the text box below it. This process is used to prevent automated spam bots.
Image Verification
(case insensitive)

Forum Jump:


Users browsing this thread:
1 Guest(s)