 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Nobody likes monday mornings, eh?
Well on monday afternoon, I got home to find that rather than rendering
all day, my PC had stopped with an unusual BSoD message. (0x0000004E)
Anyway, I rebooted it, and it was behaving quite strangely. Since the
temperature in my bedroom was an insane +30°C, I assumed it was just
malfunctioning due to overheating.
Yesterday I managed to get the room temperature down to a slightly more
sensible +24°C, but still the computer misbehaves. So I ran a memory
test... and it almost immediately spews out thousands of errors. So it
seems my PC now has an actual hardware fault somehow. (!)
What this tells me is that the program wrote some data to memory and
then read it back, and it wasn't the same. This indicates a fault with
either the RAM itself, the motherboard, the CPU cache, or just some kind
of obscure CPU fault that makes it execute instructions wrong.
I tried running tests with the CPU cache disabled, and I still get
identical errors, which leads me to believe it's not any kind of CPU
fault. (Which is just as well; I don't imagine anybody still sells
Socket-939 CPUs...)
Next trick is to start swapping RAM modules and see if that fixes it. I
don't remember what kind of RAM this sucker takes, but given that even
obsolete PC-100 RAM is still on sale, it should be possible to replace
this stuff if that's all the problem is.
I still have no idea how the hell high temperatures would make a RAM
module go faulty, but there we are... (It's not like we're talking about
being baked at 200°C or anything!)
It's a dual-channel system, which presumably means RAM modules have to
be installed in pairs. (Not sure what happens if you install them singly
- whether it falls down to single-channel, or stops working completely,
or what. And I don't imagine for a second that the motherboard manual
will tell me!) I've got 2 pairs of modules installed, so I might be able
to isolate it to one of the pairs.
Failing that... I might be upgrading my PC quite a lot sooner than I had
expected! o_O
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Invisible wrote:
> It's a dual-channel system, which presumably means RAM modules have to
> be installed in pairs. (Not sure what happens if you install them singly
> - whether it falls down to single-channel, or stops working completely,
> or what. And I don't imagine for a second that the motherboard manual
> will tell me!)
Almost unbelievably, the motherboard manual *does* tell you (in the
customary broken English).
Apparently 1 module will work, 2 modules will work (provided they're
installed in slots 1+2), and 3 modules will not work under any
circumstances. (And, obviously, 4 will work.)
I currently have 4 installed, meaning that I'll have to remove 2 of them
to see what's up.
Also, it's DDR-400 RAM, which Ebuyer no longer sells in quantities
greater than 1GB for some reason. And it's not nearly as cheap as you'd
imagine either. (~£30 for 1GB) But it at least looks like I might be
able to get my PC going again. Cos, damn, I was like a little lost lamb
last night...
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Am 26.05.2010 10:32, schrieb Invisible:
> Also, it's DDR-400 RAM, which Ebuyer no longer sells in quantities
> greater than 1GB for some reason.
As you're probably not running many memory-hungry applications in
parallel, and I guesstimate we're talking about a 32-bit system, you
won't be able to utilize much more than 4 GB anyway. So presuming that
using all 4 slots typically gives some speed advantage, 1 GB is the
maximum module size you'll want to buy.
As for the price - what do you expect from a product that is no longer
manufactured in too high quantities anymore, with the primary target
group being people desperately trying to keep their old legacy computers
running?
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
>> Also, it's DDR-400 RAM, which Ebuyer no longer sells in quantities
>> greater than 1GB for some reason.
>
> As you're probably not running many memory-hungry applications in
> parallel, and I guesstimate we're talking about a 32-bit system, you
> won't be able to utilize much more than 4 GB anyway. So presuming that
> using all 4 slots typically gives some speed advantage, 1 GB is the
> maximum module size you'll want to buy.
Actually it's a 64-bit system. And there's a difference between using
one module and using two; the former gives you single-channel memory,
the latter gives you dual-channel memory, which has twice the bandwidth
(but identical latency). Using four modules yields no advantage, other
than being able to add more memory without throwing away what you
already have. (Which is what I did, incidentally.)
