Performance Impact of kernel update (4.9.65-3+deb9u2)

Questions about hardware, drivers and peripherals.
In the Original Post please also include the output of inxi -Fzx
User avatar
ScottQuier
Posts: 1769
Joined: 18 Jul 2013 15:55
Location: Newport News, VA

Performance Impact of kernel update (4.9.65-3+deb9u2)

Postby ScottQuier » 05 Jan 2018 15:32

By now, most have read about the issues with the "more modern" Intel processors and the predicted impact the kernel patch will have on performance.

Reading this thread gave me the clue that the patch has already been made and distributed.

Reading this post gave me the idea to try to determine the impact of the patch by comparing the before and after performance on just one of my computers.

Code: Select all

$ inxi -Fzx
System:    Host: sagerk Kernel: 4.9.0-5-amd64 x86_64 (64 bit gcc: 6.3.0) Desktop: KDE Plasma 5.8.6 (Qt 5.7.1)
           Distro: SolydXK
Machine:   Device: laptop System: Notebook product: W65_67SZ
           Mobo: Notebook model: W65_67SZ BIOS: American Megatrends v: 1.03.05RLS1 date: 03/13/2014
Battery    BAT0: charge: 51.8 Wh 100.0% condition: 51.8/56.0 Wh (92%) model: Notebook BAT status: Full
CPU:       Quad core Intel Core i7-4710MQ (-HT-MCP-) cache: 6144 KB
           flags: (lm nx sse sse2 sse3 sse4_1 sse4_2 ssse3 vmx) bmips: 19954
           clock speeds: max: 3500 MHz 1: 2444 MHz 2: 2500 MHz 3: 1505 MHz 4: 1621 MHz 5: 2628 MHz 6: 1603 MHz
           7: 2550 MHz 8: 2316 MHz
Graphics:  Card: Intel 4th Gen Core Processor Integrated Graphics Controller bus-ID: 00:02.0
           Display Server: X.Org 1.19.2 drivers: modesetting (unloaded: fbdev,vesa)
           Resolution: 1920x1080@60.01hz
           GLX Renderer: Mesa DRI Intel Haswell Mobile GLX Version: 3.0 Mesa 13.0.6 Direct Rendering: Yes
Audio:     Card-1 Intel 8 Series/C220 Series High Definition Audio Controller
           driver: snd_hda_intel bus-ID: 00:1b.0
           Card-2 Intel Xeon E3-1200 v3/4th Gen Core Processor HD Audio Controller
           driver: snd_hda_intel bus-ID: 00:03.0
           Sound: Advanced Linux Sound Architecture v: k4.9.0-5-amd64
Network:   Card-1: Intel Wireless 7260 driver: iwlwifi bus-ID: 04:00.0
           IF: wlp4s0 state: up mac: <filter>
           Card-2: Realtek RTL8111/8168/8411 PCI Express Gigabit Ethernet Controller
           driver: r8169 v: 2.3LK-NAPI port: e000 bus-ID: 05:00.2
           IF: enp5s0f2 state: down mac: <filter>
Drives:    HDD Total Size: 250.1GB (30.8% used)
           ID-1: /dev/sda model: Samsung_SSD_850 size: 250.1GB
Partition: ID-1: / size: 20G used: 8.8G (48%) fs: ext4 dev: /dev/sda1
Sensors:   System Temperatures: cpu: 48.0C mobo: N/A
           Fan Speeds (in rpm): cpu: N/A
Info:      Processes: 234 Uptime: 23 min Memory: 1147.5/15978.6MB Init: systemd runlevel: 5 Gcc sys: 6.3.0
           Client: Shell (bash 4.4.121) inxi: 2.3.5

Now, I don't know if sysbench is the best tool for this or if the results I'm posting below are a false positive; I would be very interested in hearing what others might have to say. Additionally, as I can still boot into the "pre-patched" kernel, I can run other tests that might be suggested.

Code: Select all

# Before the patch/install of linux-headers-4.9.0-5-amd64 linux-headers-4.9.0-5-common linux-image-4.9.0-5-amd64

$ sysbench --batch --batch-delay=5 --num-threads=2 --max-requests=50000 --test=threads run
 sysbench 0.4.12:  multi-threaded system evaluation benchmark

 Running the test with following options:
 Number of threads: 2
 
 Doing thread subsystem performance test
 Thread yields per test: 1000 Locks used: 8
 Threads started!
 Done.

 
 Test execution summary:
     total time:                          3.8479s
     total number of events:              50000
     total time taken by event execution: 7.6911
     per-request statistics:
          min:                                  0.11ms
          avg:                                  0.15ms
          max:                                  0.32ms
          approx.  95 percentile:               0.20ms
 
 Threads fairness:
     events (avg/stddev):           25000.0000/1.00
     execution time (avg/stddev):   3.8455/0.00
 

# After the above patch
$ sysbench --batch --batch-delay=5 --num-threads=2 --max-requests=50000 --test=threads run
 sysbench 0.4.12:  multi-threaded system evaluation benchmark

 Running the test with following options:
 Number of threads: 2
 
 Doing thread subsystem performance test
 Thread yields per test: 1000 Locks used: 8
 Threads started!
 min: 0.0002  avg: 0.0003  max: 0.0012  percentile: 0.0003
 Done.

 
 Test execution summary:
     total time:                          7.0942s
     total number of events:              50000
     total time taken by event execution: 14.1833
     per-request statistics:
          min:                                  0.25ms
          avg:                                  0.28ms
          max:                                  1.17ms
          approx.  95 percentile:               0.32ms
 
 Threads fairness:                                                                         
     events (avg/stddev):           25000.0000/0.00                                       
     execution time (avg/stddev):   7.0916/0.00     

IF this is a valid test and IF I'm reading this correctly, it would appear that the patch has had a negative 50% impact on performance. But, opening files, doing forum posts, writing stuff in word processors, etc - I don't (yet) experience any issues.

It will be really interesting to see what impact I see when I start processing photos in darktable (which is hugely CPU intensive).
Scott
Quoting zerozero, "The usage of PPA's in debian-based
systems is risky at best and entails serious compatibility
problems; usually it's the best way to destroy an install"

kurotsugi
Posts: 1992
Joined: 09 Jan 2014 00:17

Re: Performance Impact of kernel update (4.9.65-3+deb9u2)

Postby kurotsugi » 06 Jan 2018 04:55

the differences are in microseconds so you'll likely won't notice it. the result will vary depends on hardware specs too.

User avatar
ScottQuier
Posts: 1769
Joined: 18 Jul 2013 15:55
Location: Newport News, VA

Re: Performance Impact of kernel update (4.9.65-3+deb9u2)

Postby ScottQuier » 06 Jan 2018 22:02

kurotsugi wrote:the differences are in microseconds so you'll likely won't notice it. the result will vary depends on hardware specs too.

Hadn't really put that into perspective. Thanks.
Scott
Quoting zerozero, "The usage of PPA's in debian-based
systems is risky at best and entails serious compatibility
problems; usually it's the best way to destroy an install"


Return to “Hardware”

Who is online

Users browsing this forum: No registered users and 3 guests