Question Burnt EPS receptacle on Asrock X299 ITX

[gecko19]

Hello All,

I was using my Asrock X299 ITX based motherboard today, running a 10980xe at stock speeds(or lower @4.4 GHZ) using a Silverstone SX800-LTI as PSU. The setup has an AMD 6600XT graphics card and is installed within a thermaltake G3 slim case. The machine has been used for about a 1000 hours since installation. Most of the time it does heavy computation work.

Today, I was running a multi core build and noticed a burning smell, the machine continued running still. I switched off the power and then noticed that the EPS receptacle was burnt. I can get a new receptacle and solder a new one on. However, my question is more on :

1. Why did it happen,
2. Will it happen again.
3. How can I ensure this does not happen.
4. Should I change the motherboard
5. Should I change the power supply?

Burnt Receptacle
Burnt Connector

 

[helpstar]

did you use the original cabling of this PSU?

did you plug in all 24 ATX pins into the PSU?
https://www.silverstonetek.com/upload/images/memo/4pin-sense.jpg

 

[drea.drechsler]

Hello All,

I was using my Asrock X299 ITX based motherboard today, running a 10980xe at stock speeds(or lower @4.4 GHZ) using a Silverstone SX800-LTI as PSU. The setup has an AMD 6600XT graphics card and is installed within a thermaltake G3 slim case. The machine has been used for about a 1000 hours since installation. Most of the time it does heavy computation work.

Today, I was running a multi core build and noticed a burning smell, the machine continued running still. I switched off the power and then noticed that the EPS receptacle was burnt. I can get a new receptacle and solder a new one on. However, my question is more on :

1. Why did it happen,
2. Will it happen again.
3. How can I ensure this does not happen.
4. Should I change the motherboard
5. Should I change the power supply?

Burnt Receptacle
Burnt Connector

I'm pretty sure that connector powers only the CPU. Considering it did that running in stock configuration, i.e. not overclocking, I'd suggest you need a new motherboard, one that's compatible with that CPU. It would have a 12 pin connector at least for that 165W TDP CPU.

Excessive current draw is what melted the connector and that results from power draw of the CPU. Or more precisely the VRM that powers the CPU. If it drew enough power to melt the connector the VRM is very likely damaged and, at best, unreliable; another reason to replace the board.

Whether you change the PSU or not depends on how badly it was overloaded...and if it's of decent enough quality to keep anyway. You DEFINITELY have to change that power cable so if it's not modular you'll need some skills or a new PSU.

 

[gecko19]

did you use the original cabling of this PSU?

did you plug in all 24 ATX pins into the PSU?
https://www.silverstonetek.com/upload/images/memo/4pin-sense.jpg

Hello Helpstar,

No, I did not connect the sense wire - I will from now on if I do get it back to working state 😊

 

[drea.drechsler]

....
No, I did not connect the sense wire - I will from now on if I do get it back to working state 😊

Definitely do so...but not sure it would help much with your problem...

[SOLVED] - Importance of the 4-Pin Sense in the PSU

Hi Everyone, I have a quick question on the importance of the 4-pin sense in the PSU. Here's a bit of a background: I've previously purchased and used my Silverstone SX800-LTI PSU on a small form factor. But due to thermal issues, I had to move all the components to a slightly bigger case. The...

forums.tomshardware.com

Your problem is way too much power-hog CPU on an ITX board not suitable for it running heavy computational work, as you put it, for long periods (1000 hours).

 

[gecko19]

Thanks all. I guess this is a lesson for me. I should have spend more time to research a better ATX case. Had the case been bigger the power supply would have been a normal one, which could then mean I could have a normal board with normal EPS pin outs for extra power. In essence, the motherboard goes to the bin. The case, the SODIMM dram, the CPU and the power supply need to be sold. Hopefully people like myself who come upon this thread will rethink their decision to provision for huge computing power in a small form fact using a small form factor motherboard.

 

[drea.drechsler]

Thanks all. I guess this is a lesson for me. I should have spend more time to research a better ATX case. Had the case been bigger the power supply would have been a normal one, which could then mean I could have a normal board with normal EPS pin outs for extra power. In essence, the motherboard goes to the bin. The case, the SODIMM dram, the CPU and the power supply need to be sold. Hopefully people like myself who come upon this thread will rethink their decision to provision for huge computing power in a small form fact using a small form factor motherboard.

I think it can be done successfully just choose a more efficient platform. You may not like considering AMD but a Ryzen 9 5950X would be a far more powerful processor and runs well on many ITX boards, drawing a lot less power in the process. It would do very well in your Thermaltake G3 slim as it fits a 240mm AIO comfortably. I'm not sure there are many AM4 ITX boards that fit SODIMM dram's, though.

 

[gecko19]

I think it can be done successfully just choose a more efficient platform. You may not like considering AMD but a Ryzen 9 5950X would be a far more powerful processor and runs well on many ITX boards, drawing a lot less power in the process. It would do very well in your Thermaltake G3 slim as it fits a 240mm AIO comfortably. I'm not sure there are many AM4 ITX boards that fit SODIMM dram's, though.

Thanks @drea.drechsler .
I worked out some numbers based on the ATX PSU spec, the EPS pin should deliver a maximum of 20A at 12v which equates to 240W. The Mainboard 24pin connector should support 12A for the 3.3V rail, 14A for the 5V rail, and 20A for the 12V rail.

On the 12V rail, my burnt out configuration should have supported 480W, though the 10980xe was at stock so maybe using ~350W . Sorry for bringing this up again, but I am trying to gauge what caused the failure(I have another machine running the same config and I am speculating changing the board with an atx board with 2 EPS pins).

Is it safe to say that the Silverstone SX800 LTI should have throttled the power delivered to the EPS pin which could have then prevented the burn out, or was it a bad cable, maybe?

 

[kerberos_20]

The 8-pin connector for CPU (the EPS12V Connector, also called EATX12V) is rated for 7 Amps per wire pair (4 pairs at 7 amps = 28 amps), and is rated for 336 Watts maximum (12V * 28 Amps = 336 Watts).

what most likely happened is that connector wasnt fully inserted

 

[BFG-9000]

In essence, the motherboard goes to the bin. The case, the SODIMM dram, the CPU and the power supply need to be sold.

I certainly wouldn't trust that PSU EPS connector anymore either as it's been overheated. And if you were going to bin the mobo and PSU anyway, why not just solder them together without any connectors? Can't get any lower resistance than that.

That motherboard was designed for 165w TDP 79xx CPUs and as mentioned the 8-pin is normally fully sufficient for that, hence why they designed it that way. However melting 4-pin P4 connectors were long a thing on <100w CPUs so this may be kind of rare but not exactly unusual, especially if the rated 50 mating cycles was exceeded. You could've just got unlucky with a bit of dirt or plastic causing enough resistance for it to start getting hot, leading to a runaway feedback loop.

 

[drea.drechsler]

...
On the 12V rail, my burnt out configuration should have supported 480W...
...

That may be, if the connections are tight and perfect. Molex connectors often use a rolled socket contact that can fit loosely on the pin with only a few mate-unmate cycles. A loose contact will be resistive which will generate a lot of heat at that point; high heat makes it more resistive which generates more heat and thermal run-away is off to the races. That's why many boards use 12 and even 16 pin connections when 4 or 8 might do "by the numbers"; greater redundancy helps prevent burn-ups.

I don't know of any PSU mechanism to throttle the CPU in cases of high hook-up connection temperature. If anything it will keep voltage constant at the input of the CPU VRM to compensate for the voltage drop of the resistive contact.