Jump to content

I have a confession to make


gt_swagger
 Share

Recommended Posts

I have a need for speed.

 

After seeing Windoze XP ****smoke**** MDK 10.0 in some distributing computing command line (DOS / BASH) tasks... I am no longer happy.

 

Could somebody provide me instructions / a link to a site with instructions for installing MDK 10.0 fully optimized from the ground up?

 

I have compiled MDK cooker kernels optimized for P4, but I had issues when switching kernels... namely disappearing hardware :wall:

 

Definately want everything to operate off these flags: -march=pentium4 -03 -pipe

 

I saw how Gentoo is geared towards this sort of thing, but fell into cardiac arrest upon seeing the installation "handbook" -- aka Crime and Punishment. (Not to mention the LiveCD cannot get my 3C940 Gigabit LOM to work) :cheesy:

 

Please help me get the true performance my penguin is capable of!

Link to comment
Share on other sites

alsa            0:off   1:off   2:on    3:on    4:on    5:on    6:off
dm              0:off   1:off   2:off   3:off   4:off   5:on    6:off
kheader         0:off   1:off   2:on    3:on    4:off   5:on    6:off
netfs           0:off   1:off   2:off   3:on    4:on    5:on    6:off
network         0:off   1:off   2:on    3:on    4:on    5:on    6:off
partmon         0:off   1:off   2:off   3:on    4:on    5:on    6:off
random          0:off   1:off   2:on    3:on    4:on    5:on    6:off
rawdevices      0:off   1:off   2:off   3:on    4:on    5:on    6:off
sound           0:off   1:off   2:on    3:on    4:on    5:on    6:off
keytable        0:off   1:off   2:on    3:on    4:on    5:on    6:off
syslog          0:off   1:off   2:on    3:on    4:on    5:on    6:off
oki4daemon      0:off   1:off   2:off   3:off   4:off   5:off   6:off
crond           0:off   1:off   2:on    3:on    4:on    5:on    6:off
xinetd          0:off   1:off   2:off   3:on    4:on    5:on    6:off
portmap         0:off   1:off   2:off   3:on    4:on    5:on    6:off
xfs             0:off   1:off   2:on    3:on    4:on    5:on    6:off
hotplug         0:off   1:off   2:off   3:off   4:off   5:off   6:off
nfslock         0:off   1:off   2:off   3:on    4:on    5:on    6:off
devfsd          0:off   1:off   2:on    3:on    4:on    5:on    6:off
atd             0:off   1:off   2:off   3:on    4:on    5:on    6:off
internet        0:off   1:off   2:off   3:on    4:off   5:on    6:off
harddrake       0:off   1:off   2:off   3:on    4:on    5:on    6:off
numlock         0:off   1:off   2:off   3:on    4:on    5:on    6:off
mtink           0:off   1:off   2:off   3:off   4:off   5:off   6:off
cups            0:off   1:off   2:on    3:on    4:on    5:on    6:off
tmdns           0:off   1:off   2:off   3:on    4:on    5:on    6:off
xinetd based services:
       rsync:  off
       fam:    on
       cups-lpd:       off
       cvs:    off

 

For comparison purposes...

OS            ||  Average Time Per Work Unit
------------  ||  --------------------------
Windoze XP    ||  15 mins, 30 secs
MDK 10.0      ||  25 mins, 10 secs
MDK 10.0*     ||  21 mins, 30 secs
MDK 10.0**    ||  ?? mins, ?? secs

*  Using Wine
** After tweeking

Edited by gt_swagger
Link to comment
Share on other sites

  • 2 weeks later...

--- Update ---

Terminated these processes:

- netfs

- xinetd

- xfs

 

--- New hardware Specs ---

Pentium IV 2.8 Ghz, 800 Mhz FSB (3,150.5 Mhz, 900 MhzFSB)

Asus P4P800 (1016) -- Turbo mode and PAT (aka MAM) enabled

Timings: 2.5 : 3 : 3 : 6 : 4 (5:4)

1.6v CPU core

1.6v AGP

2.75v DDR

Latency for PCI set to 32

DDR Core clock 320 Mhz

Legacy USB disabled

128 AGP

Disable spread spectrum

 

 

Thanks for any further tweak help that can provided.

 

I'm :screwy: for speed.

Edited by gt_swagger
Link to comment
Share on other sites

Just out of curiosity what task did Xp smoke Mandrake on? And can we test it? If its for distributed computing it could be wholy dependent on the comunications card driver. The properly writen one for windows vs the hacked one for linux. And if thats the case then you can optimize Mandrake till the cows come home and never see a real proformance increase.

