Re: OpenServer 6.0.0 hangs randomly. Please help!

Brian K. White wrote:

From: "Bill Campbell" <bill@xxxxxxxxxxxxx>

On Tue, Aug 21, 2007, Bill Vermillion wrote:

The HW vendor did NOT follow my specs and had a machine that would
not recognize nor use ECC memory.

Bela and others went into long rants when Intel and others were pushing
the idea that we don't need parity checking RAM.

Actually why weren't they right?
Yes you need more robust memory solutions, but why couldn't you impliment
the parity checking and-or error correcting in the chipset using extra
sticks of any-old ram instead of building it into each stick of ram?
Ram-raid as it were. Which I think they have actually. That would be quite a
big chunk of overhead off our backs and better for everyone if a whole type
of ram and it's design and manufacture chain went completely away and those
resources just went into making more regular ram.

Such a scheme would have met my requirements -- it's still a parity or
ECC scheme, even if the memory sticks you install into the machine don't
have extra parity bits.

But... your scheme basically wouldn't work. It would limit the
amount of RAM in the machine to the amount that could be covered by
the parity/ECC RAM bits in the chipset (so you would have to buy a
fancier chipset model for a larger machine). Worse, you would need
a different speed of chipset internal parity/ECC RAM for each speed
of external memory stick (look how much trouble ensues when you use
mismatched sticks in a single machine, even if they're the _same_
speed and possibly even the same SKU, but different batches from the
manufacturer...) Worst of all, the timing for it just wouldn't work
very well and you'd end up with a flaky design.

And all that just to save 1 bit in 9. It's not "quite a big chunk of
overhead". If parity/ECC RAM is more expensive than regular RAM, it
certainly isn't because of the 1/9 extra hardware, it's due to low
demand because the chipset and system designers don't use it.

(I was going to say something here about how a certain large
hardware-spec-setting OS vendor should push parity or ECC for all
systems (down to desktops & laptops) so they could show us how much of
the instability really isn't their fault at all, it's the memory -- then
it occurred to me that perhaps it wouldn't make much difference and
would only further highlight the true source of instability...)

Bela<
.

Relevant Pages

  • Re: P4T533-C unstable with 1.5 GB RAM
    ... chipset limitation, so I downloaded the datasheet for the 850E Memory ... P4T533-C unstable with 1.5 GB RAM ... There is a limit on the number of CHIPS that the 850e (or ...
    (alt.comp.periphs.mainboard.asus)
  • Re: D975XBX2 + 8 GB RAM = unstable :-(
    ... But since Kingston RAM tanked as well in the same setup I ... In reading the manual for the Asus 975X chipset version of this board it ... ECC and non-ECC DDR2 800/667/533 MHz memory ... "Memory Remap Feature" in the bios where the Intel boards with ...
    (comp.sys.intel)
  • Re: RAM question
    ... Computer' does show 512MB of RAM. ... Timing Settings' to an 'AGP Aperture Size' of 256MB i.e half my total ... Physical Memory ... If you plug a 512MB DIMM into a chipset ...
    (microsoft.public.windowsxp.hardware)
  • Re: Credit Cards (was :Re: HUGE plot hole in Harry Potter: DH (Spoilers))
    ... and if you still run 98 the issues with memory will be slightly less ... up the RAM to half a gig or so. ... my firefox regularly swells to over 500 megs by itself ... On the other hand, P2/P3 233-550 machines typically run with a BX chipset, ...
    (rec.arts.sf.written)
  • Re: RAM question
    ... screen and didn't know that 128MB of RAM ... Physical Memory ... If you plug a 512MB DIMM into a chipset ...
    (microsoft.public.windowsxp.hardware)
Loading