-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Thursday 2008-02-14 at 15:19 +0100, Per Jessen wrote:

...

How do you measure it?

Hola Carlos

I'm using one of these:

http://www.order.conrad.com/xl/1000_1999/1200/1250/1253/125319_AB_05_FB.EPS....

Ah, yes. I'd prefer a data sheet in English (or Spanish), but I get the idea. :-) What I'm looking for, and probably the manufacturer doesn't say much, is how well they measure spikes, and what about "reactive power" and "switch mode" power loads. A PC uses a switch mode power supply, quite big, taking power in a strange waveform, and this may confuse heavily cheap meters (and not so cheap). Difficult to say. A definitive test, in your case, would be to connect a scope on the +5 and +12 rails (perhaps 3.2? line). A spike going down in the +5 would provoke an instant reset. - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) iD4DBQFHtGdMtTMYHG2NR9URApqfAJ4/wyxoJ5nHJDXpb+NgoV8iuFth+wCY5XEz tOMOxiHUzD1bPj+1G3XqJg== =ZbPj -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Carlos E. R. wrote:

...

http://www.order.conrad.com/xl/1000_1999/1200/1250/1253/125319_AB_05_FB.EPS....

Ah, yes. I'd prefer a data sheet in English (or Spanish), but I get the idea. :-)

I think there is a datasheet on the conrad website in 8-9 different languages - http://www.conrad.de/

What I'm looking for, and probably the manufacturer doesn't say much, is how well they measure spikes, and what about "reactive power" and "switch mode" power loads.

I'm pretty certain it won't be able to measure any spikes. As for the type of load, it's almost 99% resistive. I.e. the load doesn't vary with the phase of the AC.

A PC uses a switch mode power supply, quite big, taking power in a strange waveform, and this may confuse heavily cheap meters (and not so cheap). Difficult to say.

I think the load on the mains supply is actually pretty straight.

A definitive test, in your case, would be to connect a scope on the +5 and +12 rails (perhaps 3.2? line). A spike going down in the +5 would provoke an instant reset.

Hmmm, I could do that, but I've only got a normal scope (no storage), so I'd have to sit and watch it. Maybe. /Per Jessen, Zürich -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Randall R Schulz wrote:

On Thursday 14 February 2008 08:52, Per Jessen wrote:

...

Carlos E. R. wrote:

...

I'm pretty certain it won't be able to measure any spikes. As for the type of load, it's almost 99% resistive. I.e. the load doesn't vary with the phase of the AC.

"Resistive" means the current and the voltage are in phase with each other.

But I don't see how a transformer primary, which is an inductor, can be anything _but_ reactive.

An unloaded transformer is inductive, which limits the current to a low value and lagging voltage by 90 degrees. As the load increases, the voltage and current come closer into phase. When the transformer is at maximum load, the phase angle will be 0, assuming a resistive load on the transformer. -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Randall R Schulz wrote:

Ah, right. Well, my EE courses were a very long time ago. (Programming rots the brain, it appears...)

Amen to that. (my own EE classes were 20 years back). /Per Jessen, Zürich -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Carlos E. R. wrote:

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1

The Thursday 2008-02-14 at 13:07 -0500, James Knott wrote:

...

...

But I don't see how a transformer primary, which is an inductor, can be anything _but_ reactive.

An unloaded transformer is inductive, which limits the current to a low value and lagging voltage by 90 degrees. As the load increases, the voltage and current come closer into phase. When the transformer is at maximum load, the phase angle will be 0, assuming a resistive load on the transformer.

The problem in this case is that the load is neither: it is "switched". To calculate the real power used you need to sample the waveform, at least 15 samples per period, both voltage and current, then multiply and integrate (root mean square, RMS). You need some type of cpu. Actually, there is a chip that does all this, and I guess it is why these measurement gadgets have appeared in the market.

Quite so. I was just replying to that transformer comment. With switching supplies, transformers are often not used on the input. The incoming AC is rectified, filtered and the resulting DC is used to power the switching regulator that provides the desired output voltage(s). This process results in a lot of noise, that has to be filtered from the power line & load as well as shielded, to prevent interference.. -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

On Thursday 14 February 2008 12:26, James Knott wrote:

...

Quite so. I was just replying to that transformer comment. With switching supplies, transformers are often not used on the input. The incoming AC is rectified, filtered and the resulting DC is used to power the switching regulator that provides the desired output ...

Yowsa! What about the isolation provided by a transformer on the input? How do these things get UL (or other regulatory agency) approval? Randall Schuzl -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Thursday 2008-02-14 at 13:13 -0800, Randall R Schulz wrote:

Yowsa!

What about the isolation provided by a transformer on the input? How do these things get UL (or other regulatory agency) approval?

:-) Have you seen a TV set, the CRT type, inside? One of the AC cables goes to the chassis - hopefully the neutral :-P - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) iD8DBQFHtLSQtTMYHG2NR9URAkH/AJ4hVnFaUIP5wWzrtlMs9uEYh8O1IwCbB7r0 Mwk0KmzfNQfmAAVSw7WcvrY= =15fn -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Carlos E. R. wrote:

The problem in this case is that the load is neither: it is "switched". To calculate the real power used you need to sample the waveform, at least 15 samples per period, both voltage and current, then multiply and integrate (root mean square, RMS). You need some type of cpu. Actually, there is a chip that does all this, and I guess it is why these measurement gadgets have appeared in the market.

RMS measurements are not exactly new, Carlos. What's new is that they can now be done with cheap consume-style equipment. /Per Jessen, Zürich -- http://www.spamchek.com/ - your spam is our business. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Carlos E. R. wrote:

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1

The Thursday 2008-02-14 at 21:28 +0100, Per Jessen wrote:

...

RMS measurements are not exactly new, Carlos. What's new is that they can now be done with cheap consume-style equipment.

Not new, but affordable. I only knew one multimeter that could do Vac RMS, that was the "Fluke", and I never could afford it. Not even my employers!

Yes, Fluke makes some VERY nice meters...all of them expensive. On the other hand, they're extremely durable, worth ever cent for those who need that level of capability plus durability.

And I don't know what type of circuit they used.

-- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Doug McGarrett wrote:

On Thursday 14 February 2008 12:50, Randall R Schulz wrote:

...

On Thursday 14 February 2008 08:52, Per Jessen wrote:

...

Carlos E. R. wrote:

...

I'm pretty certain it won't be able to measure any spikes. As for the type of load, it's almost 99% resistive. I.e. the load doesn't vary with the phase of the AC.

"Resistive" means the current and the voltage are in phase with each other.

But I don't see how a transformer primary, which is an inductor, can be anything _but_ reactive.

...

...

