registered memory choices... please HELP!

[krw]

i'm truly sorry, but i have no clue
to this

Why should you be sorry? It appears that I've been led off chasing
geese, and perhaps a few of you folks too.

if i run across anything that might
give half a clue, i'll post it.
(don't hold yer breath; if anything,
you'll run across a solution before i do)

i'm not happy about saying this
for a fellow Tyan S2875 owner/user!

YOu haven't a problem? Have you looked in the BIOS message queue?
That would be good information (though I'm off on other things
right now).

if SUSE 10.0 works ok for me without
problems, i'll post it in the next
6 weeks (for whatever that might
be worth to you)

SuSE 10. should work fine for you, unless you do dual-head on a
Matrox card. :-( I'm trying to get my new toy tamed with the old.
Complicated setups don't work so well anymore. DualHead and
multiple systems with a KVM used to be easy. :-(

 

[willbill]

Why should you be sorry?

LOL. i might wish it on my
worst enemy, but no one on
this n/g (of course, i haven't
gotten into any battles here. )

i mean the whole thing is too weird

and i'm glad that you are the one
who's gone thru it and not me <mild
smile, but not a slam in any way
(i mean i *own* a 2875 from the
same production run as yours)>

i'll find out when i load SUSE 10.0
in the next 10 days

i agree with your last post, that
your ECC error is highly unlikely
to be your 1.5GB of Crucial memory

It appears that I've been led off chasing
geese, and perhaps a few of you folks too.




YOu haven't a problem?

no

not a single real ECC error (reported
in the bios event log)

one or two trivia things (that really
didn't mean anything), but NO ECC errors

at least with any hard boots into DOS
(the one included with Win98SE), nor
with Win98SE gui, nor with WinXP Pro, sp2

Have you looked in the BIOS message queue?

yes, also see above

That would be good information (though I'm off on other things
right now).



SuSE 10. should work fine for you, unless you do dual-head on a
Matrox card. :-(

that's kinda what i was thinking

meaning that it might be a flaw in the
bios s/w (i assume the current 3.02)
vs. your oddball Matrox video card

and no, i never advise anyone to start
flashing their mobo bios back and forth,
nor to flash to a beta. i do as little
of it as possible

I'm trying to get my new toy tamed with the old.
Complicated setups don't work so well anymore. DualHead and
multiple systems with a KVM used to be easy. :-(

last time i ran with Linux (Redhat?) was roughly
4.5 years ago

the one lasting memory from that was what
a tricky son of a gun it was to get the
gui video right

on a related note: i'm assuming that someone
at Tyan, or here on the n/g(s), suggested that
you try running your memory single channel
to see if that would stop the reported ECC
error (when you boot SUSE 10.0)?

if you thought of that on your own, then
i'm more than a little impressed.

bill

 

[Keith]

LOL. i might wish it on my
worst enemy, but no one on
this n/g (of course, i haven't
gotten into any battles here. )

i mean the whole thing is too weird

and i'm glad that you are the one
who's gone thru it and not me <mild
smile, but not a slam in any way
(i mean i *own* a 2875 from the
same production run as yours)>

I'm going to follow up your idea on the BIOS level. This was the
best idea I've heard in a long time. The whole system is down
right now (trying to get the laptop working and it objects to the
KVM[*]) so I can't look at what I'm running.

[*] It looks like I'm going to have to go USB for the keyboard and
mouse. The mouse is no problem (Logitech MX1000), but it means a
new keyboard. :-(

i'll find out when i load SUSE 10.0
in the next 10 days

i agree with your last post, that
your ECC error is highly unlikely
to be your 1.5GB of Crucial memory

I thought it rather comforting that Crucial was ready to RMA all
four sticks at the mere suggestion that something might be wrong,
though. They may be a little more expensive, but their service is
nothing short of amazing.

no

not a single real ECC error (reported
in the bios event log)

one or two trivia things (that really
didn't mean anything), but NO ECC errors

at least with any hard boots into DOS
(the one included with Win98SE), nor
with Win98SE gui, nor with WinXP Pro, sp2

I don't get it with anything other than once when SuSE boots (it
took a while to "get" that correlation).

yes, also see above



that's kinda what i was thinking

meaning that it might be a flaw in the
bios s/w (i assume the current 3.02)
vs. your oddball Matrox video card

