Can I multi-thread the CPU tasks?

Mr P Hucker
Mr P Hucker
Joined: 12 Aug 06
Posts: 819
Credit: 481369262
RAC: 4124

Keith Myers wrote:Most good

Keith Myers wrote:

Most good power supplies will crowbar their outputs if they detect trying to drive a dead short.  So that is why the power supply won't show any output with the cards plugged in.  Probably why there was little smell, the card didn't get supplied +12V long enough to really burn things up.

I'm surprised at how quick though.  Presumably it has to allow a "short" for a short time on startup, to fill smoothing capacitors etc.  But I saw no evidence of anything happening.  No fans on the computer, or the PSU's own fan, even twitched.  Mind you, if they're connected to the same shorted 12V rail they wouldn't....

Also please explain "crowbar their output" - I thought to crowbar an output meant to short it.  What would be achieved by shorting an already shorted output?

As for the smell, I have smelt the same smell previously, while things were working.  This was about a week or so ago, and I couldn't work out if the smell was in the air coming from the GPU or the PSU.

 

If this page takes an hour to load, reduce posts per page to 20 in your settings, then the tinpot 486 Einstein uses can handle it.

Mr P Hucker
Mr P Hucker
Joined: 12 Aug 06
Posts: 819
Credit: 481369262
RAC: 4124

Update, I've now dismantled

Update, I've now dismantled it, and can see nothing wrong in what I assume is the power regulation area.  It looks like this (why can't this forum see dropbox images inline?):

https://www.dropbox.com/s/qvylgzwcaefc7nd/Broken%20card.jpg?dl=0

And I think this is the problem on the back, next to the fan connector.  Looks like the broken fan shorted out, and has melted some PCB.  This is way too small to repair, as I'm not a surgeon!:

https://www.dropbox.com/s/5u5uz4kstssh91n/Fan%20connector.jpg?dl=0

And what's more, the two 12V PCI Express power connectors are both showing 0.25 ohms (the fan has been unplugged).  But on another R9 280X (although a different make) those connectors show 7 k ohms.  I don't know how that circuit is supposed to work, but that's a major difference between the two cards.  Strangely though, the fan is at 12 k ohms, so not shorted just now.

If this page takes an hour to load, reduce posts per page to 20 in your settings, then the tinpot 486 Einstein uses can handle it.

Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 4704
Credit: 17548332938
RAC: 6430026

What is Crowbar

What is Crowbar Protection?

I would first look at the PCIe power connectors on the card for signs of melted plastic and burnt pins.  They would probably be the likely culprit for a temporary smell and the card continued to work for several more days.

 

Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 4704
Credit: 17548332938
RAC: 6430026

The driver transistors to the

The driver transistors to the left of the fan connection strip are probably shorted.  Hence your 0.25 ohm.  I can't tell if the traces from the fan lead solder pads to the driver transistors are open.  Too much crud covering them up.  Take a toothbrush and some 91% isopropyl alcohol and scrub the area around the fan solder connection points clean to see things better.

Q201, Q203, Q210 and D202 are suspects.

Without the schematic for the card, all you can do is trace +12V buses on the card and see where they go.  I would figure out which trace feeds the fan connections and the driver transistors and cut the trace leading to them.  Then measure the input PCIE connectors and see if the short is removed and now measures the same as a good card.  That way you know which area is shorting.  Then figure out which transistor is shorting from the high to low side and replace it.

If you have a curve tracer or a scopemeter with a curve tracer function, I would measure the driver transistor first off to see which are bad.  That way you don't have to cut the supply trace to isolate the short.

 

Mr P Hucker
Mr P Hucker
Joined: 12 Aug 06
Posts: 819
Credit: 481369262
RAC: 4124

Keith Myers wrote:What is

Keith Myers wrote:

What is Crowbar Protection?

Isn't that for when the power supply is giving out too many volts?  Rather than my problem of the load taking too many amps?

And why don't they make it cut off the power instead of shorting it?

Keith Myers wrote:

I would first look at the PCIe power connectors on the card for signs of melted plastic and burnt pins.  They would probably be the likely culprit for a temporary smell and the card continued to work for several more days.

I've dismantled the whole thing and all I can find is the melted PCB around the fan connector.  If it doesn't sell for a reasonable price, I'll try to repair it.