/Per Jessen, Zürich

Randall Schulz

A transformer literally does what its name implies--it transforms a given voltage and current into another voltage and current, with virtually no loss. Therefore, if the output load on a transformer is resistive, then the input to it is resistive also. The relationship between input voltage and current to output voltage and current is linear; for example, if the input is 100 volts at 1 amp, the output could be 10 volts at 10 amps. The transformer is not absolutely lossless, but it is extremely efficient. In most cases you can very comfortably put your fingers on a transformer core and feel only a little warmth.

The transformer appears resistive, only at full (resistive) load. At anything less than that, there's still some amount of inductance to limit current and that inductance will cause the current phase shift. -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Doug McGarrett wrote:

On Thursday 14 February 2008 12:50, Randall R Schulz wrote:

...

On Thursday 14 February 2008 08:52, Per Jessen wrote:

...

Carlos E. R. wrote:

...

I'm pretty certain it won't be able to measure any spikes. As for the type of load, it's almost 99% resistive. I.e. the load doesn't vary with the phase of the AC. "Resistive" means the current and the voltage are in phase with each other.

But I don't see how a transformer primary, which is an inductor, can be anything _but_ reactive.

...

...

/Per Jessen, Zürich Randall Schulz

A transformer literally does what its name implies--it transforms a given voltage and current into another voltage and current, with virtually no loss. Therefore, if the output load on a transformer is resistive, then the input to it is resistive also. The relationship between input voltage and current to output voltage and current is linear; for example, if the input is 100 volts at 1 amp, the output could be 10 volts at 10 amps. The transformer is not absolutely lossless, but it is extremely efficient. In most cases you can very comfortably put your fingers on a transformer core and feel only a little warmth.

Note that's the iron CORE, conveying the magnetic field, **NOT** the WINDINGS which are conveying the current! -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Thursday 2008-02-14 at 17:52 +0100, Per Jessen wrote:

...

A PC uses a switch mode power supply, quite big, taking power in a strange waveform, and this may confuse heavily cheap meters (and not so cheap). Difficult to say.

I think the load on the mains supply is actually pretty straight.

I'm not so sure, but never mind. Se bellow.

...

A definitive test, in your case, would be to connect a scope on the +5 and +12 rails (perhaps 3.2? line). A spike going down in the +5 would provoke an instant reset.

Hmmm, I could do that, but I've only got a normal scope (no storage), so I'd have to sit and watch it. Maybe.

I'm thinking that the power measured at the input is not important at all. The important figures are the amperage capacity on each of the DC output lines, and that's the power that has to be matched to the load, and where the measurements are important. The AC side is irrelevant, somehow. As to the scope, yes, that would be enough, but of course, you have to be looking at it. Set a slow H speed. Humm... a digital gadget (multiple A/D) connected to another PC would be _very_ nice. They are not so expensive as they were time ago... Argh. I have a good A/D, D/A card, but it is for ISA bus, and MsDOS drivers. Pity! I can not experiment with it in Linux, too late... no hw isa bus. - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) iD8DBQFHtKLytTMYHG2NR9URAtKPAJ9gKArym/lBQTXa8KHs7cR/FuWzBQCfcf1m 8xNPMtu9/2mERBN3cuCgrHg= =rYkf -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Thursday 2008-02-14 at 21:38 +0100, Per Jessen wrote:

Carlos E. R. wrote:

...

I'm thinking that the power measured at the input is not important at all. The important figures are the amperage capacity on each of the DC output lines, and that's the power that has to be matched to the load, and where the measurements are important. The AC side is irrelevant, somehow.

Well, my only reason for looking at the AC side was to see what the real load was. I'm a little disaapointed that I got convinced to buy a much bigger power-supply, when my 350W would quite obviously have been enough. Still, like I said, the new one is also much quieter, which is very pleasant.

What I think is worth noting here - a somewhat highend system like this still uses barely 250W at _full_ load.

Still, I don't know if those new measurement gadgets can be trusted. I mean I don't know, not that they are bad. Then, there is another issue. The important thing is the power needed at each of the internal DC lines. Some one said that the MB nowdays uses power from the 12 V line, and also the video card uses it (I don't know what for). Plus the fans, the high speed drives... sum all of that. If it close to 10 Amps, and the 350W unit can supply only 10 A (for instance), then you really need a higher wattage power supply: not because you need that so much power, but because one of the lines uses much more power. If this is the case, what the manufacturers will have to do is to change the power distribution on new PS so that they have more Amps at the 12 V line. Its an hypothesis only, but it might be so. Let me see... for instance, the Seagate barracuda, ST3250310AS, 250GB, a small disk nowdays (the top of the family is 750 GB, 8 heads). From the datasheet: Table 12: DC power requirements - --long_line--------------------------------------------------------------------------------- Power dissipation (4-disc values shown) Avg (watts 25° C) Avg 5V typ amps Avg 12V typ amps Spinup — — 2.8 (peak) Idle* 9.30 0.611 0.520 Idle* (with offline activity) 10.40 0.719 0.567 Operating (40% r/w, 40% seek, 20% inop.) 13.00 0.772 0.762 Seeking (random, 20% idle) 12.60 0.613 0.795 Standby 0.80 0.106 0.023 Sleep 0.80 0.106 0.023 Ie, the peak power during spin up is 2.8 Amps on the 12V line, 0.8A avg maximum during seeks. The power supply must be able to handle that, and if it is true the MB uses also that line, there will be problems. Of course, the same calculation should be done for the MB and video card. - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) iD8DBQFHtMAitTMYHG2NR9URAigCAJwPcWswy456CXmeYKteMPt6MUBCnQCfRtQq VeRq3ttPpA0XcDaeO4JQnyE= =SmbG -----END PGP SIGNATURE-----

On Thu, Feb 14, 2008 at 5:43 PM, jdd wrote:

Carlos E. R. a écrit :

...

The important thing is the power needed at each of the internal DC lines.

new power units data sheet list max amp for each power line...

That can be part of the problem. Some newer PSUs have multiple power lanes. Say one lane is designed to drive 2 peripherals, but the unsuspecting user (or PC builder) adds a couple of Y splitters and drives 4 peripherals with it. Can be trouble. I've got a big tower chassis with 15 drive slots. If I ever filled it up, I would have to really watch how the power was distributed and not just go from one Y connector to the next. I've actually been thinking how cool it would be to get a big 3ware card and fill that thing up with 1TB drives. There price has dropped to $240 or so on sale, so for a few $K, I could build a nice 10TB machine!!!! I'll save that for another thread on another day. Greg -- Greg Freemyer Litigation Triage Solutions Specialist http://www.linkedin.com/in/gregfreemyer First 99 Days Litigation White Paper - http://www.norcrossgroup.com/forms/whitepapers/99%20Days%20whitepaper.pdf The Norcross Group The Intersection of Evidence & Technology http://www.norcrossgroup.com -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Carlos E. R. wrote:

The Thursday 2008-02-14 at 21:38 +0100, Per Jessen wrote:

...

