Jetway 830 CF v3.0

[JEWilson]

Odd issue with this

Rec'd board dead - was recap city
Most of the caps required replacement

Anyway, recapped these with Pana FM

Got some memory from eBay
2 of Crucial 512MB PC133 ECC Unbuffered

Using MemTest86 v3.3 to test the RAM
1st stick inserted DIMM slot 1 - Ran Test 5 and 8 for 20 iterations
Took some time - but tested ok
2nd stick inserted DIMM slot 1 - Ran Test 5 and 8 for 20 iterations
Tests ok

Inserted both DIMMs in slot 1 and slot 2
Test 5 fails - get bad memory address recurrently

I suspect this may be more of a problem with the mobo as opposed
to memory.
Thinks I'll re-check the traces and cap ESR on the DC-DC regulator
for the RAM

Any suggestions?

Thanks

[JEWilson]

Resolved

There were six other caps on the board that I missed - 10V 1000uF
Replaced these

Checked for integrity of solder joints on old work
Funny, with this board had to pay attention to physical solder connections.

The mobo is a bitch to get a good joint!

Anyway, after the work done
Back to testing the memory - so far no errors - this is a good sign

[Big Pope]

Good job!! I also have some same experience with Jetway 830CF in my side

[JEWilson]

Hmmm...

Seems I was premature to state this mobo is problem free.
Under test, the memory is reporting errors again.

Always appears to be the first MB, errors appear at 0.6 and 0.9 MB
Odd.

I'm beginning to suspect the DIMM slots as the number of errors
change when I reseat the DIMMs but errors remain 0.6 and 0.9 MB.
The errors range from a bit to serveral bits in each byte reported
as in error.

As the memory was Crucial/Micron, I have sought to get these
replaced under warranty.

I will test it again, when I receive the replacement DIMMs.

I only want to keep this mobo for testing a SIS chipset

There was an early bug with SIS730 and MemTest86 v1.26
see further http://forum.x86-secret.com/archive/...hp?t-2404.html

I suspect this has been corrected of course.

Anybody have comments to add here?

[Fizzycapola]

On my first board recap after a few months one of the capacitors developed a wobble, which was fixed by reheating with my new 60W iron. Presumably thermal cycling had disbonded the soldering. I find some difficulty in soldering can be solder related, at the moment I'm using some professional solder I found on ebay from a home seller advertising as something from their work which contains 4% silver, copper and tin with 1% non poisonous flux, not sure if it's no clean but at 1% it doesn't seem to splooge all over the place or be visible I just leave it, happier than the 60/40 and cheap ebay silver bearing solder, I was using before but the joints don't shine atleast the solder seems to go in the right places so far. Also got some samples of eutectic solder which seems ok. All this is done with a real 60W rated medium size pencil tip which doesn't fall to pieces like some tips such as the badly shapen Silverline (fake?)ones I bought on ebay, blunt enough not to scratch the board but narrow enough to get in the places needed.

Having a look in BIOS to see voltages, you want then to stay within 5%. If they do perhaps it's just RAM has come to the end of it's useful life.

[PCBONEZ]

I know the chipset supports ECC but that doesn't mean the BIOS does.
-
There is no mention of supporting ECC in it's manual and in general that means it doesn't.
.

http://www.jetway.com.tw/jw/download...roductid=830CF

.

[gdement]

It seems if the sockets were bad then it shouldn't matter if both modules were installed. Sounds more like the motherboard just isn't handling the load from multiple modules. But maybe marginal contacts could contribute to that since essentially a bad contact would manifest itself as more load at that pin.

You could try setting manual RAM timings, just in case the BIOS is goofing them up somehow.

It's also possible one of your modules has lost reasonable tolerances. I've had that happen with an SDRAM module - it was causing errors only on certain boards when loaded with multiple modules. But it was the only module that would create this behavior, and it always worked fine by itself. So apparently that module just didn't have decent tolerances anymore.

Does the faulty address change if you reverse the position of the modules?

[JEWilson]

Thx

First, there is a setting in the BIOS to support parity checks.

ECC was working fine so far as, formerly, I was running FSB 100MHz
with both RAM slots populated - had no problems.

I'll contact Jetway to make sure ECC support is included

Will wait until get RAM replaced

Thanks again, for the input

[JEWilson]

Tried some PC133 ECC Unbuffered had lying around.

First tried a mached pair of non-branded 128MB.

No errors, none, zero.

Must have been broken Crucial RAM - probably old and out of tolerance.

These have been returned under warranty.
Expect to receive new replacement RAM soon.

[JEWilson]

Well...

Got the new replacement RAM quick! Thx Lexar-Crucial

As I am testing this RAM at the moment and those tests are
just about complete, I find the RAM is problem free.

I have tested DIMM slot 2 independently and there are no issues.

However, with both the DIMM slots populated, I recurrently get
errors in the first megabyte of RAM and these appear to be
only single or double bit errors as before.

This persists even after exercising all the permutations with modules and
DIMM population.

The number of errors reported varies also.

I'm beginning to suspect it may be a bug in MemTest86 v3.3 with
this chipset, CPU and RAM combination.

I believe v3.4 has been released for MemTest86.

Will report findings in due course
Will try this as well as alternative memory testers.

[JEWilson]

Still reporting errors even with a different memory tester

Q - If populated with 2 of matched 128MB ECC Unbuff
Both DIMMs populated - no errors, none

Q - If populated with 2 of matched new Crucial 512MB ECC Unbuff
Both DIMMs populated - errors always DIMM Slot 0
as described

Per chance, is this a power problem?
Replaced all the caps in the memory regulation area with Pana FM
It seems obvious, the power requirements for two DIMMs would
be constant irrespective of module density

I'm only guessing - it may a bad connection with a cap on the mobo
Dunno...

