Phenom II X6

Message

Atha · Post by **Atha** » Fri Oct 22, 2010 9:29 pm

Chiitoo wrote:Would be interesting to see some benchmarks like from bashmark…

How about POV-Ray?
http://www.povray.org/download/benchmark.php

Maybe also nbench? – But it's also very very old (1997?)…
[topic=93250]BYTE Benchmarks[/topic]

I'm a little concerned that old benchmark software will not be as good on multi-core CPUs as on single-core, which was state of the art when the benchmark software was new.

If interested, I can post some results…

Post by **Chiitoo** » Fri Oct 22, 2010 10:04 pm

Atha wrote:
Chiitoo wrote:Would be interesting to see some benchmarks like from bashmark…
How about POV-Ray?
http://www.povray.org/download/benchmark.php

Maybe also nbench? – But it's also very very old (1997?)…
[topic=93250]BYTE Benchmarks[/topic]

I'm a little concerned that old benchmark software will not be as good on multi-core CPUs as on single-core, which was state of the art when the benchmark software was new.

If interested, I can post some results…

Yeah, I share your feelings on old benchmarks.

And yes, I'm interested and I will look into those myself!
Need to find some things to stress the system really, because of the problems I mentioned. Could maybe compare results with another PSU or something with myself or even hope to be able to generate the crash.
I really don't know many since I've been using Gentoo, well, Linux in general only for few weeks now. ^^

I think I got phoronix-test-suite but haven't tried it yet.

Atha · Post by **Atha** » Sat Oct 23, 2010 9:00 am

Chiitoo wrote:And yes, I'm interested and I will look into those myself!
Need to find some things to stress the system really, because of the problems I mentioned. Could maybe compare results with another PSU or something with myself or even hope to be able to generate the crash.
I really don't know many since I've been using Gentoo, well, Linux in general only for few weeks now. ^^

If it is only about really stressing your system, why not use portage? If you emerge, say, OpenOffice.org, you'll give all your cores a hard time and you can measure how long it takes. Best you do this on a RAM based filesystem, like tmpfs, since this will also eliminate most of the hard drive dependend performance differences.

If it is only importaint to benchmark different systems you could also measure the time to build the Linux kernel itself.

For benchmarking I'll try to boot into the single-user runlevel, but I have yet to read how I do that. It will definitly change the results if you're having KDE running while benchmarking your system, but to make it a perfect environment the single-user mode is the best I can come up with.

Cheers,
A.

Atha · Post by **Atha** » Sat Oct 23, 2010 11:56 am

The bad news: povray and nbench are utilizing only one core! That makes 1 core work at 100% and 5 cores fall almost asleep. I am not sure what the actual clock frequency was, i.e. if Turbo Core kicked in or not. Where can I find this information? I looked through /sys/* but couldn't find anything other than that a core was either @ 800 or @ 3200 i.e. 3.2 GHz, but what about Turbo Core 3.6 GHz?

Code: Select all

BYTEmark* Native Mode Benchmark ver. 2 (10/95)
Index-split by Andrew D. Balsa (11/97)
Linux/Unix* port by Uwe F. Mayer (12/96,11/97)

TEST                : Iterations/sec.  : Old Index   : New Index
                    :                  : Pentium 90* : AMD K6/233*
--------------------:------------------:-------------:------------
NUMERIC SORT        :          1682.4  :      43.15  :      14.17
STRING SORT         :          290.97  :     130.01  :      20.12
BITFIELD            :      5.2812e+08  :      90.59  :      18.92
FP EMULATION        :          310.64  :     149.06  :      34.40
FOURIER             :           32316  :      36.75  :      20.64
ASSIGNMENT          :          40.096  :     152.57  :      39.57
IDEA                :            9964  :     152.40  :      45.25
HUFFMAN             :          3246.9  :      90.04  :      28.75
NEURAL NET          :          60.128  :      96.59  :      40.63
LU DECOMPOSITION    :          1979.2  :     102.53  :      74.04
==========================ORIGINAL BYTEMARK RESULTS==========================
INTEGER INDEX       : 106.808
FLOATING-POINT INDEX: 71.397
Baseline (MSDOS*)   : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0
==============================LINUX DATA BELOW===============================
CPU                 : 6 CPU AuthenticAMD AMD Phenom(tm) II X6 1090T Processor 3200MHz
L2 Cache            : 512 KB
OS                  : Linux 2.6.36-gentoo-4
C compiler          : x86_64-pc-linux-gnu-gcc
libc                : 
MEMORY INDEX        : 24.700
INTEGER INDEX       : 28.218
FLOATING-POINT INDEX: 39.599
Baseline (LINUX)    : AMD K6/233*, 512 KB L2-cache, gcc 2.7.2.3, libc-5.4.38
* Trademarks are property of their respective holder.

