Re: +12v Rising under load, +5v dropping.

On 2005-08-02, Johnny B Good <jcs.computers***@xxxxxxxxxxxxxxxx> wrote:
> The message <slrndet5ng.5vm.google@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>
> from Toby <google@xxxxxxxxxxx> contains these words:
>
>> I have a hardware problem!
>
>> My computer is stable if I use it casually. I can browse the web for
>> hours and
>> even do multi-track audio recording, with no problems.
>
> Those tasks are only slightly more arduous than idle.
>
>> However, if I strain the computer at all (e.g. run UT2K3, even in a
>> window at
>> 640x480, or rip a CD to MP3 using CDex) then my +12v line rises up to
>> around
>> +13.5v, and my +5v drops.
>
> That 13.5v figure is slightly over the 10% limit. How low does the 5v go?
> What's it like using CPUBurn or (slightly less stressful) Prime95'
> stress testing?

I just ran BURNK7.exe (i.e. CPUBurn for Athlon XP chips) and the +12v line
ran up to 13.376v within 10 seconds, accompanied by a burning smell from
the PSU! my +12v line normally sits at 12.928v, so that's a rise of
0.448v.

The +5v line only dropped from it's normal 4.811v down to 4.703v.
I guess that, for a smaller voltage, a proportionally smaller

The AsusProbe graph is spectacular :D
http://www.asktoby.com/miscimages/voltage2.gif
(I think you'll spot the moment I ran cpuburn!)

>> Here is a screenshot of Asus Probe after some hard activity. I've
>> marked the
>> section so you can see what I mean:
>
>> http://www.asktoby.com/miscimages/voltage.gif
>
> Presumably, the voltage figures are those prevailing at the time of the
> snapshot during close to idle conditions?

That's correct. My +12v sits at around 12.8v or 12.9v while my PC is
doing simple tasks, like posting here for example.

>> I can get away with this if I only do it for a short while, but if I
>> continue
>> straining the computer then the PSU begins to make a burning silicon smell.
>
> You're absolutely certain the smell origionates from within the PSU
> itself and not from MoBo components?

Certain.

>
>> I originally thought it may be heat, but the box is now very well
>> stocked with
>> fans. My CPU is 37 degrees C, and the case is 33 degrees C.
>
> Are these readings taken at idle after the initial half hour typically
> required to reach thermal equilibrium, i.e steady state? If so, and the
> room temperature is around 23 deg C, those temperatures are commendably
> low.

The room temperature is 21.6 deg C. At thermal equilibrium, the CPU is at
35 deg C and the mobo is at 30 deg C. I had to rule out overheating, so I
have a lot of fans.

>> I do have a lot of things plugged into the PSU:
>
>> 1 x 450W PSU ('Mercury' brand)
>> 1 x Asus A7N8X mobo
>> 2 x Creamware Pulsar II cards - each of these has
>> 6 SHARC dsps, so I guess 12 SHARCS use
>> quite a lot of power.
>> 1 x Sapphire Radeon Atlantis 9800 Pro
>> (Passive heatpipe. This card has it's own
>> power connection, direct from the PSU,
>> so I guess it's power hungry too. It is dual head.)
>> 8 x Fans of varying sizes, some run from the mobo,
>> some run directly from the PSU.
>
> That's a lot of fans. I hope, for the sake of your hearing, that
> they're large 7 volted fans to keep the noise down to a muted roar. :-)
>
>> 2 x HDDs: both IBM Deskstar 80Gb (UDMA100, 7200rpm)
>
> And the cpu is what, exactly? Obviously some model of socket A
> processor. I had to google to confirm this and the first hit took me to
> a review involving the use of an XP1900+ chip and mention of older
> Athlons and Durons so not exactly a current MoBo.

Silly me for forgetting to mention: Athlon XP 2800. Full details in my sig.

> Can you borrow a proper wattmeter to check the mains power consumption?
> I for one am intrigued by what you may discover. I typically see figures
> between 100 to 130 watts idle for most socket A systems depending on the
> system spec with 30 to 40 watt increases due to running CPUBurn stress
> tests.

Sadly I cannot. I may try pulling out some of my PCI cards and re-testing,
to see if that lowers the load significantly.

> You could try lowering the Vcore setting if you're not overclocking and
> the MoBo supports such adjustments. Some of the socket A chips can
> tolerate quite a significant reduction but it rather depends on the
> individual chip's characteristics. I'm running a Barton cored XP2500+ at
> 1.35v instead of the specified 1.65v. My particular sample of chip is, I
> have to admit, the exception (and, no, it _isn't_ a mobile version!).

I will also try lowering the VCore setting, and report back.

--
Toby.
Add the word afiduluminag to the subject
field to circumvent my email filters.
Ignore any mail delivery error.

Toby's CPU:
OPN (Ordering Part Number):AXDA2800KV4D
AMD Athlon XP 0.13µm
Type: Barton
Socket A
OPGA Package
OPN: AXDA2800DKV4D
Core: 13nm
Die size: 101mm²
Transistors: 54.3 million
Max FSB: 166/333MHz
Multiplier: 12.5
Frenquency: 2.083GHz
Cache L1 Instructions: 64KB
Cache L1 Datas: 64KB
512kb l2 cache
Operating voltage: 1.65V
Power (max): 68.6W
Die temperature (max): 85°C
CPU data bus width: 64-bit
.