The G550 isn't by any means an oddball. It's a rather standard
"business graphics" sort of card. I may have to replace it to get
back to DualHead on the Linux system, though with what I haven't a
clue. I'll dig through the SuSE compatibility notes to find one
that'll do what I need and hopefully not too expensive.

and no, i never advise anyone to start
flashing their mobo bios back and forth,
nor to flash to a beta. i do as little
of it as possible

Yepper! I flashed it once, a year or so ago, but I can't remember
the version. I thought I toasted the board then, but it was simply
a screwed BIOS. Resetting it to the defaults and them adding my
changes brought it back to life (whew!). I'll see if there is a
later version.

....

Oh, MY! It looks like there are a *ton* of fixes for the BIOS I
don't have. Including, perhaps, my SuSE issue with SATA. I'll bet
I'm back at 1.03.

last time i ran with Linux (Redhat?) was roughly
4.5 years ago

the one lasting memory from that was what
a tricky son of a gun it was to get the
gui video right

It was simple with SuSE 9.0. It was easier than WIndows, in fact.
It simply doesn't work in 10.0 (SuSE blames Matrox who ignores the
issue).

on a related note: i'm assuming that someone
at Tyan, or here on the n/g(s), suggested that
you try running your memory single channel
to see if that would stop the reported ECC
error (when you boot SUSE 10.0)?

It works fine, which is why chasing a DIMM got nowhere. There was
no way to isolate a single DIMM since I have unmatched pairs. Any
single-channel combination worked fine. DOS works fine. Win2K
boots fine (well I didn't install it successfully), but the CD
boots without ECC errors.

if you thought of that on your own, then
i'm more than a little impressed.

I've gone around the block a few times on this one. ;-/

Your BIOS thought is a good one though. The other thought I had was
a board problem that Linux uncovers then ChipKills. There was some
combination of chipkill that left me with an unbootable board (had
to reset CMOS to get it back).

The bottom line is that I'm bored with the problem. The system
works (modulo a KVM switch that doesn't), so I've left it alone. I
will flash the BIOS soon though.

 

[willbill]

Yepper! I flashed it once, a year or so ago, but I can't remember
the version. I thought I toasted the board then, but it was simply
a screwed BIOS. Resetting it to the defaults and them adding my
changes brought it back to life (whew!). I'll see if there is a
later version.

....

Oh, MY! It looks like there are a *ton* of fixes for the BIOS I
don't have. Including, perhaps, my SuSE issue with SATA. I'll bet
I'm back at 1.03.

i rather doubt "1.03" (way too old)

when you cold boot, given that you've got 1.5GB
of memory, you'll have several seconds to hit
the pause key on the keyboard

your current mobo bios version will be shown
on line 3 of the initial boot screen

mine reads:
Tyan Tiger K8W(S) BIOS 3.02

It works fine, which is why chasing a DIMM got nowhere. There was
no way to isolate a single DIMM since I have unmatched pairs. Any
single-channel combination worked fine. DOS works fine. Win2K
boots fine (well I didn't install it successfully), but the CD
boots without ECC errors.

I've gone around the block a few times on this one. ;-/

ah, of course. you took my suggestion and started
with a single 512MB dimm, which forced you into
single channel

your merit was in observing that it cured
your incidence of ECC errors on boot into SUSE

pretty doggone excellent

Your BIOS thought is a good one though. The other thought I had was
a board problem that Linux uncovers then ChipKills. There was some
combination of chipkill that left me with an unbootable board (had
to reset CMOS to get it back).

The bottom line is that I'm bored with the problem. The system
works (modulo a KVM switch that doesn't), so I've left it alone. I
will flash the BIOS soon though.

i can understand that. it clearly works, and you've
at worst got an annoying minor nuisance problem

kindly let us all know if your mobo bios
update to 3.02 solves your flakey ECC message
with boot into SUSE 10.0

if it doesn't, i'll post in maybe 9 days on
if i do or don't get an ECC error with my own
boot of SUSE 10.0

fwiw, my new S.M. dualie mobo also appears
to have a bios event log type readout on ECC
errors, but the pdf doesn't give specifics
(i'll have to wait till i get it running)

iow, if you didn't buy your "server" Tyan S2875
single for it's ECC ability (which i did), then
why did you buy it?

