Linux 3.4 Kernel Benchmarks

[phoronix]

Phoronix: Linux 3.4 Kernel Benchmarks

Here's a look at the performance of the Linux 3.4 kernel, which was recently released, compared to all major kernel releases going back to Linux 2.6.39. Multiple kernel sub-systems are being compared in this round of Intel Linux benchmarking.

http://www.phoronix.com/vr.php?view=17381

[Brazilian Joe]

What about energy-consumption-when-idle?

In other series you addressed the energy consumption issue under linux. It would be nice to see how well it scales across different kernel revisions, and if possible throw a Windows 7 and Mac OS X in the mix for us to get the full picture, i.e. If I throw a linux distro on my [note/net/ultra/mac]book, will I get equal or better battery life than other mainstream OSes?

[Michael]

What about energy-consumption-when-idle?

In other series you addressed the energy consumption issue under linux. It would be nice to see how well it scales across different kernel revisions, and if possible throw a Windows 7 and Mac OS X in the mix for us to get the full picture, i.e. If I throw a linux distro on my [note/net/ultra/mac]book, will I get equal or better battery life than other mainstream OSes?

I save it for another article for necessary page traffic...

[xir_]

I save it for another article for necessary page traffic...

I remember reading that Ubuntu enabled some extra power saving optimisation in 12.04. I'd be interested to see it compared with Fedora and CentOS. (maybe a bit of stock Debian too)

Keep up the good work Michael, you do a good job.

[frief]

3.4 git on the test selection!

"2.6.39" "3.0" "3.1" "3.2" "3.3" "3.4 Git"
100.0 102.8 102.1 101.1 103.1 104.2

By about 4.2% over 2.6.39 and about 1.1% over 3.3.
(Note, also includes some graphics tests)

Congratulations!)




R snippet:

install.packages("matrixStats")
library("matrixStats")

test <- matrix( ncol=6, byrow=TRUE,
    dimnames = list(NULL, c( "2.6.39", "3.0", "3.1", "3.2", "3.3", "3.4 Git" )),
    c(   # numbers by hand, errors possible
	  81.73,   79.20,   86.57,   80.80,   81.83,   81.10,
	  94.03,   94.07,   94.50,   94.00,   95.33,   97.87,
	  77.97,   79.17,   80.47,   77.70,   78.07,   80.85,
	151.23,  146.37,  148.48,  142.40,  147.92,  146.38,
	207.97,  206.57,  214.83,  204.47,  240.86,  242.81,
	101.13,  103.45,  105.37,  104.75,  104.85,  107.67,
	  92.35,   95.12,   95.47,   90.76,   94.57,   95.71,
	119.46,  240.53,  120.43,  116.55,  128.11,  126.90,
	  23.84,   23.79,   26.35,   26.02,   26.40,   26.47,
	  96.90,   97.27,   98.70,   94.70,   95.70,   95.50,
	  32.40,   32.40,   33.30,   33.10,   33.73,   33.73,
	  83.38,   83.26,   95.39,   95.96,   96.69,   96.78,
	5486.26, 5492.40, 5523.62, 5501.83, 5567.22, 5561.23,
	  86.43,   86.07,   86.36,   86.35,   86.25,   86.46,
	2437.40, 2440.64, 2476.31, 2514.63, 2539.99, 2537.84,
	  20.00,   19.72,   20.21,   20.41,   20.73,   20.93,
	3287.02, 3228.48, 3243.00, 3239.22, 3269.52, 3266.19,
	  15.10,   15.31,   14.53,   14.31,   13.97,   13.55,
	  60.25,   60.44,   62.36,   62.39,   63.27,   63.20,
	  9.81,    9.60,    9.74,    9.98,   10.02,    9.93,
	  56.43,   56.42,   57.57,   58.38,   58.54,   58.63,
	8196.00, 8327.00, 8190.00, 8820.00, 8912.00, 8908.00,
	185.36,  185.17,  183.95,  184.18,  182.34,  181.97,
	106.91,  106.90,  106.78,  106.79,  106.44,  131.46
    )
)

test.gavg <- colProds(test) ^ (1/nrow(test))  # geometric mean ...
test.gavg

test.gavg * 100 / test.gavg[1] # reference to first

[jumico]

Why does the compile bench becnchmark show the 3.0 kernel as 2x faster than the others with no mention of it?

[halo9en]

I remember reading that Ubuntu enabled some extra power saving optimisation in 12.04.

This? http://www.phoronix.com/scan.php?pag...tem&px=MTA0OTU
I think at least some of these optimizations (e.g. RC6) are enabled by default in 3.4.