From: Adir Abraham (adir_at_nonexisting.hamakor.org.il)
Date: Fri 03 Sep 2004 - 12:32:59 IDT
On Thu, 2 Sep 2004, Omer Zak wrote:
> Unfortunately, a bay for swapping hard disks prevents installation of
> memory module in two of the three memory sockets in the motherboard, so
> I could not try to use the PC-2100 256MB memory module. So the question
> whether they would have cooperated with the motherboard is still open.
Are you sure that you can't live without the bay? try to find a reasonable
place at the bottom of the iron (if there's nothing else stuck in there),
or even below the floppy drive (an exotic place, but works at most cases).
> By the way, according to memtest, for my 1.7GHz P4 processor, the speeds
> are as follows:
> L1 cache (8K) - 13933MB/Sec (both 256MB and 512MB configurations)
> L2 cache (256K) - 11886MB/Sec (both configurations)
L1 and L2 has nothing to do with your memory modules. They sit at the core
and near-core of your CPU.
> Memory - 819MB/Sec for the PC-2100 module
> 809MB/Sec for the PC-2700 module
>
> In other words:
> 1. I lost 1.2% of the memory bandwidth. Reasonable price to pay for the
> benefit of reducing to zero my swap space usage.
I doubt that you really lost anything. It can fall on a benchmark mistake.
1.2% difference is a reasonable one. In general, you stayed with the same
(maximum) bandwidth which you can acheive with your motherboard.
> 2. The L1 and L2 cache bandwidths are close to each other. This is
> contrary to my intuition.
They are not so close. around 2000MB/sec (~2GB/sec) differs between them.
However, I don't remember what the exact speed and relations should be.
I'll check that out later if you wish.
Adir.
>
>
> Ira Abramov wrote:
> > Quoting Omer Zak, from the post of Tue, 31 Aug:
> >
> >>Thus, if the PC-2700 cannot work with my motherboard (with or without
> >>PC-2100), I'll lose. This is why I asked in this mailing list.
> >
> >
> > well, I have ran together 66Mhz and 100 in the same board and 100 with
> > 133 and it was just fine. 133Mhz dimms in a 66Mhz bus have mixed results
> > depending on make.
> >
> > I would look for a store that would be sympathetic to your problem, and
> > let you return the module if it fails after a few minutes or hours of
> > memtest or memtest+. all other methods that do not include such an
> > empiric test are educated guesses and quoted manufacturer's best
> > practices at best.
> >
> > I say mix and match. Hail Discordia!
>
> --- Omer
> My own blog is at http://www.livejournal.com/users/tddpirate/
>
> My opinions, as expressed in this E-mail message, are mine alone.
> They do not represent the official policy of any organization with which
> I may be affiliated in any way.
> WARNING TO SPAMMERS: at http://www.zak.co.il/spamwarning.html
>
>
> =================================================================
> To unsubscribe, send mail to linux-il-request_at_linux.org.il with
> the word "unsubscribe" in the message body, e.g., run the command
> echo unsubscribe | mail linux-il-request_at_linux.org.il
=================================================================
To unsubscribe, send mail to linux-il-request_at_linux.org.il with
the word "unsubscribe" in the message body, e.g., run the command
echo unsubscribe | mail linux-il-request_at_linux.org.il
This archive was generated by hypermail 2.1.7 : Fri 03 Sep 2004 - 12:43:22 IDT