AMD's Linux Driver Will Likely See A Power Change For The Radeon RX 480 Too

[phoronix]

Phoronix: AMD's Linux Driver Will Likely See A Power Change For The Radeon RX 480 Too

By now you may have heard that there is the potential for the Radeon RX 480 to draw more power from the PCI-E bus than it's rated to provide. In rare situations, this could potentially cause problems for the system. AMD/RTG is preparing to release a Windows driver fix while I checked in with AMD about addressing this situation under Linux...

http://www.phoronix.com/scan.php?page=news_item&px=AMD-RX-480-Power-Issue

[eydee]

If you don't have such a card yet, just wait until partners release versions with 8-pin connectors. For them it's more important to keep customers than then the forced compliance to nvidia AMD has. In my opinion the power issue itself is a lot less of an issue compared to that AMD let this slide and go into production just to avoid nvidia bitching about the 8-pin connector. As if AMD management just turned into a bunch of politicians...

[L_A_G]

There's a bunch of other cards that have done exactly the same thing in the past and they're not even all AMD cards (Nvidia messed up with certain versions of the 750 Ti and 960) and none of them have broken motherboards. The only confirmed case of a PCIe bus being ruined when powering an RX 480 was with an Abit board that was downright filthy with liquid damage and loads of dust, so it could be that it wasn't even the 480's fault.

Don't get me wrong however, I'd still recommend waiting for partner cards with 8 pin connectors and better cooling as the 6 pin reference cards are power limited and running really hot due to a really crappy reference cooler.

[lbalbalba]

Sorry, but this article is about overclocking, which - by definition - means pushing the hardware beyond the specifications of the vendor. You choose to do something with it, that the vendor did not intend. If you choose to push the hardware beyond the specifications provided by the vendor, then the results are solely your responsibility too. You may feel that, because you bought and own the hardware, it's up to you to do whatever you wish with it: but that only goes so far. If you choose to set the hardware on fire - because you feel that it's your unalienable right to do with it whatever you want - and then you proceed to stick your hands in it (ouch, that hurts!), then that's entirely your fault, and not the fault of the hardware vendor.

[pipe13]

Sorry, but this article is about overclocking, which - by definition - means pushing the hardware beyond the specifications of the vendor. You choose to do something with it, that the vendor did not intend. If you choose to push the hardware beyond the specifications provided by the vendor, then the results are solely your responsibility too. You may feel that, because you bought and own the hardware, it's up to you to do whatever you wish with it: but that only goes so far. If you choose to set the hardware on fire - because you feel that it's your unalienable right to do with it whatever you want - and then you proceed to stick your hands in it (ouch, that hurts!), then that's entirely your fault, and not the fault of the hardware vendor.

I'm not sure I agree with that. I've never been an overclocker and never will be But I'm not sure I agree. AMD or Nvidia or Intel sell an unlocked card or CPU with an explicit "Here's your new toys, kiddies: bang on 'em to your hearts content!!!" With the implicit assumption that the worst that can happen is frame tearing or BSOD -- because that's all ten thousand OC sites have ever seen happen -- then its best to live down to that expectation. And if that means a minor firmware clock limit to prevent physical damage to the customers' PC when customer does something he or she has been lead -- by you -- to believe is safe, well then, its a simple software fix. Warn the early adopters, and fix it as soon as practical.

Just as AMD is doing. Good on them. Sheeesh.

[Edit. Bus limit, clock limit, power regulator limit, whatever...]

[zamadatix]

Sorry, but this article is about overclocking, which - by definition - means pushing the hardware beyond the specifications of the vendor. You choose to do something with it, that the vendor did not intend. If you choose to push the hardware beyond the specifications provided by the vendor, then the results are solely your responsibility too. You may feel that, because you bought and own the hardware, it's up to you to do whatever you wish with it: but that only goes so far. If you choose to set the hardware on fire - because you feel that it's your unalienable right to do with it whatever you want - and then you proceed to stick your hands in it (ouch, that hurts!), then that's entirely your fault, and not the fault of the hardware vendor.

No, actually, it's about the stock card at stock clocks and the amount of power drawn over the PCIe bus as found by multiple reviewers and buyers. At least find out what the story is before you start spouting your irrelevant opinions.

480 was a good release but it is odd they went with 1 6 pin power connector. You'd think they'd want the reference card to have OC potential for reviews on release not be at (over) max safe power draw. Non stock cards will have more power of course.

[Veerappan]

No, actually, it's about the stock card at stock clocks and the amount of power drawn over the PCIe bus as found by multiple reviewers and buyers. At least find out what the story is before you start spouting your irrelevant opinions.

This. The issue with the reference design of the RX 480 is that it is drawing over 75w over the PCIe bus (the max the spec allows), and is also drawing over 75w using the 6-pin power connector WHEN RUN AT STOCK CLOCKS.

The driver that AMD is working on for windows will reduce the amount of power that is drawn from the motherboard, and balance that by drawing more over the 6-pin connector (which usually have thicker guage wires and can handle more current than the traces on the motherboard). It'll still be out of spec, but less likely to damage components (not that it was likely to happen in the first place). The optional setting for the windows driver (Compatibility Mode) will force the card to adhere to the actual specifications (75w max over PCIe + another 75w over 6-pin).

The Compatibility Mode setting will probably force slightly lower voltages and memory clocks and the performance lost has been more than balanced by efficiency improvements in the driver logic itself (or so AMD claims), so in all the card will possibly run faster with this driver update than before it.

[johnc]

There's a bunch of other cards that have done exactly the same thing in the past and they're not even all AMD cards (Nvidia messed up with certain versions of the 750 Ti and 960) and none of them have broken motherboards.

That's actually not correct. The 750 Ti and 960 went out of spec for very very brief spikes (like microseconds), not for sustained periods. In effect, it isn't the same thing.

I'm not saying the RX 480 would fry motherboards, but who knows what could happen after a few years of use.

[jaguar]

That's actually not correct. The 750 Ti and 960 went out of spec for very very brief spikes (like microseconds), not for sustained periods. In effect, it isn't the same thing.

I'm not saying the RX 480 would fry motherboards, but who knows what could happen after a few years of use.

That's actually not correct.

http://www.tweaktown.com/reviews/6148/msi-geforce-gtx-750-ti-2gb-twin-frozr-gaming-oc-video-card-review/index23.html

750ti pulling the same watts as a 270x a card that comes with 6 pin connector .
The truth is that nvidia is spreading fud all over the place
and as michael said his card is running fine

[soulsource]

I'm just hoping that the driver optimizations don't negatively affect performance of third party cards which are designed with a sufficient power connection...