Carlos E. R. wrote:

...

...

I'm thinking that the power measured at the input is not important at all. The important figures are the amperage capacity on each of the DC output lines, and that's the power that has to be matched to the load, and where the measurements are important. The AC side is irrelevant, somehow.

...

Well, my only reason for looking at the AC side was to see what the real load was. I'm a little disaapointed that I got convinced to buy a much bigger power-supply, when my 350W would quite obviously have been enough. Still, like I said, the new one is also much quieter, which is very pleasant.

...

What I think is worth noting here - a somewhat highend system like this still uses barely 250W at _full_ load.

Still, I don't know if those new measurement gadgets can be trusted. I mean I don't know, not that they are bad.

Then, there is another issue.

The important thing is the power needed at each of the internal DC lines. Some one said that the MB nowdays uses power from the 12 V line, and also the video card uses it (I don't know what for). Plus the fans, the high speed drives... sum all of that. If it close to 10 Amps, and the 350W unit can supply only 10 A (for instance), then you really need a higher wattage power supply: not because you need that so much power, but because one of the lines uses much more power.

If this is the case, what the manufacturers will have to do is to change the power distribution on new PS so that they have more Amps at the 12 V line.

Its an hypothesis only, but it might be so.

Let me see... for instance, the Seagate barracuda, ST3250310AS, 250GB, a small disk nowdays (the top of the family is 750 GB, 8 heads). From the datasheet:

Table 12: DC power requirements

--long_line--------------------------------------------------------------------------------- Power dissipation (4-disc values shown) Avg (watts 25° C) Avg 5V typ amps Avg 12V typ amps Spinup   2.8 (peak) Idle* 9.30 0.611 0.520 Idle* (with offline activity) 10.40 0.719 0.567 Operating (40% r/w, 40% seek, 20% inop.) 13.00 0.772 0.762 Seeking (random, 20% idle) 12.60 0.613 0.795 Standby 0.80 0.106 0.023 Sleep 0.80 0.106 0.023

Ie, the peak power during spin up is 2.8 Amps on the 12V line, 0.8A avg maximum during seeks. The power supply must be able to handle that, and if it is true the MB uses also that line, there will be problems.

Of course, the same calculation should be done for the MB and video card.

-- Cheers, Carlos E. R.

Due to low vcore voltages for cpu and gpu and the load on the 3.3v (5v vcc is a thing of the past) they are derived from the 12v supply nowadays. The lower the operating voltage the faster the logic switches and less power is needed. That is why you have extra 12v connections on MB. Dave -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Carlos E. R. wrote:

The Friday 2008-02-15 at 07:27 +0200, Dave Plater wrote:

...

Due to low vcore voltages for cpu and gpu and the load on the 3.3v (5v vcc is a thing of the past) they are derived from the 12v supply nowadays. The lower the operating voltage the faster the logic switches and less power is needed. That is why you have extra 12v connections on MB. Dave

But I don't see why derive it from the 12 V line instead of the 5 V line, leaving it unused and overloading the 12 V line instead.

-- Cheers, Carlos E. R.

Search me, thats why multiple rail PSUs came about. Isolating noisy fans and disks etc from cpu & gpu. Dave -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Dave Plater wrote:

Carlos E. R. wrote:

...

The Thursday 2008-02-14 at 21:38 +0100, Per Jessen wrote:

...

Carlos E. R. wrote:

...

I'm thinking that the power measured at the input is not important at all. The important figures are the amperage capacity on each of the DC output lines, and that's the power that has to be matched to the load, and where the measurements are important. The AC side is irrelevant, somehow.

Well, my only reason for looking at the AC side was to see what the real load was. I'm a little disaapointed that I got convinced to buy a much bigger power-supply, when my 350W would quite obviously have been enough. Still, like I said, the new one is also much quieter, which is very pleasant.

What I think is worth noting here - a somewhat highend system like this still uses barely 250W at _full_ load.

Still, I don't know if those new measurement gadgets can be trusted. I mean I don't know, not that they are bad.

Then, there is another issue.

The important thing is the power needed at each of the internal DC lines. Some one said that the MB nowdays uses power from the 12 V line, and also the video card uses it (I don't know what for). Plus the fans, the high speed drives... sum all of that. If it close to 10 Amps, and the 350W unit can supply only 10 A (for instance), then you really need a higher wattage power supply: not because you need that so much power, but because one of the lines uses much more power.

If this is the case, what the manufacturers will have to do is to change the power distribution on new PS so that they have more Amps at the 12 V line.

Its an hypothesis only, but it might be so.

Let me see... for instance, the Seagate barracuda, ST3250310AS, 250GB, a small disk nowdays (the top of the family is 750 GB, 8 heads). From the datasheet:

Table 12: DC power requirements

--long_line--------------------------------------------------------------------------------- Power dissipation (4-disc values shown) Avg (watts 25° C) Avg 5V typ amps Avg 12V typ amps Spinup   2.8 (peak) Idle* 9.30 0.611 0.520 Idle* (with offline activity) 10.40 0.719 0.567 Operating (40% r/w, 40% seek, 20% inop.) 13.00 0.772 0.762 Seeking (random, 20% idle) 12.60 0.613 0.795 Standby 0.80 0.106 0.023 Sleep 0.80 0.106 0.023

Ie, the peak power during spin up is 2.8 Amps on the 12V line, 0.8A avg maximum during seeks. The power supply must be able to handle that, and if it is true the MB uses also that line, there will be problems.

Of course, the same calculation should be done for the MB and video card.

-- Cheers, Carlos E. R.

Due to low vcore voltages for cpu and gpu and the load on the 3.3v (5v vcc is a thing of the past) they are derived from the 12v supply nowadays. The lower the operating voltage the faster the logic switches and less power is needed. That is why you have extra 12v connections on MB. Dave

Actually, I think you've got that a bit backward. Generally lower power results from lower voltage and that lower power allows faster switching. To get fastest switching from a device, ignoring power dissipation, requires higher voltage. -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Carlos E. R. wrote:

...

What I think is worth noting here - a somewhat highend system like this still uses barely 250W at _full_ load.

Still, I don't know if those new measurement gadgets can be trusted. I mean I don't know, not that they are bad.

I don't know if they can be trusted, but I can't really think of a reason not to. I guess I could find a well-defined load (a large wattage resistor for instance) to check it.

The important thing is the power needed at each of the internal DC lines. Some one said that the MB nowdays uses power from the 12 V line, and also the video card uses it (I don't know what for). Plus the fans, the high speed drives... sum all of that. If it close to 10 Amps, and the 350W unit can supply only 10 A (for instance), then you really need a higher wattage power supply: not because you need that so much power, but because one of the lines uses much more power.

