A7V rev 1.02 memory issues

G

Guest

Guest
Archived from groups: alt.comp.periphs.mainboard.asus (More info?)

My A7V v1.02 is having an odd problem, I can use 1 stick of memory
without an issue, but when I add a second (or third), memtest throws
errors like crazy. I've read 30+ pages back in this forum, tried
searching it, and have still not found the answer yet. I've tried all
3 VIO jumper settings (no difference) and I have 6 sticks of PC133
memory at my disposal to try. 2 Kingston, 2 PNY, 1 infineon and 1
micron. I flashed the BIOS to 1012beta just to see if it would help
(it didn't). All 6 sticks are known good, they all test perfect in
another system (together and individually). I tried memtest 3.1a and
memtest+ v1.26 (doesn't matter, they both throw the errors). Tried
setting it to PC100 or PC133, doesn't make a difference. Manually set
the timings to 3T/3T/3T also. CPU is a Duron 1300.

I'm at the end of my rope here, I'm about ready to throw this board
out the window and swear off Asus for good unless someone has some
suggestions for me :)

-GV
 

Paul

Splendid
Mar 30, 2004
5,267
0
25,780
Archived from groups: alt.comp.periphs.mainboard.asus (More info?)

In article <93hck09e4hgj0neku8liskev3pnot35gfv@4ax.com>, GlassVial
<glassvial@NOSPAM.hotmail.com> wrote:

> My A7V v1.02 is having an odd problem, I can use 1 stick of memory
> without an issue, but when I add a second (or third), memtest throws
> errors like crazy. I've read 30+ pages back in this forum, tried
> searching it, and have still not found the answer yet. I've tried all
> 3 VIO jumper settings (no difference) and I have 6 sticks of PC133
> memory at my disposal to try. 2 Kingston, 2 PNY, 1 infineon and 1
> micron. I flashed the BIOS to 1012beta just to see if it would help
> (it didn't). All 6 sticks are known good, they all test perfect in
> another system (together and individually). I tried memtest 3.1a and
> memtest+ v1.26 (doesn't matter, they both throw the errors). Tried
> setting it to PC100 or PC133, doesn't make a difference. Manually set
> the timings to 3T/3T/3T also. CPU is a Duron 1300.
>
> I'm at the end of my rope here, I'm about ready to throw this board
> out the window and swear off Asus for good unless someone has some
> suggestions for me :)
>
> -GV

You should be using groups.google.com to research the issue. I tried
A7V and "three sticks" as two search terms (incl. the double quotes
so that the two words three and sticks are found together).

http://groups.google.com/groups?threadm=Xns917FBA9956FBbobmarleycom%40216.221.81.119

You could also try private forums specializing in your board, like
a7vtroubleshooting.com .

Experience with RAM will vary from board to board, and with different
brands of memory DIMMs. The basic problem is, a shared memory bus has
poor signal integrity, so the speed versus number of sticks problem
will _always_ be an issue. (The upcoming Intel FB memory stick will
solve this problem once and for all, as Intel has some kind of
hub architecture for memory in the works.)

First of all, try to get two sticks running, using slot1 and slot3.
Leave slot2 blank for now. You may be able to run at full speed that
way.

To run three sticks, might require using single sided RAM (which of
course does nothing good for you in terms of total RAM on board - two
double sided sticks will give more total RAM). Using a double sided
in the slot furthest from the processor, and two single sided nearest
the processor, is another potential config, again doing nothing for
you in terms of max memory.

I have read accounts of this problem with later board models, and
occasionally someone will manage to get a board full of RAM running
at full speed, but it doesn't happen that often.

There are a couple of solutions to the problem.

One is boards that use registered memory. This is typically server
boards, but there are some desktop boards that can use registered
as well. Registered memory has a latch on the address bus, and its
main benefit, is the fact that the latch drives the address to the
memory chips, removing the heavy capacitive load from the address
bus. This allows the Northbridge to drive more memory loads without
a problem. The single electrical load of the latch chip causes much
less transmission line discontinuity than does 16 memory chips in
parallel sitting on the address bus.

A second solution, is for a motherboard to have multiple address
busses with the same info on them. The Athlon64 S754 processor does
this, and there have been a number of chipsets in the past that
have done this as well (Nforce2 is an example).

So, I wouldn't single out Asus as the only maker suffering from the
problem. If you know a bit more about the electrical nature of the
problem and its solutions, that can make finding a board that can
take a full load of DIMMs a bit easier.

For example, dual channel boards with four sockets, have two
DIMMs on one bus and two DIMMs on the other bus. That is one way
of buying a product with two address busses. However, you still
have to check Google, and read the manual for a given board,
before buying, because some chip designers still don't get it.
For example, the ATI chipsets don't seem to be up to the job,
based on some comments I've read yesterday in a Corsair forum.

