On 18 Apr 00, at 0:34, Siegmar Szlavik wrote:

> I don't say that prime95/ntprime has a problem, just that it can 
> make (serious) problems on some systems and they don't go away until
> it gets uninstalled. I don't think it has to do with a memory leak,
> because if it comes to the point where the system is no longer
> useable (for example it takes 20-30 seconds to start a program like
> the NT explorer) in the majority of the cases the system is already 
> not useable directly after reboot.

Ah, this _is_ a bit different.

I take it you're running Win NT WS 4.0 SP 4+ (as I said before, you 
really _should_ be running SP 6a, but anything previous to SP 4 is 
definitely going to leave you open to numerous problems).

Did you try disabling NTPrime (Control Panel/Services/Startup) so 
that it doesn't start automatically at boot time? Does this cure the 
problem? Does the problem reappear if you then manually start NTPrime 
with the system running "normally"? If the answer to _either_ of the 
last two questions is "no", then NTPrime isn't the problem.

Does the incidence of the problem coincide with a great deal of disk 
activity, which doesn't die away after at most one minute but 
continues indefinitely? If so, this is a clear indication that the 
problem is being caused by a shortage of memory.

How much memory is there in your system? If you're running NTPrime, 
the P-1 memory allocation isn't a factor since NTPrime v20 isn't yet 
released. However, if you're running LL tests on _big_ exponents, 
memory shortage could be crucial. NTPrime is going to use somewhere 
between 16 and 20 megabytes of _physical_ memory to run LL tests on 
"10 million digit" exponents. I wouldn't try to run LL tests on 
exponents much bigger than about 10 million on a system running NT 
with 64 MB (or less) of memory, unless there really was only a very 
occasional need for the system to run anything else. If you're tight 
on memory, you may find that switching to double-checking assignments 
cures your problem.

I take it that you're not trying to do anything daft like running 
more instances of NTPrime/Prime95 (in total) than you have processors 
on the system. Running like this causes inefficiencies due to excess 
task switching and also consumes memory to no good purpose.

Also I still think it's worth checking for unwanted intrusions 
(BackOrifice seems fashionable at the moment) and also removing any 
services like FindFast which appear to have very little value - 
certainly FindFast can cause so much disk activity that things pretty 
well grind to a halt for a while - _except_ CPU soak programs like 
Prime95/NTPrime which can use the cycles which would otherwise be 
wasted since the other, higher-priority processes are all waiting for 
disk activity to finish. This is a _big_ price to pay for something 
which _might_ (if you're exceptionally fortunate) help you to open a 
MS Office document a split second faster.

BTW it's perfectly normal for the first instance of opening e.g. 
Windows Explorer to be slower than usual (though 20 seconds certainly 
sounds excessive!), since none of the various DLLs etc. required will 
be already in memory. On my system, during the first 5 minutes or so 
following a system boot, everything is unusually slow due to the 
activity of an anti-virus scanner. I'm prepared to live with this for 
the extra security it offers.

Defragmenting the disk can make a major improvement to the speed at 
which applications open - you really need to use a quality 
defragmenter like Diskeeper (Executive Software) since you want to 
get the swap/page file and the index (directory) files (plus the MFT 
on a NTFS volume) properly organized, as well making the data files 
contiguous. I also find that it helps to fix the page/swap file size 
(Control Panel/System/Performance) so that the minimum & maximum 
sizes are the same; the maximum of 128 MB and twice the system memory 
seems to be a sensible value to start with, if your system really 
needs more it will let you know! The idea is to prevent the page/swap 
file becoming excessively fragmented by cumulative allocation of 
extra chunks.

Hope this is of some use to you!


Regards
Brian Beesley
_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ      -- http://www.tasam.com/~lrwiman/FAQ-mers

Reply via email to