Or a powersupply with a higher output rating for the 12V line. But you're right, point taken.

Ie, the peak power during spin up is 2.8 Amps on the 12V line, 0.8A avg maximum during seeks. The power supply must be able to handle that, and if it is true the MB uses also that line, there will be problems.

Well, although the measurement device cannot measure peaks, the harddisk spin-up time is 3-4-5 seconds or so, so I think I would have seen any excess load. During startup, the machine used up to 175W. /Per Jessen, Zürich -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Dave Plater wrote:

Going by your temperature info the machine bombs due to a temperature spike when all cores are working hard. Nothing wrong with PSU.

Agree, the PSU has been exonerated long ago. Temperature spike? Is there such a thing? It would possibly explain the situation. The lm-sensors output (that's all I have for now) will occasionally jump a degree up and down, but I'm taking that to be a rounding thing. Is it really possible for the CPU temperature to jump e.g. 5 or 10 degrees for a second and then settle back down again? /Per Jessen, Zürich -- http://www.spamchek.com/ - your spam is our business. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Greg Freemyer wrote:

...

Temperature spike? Is there such a thing? It would possibly explain the situation. The lm-sensors output (that's all I have for now) will occasionally jump a degree up and down, but I'm taking that to be a rounding thing. Is it really possible for the CPU temperature to jump e.g. 5 or 10 degrees for a second and then settle back down again?

I don't know by how many degrees, but a CPU that is fully utilizing all its circuitry will be hotter that one that is not. And yes the spike can be pretty short, but I doubt just a second or two.

It's been too long since I looked at energy, heat, heat conduction and all that (more than 20 years), but the idea of a temperature spike just doesn't sound right in the context of a PC. Like you say, a spike shorter than a second or two just doesn't seem realistic. Something to do with the amount of energy it would take to make a 5degree increase in temperature in such a short timespam. /Per Jessen, Zürich -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

On 02/16/2008 02:25 AM, Per Jessen wrote:

It's been too long since I looked at energy, heat, heat conduction and all that (more than 20 years), but the idea of a temperature spike just doesn't sound right in the context of a PC. Like you say, a spike shorter than a second or two just doesn't seem realistic. Something to do with the amount of energy it would take to make a 5degree increase in temperature in such a short timespam.

I don't think it is a temperature spike per se, but getting too close to the tolerance of the transistors inside the CPU. It is possible they they get close enough to the max temps at full load they go into thermal overload and protection circuitry inside the CPU kicks in to protect the processor. All to say it may not be the board but the CPU cooling. I personally thought 65 was too hot. If the max is 70, 65 is too close for comfort for any sustained load. -- Joe Morris Registered Linux user 231871 running openSUSE 10.3 x86_64 -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Carlos E. R. wrote:

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1

Content-ID:

The Saturday 2008-02-16 at 08:05 +0800, Joe Morris wrote:

...

protection circuitry inside the CPU kicks in to protect the processor. All to say it may not be the board but the CPU cooling. I personally thought 65 was too hot. If the max is 70, 65 is too close for comfort for any sustained load.

It depends on the particular cpu model, some can run quite hot. So, better find out what's the maximum temp for that model first, and adjust the trip point. Certainly, if it is running at 65° and the trip point is 70°, it is certain it will trip on high load. No room!

There is a list of cpu temps somewhere... here: http://www.heatsink-guide.com/

Mmm, too old, the "Phenom" is not listed. Somewhere! I'm off to bed.

http://www.amd.com/us-en/ Click on Support & Download http://www.amd.com/us-en/Processors/ProductInformation/0,,30_118,00.html Click on AMD Phenom * System Component Information http://www.amd.com/us-en/Processors/TechnicalResources/0,,30_182_869_15345,0... On the right hand side of the page, click on AMD Phenom(TM) Thermal Solutions Heatsinks http://www.amd.com/us-en/thermal/ResultsHandler/1,,30_182_869_15345%5E15353,... -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Friday 2008-02-15 at 08:44 +0100, Per Jessen wrote:

...

Still, I don't know if those new measurement gadgets can be trusted. I mean I don't know, not that they are bad.

I don't know if they can be trusted, but I can't really think of a reason not to. I guess I could find a well-defined load (a large wattage resistor for instance) to check it.

Because I distrust cheap meters. Personal thing. :-) You can measure a stove, it will measure all right. But an oven or a motor with a triac regulator... I wonder. Maybe it works fine.

...

really need a higher wattage power supply: not because you need that so much power, but because one of the lines uses much more power.

Or a powersupply with a higher output rating for the 12V line. But you're right, point taken.

Yes, that's what I meant. - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) iD8DBQFHtXoUtTMYHG2NR9URAsu/AJ9dkQdrDnPd3erFkm+H7SWRnHdMwgCdFGYU 91tGc4umTU1DOLje8HgWZAo= =Q8fZ -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Carlos E. R. wrote:

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1

The Thursday 2008-02-14 at 17:52 +0100, Per Jessen wrote:

...

...

A PC uses a switch mode power supply, quite big, taking power in a strange waveform, and this may confuse heavily cheap meters (and not so cheap). Difficult to say.

I think the load on the mains supply is actually pretty straight.

I'm not so sure, but never mind. Se bellow.

...

...

A definitive test, in your case, would be to connect a scope on the +5 and +12 rails (perhaps 3.2? line). A spike going down in the +5 would provoke an instant reset.

Hmmm, I could do that, but I've only got a normal scope (no storage), so I'd have to sit and watch it. Maybe.

I'm thinking that the power measured at the input is not important at all. The important figures are the amperage capacity on each of the DC output lines, and that's the power that has to be matched to the load, and where the measurements are important. The AC side is irrelevant, somehow.

As to the scope, yes, that would be enough, but of course, you have to be looking at it. Set a slow H speed. Humm... a digital gadget (multiple A/D) connected to another PC would be _very_ nice. They are not so expensive as they were time ago...

Argh. I have a good A/D, D/A card, but it is for ISA bus, and MsDOS drivers. Pity! I can not experiment with it in Linux, too late... no hw isa bus.

Could be worse...could be S-100! :-) Carlos, I'll bet you could make a USB<==>ISA bus adapter. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Carlos E. R. wrote:

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1

The Thursday 2008-02-14 at 15:58 -0500, Aaron Kulkis wrote:

...

...

Argh. I have a good A/D, D/A card, but it is for ISA bus, and MsDOS drivers. Pity! I can not experiment with it in Linux, too late... no hw isa bus.

Could be worse...could be S-100! :-)

Carlos, I'll bet you could make a USB<==>ISA bus adapter.

I have not seen such a thing :-? Very complex: data bus, addr bus, irq control, dma control...