i mean, your single best reason

all ears, bill

 

[krw]

i rather doubt "1.03" (way too old)

Don't bet on it (like any rational person a BIOS upgrade is the
last thing one wants to do), though have no easy way of checking
tonight.

when you cold boot, given that you've got 1.5GB
of memory, you'll have several seconds to hit
the pause key on the keyboard

Good idea, but with none of the KVM stuff it's kinda hard to see.
;-)

ah, of course. you took my suggestion and started
with a single 512MB dimm, which forced you into
single channel

No, umm, I actually worked backwards removing a DIMM at a time
(less impact on my performance), and everything worked; flawlessly.
Then I tried removing another DIMM, same - same - same... As long
as it's single channel, no prob.

your merit was in observing that it cured
your incidence of ECC errors on boot into SUSE

pretty doggone excellent

No, pretty dumb, actually. It took months of part-time fiddling
and likely wearing out the DIMM mechanicals. I've since stopped
since I got nowhere. ...and it all works.

i can understand that. it clearly works, and you've
at worst got an annoying minor nuisance problem

Yep. It's getting into "carpentry season" so I'll have even less
time to fiddle. ;-)

kindly let us all know if your mobo bios
update to 3.02 solves your flakey ECC message
with boot into SUSE 10.0

Sure. I'll be at least next week. Gotta get a new keyboard and
KVM. :-(

if it doesn't, i'll post in maybe 9 days on
if i do or don't get an ECC error with my own
boot of SUSE 10.0

I doubt it now, with what you've said. Do let me know though.

fwiw, my new S.M. dualie mobo also appears
to have a bios event log type readout on ECC
errors, but the pdf doesn't give specifics
(i'll have to wait till i get it running)

iow, if you didn't buy your "server" Tyan S2875
single for it's ECC ability (which i did), then
why did you buy it?

It's not a "server" board, IMO. It's only a single processor board
(S2875S) and even the dual processor board only has memory on one.
It's a nice "workstation" board though. I likely wouldn't have
bought it if there were a better choice at the time (S939 wasn't
available).

I rather liked Tyan as a company (I saw some nice stuff from them
when I was doing this sort of thing) and my 1598C2 has been a
workhorse Windows system since '99. It's likely to be retired to
the basement since I've upgraded the Windows machine to a T60.

i mean, your single best reason

Tyan had (has?) the best after-purchase support of 'em all. I'd
have bought another recently, but they don't sell ThinkPads. ;-)

 

[willbill]

Tyan had (has?) the best after-purchase support of 'em all.

imho, they still do. they weren't able to update
the 2875 (single or multi) to Opteron dual core,
but they did provide new bios support up to 250
(90 nm only)

you have to read the supported CPU list carefully

i'm not totally sure if that applies to the 2875 dualie

I'd
have bought another recently, but they don't sell ThinkPads. ;-)

<smiling>

can't help you with laptops of any kind
coz they go to zero in less than half the
time it takes for desktops

bill

 

[willbill]

ok. if anything i owe you on a couple of counts
at this point. figure maybe a week for some input
(on chips n/g)

it's the SuperMicro H8DCE mobo, with the nVidia
nForce Pro (2200/2050) chipset, with two PCI
Express x16, two PCI Express x4, and 3 standard PCI
(lots of slots)

one other thing, it's one of 3 similar
mobos that S.M. offers. i'd never have
seen the other two (key feature is various
PCI-X slots) without having seen that
post by stow for the refs. for that matter,
stow is the one likely to post with real info
about using it when running it as a dualie
coz i'm only running with a single, single core
for maybe the next 6 months

anyhow, i got it running today. took an
existing 250GB HD with 29GB/4 partitions
of stuff on it, pruned it to 9GB/3 partitions;
kept DOS on it, got rid of Win98SE gui, and
loaded WinXP sp2

went ok

had a hang up with loading video drivers
for it (don't know why; error msg; no BSD),
and decided to do some maintenance before
going further (full backup, swap the new
back up disk in to make 100% sure that my
full backups are really working, add that 3rd
GB of memory, try switching to a different CPU
fan (running 61C with an Opty 248 (90nm);
37C with an Opty 142 (130 nm)) on my Tyan))

in general, i'm happy to say that i like this
S.M. board better than my Tyan S2875.