I stripped the povray output:

Code: Select all

Persistence of Vision(tm) Ray Tracer Version 3.6.1 (x86_64-pc-linux-gnu-g++
 4.4.4 @ x86_64-pc-linux-gnu)
This is an unofficial version compiled by:
 Portage (Gentoo Linux) on HexaBlack.thnet
 The POV-Ray Team(tm) is not responsible for supporting this version.

[…]

Support libraries used by POV-Ray:
  ZLib 1.2.3, Copyright 1995-1998 Jean-loup Gailly and Mark Adler
  LibPNG 1.4.3, Copyright 1998-2002 Glenn Randers-Pehrson
  LibJPEG 8, Copyright 1998 Thomas G. Lane
  LibTIFF 3.9.4, Copyright 1988-1997 Sam Leffler, 1991-1997 SGI
Redirecting Options
  All Streams to console..........On 
  Debug Stream to console.........On 
  Fatal Stream to console.........On 
  Render Stream to console........On 
  Statistics Stream to console....On 
  Warning Stream to console.......On 
Parsing Options
  Input file: benchmark.pov (compatible to version 3.50)
  Remove bounds........On 
  Split unions.........Off
  Library paths:
    /usr/share/povray
    /usr/share/povray/ini
    /usr/share/povray/include
Output Options
  Image resolution 384 by 384 (rows 1 to 384, columns 1 to 384).
  Graphic display......Off
  Mosaic preview.......Off
  CPU usage histogram..Off
  Continued trace......Off
Tracing Options
  Quality:  9
  Bounding boxes.......On   Bounding threshold: 3
  Light Buffer.........On 
  Vista Buffer.........On   Draw Vista Buffer....Off
  Antialiasing.........On  (Method 1, Threshold 0.300, Depth 3, Jitter 1.00)
  Clock value:    0.000  (Animation off)

  0:00:00 Parsing


 Building mesh2: 
   - vertex_vectors
   - normal_vectors
   - uv_vectors
   - face_indices

  0:00:00 Creating bounding slabs
  0:00:00 Creating vista buffer
  0:00:00 Creating light buffers
  0:00:00 Creating light buffers 2299K tokens
Scene Statistics
  Finite objects:          171
  Infinite objects:          3
  Light sources:             2
  Total:                   176

[…]

Render Statistics
Image Resolution 384 x 384
----------------------------------------------------------------------------
Pixels:           147840   Samples:          573672   Smpls/Pxl: 3.88
Rays:            1860849   Saved:             23713   Max Level: 12/12
----------------------------------------------------------------------------
Ray->Shape Intersection          Tests       Succeeded  Percentage
----------------------------------------------------------------------------
Box                           79514410         9416625     11.84
Cone/Cylinder                 78068415         6602507      8.46
CSG Intersection             169535597        58625425     34.58
CSG Merge                       816474           33903      4.15
Fractal                        1818412          105934      5.83
Height Field                   3557273          102567      2.88
Height Field Box               3557273          685370     19.27
Height Field Triangle          3239597          105862      3.27
Height Field Block             5678772         1670066     29.41
Height Field Cell             22370353         1778819      7.95
Isosurface                    11941322          730194      6.11
Isosurface Container          12429502        11941884     96.08
Isosurface Cache                184113           42170     22.90
Mesh                          15214896           64258      0.42
Plane                         91963606         1292941      1.41
Sphere                       280662882       164061129     58.45
Superellipsoid                  608983           44255      7.27
Torus                          2969319          420426     14.16
Torus Bound                    2969319          484614     16.32
True Type Font                  794568           80877     10.18
Clipping Object                2581333         1535370     59.48
Bounding Box                 520147396       148819720     28.61
Vista Buffer                  22395267        12889381     57.55
----------------------------------------------------------------------------
Isosurface roots:          11935960
Function VM calls:        172757012
----------------------------------------------------------------------------
Roots tested:                484614   eliminated:               277015
Calls to Noise:          4823836177   Calls to DNoise:      2612354843
----------------------------------------------------------------------------
Media Intervals:           39562150   Media Samples:         356906016 (9.02)
Shadow Ray Tests:         128411752   Succeeded:              52349586
Reflected Rays:              224745   Total Internal:             1062
Refracted Rays:              144400
Transmitted Rays:            621932
----------------------------------------------------------------------------
Number of photons shot:           74025
Surface photons stored:           65040
Priority queue insert:          6000933
Priority queue remove:          1464268
Gather function called:          669739
----------------------------------------------------------------------------
Smallest Alloc:                   9 bytes
Largest  Alloc:             2560008 bytes
Peak memory used:           7169485 bytes
Total Scene Processing Times
  Parse Time:    0 hours  0 minutes  0 seconds (0 seconds)
  Photon Time:   0 hours  0 minutes 19 seconds (19 seconds)
  Render Time:   0 hours 16 minutes  4 seconds (964 seconds)
  Total Time:    0 hours 16 minutes 23 seconds (983 seconds)