However, for reasons unknown, this specific motherboard only supports
4GB anyway. I guess because it was one of the first 64-bit boards...
> As for the price - what do you expect from a product that is no longer
> manufactured in too high quantities anymore, with the primary target
> group being people desperately trying to keep their old legacy computers
> running?
Yeah, figures. You wanna look at the prices for PC-100! ;-)
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Invisible <voi### [at] dev null> wrote:
> What this tells me is that the program wrote some data to memory and
> then read it back, and it wasn't the same. This indicates a fault with
> either the RAM itself, the motherboard, the CPU cache, or just some kind
> of obscure CPU fault that makes it execute instructions wrong.
When your PC starts misbehaving clearly due to some hardware failure,
the very first thing you do is check the PSU. If possible, try changing
it to another one which is known to work (and has enough wattage to run
your PC). Only then, if the problems persist, should you test other things.
In my experience 90% of PC hardware failures are caused by a failing PSU.
The symptoms can be extremely varied, all the way from the diverse BIOS test
failing (such as the CPU or RAM tests) to random crashes and reboots, as well
as the booting process failing seeminlgy due to some disk error.
If your PSU is failing, hardware tests (performed by BIOS or third-party
software) are completely unreliable and tell you absolutely nothing. Often
replacing the faulty PSU is enough to solve all the problems.
(There might be a few cases where the faulty PSU actually damaged some
other hardware in the process due to its builtin failsafes breaking. This
can be extremely confusing because neither replacing the PSU only or the
other hardware only will solve the problems, and it can be difficult to
diagnose that you need to replace *both*.)
Yes, I do have experience on this.
--
- Warp
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Am 26.05.2010 12:33, schrieb Invisible:
> However, for reasons unknown, this specific motherboard only supports
> 4GB anyway. I guess because it was one of the first 64-bit boards...
Looks like that, yes; surprises me that there do exist 64-bit boards
that don't support at least DDR2.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Warp wrote:
> When your PC starts misbehaving clearly due to some hardware failure,
> the very first thing you do is check the PSU.
That's because computers are really analog devices. If the PSU isn't
delivering enough voltage or current to keep the thing in spec, it goes back
to being an analog device (i.e., the circuits aren't driven to saturation),
and everything that assumes it's behaving digitally stops working reliably.
That's what I mean when I say at the lowest levels, everything on the wires
and in the computer is analog. :-)
--
Darren New, San Diego CA, USA (PST)
Ada - the programming language trying to avoid
you literally shooting yourself in the foot.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Invisible wrote:
> I still have no idea how the hell high temperatures would make a RAM
> module go faulty,
http://en.wikipedia.org/wiki/Thermal_runaway#Semiconductors
--
Darren New, San Diego CA, USA (PST)
Ada - the programming language trying to avoid
you literally shooting yourself in the foot.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On 5/26/2010 11:55 AM, Darren New wrote:
> That's because computers are really analog devices. If the PSU isn't
> delivering enough voltage or current to keep the thing in spec, it goes
> back to being an analog device (i.e., the circuits aren't driven to
> saturation), and everything that assumes it's behaving digitally stops
> working reliably.
A failing power supply causes a system to act erratically more often due
to switching noise not being filtered out than to low voltage or
current. Typically, the electrolytic filter capacitors start to
degrade, which allows the switching noise through. This can cause a bit
of havoc to ensue in a system.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
> A failing power supply causes a system to act erratically more often due
> to switching noise not being filtered out than to low voltage or current.
Switching noise (eg at 50 Hz) is just out-of-spec voltage for a few million
CPU cycles :-)
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
scott wrote:
>> A failing power supply causes a system to act erratically more often
>> due to switching noise not being filtered out than to low voltage or
>> current.
>
> Switching noise (eg at 50 Hz) is just out-of-spec voltage for a few
> million CPU cycles :-)
I would imagine it's more likely to upset the video or audio systems
than the CPU itself...