they both use AMI bios. H8DCE seems to have
a few more options than my older S2875.
of course it is the more expensive board

e.g. one that i like (and actually understand ),
is that i have 8 fan connectors (!) on the mobo
that can be monitored in the bios (on my Tyan
i've got 3 and only 1 works). plus, it gives
me 6 options to set fan speeds:
1) full speed 12V
2) optimized server w/3-pin fans
3) " w/s " " "
4) " server w/4-pin fans
5) " w/s " " "
6) " quiet

only thing i'm not sure about with the
S.M. is how any ECC errors are reported

with the Tyan there's resident space for
63 events. the S.M. bios allows me to
enable the option:
'MCA DRAM ECC Logging/Reporting'

odds are i'll try e-mailing S.M. in hopes
of getting some input on how any ECC errors
actually show up. i mean, is it a hidden
event log that only shows up when there
are errors? or maybe the thing is too new?
i.e. bios ver. 1.1, dated 12/31/'05, with
the board marked with a sticker on one
of the slots: "mfgd 3/'06"

bill

 

[George Macdonald]

one other thing, it's one of 3 similar
mobos that S.M. offers. i'd never have
seen the other two (key feature is various
PCI-X slots) without having seen that
post by stow for the refs. for that matter,
stow is the one likely to post with real info
about using it when running it as a dualie
coz i'm only running with a single, single core
for maybe the next 6 months

anyhow, i got it running today. took an
existing 250GB HD with 29GB/4 partitions
of stuff on it, pruned it to 9GB/3 partitions;
kept DOS on it, got rid of Win98SE gui, and
loaded WinXP sp2

went ok

had a hang up with loading video drivers
for it (don't know why; error msg; no BSD),
and decided to do some maintenance before
going further (full backup, swap the new
back up disk in to make 100% sure that my
full backups are really working, add that 3rd
GB of memory, try switching to a different CPU
fan (running 61C with an Opty 248 (90nm);
37C with an Opty 142 (130 nm)) on my Tyan))

Wow that's hot for a 90nm. Was that an AMD fan from a boxed retail unit?

in general, i'm happy to say that i like this
S.M. board better than my Tyan S2875.

they both use AMI bios. H8DCE seems to have
a few more options than my older S2875.
of course it is the more expensive board

e.g. one that i like (and actually understand ),
is that i have 8 fan connectors (!) on the mobo
that can be monitored in the bios (on my Tyan
i've got 3 and only 1 works). plus, it gives
me 6 options to set fan speeds:
1) full speed 12V
2) optimized server w/3-pin fans
3) " w/s " " "
4) " server w/4-pin fans
5) " w/s " " "
6) " quiet

only thing i'm not sure about with the
S.M. is how any ECC errors are reported

with the Tyan there's resident space for
63 events. the S.M. bios allows me to
enable the option:
'MCA DRAM ECC Logging/Reporting'

odds are i'll try e-mailing S.M. in hopes
of getting some input on how any ECC errors
actually show up. i mean, is it a hidden
event log that only shows up when there
are errors? or maybe the thing is too new?
i.e. bios ver. 1.1, dated 12/31/'05, with
the board marked with a sticker on one
of the slots: "mfgd 3/'06"

Sounds interesting - thanks for the info.

Now I'm wondering where Dell is going to get their 4S mbrds?

 

[willbill]

fwiw, that went ok. the corsair DDR 400
that i got isn't on the S.M. approved list
(for the H8DCE), but it is on Tyan's list.
the bios sees the 2GB of 400 1st, but also
notices the 1GB of 333 and during boot
the post messages show that the 3GB of
memory is running at 333

Wow that's hot for a 90nm.

yes, that's what i've been thinking too,
but i haven't noticed what kind of cpu temps
people post about for their 90nm Opty 940 cpu's

fwiw, my case temp is a fairly reasonable 39C
(as reported by the SuperMicro supplied monitoring
program); and the 550w power supply's cheap watt
meter shows that i'm drawing roughly 105-110w
(with so far: 1 HD, 1 nVidia 7600GT video board,
and the 3GB of low profile registered memory)

Was that an AMD fan from a boxed retail unit?