For the statistics: uname -a

Code: Select all

Linux HexaBlack 2.6.36-gentoo-4 #1 SMP Thu Oct 21 21:12:39 CEST 2010 x86_64 AMD Phenom(tm) II X6 1090T Processor AuthenticAMD GNU/Linux

HexaBlack is the name I gave to my X6. -4 is my personal configuration iteration; this one is actually 4.2 – I'm still experimenting with it…

Post by **Chiitoo** » Sat Oct 23, 2010 7:13 pm

Yeah, I have been using emerge a lot and been building a lot of kernels.
Been even using --jobs for emerge which as far as I understand it, should make more use of the thing and I surely get some slowdowns then but no crashing... yet.

For the BYTEmark, I noticed the single core thing as well and my results are pretty much identical to yours, no surprise there really. I ran this with KDE and several other things running while for the bashmark I tried running it before and after launching KDE (after POST and BOOT my computer is in the command-line with nothing much running). The results were the same, I think.
If you check the results from bashmark in [post=6461893]the other thread[/post], there are some serious differences with another Phenom II X6 user. Could be just because it doesn't seem to actually use my CPU much at all, as to why this is, no idea as of yet.

As to how to get the best readings for the frequencies and such, or even temps, I have not yet looked into that enough. I have sensors installed and going but it's not really configured even. Interesting thing is, that for me everything says it's at 3.6+ all the time. :S
I think it's because of a BIOS setting, I can't remember which, maybe the Cool and Quiet?
I remember trying it on and it lowered the performance a lot...

There are many things I should look into regarding this, still.

Anyways, some results here for BYTEmark, I will look into Pov-Ray a bit later as it's not as simple it seems. :S
I had KDE and some other things running again, will see if there is a big difference after next boot, without starting KDE etc.

Code: Select all

BYTEmark* Native Mode Benchmark ver. 2 (10/95)
Index-split by Andrew D. Balsa (11/97)
Linux/Unix* port by Uwe F. Mayer (12/96,11/97)

TEST                : Iterations/sec.  : Old Index   : New Index
                    :                  : Pentium 90* : AMD K6/233*
--------------------:------------------:-------------:------------
NUMERIC SORT        :          1706.9  :      43.77  :      14.38
STRING SORT         :           290.8  :     129.94  :      20.11
BITFIELD            :      5.9362e+08  :     101.83  :      21.27
FP EMULATION        :          314.72  :     151.02  :      34.85
FOURIER             :           32735  :      37.23  :      20.91
ASSIGNMENT          :          40.503  :     154.12  :      39.98
IDEA                :           10108  :     154.60  :      45.90
HUFFMAN             :            3300  :      91.51  :      29.22
NEURAL NET          :           61.48  :      98.76  :      41.54
LU DECOMPOSITION    :          2014.1  :     104.34  :      75.34
==========================ORIGINAL BYTEMARK RESULTS==========================
INTEGER INDEX       : 109.661
FLOATING-POINT INDEX: 72.659
Baseline (MSDOS*)   : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0
==============================LINUX DATA BELOW===============================
CPU                 : 6 CPU AuthenticAMD AMD Phenom(tm) II X6 1090T Processor 3616MHz
L2 Cache            : 512 KB
OS                  : Linux 2.6.34-gentoo-r11
C compiler          : x86_64-pc-linux-gnu-gcc
libc                : 
MEMORY INDEX        : 25.763
INTEGER INDEX       : 28.631
FLOATING-POINT INDEX: 40.300
Baseline (LINUX)    : AMD K6/233*, 512 KB L2-cache, gcc 2.7.2.3, libc-5.4.38
* Trademarks are property of their respective holder.

hielvc · Post by **hielvc** » Tue Oct 26, 2010 12:26 am

bogomips=2*cpu_freq or cpu_Hz=bogo/2

Code: Select all

uname -a
Linux amd_1055 2.6.36-gentoo #4 SMP Sat Oct 23 09:42:41 PDT 2010 x86_64 AMD Phenom(tm) II X6 1055T Processor AuthenticAMD GNU/Linux

grep "bogo" /proc/cpuinfo
bogomips	: 6440.05
bogomips	: 6439.90
bogomips	: 6439.90
bogomips	: 6439.88
bogomips	: 6439.92
bogomips	: 6439.92

Cpu 3220 MHz. Standard AMD fan no voltage tweaking. FSB 230
~amd64
CFLAGS="-march=amdfam10 -O2 -fomit-frame-pointer -pipe"
MAKEOPTS="-j6"

What we all can relate to is compile times. Here is my kernel, gcc, glibc, and openoffice genlop times.