Keith Myers wrote:

The driver transistors to the left of the fan connection strip are probably shorted.  Hence your 0.25 ohm.  I can't tell if the traces from the fan lead solder pads to the driver transistors are open.  Too much crud covering them up.  Take a toothbrush and some 91% isopropyl alcohol and scrub the area around the fan solder connection points clean to see things better.

Q201, Q203, Q210 and D202 are suspects.

Without the schematic for the card, all you can do is trace +12V buses on the card and see where they go.  I would figure out which trace feeds the fan connections and the driver transistors and cut the trace leading to them.  Then measure the input PCIE connectors and see if the short is removed and now measures the same as a good card.  That way you know which area is shorting.  Then figure out which transistor is shorting from the high to low side and replace it.

If you have a curve tracer or a scopemeter with a curve tracer function, I would measure the driver transistor first off to see which are bad.  That way you don't have to cut the supply trace to isolate the short.

Thanks, I didn't think of the transistors - are those to control fan speed?  Since the card is just for Boinc and needs a lot of cooling, I'd probably just do without those transistors and connect the fans straight to 12V (which is what I've done with other cards where the fans have failed and I happened to have some non-4-pin fans kicking about - and they don't have to be the right size either, I've strapped 140mm fans to cards with cable ties).

I only have a basic scope made from an Arduino.  If there's an easy way of using that....

If this page takes an hour to load, reduce posts per page to 20 in your settings, then the tinpot 486 Einstein uses can handle it.

Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 4704
Credit: 17548332938
RAC: 6430026

A crowbar circuit is just a

A crowbar circuit is just a generalization of a protection circuit.  Not exclusive to just an overvoltage condition.  The circuit is designed to protect the device under power and also the power supply from an abnormal condition.

From your backside image, I don't see anything rudely burnt up.  Should be able to repair it with just a few components replaced and the fan.

 

Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 4704
Credit: 17548332938
RAC: 6430026

Yes, the transistors provide

Yes, the transistors provide the PWM switching of the fan rpm. I'm guessing those transistors are connected to the fan tag strip.  But you would have to visually trace their inputs and outputs and ohm the transistor output to fan tag strip points to be sure. Normally would have the PWM control circuit somewhere close to the vicinity of the fan connection point.

You would need a curve tracer box to connect to the scope and set up for X-Y connections.  A curve tracer is just a circuit that applies a low voltage AC signal to the device under test.  Or better ones use a sweep or sawtooth waveform.  You use it to plot the I-V  curve on the X-Y axis of the scope to look at a semiconductor junction.  eBay search on curve tracer and you will find tons of hits on either kits or finished projects from basic to full-featured.

 

Mr P Hucker
Mr P Hucker
Joined: 12 Aug 06
Posts: 819
Credit: 481369262
RAC: 4124

Keith Myers wrote: From your

Keith Myers wrote:

From your backside image, I don't see anything rudely burnt up.  Should be able to repair it with just a few components replaced and the fan.

Aren't these PCBs multilayer?  I can't get at any tracks in the middle....

Assuming I can repair the tracks, I'll just remove the transistors.  I don't need fan speed control.  Or would the rest of the circuit get upset that I've stolen it's accelerator pedal?

If this page takes an hour to load, reduce posts per page to 20 in your settings, then the tinpot 486 Einstein uses can handle it.

Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 4704
Credit: 17548332938
RAC: 6430026

I'm sure the PCB is

I'm sure the PCB is multilayer.  Impossible to say whether the 12V bus is on the surface or buried in the substrate and brought to the surface via vias and through holes. You'll have to figure that out with probing with a multimeter.

I would just remove the short on the 12V bus from whatever component is shorted and find the nearest 12V exposure on the board and bridge it over to the fan tag strip area and just have straight 12V power the fans at full speed.  The PWM circuit could care less that it is not driving anything.

 

Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 4704
Credit: 17548332938
RAC: 6430026

I would hold the PCB up to a

I would hold the PCB up to a very strong backlight and you should be able to see the 12V traces from the PCIE input connector to the fan area. The traces will be fairly wide compared to signal traces. Ohm the likely suspects out for your connection point.

But first you will need to determine which component(s) are shorted from 12V bus to ground.

 

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.