Kolivas Pushes New Kernel Responsiveness Patches

[karl]

I've seen the behavior in Windows 7 64bit. In fact I am starting to believe that 64bit is plagued with problems. I speculate that developers just don't understand how to problem for 64bit yet. Perhaps it could even be just the AMD processors. I unlocked an extra core on my processor. I noticed the system was more responsive when I kept the core locked (Windows / Linux) in the 64bit environment.

1. Has anyone with IA64 had any problems?
2. If you have AMD64, have you unlocked a locked core?
3. Sun and DEC have had 64bit operating systems for years. Anyone with experience under those environments got any input?

I'm 100% sure this has nothing to do with 64 bit nor with AMD cpu's.

[deanjo]

I'm 100% sure this has nothing to do with 64 bit nor with AMD cpu's.

Exactly, I've been running 64-bit since 2003 in linux and 2005 with windows on AMD processors and have not had the issues you describe. You problem is more then likely that the "disabled core" is a defective one and that is why it became an X3. There are no guarantees that just because an X3's disabled core is unlockable that it is a good core. With all X3's it's the luck of the draw if you got one that is just disabled to fill demand or disable because of a defect.

[squirrl]

Don't start the fires just yet

Understood but I was trying to make sure we know most of the parameters at play here.

I'm testing things myself.

I have:
Gigabyte GA-MA785GM-US2H
AMD SEMPRON 140 UNLOCKED to an AMD Athlon(tm) II X2 4400e
1024 Megabytes of ram (initially)
128 Megabytes shared onboard ATI 4200HD Video Chipset
160 Gigabyte Western Digital IDE hard-drive (Walmart discount)
DVD/Burner IDE (Walmart discount)

Ubuntu 9.10 Karmic Koo'lala 64bit latest updates 2.6.31-20 kernel.

Essentially there is only one IDE connector on this motherboard so the bandwidth is shared.

I had the sluggish response with copying files to and from the hard-drive. What really hit performance was copying from the DVD-Rom.

I suspect it's the IDE port. I've encountered this in the past and used a raid controller / IO card to hook up the hard-drive which alleviated the IO problems somewhat.

Today I installed 4 Gigabytes of memory and the performance magically picked up. The user interface, Gnome has improved when copying a large amount of files. I'm using my Wine directory which is full of ISO and huge files as copying test.

I know the Intel guy just mentioned the VM and I truly believe he is correct. There are some major performance issues with the VM after 2.6.24.

I'm going to make SVG's myself.

[deanjo]

Understood but I was trying to make sure we know most of the parameters at play here.

I'm testing things myself.

I have:
Gigabyte GA-MA785GM-US2H
AMD SEMPRON 140 UNLOCKED to an AMD Athlon(tm) II X2 4400e
1024 Megabytes of ram (initially)
128 Megabytes shared onboard ATI 4200HD Video Chipset
160 Gigabyte Western Digital IDE hard-drive (Walmart discount)
DVD/Burner IDE (Walmart discount)

Ubuntu 9.10 Karmic Koo'lala 64bit latest updates 2.6.31-20 kernel.

Essentially there is only one IDE connector on this motherboard so the bandwidth is shared.

I had the sluggish response with copying files to and from the hard-drive. What really hit performance was copying from the DVD-Rom.

I suspect it's the IDE port. I've encountered this in the past and used a raid controller / IO card to hook up the hard-drive which alleviated the IO problems somewhat.

Today I installed 4 Gigabytes of memory and the performance magically picked up. The user interface, Gnome has improved when copying a large amount of files. I'm using my Wine directory which is full of ISO and huge files as copying test.

I know the Intel guy just mentioned the VM and I truly believe he is correct. There are some major performance issues with the VM after 2.6.24.

I'm going to make SVG's myself.

If a raid controller started alleviating your issues then it's quite possible it is a chipset driver issue. In windows land it is well known that the best performance when it comes to AMD/ATI chipsets that you are better off using the built in windows drivers then the AMD ones.

[squirrl]

Quick question, where are the perf tools located?

[squirrl]

I was just thinking about FreeBSD, back in 1999 I was using 4.3. The file operations were slow but the system response was quick.

Everything felt double buffered.

[kraftman]

Quick question, where are the perf tools located?

Download the Linux source package and they're in there - tools/perf or something like this.

[Yfrwlf]

Maybe it was already suggested but it might be prudent to make a test in PTS which runs two or more different programs at the same time in order to attempt to benchmark any noticeable improvements that each scheduler may provide. These need to of course be separated out enough so that they don't fall victim to the scheduler lumping them together, and of course one should stress I/O and the other, say, audio or graphics or something like that.

I'm sure readers would be interested in such a project in response to this scheduler.

I.e., does it really work, and is it really better? Tests will help determine that.

Ah, the scientific method...

[RealNC]

Maybe it was already suggested but it might be prudent to make a test in PTS which runs two or more different programs at the same time in order to attempt to benchmark any noticeable improvements that each scheduler may provide.

It has been mentioned already that the point of the new scheduler is NOT performance. The vanilla scheduler doesn't have performance problems. This is not even the point.

[Yfrwlf]

It has been mentioned already that the point of the new scheduler is NOT performance. The vanilla scheduler doesn't have performance problems. This is not even the point.

Um, unless I totally read the article wrong, I'm pretty sure the point is to give desktop users priority access to things which are critical to them, like being able to continue moving their mouse, and not making audio and graphics choppy, all while performing some disk I/O. In other words, don't equally share CPU with all programs like a server might do, but instead have priorities. This has been a problem plaguing Linux desktop users for a long time. Windows 2000 performs better sadly while trying to play music for instance while multitasking other intensive activities. If it is simply a matter of CPU niceness settings not being configured, so be it, but I thought the scheduler played a role in this issue.