Code: Select all

amd_1055 linux # make clean; time make -j6

real	1m23.091s
user	6m51.253s
sys	0m36.009s

hielvc@amd_1055 ~ $ genlop -t gcc
Mon Oct 25 12:38:47 2010 >>> sys-devel/gcc-4.4.5
       merge time: 10 minutes and 47 seconds.

hielvc@amd_1055 ~ $ genlop -t glibc
Mon Oct 25 12:03:28 2010 >>> sys-libs/glibc-2.12.1-r1
       merge time: 8 minutes and 21 seconds.

hielvc@amd_1055 ~ $ genlop -t openoffice
Sun Oct 24 23:25:58 2010 >>> app-office/openoffice-3.2.1
       merge time: 39 minutes and 41 seconds.

emerge -eqp --nodeps @system |  genlop -p
Estimated update time: 45 minutes.

monsm · Post by **monsm** » Mon Nov 01, 2010 1:50 pm

hielvc wrote:
hielvc@amd_1055 ~ $ genlop -t openoffice
Sun Oct 24 23:25:58 2010 >>> app-office/openoffice-3.2.1
merge time: 39 minutes and 41 seconds.

emerge -eqp --nodeps @system | genlop -p
Estimated update time: 45 minutes.[/code]

Wao, thats fast! I think its time I replaced my old K7 box with one of these.
Last time I did openoffice it took more than 12 hours...

Mons

jannis · Post by **jannis** » Tue Dec 28, 2010 7:20 am

I have a Phenom II X6 1055T and I know how to find out if TurboCore kicks in. The only way is to use "cpufreq-aperf" as root and take a look the the col "Average freq(KHz)". There you can see the actual frequency the cores are running at.
If that does not get over the rated clock, check your dmesg. There should be a line like "[ 2.218052] powernow-k8: Core Performance Boosting: on."

foomor · Post by **foomor** » Sat Feb 05, 2011 5:37 am

I've ordered Phenom II x6 1090T Black Edition and i am interested if linux kernel does support it entirely now? Will i have all those features like cool'n'quiet and turbo core working properly?

jannis · Post by **jannis** » Sat Feb 05, 2011 6:44 am

The Phenom II X6 1055T is completely support. The Black Edition should be as well, I just wouldn't know how to modify the multiplicator there.

hielvc · Post by **hielvc** » Sat Feb 05, 2011 6:42 pm

For your 1055 you cant change the multiplier. You have to change your FSB which is set 200, stock. Ive pushed mine to 240 with stock fan and no voltage tweaking. But I prefer to keep my max temperature at 60C or less so I generally run at FSB=230 . Watch your Memory speeds and adjust their multiplier lower to keep them from over clocking to high.

Good Luck

AaronPPC · Post by **AaronPPC** » Sun Feb 06, 2011 3:59 am

Any motherboard recommendations?

I currently use use an MSI board on my current X2 system. A quick Google search says that Gigabyte offers the best Linux support. I hope that means that sleep works, which is my first requirement (and a reaction to my MSI). I'm also want to be able to flash the BIOS easily. I don't really care about USB3, but SATAIII would be nice.

I looked at a couple Gigabyte motherboards on Newegg and there were quite a few complaints about failed audio and LAN.

What are you using?

firephoto · Post by **firephoto** » Sun Feb 06, 2011 4:52 am

AaronPPC wrote:Any motherboard recommendations?

I currently use use an MSI board on my current X2 system. A quick Google search says that Gigabyte offers the best Linux support. I hope that means that sleep works, which is my first requirement (and a reaction to my MSI). I'm also want to be able to flash the BIOS easily. I don't really care about USB3, but SATAIII would be nice.

I looked at a couple Gigabyte motherboards on Newegg and there were quite a few complaints about failed audio and LAN.

What are you using?

I'm running a gigabyte GA-890GPA-UD3H with a 1090T and everything seems to be working. Sleep and resume works but it was an accident when it happened because I don't use that but it worked without any issue that I noticed. I've been running with the ATI graphics too and they're ok too and so far haven't frustrated me enough to put the nvidia card in.

