Heavy emphasis on serving in the benchmarks:
http://www.2cpu.com/articles/98_1.html?73782&28295
Definitely worth the upgrade if you're into MySQL.
Comparison Of Kernel 2.6.4 vs. 2.4.25
Moderator: [PX]Kal Bob
- THE JEW (RaVeN)
- Posts: 12499
- Joined: Sun Mar 23, 2003 2:52 pm
- Location: Bethlehem ;)
- Contact:
Comparison Of Kernel 2.6.4 vs. 2.4.25
******* /=========\
****(_]/_____________\[_)
***** /(__)==JEW==(__)\
***** |=o__________o=|
***** |_|====---====|_|

¨˜”°º•[K]•º°”˜¨
****(_]/_____________\[_)
***** /(__)==JEW==(__)\
***** |=o__________o=|
***** |_|====---====|_|

¨˜”°º•[K]•º°”˜¨
-
[PX]Kal Bob
- Site Admin
- Posts: 1959
- Joined: Sun Mar 23, 2003 1:41 am
- Location: Stalinist Russia!
- Contact:
A few points:
First, 2.6.4 is still considered by most to be a BETA, a very good BETA but still just that. many production line people won't touch a kernel untill it is in the teens. Debain waited untill 2.4.18 to add it. I exspect better results as time goes on.
Seconed, the tester used the -j4 switch, this means do 4 simulanius actions. problem, the system is a Dual machine. Yah it has hypertreding which is crap, very inefficent. I think they would have gotten better results with -j2.
third, the 2.6 kernel is more, from what I have seen, better support for new hardware, and making user moddifing simpilar. I am not supprised that there are less then phenominal gains considering this. Also much of the improvements targeted the Desktop market.
Fourth, damn I loved thouse Samba Stats eh? that was sweet!!!
PS. I am tired and in the Middle of an Essay right now, I will spellcheck this later.
PPS. For the Jew, I was working on AJ's Quad Itanium and the AI64 2.6.4 kernel can support up to 256 proccessors. We talked about this before, just though I would give you the update. Cudows on the great benchmarking post :=D>:
First, 2.6.4 is still considered by most to be a BETA, a very good BETA but still just that. many production line people won't touch a kernel untill it is in the teens. Debain waited untill 2.4.18 to add it. I exspect better results as time goes on.
Seconed, the tester used the -j4 switch, this means do 4 simulanius actions. problem, the system is a Dual machine. Yah it has hypertreding which is crap, very inefficent. I think they would have gotten better results with -j2.
third, the 2.6 kernel is more, from what I have seen, better support for new hardware, and making user moddifing simpilar. I am not supprised that there are less then phenominal gains considering this. Also much of the improvements targeted the Desktop market.
Fourth, damn I loved thouse Samba Stats eh? that was sweet!!!
PS. I am tired and in the Middle of an Essay right now, I will spellcheck this later.
PPS. For the Jew, I was working on AJ's Quad Itanium and the AI64 2.6.4 kernel can support up to 256 proccessors. We talked about this before, just though I would give you the update. Cudows on the great benchmarking post :=D>:
KalNet SysAdmin.
Elitism- It's Lonely At The Top. But It's Comforting To Look Down Upon Everyone At The Bottom.
Elitism- It's Lonely At The Top. But It's Comforting To Look Down Upon Everyone At The Bottom.
- THE JEW (RaVeN)
- Posts: 12499
- Joined: Sun Mar 23, 2003 2:52 pm
- Location: Bethlehem ;)
- Contact:
Yeah, the Hyperthreading thing is up for debate. But if I remember correctly, he did it for both so that rules it out being a factor for one and not the other. Programs do work much better with Hyperthreading if they're optimized for it. Essentially, it makes the program think it has a complete cpu cycle when it's actually only using what's left over. Great for multi-tasking with winamp, word, etc. Not so good as an additional thread encoding, but it still does help.
******* /=========\
****(_]/_____________\[_)
***** /(__)==JEW==(__)\
***** |=o__________o=|
***** |_|====---====|_|

¨˜”°º•[K]•º°”˜¨
****(_]/_____________\[_)
***** /(__)==JEW==(__)\
***** |=o__________o=|
***** |_|====---====|_|

¨˜”°º•[K]•º°”˜¨