It's a joke, Carlos. Laugh. ve haff vays uff making you smile. </cheesyWW2movie> -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

On Sun, 2008-02-17 at 23:38 +0100, Carlos E. R. wrote:

-----BEGIN PGP SIGNED MESSAGE-----

Ah...! You could have used a smiley - like :-p

What a pity. Just for a moment I thought I could use my daq card.

shouldn't be too hard to make an old pc into a network server for your daq...(smiley face) Tom in NM -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Sunday 2008-02-17 at 22:24 -0700, Tom Patton wrote:

...

What a pity. Just for a moment I thought I could use my daq card.

shouldn't be too hard to make an old pc into a network server for your daq...(smiley face)

Yes, could be... - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) iD8DBQFHuW0VtTMYHG2NR9URAuiNAJ4t1Zd8HsWo630P2M8Wmh3FTqaNCwCdENIH ilzrS2DSVHK7mu10GFbW5z0= =4Ay7 -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Thursday 2008-02-14 at 12:06 -0500, Aaron Kulkis wrote:

...

I'm using one of these:

http://www.order.conrad.com/xl/1000_1999/1200/1250/1253/125319_AB_05_FB.EPS....

I would look for a device with a needle rather than a digital readout.

I don't think you can measure true rms with a coil and needle. :-? It was measured by thermal effects... - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) iD8DBQFHtZCLtTMYHG2NR9URAlnhAJ0R9Gtq/CU8JbPQ037xqOdgFNcMMQCfY7FD awIpr0Edt9dSA+nNqvzIVUQ= =yy55 -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Per Jessen wrote:

Carlos E. R. wrote:

...

...

...

I'm using one of these:

http://www.order.conrad.com/xl/1000_1999/1200/1250/1253/125319_AB_05_FB.EPS.... I would look for a device with a needle rather than a digital readout. I don't think you can measure true rms with a coil and needle. :-? It was measured by thermal effects...

Carlos, Aaron only said "readout", which is unrelated to how it is measured. The advantage of an analog readout over a digital is that it is possible for a human to see spikes/changes more clearly.

/Per Jessen, Zürich

Of course, depending on how fast that spike occurs, the needle won't be able to follow it. There are some digital meters that can be configured to display the peak value. -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Carlos E. R. wrote:

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1

The Friday 2008-02-15 at 16:07 +0100, Per Jessen wrote:

...

...

...

I would look for a device with a needle rather than a digital readout.

I don't think you can measure true rms with a coil and needle. :-? It was measured by thermal effects...

Carlos, Aaron only said "readout", which is unrelated to how it is measured. The advantage of an analog readout over a digital is that it is possible for a human to see spikes/changes more clearly.

I know that. I've always liked analog multimeters for that reason.

But I don't think you can get a needle movement with rms voltage measurement. And the amp meter "tweezers" I have used did certainly work badly with triac controlled motors and PC power supplies. So allow me some doubt that a needle thing could measure ac power properly nowdays.

Not so much for "accurate" reading but looking for fluctuations, which are much easier to make sense of with a needle than with a digital readout.

Your gadget is probably using a new chip that made the market probably recently designed for the purpose of power measurement with complex waveform. I wonder if power distributors are thinking on using such chips on house meters instead of the mechanic meters they use now.

I doubt it. The current meters measure total power (real power + imaginary power). Since the generators see both the real power and the imaginary power, it is in everyone's best interest that power continue to be metered using the current method which bills for both real and imaginary power. [Because of this, appliance manufacturers, especially electric motor manufacturers, are motivated to design things so that imaginary power is kept to a minimum]. Real power is current which is in phase with the supply voltage. Imaginary power is current which is out of phase with the supply voltage. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

On Friday 15 February 2008 08:15, Carlos E. R. wrote:

The Thursday 2008-02-14 at 12:06 -0500, Aaron Kulkis wrote:

...

...

I'm using one of these:

http://www.order.conrad.com/xl/1000_1999/1200/1250/1253/125319_AB_05_FB. EPS.jpg

I would look for a device with a needle rather than a digital readout.

I don't think you can measure true rms with a coil and needle. :-?

It was measured by thermal effects...

-- Cheers, Carlos E. R. There are electronic true-RMS voltmeters with analog readouts. For example, the Balantine 323 has a nice big dial. You would have to measure the voltage drop across a known value shunt resistor in series with the load, and calculate the power from voltage and current. --doug, wa2say

Blessed are the peacemakers ... for they shall be shot at from both sides. --A.M. Greeley -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Friday 2008-02-15 at 15:56 -0500, James Knott wrote:

...

...

I don't think you can measure true rms with a coil and needle. :-?

It was measured by thermal effects...

There are electronic true-RMS voltmeters with analog readouts. For example, the Balantine 323 has a nice big dial. You would have to measure the voltage drop across a known value shunt resistor in series with the load, and calculate the power from voltage and current.

Ummm... He was talking about RMS, which implies voltage or current. There's no such thing as RMS power.

Well... you have to multiply voltage with current, around the cycle, and integrate. If they are 90° out of phase, result is zero. If current is switched, who knows the result! I mean, those chips should know, but I don't know if they know, if you follow me :-) - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) iD8DBQFHthWitTMYHG2NR9URAsCtAJ9kLQAvnFj3LHGbmfZ4opQcRDObkgCffAU/ ufQopS1zIK4C4ueEW2NxirE= =/4Pv -----END PGP SIGNATURE-----

Aaron Kulkis wrote:

James Knott wrote:

...

Doug McGarrett wrote:

...

On Friday 15 February 2008 08:15, Carlos E. R. wrote:

...

The Thursday 2008-02-14 at 12:06 -0500, Aaron Kulkis wrote:

...

...

I'm using one of these:

http://www.order.conrad.com/xl/1000_1999/1200/1250/1253/125319_AB_05_FB.

EPS.jpg I would look for a device with a needle rather than a digital readout. I don't think you can measure true rms with a coil and needle. :-?

It was measured by thermal effects...

-- Cheers, Carlos E. R. There are electronic true-RMS voltmeters with analog readouts. For example, the Balantine 323 has a nice big dial. You would have to measure the voltage drop across a known value shunt resistor in series with the load, and calculate the power from voltage and current.

Ummm... He was talking about RMS, which implies voltage or current. There's no such thing as RMS power.

RMS is short for Root-mean-squared i.e. _______ / _ 2 V X

RMS can apply to any function vs. time (voltage, current, power, etc.)

It's been a while since I've done it, but if you try calculating RMS power, you soon find your calculations are meaningless. The purpose of the RMS calculation is to determine the equivalent DC voltage or current that would produce the same amount of power. So, when you square a voltage or current sample you are in effect making an instantaneous power calculation P=E^2/R or P=I^2R, then taking the mean power value over the cycle and then converting back to a voltage or current value. -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Aaron Kulkis wrote:

RMS is just as statistical method, which is useful for making sense of any time-variant function.

