Re: More memory = slower system???? WTF?
- From: Aragorn <stryder@xxxxxxxxxxxxxxx>
- Date: Fri, 27 Jan 2006 02:43:46 GMT
On Friday 27 January 2006 02:31, Captain Dondo stood up and spoke the
following words to the masses in /comp.os.linux.misc...:/
> I just upgraded my system from 1 GB RAM to 4 GB RAM. You would think
> this would speed it up....
That largely depends on the hardware and on the kernel used, actually.
> Nope. It is noticeably slower than before....
That is possible, yes...
> I mean really slower. A compile that took 30 minutes now takes an
> hour....
>
> Click on a xterm and it may take 30 seconds for a response.
>
> Try to highlight text in an xterm and it may take 5 or 20 seconds to
> respond....
>
> It feels like it's made disk access much slower, but there is no disk
> access - everything is in RAM and there is very little disk
> activity...
>
> No errors or messages anywhere. No instability that I've noticed.
> System runs cooler doing compiles, which would seem to say CPUs are
> less taxed but gkrellm shows CPU usage at 99%, same as before.
>
> The system is a 3.2GHz Intel, hyperthreaded, on an Asus mobo...
>
> So... Any ideas? Suggestions? /proc tweaks?
First of all, Asus is not too supportive in respect to GNU/Linux.
Secondly, having hyperthreading enabled may help increase speed up to a
30% gain, provided that your kernel supports SMP - and if the machine
physically *is* an SMP - SMT as well.
Thirdly, the amount of cache matters a great deal in respect to the
installed physical memory. If your physical memory increases that much
in capacity, the hit-to-miss ratio of the cache drops significantly.
> System is running 2.6.8 kernel; tomorrow I will compile a more recent
> kernel and see if that will help...
If you really _do_ mean 2.6.8 and not 2.6.8.1, then I would strongly
suggest using a more recent kernel. 2.6.8.1 was released only *one*
*day* after 2.6.8 was, and with good reason: there are some serious
flaws in 2.6.8. I experienced this myself - I actually had a kernel
crash with that one.
> But I am open to suggestions at this point....
I'm afraid that there's not all that much you can do. Check the
following:
- Is hyperthreading enabled?
- Are you compiling with /-j3/ - the amount of CPU's plus one?
- Is the installed memory type of the same latency and frequency as the
original memory?
- How large is your L2 cache, and do you have an L3? Are they enabled?
- What compiler are you using, and what compiler was used to build the
system? (/gcc/ 4.0.x appears to generate slower code than 3.5.)
- Does your kernel match the /glibc/ that your system was compiled
against? (e.g. Are you using a 2.6 kernel in a system that was compiled
for a 2.4 kernel?)
- Does your kernel _support_ multiprocessing and multithreading - i.e.
SMP and SMT? - and does it "see" all your memory? (You need a
HighMem-enabled kernel in order to be able to access 4 GB.)
That's about all I can bring up... The rest is up to you, I'm afraid.
;-)
--
With kind regards,
*Aragorn*
(Registered GNU/Linux user #223157)
.
- References:
- More memory = slower system???? WTF?
- From: Captain Dondo
- More memory = slower system???? WTF?
- Prev by Date: Re: How to save setting onto HDD when using Linux Live CD (Version: Stanix RC3)?
- Next by Date: Re: More memory = slower system???? WTF?
- Previous by thread: Re: More memory = slower system???? WTF?
- Next by thread: Re: More memory = slower system???? WTF?
- Index(es):
Relevant Pages
|