no. it's an oem AMD cpu, and i put an ASUS
X-MARS combo heatsink/fan on it. when it showed
up, the fan is only .5" thick, 70mmx2 at base,
80mmx2 at top, blows the air down into the
heatsink, and the fins position vertically
when in place on the cpu1 socket, so that half
the warm air hits the 1st stick of cpu1 memory

btw nice ASUS heat sink, with cheap, but effective,
threaded metal holes for the fan screws (in the
metal skirt on the heatsink top)

so i made the mistake of reversing the fan.
(so that the *memory* would run cool! <LOL at myself>)

when i reversed the ASUS fan, i had to file off
about 1/16" of plastic on one side of the 80mm side,
of the ASUS fan, to get it to fit the heatsink's
metal clamp down (the fan sits on top of it all)

the new fan (70mmx2 at base and top, .75" thick)
runs slightly faster and although not as wide as
the ASUS fan appears to move slightly more air
than the ASUS fan, but has only reduced the temp
by roughly 1-to-2C

so ok already, i'll go back to driving the air
down into the heat sink and see what that does
for the CPU's temp

good thing i went to an Ace hardware store,
after getting the 70mm fan, and got two extra
sets of different length screws (just in case
things came to playing around with fans and
which way to blow the air), coz at this point
i've used all 3 lengths



for one thing, slightly better fit and finish.
little things like the heatsink/fan on the
North Bridge being spring loaded (i didn't
notice it at first or i'd have taken it off
for a better look), as well as the heatsink
on the South Bridge being spring loaded

and the dirver CD set up better for Win XP (than
what Tyan's did for XP on my S2875); and the
s/w monitoring pgm being AOK (i had an OK one
on my K6-III Tyan board, but the one provided
by Tyan for their S2875 Opty board is poor)

Now I'm wondering where Dell is going to get their 4S mbrds?

the next 6 months should be especially interesting,
with lots of nice cpu improvements from both Intel
as well as some from AMD

ah, competition!

bill

 

[George Macdonald]

fwiw, that went ok. the corsair DDR 400
that i got isn't on the S.M. approved list
(for the H8DCE), but it is on Tyan's list.
the bios sees the 2GB of 400 1st, but also
notices the 1GB of 333 and during boot
the post messages show that the 3GB of
memory is running at 333

Hmm, I'm curious about this H8DCE - it says it requires an ATX 12V P/S
*but* with both a 4-pin *and* 8-pin 12V conenctors. I've never seen an ATX
P/S with the 8-pin connectro and the EPS12V ones seem to have just the
8-pin and no 4-pin - no?

 

[Keith]

Hmm, I'm curious about this H8DCE - it says it requires an ATX 12V P/S
*but* with both a 4-pin *and* 8-pin 12V conenctors. I've never seen an ATX
P/S with the 8-pin connectro and the EPS12V ones seem to have just the
8-pin and no 4-pin - no?

The SSI EPS12V spec requires the 24-pin "baseboard power
connector", the 8-pin "Processor power connector", and the 4-pin
"12V4 baseboard power connector" for systems requiring more than
16A of 12V. [*]

"Systems that require more then [sic] 16A of +12V current [sic]
to the baseboard will require this additional 2x2 power
connector.. This is due to the limited +12V capability of
the 2x12 baseboard power connector. +12V4 will power this
2x2 connector.

http://ssiforum.org/Power Supplies/EPS12V Spec 2_91.pdf

[*] The Tyan S2875 uses the 24-pin, 8-pin, and a Molex disk drive
connector. I have an ATX12V (20-pin and 4-pin) ATX-12V supply in
mine.

 

[George Macdonald]

Hmm, I'm curious about this H8DCE - it says it requires an ATX 12V P/S
*but* with both a 4-pin *and* 8-pin 12V conenctors. I've never seen an ATX
P/S with the 8-pin connectro and the EPS12V ones seem to have just the
8-pin and no 4-pin - no?