The square root of the square of power is just as much related to power as the square root of the square of voltage is related to voltage (NOT power).

Judging from this response and the one to Randall, I suspect it might be a good idea for you to review your EE texts. RMS is meaningless in the context of power. Sure you can do the calculation, but what does it get you? With current or voltage, RMS derives the equivalent DC value. Can you explain what you'd obtain by calculating RMS power? -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

James Knott wrote:

Aaron Kulkis wrote:

...

RMS is just as statistical method, which is useful for making sense of any time-variant function.

The square root of the square of power is just as much related to power as the square root of the square of voltage is related to voltage (NOT power).

Judging from this response and the one to Randall, I suspect it might be a good idea for you to review your EE texts. RMS is meaningless in the context of power. Sure you can do the calculation, but what does it get you? With current or voltage, RMS derives the equivalent DC value. Can you explain what you'd obtain by calculating RMS power?

Yes. You have some idea how much "average" power is going through resistive elements, or anything else which is heat/power sensitive. A 1-W resistor can handle (momentarily) instants of 5-W if the duration is short enough (say, the voltage spike from a drum beat) What it can't handle is 5-W RMS. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Tuesday 2008-02-19 at 07:17 -0500, James Knott wrote:

Aaron Kulkis wrote:

...

What it can't handle is 5-W RMS.

And what, precisely, is 5W RMS? If all you need is average power, then simply take the average of all the instantaneous power levels. There's no need for RMS. With RMS voltage or current, you're taking instantaneous values, converting them to power, taking the mean and converting back to equivalent DC voltage or current.

RMS is just a formula, you can apply it to any waveform. However, applied to power doesn't make sense, because the square is always positive, and power in a non resistive circuit will be both positive and negative at times, reducing the "really" used power even to zero. Thus RMS power makes no sense, although it is calculable. - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) iD8DBQFHutSytTMYHG2NR9URAnYRAJ0V2Lli4dKq+/VcXFlRAFmCLcrAMQCgiJmR YohkyJXpufUohin9yDlb4ps= =cSbk -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

On Friday 15 February 2008 22:29, Aaron Kulkis wrote:

...

...

Ummm... He was talking about RMS, which implies voltage or current. There's no such thing as RMS power.

RMS is short for Root-mean-squared i.e. _______ / _ 2 V X

RMS can apply to any function vs. time (voltage, current, power, etc.)

Of course you can blindly apply that formula to any time-varying function (if the function is constant rather than time-varying, the result is just the constant value of the "waveform") However, the point is that by squaring the primitive (and time-varying) quantity first (voltage or current, but not power which depends on both current and voltage)) then taking the average over a cycle and then taking the square root of that value, you'll get the average power. That's because the power varies with the square of the current or the voltage. So the RMS voltage and the resistance allow you to compute the average power over a full cycle. (Note that power is voltage squared divided by resistance and that this is the real reason that RMS is the way to compute the average power dissipation in a resistive circuit.) But if you have an instantaneous power function and apply the RMS calculation to that, you'll get a value with no physical meaning. Randall Schulz -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

On Sunday 17 February 2008 19:58, Aaron Kulkis wrote:

Randall R Schulz wrote:

...

...

However, the point is that by squaring the primitive (and time-varying) quantity first (voltage or current, but not power which depends on both current and voltage)) then taking the average over a cycle and then taking the square root of that value, you'll get the average power.

This is clearly incorrect. The result of applying the RMS algorithm to voltage is a value with units of voltage. This value squared and divided by the resistance into which the voltage drives the current resulting in some power dissipation gives that power.

...

...

The units of RMS voltage is the volt, not volt^2.

True. But another correction (to one of your earlier statements) is required. On Sunday 17 February 2008 19:56, Aaron Kulkis wrote:

... RMS is just as statistical method, which is useful for making sense of any time-variant function.

...

While RMS is the name for a statistical technique, as it's used it circuit analysis, it's an analytical technique. E.g., if the waveform is a sine wave, then one can compute the definite integral analytically for the function sin(x)^2. Taking the square root of this value gives the RMS value. For hard-to-integrate functions or those with no a priori analytical characterization, a numeric technique can be used. Randall Schulz -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

...

On Sunday 17 February 2008 19:56, Aaron Kulkis wrote:

...

... RMS is just as statistical method, which is useful for making sense of any time-variant function.

For the love of god and of the engineering I've read the absurd above, and I can see clearly its out of control. Stop the show-off, because you are starting to say completely incorrect things, not to mentio its an absurd of offtopic. Time variance or time invariance is NOT what you think it is. Take some time to read a basic text about time invariance in http://en.wikipedia.org/wiki/Time_invariance Seriously go back to suse linux topics... For those who are not just trying to show off, saying things like "power factor of a resistor is 1", I recommend a reading about Norton / Thevenin theorem... Best Regards Marcio Ferreira --- Druid -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

On Sunday 17 February 2008 22:58, Aaron Kulkis wrote:

Randall R Schulz wrote:

...

On Friday 15 February 2008 22:29, Aaron Kulkis wrote:

...

...

...

Ummm... He was talking about RMS, which implies voltage or current. There's no such thing as RMS power.

RMS is short for Root-mean-squared i.e. _______ / _ 2 V X

RMS can apply to any function vs. time (voltage, current, power, etc.)

Of course you can blindly apply that formula to any time-varying function (if the function is constant rather than time-varying, the result is just the constant value of the "waveform")

However, the point is that by squaring the primitive (and time-varying) quantity first (voltage or current, but not power which depends on both current and voltage)) then taking the average over a cycle and then taking the square root of that value, you'll get the average power. That's because the power varies with the square of the current or the voltage.

So the RMS voltage and the resistance allow you to compute the average power over a full cycle. (Note that power is voltage squared divided by resistance and that this is the real reason that RMS is the way to compute the average power dissipation in a resistive circuit.)

But if you have an instantaneous power function and apply the RMS calculation to that, you'll get a value with no physical meaning.

The units of RMS voltage is the volt, not volt^2. Yes. --dm

-- Blessed are the peacemakers ... for they shall be shot at from both sides. --A.M. Greeley -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Carlos E. R. wrote:

The Sunday 2008-02-17 at 15:09 -0800, Randall R Schulz wrote:

...

So the RMS voltage and the resistance allow you to compute the average power over a full cycle. (Note that power is voltage squared divided by resistance and that this is the real reason that RMS is the way to compute the average power dissipation in a resistive circuit.)

For power you have to consider the power factor, the cosine of the phase angle - which for a resistor is 1, of course.

...

But if you have an instantaneous power function and apply the RMS calculation to that, you'll get a value with no physical meaning.

which is why I said the first day that I don't know how good are those measurement gadgets when applied to a swith mode power supply.