First, I'll check the replaced caps for mechanical integrity (again)
Next, I'll check the ESR on the original caps that remain

Anything else which could be considered suspect?

Thanks for your consideration

[zandrax]

I think there is something preventing memory access, perhaps dust/oxide in ram slot: clean them with a spray deoxidiser and insert and remove ram modules 2-3 times in order to remove oxide through friction. Clean ram contacts with a vinil eraser, such as a Staedtler Mars Plastic one.
If errors still appers after cleaning, the motherboard has trouble addressing all ram: I don't know if you should blame capacitors or bad pcb tracks, but try moving the module from first to, e.g., third ram slot and watch what happens. Set the most conservative timings in bios too.

Zandrax

[JEWilson]

The mobo only has two slots for DIMMs

As to the matter of power;
The BIOS reports all rails ok in terms of voltage with the
exception of +5V. This hovers around 4.84 - 4.86V but is within regulation.

Please note, these are readings from the systems health monitor and
have not been measured with a mmeter.

The board is populated thus;

1 of AGP4x 64MB ATI FireGL8700 Graphics
1 of Highpoint ATA IDE RAID PCI Card
1 of SoundBlaster Live 5.1! PCI Card
1 of SoundBlaster O/P Card Riser
1 of CD-R
No HDDs yet

PSU is;
1 of new X-Power ATX-500TD rail O/P as found on PSU label;
+12.0V : 20.0 A max.
+ 5.0V : 36.0 A max.
+ 3.3V : 36.0 A max.
- 5.0V : 1.0 A max.
-12.0V : 1.0 A max.
+ 5.0V SB : 2.0 A max.

Q - is the low (relatively speaking) +5V reading indicative of, perhaps
a problem with low load, that is no HDDs etc?

[PCBONEZ]

Most PSU's add +5v and +12v feedback together to make one feedback signal to controller IC.
Controller IC controls both together (one correct signal) and tries to keep both in spec..

That MoBo probably powers CPU off +5v. (Guess)
+12v not loaded. (No drives) So stays normal.
+5v is loaded by CPU.
+5v goes lower due to being loaded and IC raising back it is limited because +12v is not low too. (Total error signal not that big with just 5v a little bit low and 12v not low at all.)

.

[JEWilson]

My further adventures with this mobo...

It appears, now, to me, to be an addressing issue

How have I come to this conclusion?
I have just tested the following setup with the DIMM Slots.
1 of 128MB unbranded ECC Unbuff in Slot 1.
1 of 512MB new Crucial ECC Unbuff in Slot 2.
Ran test 5 & 8 for 20 iterations (takes a long time).
No errors reported and I decided to use MemTest86 v3.4.

Q - Might this be an addressing problem?

I had an issue such as this back in the 80's when I was doing
hardward R & D with IBM PC-XT clones (y - that long ago).
The problem in that situation was the system RAM was
including the Video RAM Card as its addressable memory.

It may even be memory organisation is a problem when
combined in the problematic manner. My intuition furthermore,
finds me wondering about the AGP cache set aside for
the AGP adapter (not the video memory of the card itself).

I'm not suggesting this is the case here, rather, it
appears it does not like the highest address (1st MB)
when populated with 2 x 512MB ECC Unbuff - odd....

As I don't have a logic analyser to hand (tsk.. tsk..) or a
high speed scope (bah... humbug), I might have a chat with the tech at
Jetway (who confirmed the mobo supports ECC Unbuff) and
redirect him to this thread to see if he/she can offer any
input.

Any views?

[PCBONEZ]

Wild guess,
... and I'm not much for key words and tricky phrases today....

Along the lines of density,
,, the chipset supports 6 'rows' of memory max.

Perhaps the 512MB are configured as 4 rows each.

.

[JEWilson]

Sigh...

The tech from Jetway who originally told me the
mobo supports ECC Unbuff is now relating to
me the mobo does not support ECC Unbuff.

Will require to get some PC133 non-ECC Unbuff or registered!

[PCBONEZ]

Originally Posted by JEWilson

- If populated with 2 of matched 128MB ECC Unbuff
Both DIMMs populated - no errors, none

From what I've heard/read others say about Jetway support they'll tell you anything that comes to mind (without checking) just to get you off the phone. If it doesn't work then the next time you call they reverse it.

- You've already tested with ECC and ECC worked fine.

I think it's the configuration of those 512Mb modules.
(Memory width and depth.)
,, But either way you need different modules so play safe and get standard/universal low density.

.

[JEWilson]

Thx for that

That is unbelievable - no wonder they have a bad reputation.
If that is a how they provide service desk, then they will get
nowhere!

Collecting fault infornation is part and parcel of decent problem
resolution.

Well... won't be buying Jetway again, ever, period.

[gonzo0815]

Below is what Crucial.com adviser would suggest:


# 168-pin DIMM Banking: 2 (2 banks of 1)
# Chipset: SiS 730S
# Error Detection Support: Non-ECC only
# Max Unbuffered SDR SDRAM: 1024MB
# Module Types Supported: Unbuffered only
# Max Component Density: 256Mb
# SDR SDRAM Frequencies: PC100 and PC133
# Graphics Support: AGP 4X
# Supported DRAM Types: SDR SDRAM only
# USB Support: 1.x Compliant
I thin most important issue is chip density, but i think the bios ´would not recognize the full memory if that is the problem.