The SSI EPS12V spec requires the 24-pin "baseboard power
connector", the 8-pin "Processor power connector", and the 4-pin
"12V4 baseboard power connector" for systems requiring more than
16A of 12V. [*]

"Systems that require more then [sic] 16A of +12V current [sic]
to the baseboard will require this additional 2x2 power
connector.. This is due to the limited +12V capability of
the 2x12 baseboard power connector. +12V4 will power this
2x2 connector.

http://ssiforum.org/Power Supplies/EPS12V Spec 2_91.pdf

Ahah - thanks. Now if the mbrd & P/S mfrs would be a bit more explicit,
we'd know what to buy. I mean: SuperMicro just says ATX with the obviously
non-ATX combo of 4-pin + 8-pin connectors. I guess they'd rather you just
buy their chassis with the correct P/S.

[*] The Tyan S2875 uses the 24-pin, 8-pin, and a Molex disk drive
connector. I have an ATX12V (20-pin and 4-pin) ATX-12V supply in
mine.

Yeah I remember you saying something about the 4-pin molex disk drive
connector when you built it. So you have half of the 8-pin socket occupied
by the 4-pin?

 

[willbill]

Hmm, I'm curious about this H8DCE - it says it requires an ATX 12V P/S
*but* with both a 4-pin *and* 8-pin 12V conenctors. I've never seen an ATX
P/S with the 8-pin connectro and the EPS12V ones seem to have just the
8-pin and no 4-pin - no?

the whole power supply thing is confusing

as you point out in your 1st line, the H8DCE requires
"an ATX 12V P/S"

meaning they say nothing about needing a 24 pin
and an 8 pin and a 4 pin, but that's somewhat
built into the "12V" ATX and EPS specs

start at www.wikipedia.org with: atx

at the bottom of the def, it shows the history of
the different power supply names (some 5 or 6 of
them) and a line on the basic general specs as well
as several useful external refs if you want a good
pdf file on the main recent ones. you'll note that
the ATX 12V line only mentions a 20 pin <grin>

i simply went to newegg, pulled up all of the power
supplies, saved it to my PC as an htm file and (off line)
decided i wanted a "12V" one with a single 120mm fan
and 24/8/4 connectors and went with CoolerMaster 550
www.newegg.com/Product/Product.asp?Item=N82E16817171009

it meets these specs:
EPS 12V V2.1 and ATX 12V V2.01

per SuperMicro on page 1-12 of the manual, the power
supply has to meet ATX 12V V1.1 or above and have
adequate power

too bad they don't mention the 24/8/4 connectors
on that page.

bill

 

[krw]

George Macdonald wrote:


willbill wrote:


add that 3rd GB of memory,


fwiw, that went ok. the corsair DDR 400
that i got isn't on the S.M. approved list
(for the H8DCE), but it is on Tyan's list.
the bios sees the 2GB of 400 1st, but also
notices the 1GB of 333 and during boot
the post messages show that the 3GB of
memory is running at 333

Hmm, I'm curious about this H8DCE - it says it requires an ATX 12V P/S
*but* with both a 4-pin *and* 8-pin 12V conenctors. I've never seen an ATX
P/S with the 8-pin connectro and the EPS12V ones seem to have just the
8-pin and no 4-pin - no?

The SSI EPS12V spec requires the 24-pin "baseboard power
connector", the 8-pin "Processor power connector", and the 4-pin
"12V4 baseboard power connector" for systems requiring more than
16A of 12V. [*]

"Systems that require more then [sic] 16A of +12V current [sic]
to the baseboard will require this additional 2x2 power
connector.. This is due to the limited +12V capability of
the 2x12 baseboard power connector. +12V4 will power this
2x2 connector.

http://ssiforum.org/Power Supplies/EPS12V Spec 2_91.pdf

Ahah - thanks. Now if the mbrd & P/S mfrs would be a bit more explicit,
we'd know what to buy. I mean: SuperMicro just says ATX with the obviously
non-ATX combo of 4-pin + 8-pin connectors. I guess they'd rather you just
buy their chassis with the correct P/S.

[*] The Tyan S2875 uses the 24-pin, 8-pin, and a Molex disk drive
connector. I have an ATX12V (20-pin and 4-pin) ATX-12V supply in
mine.

Yeah I remember you saying something about the 4-pin molex disk drive
connector when you built it. So you have half of the 8-pin socket occupied
by the 4-pin?

Yes, and a 24-pin socket occupied by a 20pin ATX plug, to boot.
Confused the hell outta me when I bought the damned thing too.
Pluging a disk drive connector into the motherboard took quite a
leap of faith.