Here:

http://www.analog.com/en/

Search for the ADE7756AN (Active Energy Metering IC with Serial Interface)

-- Cheers, Carlos E. R.

It looks as good as the designer that uses it, besides a switch mode supply still converts the mains to DC, and smoothed before switching it through a ferrite core transformer. There are rules about sending high frequencies back up the mains supply. Both the power going in and out to MB is buffered so to speak. Dave -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Greg Freemyer wrote:

Per,

I don't know how you measured the load, but it is important to measure very short power surges. In particular, I'm suspicious of that load during raid1 resync. Could your technique catch a 10 msec spike etc?

Nope, not a chance. See my reply to Carlos.

I think it is the spikes that are causing trouble with Sata drives. Don't know that, but I do know that new power supplies were not a major source or error reports on lkml. Over the last year, it is extremely routine, even with new 650W supplies.

Interesting - the SATA drives have not been causing me any problems (so far), but the system is still less than stable under load. /Per Jessen, Zürich -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

On Thursday 14 February 2008, Per Jessen wrote:

Greg Freemyer wrote:

...

Per,

I don't know how you measured the load, but it is important to measure very short power surges. In particular, I'm suspicious of that load during raid1 resync. Could your technique catch a 10 msec spike etc?

Nope, not a chance. See my reply to Carlos.

...

I think it is the spikes that are causing trouble with Sata drives. Don't know that, but I do know that new power supplies were not a major source or error reports on lkml. Over the last year, it is extremely routine, even with new 650W supplies.

Interesting - the SATA drives have not been causing me any problems (so far), but the system is still less than stable under load.

/Per Jessen, Zürich

Have you made certain you have Good thermal connection between the CPU and you heatsink this sounds like a problem i had with this machine turned out it need a cleanup and new thermal compound reseat the heatsink had no problems since might be worth a look one spec of dust just fractionally too big could cause you untold hassle Pete . -- SuSE Linux 10.3-Alpha3. (Linux is like a wigwam - no Gates, no Windows, and an Apache inside.) -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Per Jessen wrote:

peter nikolic wrote:

...

Have you made certain you have Good thermal connection between the CPU and you heatsink

Only as far as I'm using the thermal material that was attached to the heatsink, and I've made sure the CPU and heatsink are both seated properly. Also, the CPU-temperature only rarely exceeds 65C.

In that case, it's probably your north bridge that it getting so hot. Zalman has a VERY nice north bridge heat sink.

...

might be worth a look one spec of dust just fractionally too big could cause you untold hassle

Yeah, it's probably worth checking.

/Per Jessen, Zürich

-- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

On Sat, 16 Feb 2008 13:28:38 +0100, Per Jessen wrote:

copper-coloured heatpipe/-sink on top. AFAICS, it's not really meant to be augmented by additional fans etc.

Normally it's meant to be cooled by the CPU fan. Some manufacturers include a smal clip-on fan for those cases where the CPU is cooled either passively or by water, in which case the heatsinks wouldn't get air. For this kind of heatpipe cooling you definitely need a good CPU cooler and it has to be of the kind that has the fan mounted on top and blowing towards the motherboard. I've made good experiences with the Thermalright XP 120 and SI 128 together with a quiet 120 mm fan like Pabst or similiar. Philipp -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Thursday 2008-02-14 at 17:47 +0100, Per Jessen wrote:

...

Have you made certain you have Good thermal connection between the CPU and you heatsink

Only as far as I'm using the thermal material that was attached to the heatsink, and I've made sure the CPU and heatsink are both seated properly. Also, the CPU-temperature only rarely exceeds 65C.

If the temperature is correct then the heatsink is not the problem. I know of someone that had a problem with a faulty heatsink: some very efficient models are gas filled (actually vapor at the boiling point which is selected to be that of the working temperature). The problem was that the gas had leaked out and the heatsink was very inefficient as a result. Quite difficult to guess... because as the heatsink doesn't work well it remains only warm, so it appears to work well. - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) iD8DBQFHtOCHtTMYHG2NR9URAiFcAJ45iVlJoO7ZpLBBT9WHb7PzeBDkUgCdFp2Z krRdrByag8OgCZEXvsdblKU= =Kx1u -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Friday 2008-02-15 at 09:16 +0100, jdd wrote:

Carlos E. R. a écrit :

...

If the temperature is correct then the heatsink is not the problem.

only if the sensor tests the cpu itself and not some radiator part. Where is the sensor located?

I believe some cpus have an internal sensor. Isn't that standard now? - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) iD8DBQFHtXf8tTMYHG2NR9URAsGVAJ454kzADAcxDmGiIrSaTAbiIGWynACeObk4 TQvkiBrbYgigOKHMz0CecnE= =ta8c -----END PGP SIGNATURE-----

Carlos E. R. wrote:

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1

The Friday 2008-02-15 at 09:16 +0100, jdd wrote:

...

Carlos E. R. a écrit :

...

If the temperature is correct then the heatsink is not the problem.

only if the sensor tests the cpu itself and not some radiator part. Where is the sensor located?

I believe some cpus have an internal sensor. Isn't that standard now?

Internal thermal cut-offs are standard now, but I'm not sure if internal sensors which will give accurate temps are standard yet. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Per Jessen wrote:

Greg Freemyer wrote:

...

Per,

I don't know how you measured the load, but it is important to measure very short power surges. In particular, I'm suspicious of that load during raid1 resync. Could your technique catch a 10 msec spike etc?

Nope, not a chance. See my reply to Carlos.

...

I think it is the spikes that are causing trouble with Sata drives. Don't know that, but I do know that new power supplies were not a major source or error reports on lkml. Over the last year, it is extremely routine, even with new 650W supplies.

Interesting - the SATA drives have not been causing me any problems (so far), but the system is still less than stable under load.

And I'll bet that this instability doesn't come immediately... but after some fairly consistant amount of time after you start the load... Which would take us back to my earlier suspicion of HEAT and that 85 C temperature right before every crash. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Saturday 2008-02-16 at 13:14 +0100, Per Jessen wrote:

If it's the Northbridge getting too hot, it's still likely Gigabytes problem. Maybe the heatpipe/sink was poorly mounted or assembled. I don't see much chance of improving on the cooling for the Northbridge.

If the heat pipe looses the gas, it works very bad. Worse than a solid metal bar. - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) iD8DBQFHttertTMYHG2NR9URAgMdAJ0ZOfHaK/tezQVYg1uMZN8qZyyfzwCdHJvC JXt3p2sMzHsmNRu5/45pgmY= =Ks7s -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

On Thu, Feb 14, 2008 at 9:59 AM, Randall R Schulz wrote:

On Thursday 14 February 2008 06:18, Greg Freemyer wrote:

...

...

