RX-480 pulling more power than spec

[Artorius]

They're pulling this much from it because they overvolted the chips as usual, and 8gbs of GDDR 5 consumes a lot of power.

AMD did the same crap at the 290X, Nano, Fury/X and the 480. They're apparently afraid of some of the chips not being able to keep themselves stable at the boost clock so they overvolt all of them... You can easily alleviate the 480's power draw by 20~30W without losing a single frame of performance with a undervolt, it's laughable.

[Remilia]

Yup, they really should find a consistent less overvolting on their cards. I can set my R9 290's power limit to -30% and it won't lose any performance, and setting it to -50% it'd lose about 10% performance but use a lot less power. Same issue here sadly.

[Deleted]

Yup, they really should find a consistent less overvolting on their cards. I can set my R9 290's power limit to -30% and it won't lose any performance, and setting it to -50% it'd lose about 10% performance but use a lot less power. Same issue here sadly.

Not really. I don't think the power level affects the voltages at all, only throttling.

Say the default limit of you card is 200W. If you set it to -30% the card will start throttling down if it draws more than 140W. If you don't notice any performance difference then your card drew less than 140W all the time during your performance test, so you did not save any power by reducing the setting to -30%.

Accordingly the -50% setting didn't save 50% power since that card didn't draw the maximum 200W to begin with.

[kail]

What are the chances of non-reference 480s having the same issue?

[mrgreenthump]

What are the chances of non-reference 480s having the same issue?

Most of them have 8 pins for advanced overclocking, so probably none.

[kail]

Most of them have 8 pins for advanced overclocking, so probably none.

Cool, 480 seems like a buy from me once after-market coolers become available. After seeing the expected price of the 1060, that card will have to bench pretty well for me to reconsider.

[Deleted]

I've dealt with a number of people on this issue and there's a lot of questionable things going on. Firstly, there's a YouTuber who tested the RX 480 on a budget AMD build with a AM2 Foxconn motherboard. This machine would shut down as if too much power was being pulled. He did the same test on the 980 Ti, and no problems. This is a really old motherboard, like 2006 old. He bought it used of course. While Foxconn does make everything, but their brand boards are built on the cheap. Worst quality stuff, with worst designed motherboards. You never buy them. Secondly, a 10 year old motherboard is going to have weak components. Capacitors age, and so do VRMs. I asked the YouTuber to find another AM2 motherboard but of brand name quality like Asus or Gigabyte, with the same chipset. That way you can be sure to say it wasn't the RX 480, but low quality motherboard. The YouTuber didn't have nice things to say to me, and this is a guy with 50k subscribers.

Another person had yellow goo on the board on what looked like an exposed case. You know a PC with missing metal PCI Slot covers. Looks like someone spilled a drink near the PC, and some liquid splashed inside onto the motherboard.

Also, why now people are interested in PCIE power consumption? When did people start testing for this, right when the RX 480 was released? Just seems like people are stretching this very far like they're reaching for something. I'd also like to see other graphic cards tested as far back as 5 years to see if the RX 480 is the only graphics card doing this, cause I really doubt it. I know why AMD is pulling so much power, because they wanted that 6-pin connector, but the situation is exaggerated.

Well youtubers yes but then there's Miners and Mining and a Low Spec Motherboard = Fun. It's maybe Extremes like affecting 0.01% of people yet still it's the 1st card that did it.

https://bitcointalk.org/index.php?to...55#msg15438155

[Vash The Stampede]

New drivers are out to fix the issue. Waiting to see if the power did go down.

[Deleted]

New drivers are out to fix the issue. Waiting to see if the power did go down.

It did. Marginally higher power draw overall, but less via mainboard and more vie 6-pin:

http://www.tomshardware.com/reviews/...-fix,4668.html

Should be absolutely harmless now; even given that earlier reports about problems might have been exagerated, fake or extreme outliers.

[Stormspark]

I watched an in-depth technical analysis of this problem, and it looks like AMD did it intentionally to make it appear that the card draws less power than it actually does. AMD is apparently scared of Pascal.

But yeah, the only motherboards that would've been ok with the card as shipped, would be high end ones. And well, people that have high end systems aren't going to be getting 480's.

[Vash The Stampede]

I watched an in-depth technical analysis of this problem, and it looks like AMD did it intentionally to make it appear that the card draws less power than it actually does. AMD is apparently scared of Pascal.

Doesn't make sense when reviewers measure overall power draw from the outlet. Regardless where the RX 480 pulls power from, you'll see it from the outlet. It's my belief that AMD did it to avoid going beyond a 6-pin connector Either that or the engineers weren't giving enough time to correct the issue.

[Deleted]

It appears that the problem is completely solved by the new driver, and manages to eek out 1/2 extra fps in most of the games.

Having tested a reference 480, cannot wait now for the custom aib with 8 pin and 8+6 pin connector.

You can make fun of this comment later this month if this cheap card wont surpass the Gtx 980, funny how solid the overclock is even with a 6 pin and modest cooling.

[Artorius]

Doesn't make sense when reviewers measure overall power draw from the outlet. Regardless where the RX 480 pulls power from, you'll see it from the outlet. It's my belief that AMD did it to avoid going beyond a 6-pin connector Either that or the engineers weren't giving enough time to correct the issue.

Normal. Their uarch engineers do their jobs properly but the lower ones in the hierarchy can't design a PCB correctly nor tune the manufactured chips with sane values. It's done to increase the number of usable chips, but they could simply send the best ones to their partners to be sold at higher clocks while using the bad chips on the reference cards. It isn't an engineering problem to be honest, it's a logistic problem.