Re: registered memory choices... please HELP!

On Fri, 05 May 2006 07:57:13 -0500, willbill <trek@xxxxxxxxxxxxx> wrote:

George Macdonald wrote:

On Thu, 04 May 2006 15:38:18 -0500, willbill <trek@xxxxxxxxxxxxx> wrote:

Crucial's got a good name, Tyan has it on their
approved 333 list (Micron chips) and the price
is right. i'm going with 2 of their $106 1GB
sticks (3GB total on the machine)


I assume you meant Corsair above...


you are right. good catch


and are you sure it's the regular
Corsair stuff which is approved by Tyan


yes, it's the regular stuff. the model number
on newegg identically matches what Tyan shows
on their approved memory list


and not the Corsair "System Select"
modules, which have a price more in line with Crucial's and Kingston's.

Note that Crucial has a higher price on DDR333 ($200.) than for their
DDR400 ($176.). I guess DDR333 demand is so low and it's easier to just
produce one speed grade, DDR400, which would work OK at DDR333.

The Corsair at $106. seems like an unusually good buy but in that case I
think I'd still go for the Corsair DDR400 at $107. and run it at DDR333 -
might even have a bit more leeway on timing. As long as you can get it
exchanged if you have problems, looks like maybe the way to go.


that's an interesting thought, George.
thank you for that. :)

in fact, it's so interesting that i'm going with it. :)

i didn't even think to look at the pricing
on the DDR400. my bad

when i went back (this past few minutes),
i noted their DDR numbers: 266 has 40 choices,
333 has 84 choices, and 400 has 250 (!)

bill

p.s.

i've noted that if one wants to run huge
amounts of memory (i don't), you have to run
at 266 (for some reason that i don't fully
understand)

Yep it's just the loading on the bus signals: more chips means more load
which can only be compensated up to a point and eventually you just have to
lower the clock as signals degrade. Note that later AMD64 chips, starting
with rev E IIRC, have an improved memory controller which is better in this
respect. It's also true that some people have reported exceeding AMD's
recommendations and have managed to run at DDR400 with all ranks full. You
can check AMD's specs in Chapter 4 of the BIOS & Kernel Developer's Guide,
26094.PDF.

fwiw, my new S.M. dualie mobo is spec'ed at
running up to 32GB (with with both CPU
sockets running). but i don't ever plan
to use the 2nd CPU socket, so i'll never
get there. the only reason i can think
of for running with truly large amounts
of memory is if you're running the machine
as a true server. to me, 3GB on a personal
use machine is a lot of memory

I'll be interested to hear how the SuperMicro mbrd works out. Which
chipset is it? They're still oddly umm, shy about revealing their mbrd
"support" of AMD CPUs - I guess they don't want to lose Intel
"benefits".:-)

--
Rgds, George Macdonald
.

Relevant Pages

  • Re: nfs-server silent data corruption
    ... because of the memory you added. ... Each CPU contains its own memory controller (just like all AMD's ... More like one chipset consisting of several physical chips. ... Since it is always the case that each PCI-E slot or PCI-E device sits ...
    (freebsd-net)
  • Re: nfs-server silent data corruption
    ... because of the memory you added. ... Each CPU contains its own memory controller (just like all AMD's ... More like one chipset consisting of several physical chips. ... Since it is always the case that each PCI-E slot or PCI-E device sits ...
    (freebsd-stable)
  • Re: 64-bit PolyML
    ... Ah -- so you weren't running the 62-bit code in the same CPU. ... bandwidth in the CPU and on the data bus and cache memory, ... A factor of 1.25 (MLton) is ... If the price is too high for someone - ...
    (comp.lang.functional)
  • Re: EFI/console general question
    ... >VAX board using already fabricated Microvax II chips, but put in an EFI ... branch to a specific memory location is a pretty common way of starting. ... This means the CPU HW doesn't need to know how ...
    (comp.os.vms)
  • Re: Which Sun Blade?
    ... big difference in price between ... the 900 and 1.2 CPU ... (some with memory some without, ... all without hard disks and most had ...
    (comp.sys.sun.hardware)