Per,

I don't know how you measured the load, but it is important to measure very short power surges. In particular, I'm suspicious of that load during raid1 resync. Could your technique catch a 10 msec spike etc?

Surely the capacitors in the power supply and elsewhere would absorb such brief demand spikes, no?

And what is their source? Logic circuits active only transiently or the voice coil? Obviously the spindle motors have uniform demand once the drive is spun up (but much higher while starting up, of course).

I'm guessing disk seek activity. As the seek activity is getting faster and faster, it takes more and more power to start/stop the heads as they skip around the drive. Apparently if you don't have a very solid PSU, the sata drive starts encountering problems. The trouble with that theory is I would have hoped the drive motors, etc. would be 12v and the electronics 5v, so I really don't understand how the issue happens either. Especially since disk seeks are on the order of 1 or 2 millisecs and a relatively small cap should handle that fine, I did just go back and re-read one of the posts and Tejun Heo said the issue is with the 12v supply and the new video cards that are sucking up all the 12v power. Regardless, most of the reports I have seen have been with 4 or more drives, especially in a raid 5. So it seems that the actively of one drive is causing problems for other drives sharing the PSU. On the lkml-ide list, the very first troubleshooting question for this type of issue is to power up the drives from a separate PSU. I saw one of the lead developers say that 90% of the reported problems currently are caused by power issues, not the kernel code itself. If you want to experiment with using a cheap external PSU: http://modtown.co.uk/mt/article2.php?id=psumod FYI: Sata apparently does not have a ground on its cable, so you don't have to worry about ground loops. This too is just repeating what I've read on the lkml list, so check it out yourself if your curious. Greg -- Greg Freemyer Litigation Triage Solutions Specialist http://www.linkedin.com/in/gregfreemyer First 99 Days Litigation White Paper - http://www.norcrossgroup.com/forms/whitepapers/99%20Days%20whitepaper.pdf The Norcross Group The Intersection of Evidence & Technology http://www.norcrossgroup.com -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Thursday 2008-02-14 at 16:40 +0100, Per Jessen wrote:

One thing I don't understand is why SATA drives should cause a problem when PATAs dont. SATA and PATA is about the interface - the drive mechanics are the same after all. Aren't they?

True. But you only find the fastest and largest models on sata. If it were the 12V spikes, a large capacitor would solve the issue. And even so, I don't see why it would crash: simple the voice coil would be underpowered and run slower, but crash? Weird. - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4-svn0 (GNU/Linux) iD8DBQFHtGT2tTMYHG2NR9URAnNEAJ9OmBIEI4xX5DXgmo6edQqhNDQSHgCglLCb nJpyWAuxWkyHtJhFNxnQ9PQ= =ZewO -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

On Thursday 14 February 2008 11:05, Aaron Kulkis wrote:

Randall R Schulz wrote:

...

On Thursday 14 February 2008 06:18, Greg Freemyer wrote:

...

...

Per,

I don't know how you measured the load, but it is important to measure very short power surges. In particular, I'm suspicious of that load during raid1 resync. Could your technique catch a 10 msec spike etc?

Surely the capacitors in the power supply and elsewhere would absorb such brief demand spikes, no?

Not so much absorbing spikes as much as to provide extra current needed for the spike demanded by system components -- without the current spike the sudden increase in demand results in voltage spike which sags (i.e. goes to ground instead of higher potential).

I meant "absorb the spike in demand by supplying the transient requirement."

...

...

Motor activation is a big one. They all demand power surges to get moving.

But they stay spinning as long as the device is on (excepting a software-driven shutdown of the driver motor, of course). If we're talking about demand spikes on 10-millisecond timescale, they cannot be coming from the spindle motors. Randall Schulz -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

John Andersen wrote:

The power supply the OP mentioned was adequate for everything mentioned, and the only scenario that might draw more would be burning a DVD while doing all those other tasks at the same time as stressing the GPU.

The fact that the OP's problem persists suggests that the power supply was not an issue.

Thanks for confirming my thinking.

I would suggest the OP put on a UPS capable of running the machine for 10 minutes at least. Also, be careful about running the Voltcraft device when plugged into a UPS. Lots of these cheap units are not designed to handle the near square wave produced by the typical UPS.

I happen to have a spare 3kw UPS - not sure if it'll last 10mins, but it might. What would such a test this show, John?

I'd also start digitally measuring temperature of surface mounted components on the mobo. (Digitally as with a finger). The fact that a fan helps but a PSU does not suggests something is poorly heat sinked.

Also what I've been thinking - it's just that I'm now left with what was heatsinked/mounted by Gigabyte. I'll check the CPU tomorrow, but as it hasn't been showing more than 65C, I doubt if there's a problem there. /Per Jessen, Zürich -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

John Andersen wrote:

Also, switch to Console 10 and just watch it while your machine is working hard. You might see some messages come shooting thru when something triggers a protective shutdown.

I've had a serial console hooked up all the time to see if I could catch anything kernel-related - nothing. If only it was a kernel-issue, then I'd have _something_.

You can make some settings to eliminate overly protective shutdown sequences, some in KDE, and some in BIOS.

I've got the CUP-temp warning set at 70C. No other limits that I can see anywhere. The Phenom manual does talk about a TEMPTRIP setting which would make the machine go straight into some down state, from which only a hard reset would bring it out. Doesn't sound like my situation though.

(Oh, check the manufacturer to see if there is later bios available).

Yep, already spoke to Gigabyte and got their latest beta-BIOS. /Per Jessen, Zürich -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Aaron Kulkis wrote:

...

...

(Oh, check the manufacturer to see if there is later bios available).

Yep, already spoke to Gigabyte and got their latest beta-BIOS.

What was the procedure on that?

I wrote to their support line, and asked how to unpack the selfextracting Windows-only executable - they then made the beta available for download. Otherwise it could be unpacked with Wine. The file is copied onto a floppy, then you restart and openthe Q-Flash menu. Pretty painless I have to say. /Per Jessen, Zürich -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Carlos E. R. wrote:

The Thursday 2008-02-14 at 10:19 -0800, John Andersen wrote:

...

Even the cheapest of the cheap powersupplies can handle spikes of the duration you mention. In fact these are designed to supply of capacity tp park heads on disk drives in the event of a total power failure, and this extra capacity is available to handle spikes.

Actually, voice coil type heads park with no power at all. With no current the spring takes the heads to the park position automatically. What you say was true for stepper motor type heads, which have been out of use in PCs for... perhaps 20 years.

Actually, the drives are designed so that the heads can land safely in the data area and so there's no need for such a spring. A few years ago, I pulled the cover off a drive and I don't recall any spring pulling the heads back in it. On the other hand, the old disk pack drives certainly had a retract mechanism, consisting of a large capacitor and a relay that would connect it to the linear motor, in the event of a power failure. -- Use OpenOffice.org http://www.openoffice.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org