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Warp wrote:
> When your PC starts misbehaving clearly due to some hardware failure,
> the very first thing you do is check the PSU.
>
> In my experience 90% of PC hardware failures are caused by a failing PSU.
This does not match my experience.
I have seen PSUs die, and computers do very, very strange things as a
result. But it's fairly rare. Rarer than, say, HD failures or fan failures.
(And this of course is with those cheap £6 white-box PSUs. My home PC
has an expensive brand-name "performance" PSU - not that that
conclusively proves anything.)
> The symptoms can be extremely varied, all the way from the diverse BIOS test
> failing (such as the CPU or RAM tests) to random crashes and reboots, as well
> as the booting process failing seeminlgy due to some disk error.
>
> If your PSU is failing, hardware tests (performed by BIOS or third-party
> software) are completely unreliable and tell you absolutely nothing.
I ran a RAM check, and it consistently tells me that one specific area
of RAM isn't working. Every time I run the test, it complains about the
same chunk, fairly high up in the computer's address space.
If the power supply was unstable, I'd expect to see random glitches all
over the place, not just in one confined area.
The OS behaviour I'm seeing is also consistent with a RAM fault;
initially the system works fine, until it tries to use that chunk, at
which point it starts to malfunction spectacularly.
Perhaps more importantly, it will take me about 30 seconds to determine
if a fried RAM module is the problem. Rewiring the PSU would take far
longer, and the only PSU I have to test with *is* insufficiently
powerful to run the PC.
So in summary, I think I'm looking at a broken RAM module, not a PSU
fault. Although I guess it's possible a faulty PSU might be what _broke_
the RAM module in the first place...
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
> Perhaps more importantly, it will take me about 30 seconds to determine
> if a fried RAM module is the problem.
So is it then?
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
scott wrote:
>> Perhaps more importantly, it will take me about 30 seconds to
>> determine if a fried RAM module is the problem.
>
> So is it then?
I don't know; ask me when I get home. ;-)
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Invisible <voi### [at] devnull> wrote:
> I have seen PSUs die, and computers do very, very strange things as a
> result. But it's fairly rare. Rarer than, say, HD failures or fan failures.
So far I have had 4 PSUs break down in my computer, plus one in an iMac
at work. Three of these cases presented with random symptoms which at first
didn't raise suspicion that the PSU was the culprit (in the other two cases
the PSU simply outright stopped working, one of them with a loud snap).
So far the only other hardware failure I have ever had is a bunch of bad
sectors in one HD.
> I ran a RAM check, and it consistently tells me that one specific area
> of RAM isn't working. Every time I run the test, it complains about the
> same chunk, fairly high up in the computer's address space.
I tell you: Check your PSU first, *then* consider other possibilities.
It doesn't matter how random or how consistent the errors might be.
It might well not be the PSU, but it doesn't hurt checking it, and if it
so happens that it is indeed the PSU, it will save you a lot of work and
wasted money when you don't have to test anything else.
--
- Warp
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On 5/27/2010 3:39 AM, scott wrote:
>
> Switching noise (eg at 50 Hz) is just out-of-spec voltage for a few
> million CPU cycles :-)
>
>
Switching noise is in the kHz range, not 50/60 Hz. Google "switched
mode power supply" for more details on how they operate.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
>> Switching noise (eg at 50 Hz) is just out-of-spec voltage for a few
>> million CPU cycles :-)
>>
>
> Switching noise is in the kHz range, not 50/60 Hz.
My point was that any "noise" from the PSU is the same thing as out-of-spec
voltage.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
>> Perhaps more importantly, it will take me about 30 seconds to
>> determine if a fried RAM module is the problem.
>
> So is it then?
Yes.
--
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
I unclipped the two 1GB memory modules, and all of the memory test
errors vanished. I have now booted up Windows, and all of the bizare
behaviour has vanished. My PC works like normal again.
Woo!
Now, 1GB of DDR-400 is apparently about £30. If I'm lucky, it may be
that only *one* of these two modules is defective, and I can keep the
other one...
--
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
