RX 480 power gate Experiment

[gilang01]

OK here we go. 

As we know the RX 480 has been getting a ton of scrutiny due to the PCIE power gate. It started off on Tomshardware, then reddit, then here. 

There's a post on amd's forum (https://community.amd.com/thread/202410) 

In light of this issue, JagatReview (an Indonesian review site) has taken a direct step to test the power gate. 

The resident overclocker (hwbot 2015 world tour Asia champion) strapped a 1425 mHz rx480 (there's a reddit post on this gpu)(1150mV) to a cheap Msi H81 motherboard. 

They ran heaven benchmark for 2 hours (and still going) and found no issue with the pcie connectors (again, still running).

 

I will keep you updated if anything interesting comes up

[Jacktastic-Mofo]

Wait they have a beefy aftermarket CPU cooler attached to it?

[gilang01]

2 minutes ago, Jacktastic-Mofo said:

Wait they have a beefy aftermarket CPU cooler attached to it?

It's a gpu cooler. Pc cooler ks120se

[Starelementpoke]

2 minutes ago, gilang01 said:

It's a gpu cooler. Pc cooler ks120se

Google doesn't bring anything up when name is googled?

[Jacktastic-Mofo]

2 minutes ago, gilang01 said:

It's a gpu cooler. Pc cooler ks120se

I think there is potential for that to skew the possible results. If these guys know anything about conducting an experiment is that you want to do it without any factors that could potentially change the outcome. An aftermarket cooler on a stock GPU could potentially do that.

[Starelementpoke]

2 minutes ago, Jacktastic-Mofo said:

I think there is potential for that to skew the possible results. If these guys know anything about conducting an experiment is that you want to do it without any factors that could potentially change the outcome. An aftermarket cooler on a stock GPU could potentially do that.

To be fair, this is to see if the 480's power draw from the PCIE lane would damage modern low end boards. Don't see how making sure it keeps drawing that power would skew it.

[Jacktastic-Mofo]

1 minute ago, Starelementpoke said:

To be fair, this is to see if the 480's power draw from the PCIE lane would damage modern low end boards. Don't see how making sure it keeps drawing that power would skew it.

Experiments should be performed without any factors that could effect the outcome. This test is to see if this will occur with the stock GPU, correct? It should be left as stock to see if anything may occur.

[SImoHayha]

5 minutes ago, Jacktastic-Mofo said:

I think there is potential for that to skew the possible results. If these guys know anything about conducting an experiment is that you want to do it without any factors that could potentially change the outcome. An aftermarket cooler on a stock GPU could potentially do that.

probably wouldn't do anything its a cooler, if anything it would increase power draw or not. But who knows.

[stconquest]

9 minutes ago, Jacktastic-Mofo said:

I think there is potential for that to skew the possible results. If these guys know anything about conducting an experiment is that you want to do it without any factors that could potentially change the outcome. An aftermarket cooler on a stock GPU could potentially do that.

The experiment is not to test cooling.  It is to test maximum power draw through the PCIE slot.  The cooler is fine, it allows for more voltage due to the fact it cools better.

 

At 1425MHz, the 480 is drawing 243W.  The reference cooler would not deal with that heat well.

[Starelementpoke]

1 minute ago, Jacktastic-Mofo said:

Experiments should be performed without any factors that could effect the outcome. This test is to see if this will occur with the stock GPU, correct? It should be left as stock to see if anything may occur.

No, just to see if the power draw of the 480 will damage a modern low end mobo's PCIE slot. Doesn't really matter the cooler, just to see if it'll happen. 

[Biggerboot]

Not sure if the 'grade' of the motherboard is a factor on how likely it would be for the card to overdraw - it's not like better ESD protection or capacitors would be a deciding factor (or whatever minute differences there are between the H, B, and Z boards besides features), although I'm sure some boards are more tolerant than others for other reasons.  The board shouldn't be designed to put out more than 75w on the PCIe though. 

 

I would cap the peak performance on my RX for now until it's been a while and we either have a fix, or know what boards it happens on.

[stconquest]

3 minutes ago, Biggerboot said:

Not sure if the 'grade' of the motherboard is a factor on how likely it would be for the card to overdraw - it's not like better ESD protection or capacitors would be a deciding factor (or whatever minute differences there are between the H, B, and Z boards besides features), although I'm sure some boards are more tolerant than others for other reasons.  The board shouldn't be designed to put out more than 75w on the PCIe though. 

 

I would cap the peak performance on my RX for now until it's been a while and we either have a fix, or know what boards it happens on.

...also, signs of motherboard failure may not start appearing for quite a while....weeks... months even.

[Starelementpoke]

2 minutes ago, stconquest said:

...also, signs of motherboard failure may not start appearing for quite a while....weeks... months even.

Could take years. Decades. etc. etc.

[Kinda Bottlenecked]

1 hour ago, Jacktastic-Mofo said:

I think there is potential for that to skew the possible results. If these guys know anything about conducting an experiment is that you want to do it without any factors that could potentially change the outcome. An aftermarket cooler on a stock GPU could potentially do that.

 

Its probably there to keep the core cool while the card is clocked at 1425mhz. The underwhelming cooler on tge reference probably wouldn't allow that. This'll allow us to see overclocked scenarios with the highest peak power draw. 

 

Better than running a card on stock where it doesn't draw as much power.

[gilang01]

This is an ongoing experiment to see whether the high draw and voltage can ruin the port over time (7 hours of gaming according to the amd community post)