Link to comment
Share on other sites

Plenty of scope here :D

Like Pzatch says what areas.

 

Just a forethought. Those guys with 4x4's will beat me hands down racing across the desert .... but Ill smoke em on any highway.

 

So with that in mind....

disk access etc. check the hdparm and the filesystems.

Remember ext3 is journaled and so slower than ext2 but safer.

 

Secondly memory/swap config ???

 

Thirdly some stuff like multimedia will benfit from correct compiliotion parameters. Like Pzatch says what were the tasks ?

Link to comment
Share on other sites

ext2 is my filesystem.

 

1024 MB CORSAIR XMS3200 (no 4 GB support kernel compilation though)

250 MB Swap (hardly ever used)

 

XP command line is beating down Mandrake by a good margain in Lifemapper... and I'd imagine in SETI@Home too. My SETI@Home efficiency is around 9.75 CPU cycles / FLOP ... it lists the optimal performance of a PIV Northwood at about 6.0 to 6.2 CPU cycles / FLOP depending on RAM type.

 

I have passed a few parameters using hdparm that got a small increase in hard drive I/O ... but I think my speed is hurt by a really bad latency hiding somewhere (??), inefficient use of data after it is taken from the hd, or lack of architecture specific optimization. I know it will go faster than this. I got Linux to be absolutely outrageously smoking fast after compiling 2.6.3-7mdk to my needs (4 GB memory... architecture optimized... "fat" removed...) ... however installing the nVidia driver ruined by day by ruining the XFree86 installation :( Also had issues with "vanishing hardware" when booting from kernel 2.6.3-7 back to 2.6.3-4

Edited by gt_swagger
Link to comment
Share on other sites

Update: I think I may have found the potential hang up... a stick of the CORSAIR RAM looks like it suffered an electrical surge and was fried along with the RAM slot on the mobo ... going to get a free replacement per my warranty and service plan... will see if I get that nice performance when I get the new box.

Link to comment
Share on other sites

I hope that is the problem as those numbers you gave were a little hard for me to accept being a faithful linux user.

 

I'm still amazed at how well M$ prossess data coming out of bad hardware. And why can't linux incorporate this amazing technology into its kernel so everyone could do it?

Link to comment
Share on other sites

I hope that is the problem as those numbers you gave were a little hard for me to accept being a faithful linux user.

 

I'm still amazed at how well M$ prossess data coming out of bad hardware. And why can't linux incorporate this amazing technology into its kernel so everyone could do it?

Well.. MS may be able to process data, but is it good data? Also, as far as I know, MS always try to minimize the memory usage while Linux takes all available memory for cache. MS programs may be not touching the broken hardware at all while Linux uses the broken hardware regularly.

 

You see.. using linux is the best way to know whether your hardware is broken or not.. so it is good to prevent "lemons" :)

Link to comment
Share on other sites

I also have found that linux is beter to find out if you really have a hardware issue. Windex, as we all know, can have odd problems that seem mysterious. It might just be that windex runs on bad hardware, which really means you don't know how reliable the operation is!

Link to comment
Share on other sites

ext2 is my filesystem.

 

1024 MB CORSAIR XMS3200 (no 4 GB support kernel compilation though)

250 MB Swap (hardly ever used)

 

XP command line is beating down Mandrake by a good margain in Lifemapper... and I'd imagine in SETI@Home too. My SETI@Home efficiency is around 9.75 CPU cycles / FLOP ... it lists the optimal performance of a PIV Northwood at about 6.0 to 6.2 CPU cycles / FLOP depending on RAM type.

 

I have passed a few parameters using hdparm that got a small increase in hard drive I/O ... but I think my speed is hurt by a really bad latency hiding somewhere (??), inefficient use of data after it is taken from the hd, or lack of architecture specific optimization. I know it will go faster than this. I got Linux to be absolutely outrageously smoking fast after compiling 2.6.3-7mdk to my needs (4 GB memory... architecture optimized... "fat" removed...) ... however installing the nVidia driver ruined by day by ruining the XFree86 installation :( Also had issues with "vanishing hardware" when booting from kernel 2.6.3-7 back to 2.6.3-4

to be someone who is interested in speed, that filesystem is not very fast...

 

i think you should try

reiserfs or XFS (which i have been using for quite some time)

 

now to test your disks on a console write

hdparm -tT /dev/hdx

 

well...

 

bye hope you can speed-up your box ;)

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
 Share

×
×
  • Create New...