HTH,
Paul
 
G

Guest

Guest
Archived from groups: alt.comp.periphs.mainboard.asus (More info?)

>You should be using groups.google.com to research the issue. I tried
>A7V and "three sticks" as two search terms (incl. the double quotes
>so that the two words three and sticks are found together).
>
>http://groups.google.com/groups?threadm=Xns917FBA9956FBbobmarleycom%40216.221.81.119

Yeah, looking, not really finding much useful info yet.

>You could also try private forums specializing in your board, like
>a7vtroubleshooting.com .

Already done, nothing useful yet.

>Experience with RAM will vary from board to board, and with different
>brands of memory DIMMs. The basic problem is, a shared memory bus has
>poor signal integrity, so the speed versus number of sticks problem
>will _always_ be an issue. (The upcoming Intel FB memory stick will
>solve this problem once and for all, as Intel has some kind of
>hub architecture for memory in the works.)

Lets hope so :)

>First of all, try to get two sticks running, using slot1 and slot3.
>Leave slot2 blank for now. You may be able to run at full speed that
>way.

I had a combination working that way, I forget what it is now, if I
find it again I'll write it down. It doesn't solve my problem of
needing to get 3 sticks working in this mobo, however.

>To run three sticks, might require using single sided RAM (which of
>course does nothing good for you in terms of total RAM on board - two
>double sided sticks will give more total RAM). Using a double sided
>in the slot furthest from the processor, and two single sided nearest
>the processor, is another potential config, again doing nothing for
>you in terms of max memory.

Trying that as I type this. Perhaps this board prefers single-sided
memory...I'm not certain. Unfortunately, if that's the case, I don't
have enough single-sided sticks to make a total of 3. I only have 2
single sided sticks the rest are double-sided. Right now it's on test
#4 with the single sided stick closest to the CPU and hasn't thrown a
million errors yet...interesting. It would have by now in any other
configuration, however I'm not holding my breath! It'll probably just
fail at a later test. I'll post again after I test some more.

>I have read accounts of this problem with later board models, and
>occasionally someone will manage to get a board full of RAM running
>at full speed, but it doesn't happen that often.

Not encouraging news :p

>There are a couple of solutions to the problem.
>
>One is boards that use registered memory. This is typically server
>boards, but there are some desktop boards that can use registered
>as well. Registered memory has a latch on the address bus, and its
>main benefit, is the fact that the latch drives the address to the
>memory chips, removing the heavy capacitive load from the address
>bus. This allows the Northbridge to drive more memory loads without
>a problem. The single electrical load of the latch chip causes much
>less transmission line discontinuity than does 16 memory chips in
>parallel sitting on the address bus.

Well I'm not sure if this board can take ECC or not, I've never tried
it, and none of my sticks are ECC. I'd also rather not pour any money
into this old system if I don't have to.

>A second solution, is for a motherboard to have multiple address
>busses with the same info on them. The Athlon64 S754 processor does
>this, and there have been a number of chipsets in the past that
>have done this as well (Nforce2 is an example).
>
>So, I wouldn't single out Asus as the only maker suffering from the
>problem. If you know a bit more about the electrical nature of the
>problem and its solutions, that can make finding a board that can
>take a full load of DIMMs a bit easier.
>
>For example, dual channel boards with four sockets, have two
>DIMMs on one bus and two DIMMs on the other bus. That is one way
>of buying a product with two address busses. However, you still
>have to check Google, and read the manual for a given board,
>before buying, because some chip designers still don't get it.
>For example, the ATI chipsets don't seem to be up to the job,
>based on some comments I've read yesterday in a Corsair forum.
>
>HTH,
> Paul

Just for what it's worth, I've put all these sticks in my Abit board
(same chipset, too I might add!) in any combination and run full
memtests without incident.

Thanks for the info :)

-GV
 
G

Guest

Guest
Archived from groups: alt.comp.periphs.mainboard.asus (More info?)

>First of all, try to get two sticks running, using slot1 and slot3.
>Leave slot2 blank for now. You may be able to run at full speed that
>way.
>
>To run three sticks, might require using single sided RAM (which of
>course does nothing good for you in terms of total RAM on board - two
>double sided sticks will give more total RAM). Using a double sided
>in the slot furthest from the processor, and two single sided nearest
>the processor, is another potential config, again doing nothing for
>you in terms of max memory.

This is absolutely maddenning. I ran the test and it passed 7 times
in a row. I reboot the machine and run the same test, and memtest has
errors. WTF?!!!!

-GV
 

Paul

Splendid
Mar 30, 2004
5,267
0
25,780
Archived from groups: alt.comp.periphs.mainboard.asus (More info?)