Bios update you just have to hit the right key(s) and pick the right file from a flash drive when you boot.

hielvc · Post by **hielvc** » Sun Feb 06, 2011 6:05 am

As firephoto said a gigabyte GA-890GPA-UD3H with 1055t. No bios update. Also running sys-kernel/amd-ucode and x11-drivers/radeon-ucode. I just did a planed, my sysadmin is a dumbass, without a good backup < dont copy or plain rsync to an ntfs drive. Compress it so that attributes are inside with data. Twice that dumbass has done it. His motto is "I hate Neddyseadragon "

> so full install there I went. Using -j8 here are my genlops

Code: Select all

 $ genlop -t gcc glibc openoffice
 * sys-devel/gcc

     Thu Feb  3 17:37:07 2011 >>> sys-devel/gcc-4.5.2
       merge time: 12 minutes and 5 seconds.

     Fri Feb  4 11:00:12 2011 >>> sys-devel/gcc-4.5.2
       merge time: 12 minutes.

     Fri Feb  4 11:43:51 2011 >>> sys-devel/gcc-4.5.2
       merge time: 12 minutes and 2 seconds.

 * sys-libs/glibc

     Thu Feb  3 17:23:09 2011 >>> sys-libs/glibc-2.12.2
       merge time: 13 minutes and 31 seconds.

     Fri Feb  4 10:48:12 2011 >>> sys-libs/glibc-2.12.2
       merge time: 8 minutes and 48 seconds.

     Sat Feb  5 10:07:28 2011 >>> sys-libs/glibc-2.12.2
       merge time: 9 minutes and 50 seconds.

 * app-office/openoffice

     Fri Feb  4 22:34:40 2011 >>> app-office/openoffice-3.2.1-r1
       merge time: 35 minutes and 26 seconds.

firephoto what your genlop times?

firephoto · Post by **firephoto** » Sun Feb 06, 2011 6:55 am

hielvc wrote: firephoto what your genlop times?

Code: Select all


$ genlop -t gcc glibc
 * sys-devel/gcc

     Mon Jan 10 13:27:52 2011 >>> sys-devel/gcc-4.5.2
       merge time: 17 minutes and 3 seconds.

 * sys-libs/glibc

     Mon Jan 10 13:10:49 2011 >>> sys-libs/glibc-2.12.2
       merge time: 10 minutes and 53 seconds.

[ebuild   R    ] sys-devel/gcc-4.5.2  USE="fortran gtk mudflap (multilib) nls nptl openmp (-altivec) -bootstrap -build -doc (-fixed-point) -gcj -graphite (-hardened) (-libffi) -lto -multislot (-n32) (-n64) -nocxx -nopie -nossp -objc -objc++ -objc-gc -test -vanilla" 0 kB

Listed my gcc flags since I wonder if they're a bit different than yours or maybe it's just from my non-overclocked setup.

jannis · Post by **jannis** » Sun Feb 06, 2011 10:23 am

I have the MSI 890FXA-GD70. Quite expensive but has stuff like USB3.0 and SATA3.
The USB 3.0 controller on it is detected by the XHCI driver but I don't have any USB 3 devices to test. For the hardware-monitoring chip, you need a small kernel patch but it works fine then. Everything else runs jut perfect.

Anon-E-moose · Post by **Anon-E-moose** » Sun Feb 06, 2011 11:45 am

I'm running an old MSI NF750 (the last nvidia board they made)
with a 1055 overclocked to 3.5 Ghz
using stock fan, ~30 deg F at idle.
My case is the antec 900-2 with an extra side fan so I get good airflow for all the components

Code: Select all

Loadavg: 0.12, 0.08, 0.05

core - 30.5°C
case  - 26.0°C
---
/dev/sda: ST3500320AS: 27°C
/dev/sdb: Hitachi HDS722020ALA330: 29°C
/dev/sdc: Hitachi HDS722020ALA330: 31°C
---
Temperature Gpu: 40

Code: Select all

     Mon Dec  6 18:45:45 2010 >>> sys-libs/glibc-2.11.2-r3
       merge time: 7 minutes and 36 seconds.

     Thu Dec  2 11:06:40 2010 >>> sys-devel/gcc-4.5.1-r1
       merge time: 12 minutes and 44 seconds.

     Tue Dec  7 07:09:33 2010 >>> sys-devel/gcc-4.4.4-r2
       merge time: 9 minutes and 36 seconds.

I don't run open office so no times for that. I run with "-j6" to leave a little room for other things going on.

hielvc · Post by **hielvc** » Wed Feb 09, 2011 3:15 am

firephoto the only difference is I have " -nls "

Phenom II X6

[1.EDIT] Benchmark