In article <7koek05bbfqttkel08i4dboqeb7ngh10t7@4ax.com>, GlassVial
<glassvial@NOSPAM.hotmail.com> wrote:

> >First of all, try to get two sticks running, using slot1 and slot3.
> >Leave slot2 blank for now. You may be able to run at full speed that
> >way.
> >
> >To run three sticks, might require using single sided RAM (which of
> >course does nothing good for you in terms of total RAM on board - two
> >double sided sticks will give more total RAM). Using a double sided
> >in the slot furthest from the processor, and two single sided nearest
> >the processor, is another potential config, again doing nothing for
> >you in terms of max memory.
>
> This is absolutely maddenning. I ran the test and it passed 7 times
> in a row. I reboot the machine and run the same test, and memtest has
> errors. WTF?!!!!
>
> -GV

Do you have any reason to suspect the board has other problems ?
How about the power supply ? Is it possible the power supply
is not behaving the same way after the reboot ? There is no
reason for the test results to change between runs, unless some
external stimuli is changing, and that could be temperature
(unlikely), voltage (possible), or a change in the way that
the BIOS is initializing the system (huh?). Only voltage is
a reasonable explanation, either PSU voltage or an onboard
regulator is not stable or is not meeting the required
voltage value.

Do you have a way of measuring Vdimm ? Maybe it is dropping
in value when three DIMMs are installed.

HTH,
Paul
 
G

Guest

Guest
Archived from groups: alt.comp.periphs.mainboard.asus (More info?)

>Do you have any reason to suspect the board has other problems ?
>How about the power supply ? Is it possible the power supply
>is not behaving the same way after the reboot ? There is no
>reason for the test results to change between runs, unless some
>external stimuli is changing, and that could be temperature
>(unlikely), voltage (possible), or a change in the way that
>the BIOS is initializing the system (huh?). Only voltage is
>a reasonable explanation, either PSU voltage or an onboard
>regulator is not stable or is not meeting the required
>voltage value.
>
>Do you have a way of measuring Vdimm ? Maybe it is dropping
>in value when three DIMMs are installed.
>
>HTH,
> Paul

I don't have any reason to believe the board has other problems,
ironically Windows itself seems to be fine. I put a brand new PSU in
as part of my troubleshooting methods, made no difference. I'm
beginning to think that perhaps these memtest errors are like "false
positives" because it seems that Windows itself is running fine.
Measuring Vdmm, where, exactly? With a multimeter? If so, where?

-GV
 

Paul

Splendid
Mar 30, 2004
5,267
0
25,780
Archived from groups: alt.comp.periphs.mainboard.asus (More info?)

In article <6nggk0908gev37k2omhuue723sqcpdp8rb@4ax.com>, GlassVial
<glassvial@NOSPAM.hotmail.com> wrote:

> >Do you have any reason to suspect the board has other problems ?
> >How about the power supply ? Is it possible the power supply
> >is not behaving the same way after the reboot ? There is no
> >reason for the test results to change between runs, unless some
> >external stimuli is changing, and that could be temperature
> >(unlikely), voltage (possible), or a change in the way that
> >the BIOS is initializing the system (huh?). Only voltage is
> >a reasonable explanation, either PSU voltage or an onboard
> >regulator is not stable or is not meeting the required
> >voltage value.
> >
> >Do you have a way of measuring Vdimm ? Maybe it is dropping
> >in value when three DIMMs are installed.
> >
> >HTH,
> > Paul
>
> I don't have any reason to believe the board has other problems,
> ironically Windows itself seems to be fine. I put a brand new PSU in
> as part of my troubleshooting methods, made no difference. I'm
> beginning to think that perhaps these memtest errors are like "false
> positives" because it seems that Windows itself is running fine.
> Measuring Vdmm, where, exactly? With a multimeter? If so, where?
>
> -GV

First, start with a SDRAM module datasheet, as that will give you
the pinout of the DIMM slot.
http://download.micron.com/pdf/datasheets/modules/sdram/SD8_16C32_64x64AG.pdf

6, 18, 26, 40, 41, 49, 59, 73, 84, 90, 102, 110, 124, 133, 143, 157, 168
VDD Supply Power Supply: +3.3V ±0.3V.

On page 3 of the document, the red pins are +3.3V.

With the computer unplugged, remove a DIMM from its socket, then use
the ohmmeter scale on your meter, and trace from a VDD pin, to the
circuit with the two toroids in the upper right hand corner of the
motherboard. I think that could be VIO, and the output filter
consists of a toroid and one or more electrolytic caps. The
output would look like this:

Push/Pull ---Toroid---+-----------> VIO - To VDD on DIMM,
MOSFETs | AGP and PCI etc.
Cap.
|
GND

The only reason for using the ohmmeter, is to probe for a point
other than the DIMM socket, so you can measure VIO while all
three DIMM slots are populated, and there is no danger of
shorting something on the DIMM. The VIO should be available
somewhere in that pile of regulation circuitry in the upper right
hand corner of the board. That is where I would start.

I wish I had some kind of Athlon reference schematic, but I
don't have one in my collection.

Paul