The 480: power consumption, PCI-E powerdraw

[zlatan]

The PCI Express slot specifies +/-8% tolerance. Also a lot of multimeter is not perfectly accurate. Are these calculated in the results?

 

[sm625]

ya, see Glo's post in the big thread

http://forums.anandtech.com/showpost.php?p=38327859&postcount=643

That review shows tests that point to what was expected--max ~128 W draw at 4k in Ashes, averaging at ~110.

Performance was even showing Fury and better at 1440p, coming really close to Titan X in some titles as well...which I think no one expected? It's strange, because I haven't seen that in any of yesterday's reviews.

If every review site posted results like those, I'm sure our impression of the RX480 would be completely different. Maybe this review has time traveled and is from 3 months in the future?

 

[xthetenth]

^ And that's why I don't believe their numbers but desperately want them to be true.

Does anyone have connections to a motherboard engineer who could give their thoughts on this?



i used to think launching AIB custom first was great, but with this it might help AMD to launch them later. Because now the 1060 is coming out around the same time. If the custom cards gain a boost in performance, they will make the RX 480 look better than the 1060 when the 1060 launches.

I'd laugh so hard if NV tried to preempt AMD's launch with a picture of a cooler, and AMD preempted NV's launch by sending out new cards with good coolers.

 

[Elixer]

Slot power?

 

[zinfamous]

If every review site posted results like those, I'm sure our impression of the RX480 would be completely different. Maybe this review has time traveled and is from 3 months in the future?

That review does seem more like an outlier than most, though. While not all had the power issues, as reflected here, none seemed to report performance like this.

So, some consistency here in that power efficiency is either good or bad (seems like a lottery issue--the scale of this problem remains to be seen and what it means for the rest of AMD's line in the next year), OC is really quite terrible everywhere (could be drivers but even so--looks like this is really limited by architecture, process, chip size)...and performance is just all over the place.

The one commonality with performance is that it really shines in DX12, and this is overlooked in reviews, imo (yeah, I guess games are still limited, but this has been consistent with AMD so it should neither be surprising or ignored)

 

[Stuka87]

Slot power?

Thats total power for a device that is plugged into a slot.

This was an issue with the 295x2 came out and went past the 300W peak slot rating, which AMD admitted was past the rating.

 

[railven]

Haha this topic has been dubbed PowerGate, at least over on r/AMD.

People are digging up specifications and documentations on this one. Haha.

One guy made a comment, it spread like wildfire, he updated his comment sort of redacting his prior comment, crickets to his update.

I'm sure AMD is safe, I doubt they'd release a product knowing it was out of spec. Now, them releasing a buggy driver - that I believe.

This could all be Wattmans fault for all we know.

AMD introduces new power management feature with new cards, new cards go ape on Power draw. It makes sense to me.

 

[Final8ty]

It's actually worse on the 960 since it spikes much higher but no one complained or had their mb destroyed yet.

Question. why is it such a bad thing for the RX 480 to briefly draw 80 Watts through the 75 Watt PCIe but ok for the 750TI to briefly draw 140 Watts through the PCIe?

In fact acording to this chart the 750TI is drawring about 100 Watts through the PCIe almost as much as it is 60 Watts.

I asked in the other thread if all this is a real issue or just hyperbole.

The 750TI peaks at 140 Watts on the board PCIe, its spends about 50% of its time at 100 Watts on the board PCIe.

Toms Hardware did not see that as an issue, Nvidia did not see that as an issue, they designed and built the GPU.

Why are 80 to 90 Watt power spikes at the board PICe a catastropy when it comes to an AMD card?

https://forums.overclockers.co.uk/showthread.php?t=18738173

 

[Azix]

Because its AMD. There had to be something to dampen the RX 480 fever.

 

[LTC8K6]

We skipped long-term overclocking and overvolting tests, since the Radeon RX 480’s power consumption through the PCIe slot jumped to an average of 100W, peaking at 200W.

Tom's description of why they didn't overclock.
And we see an average of 80 watts during the Metro run for the 480?

That looks considerably worse than the 750ti? Average of 64 watts?

 

[96Firebird]

Do people not understand the difference between average and peak?

PCI specs use averages, because peaks are apart of electrical components.

 

[Despoiler]

Has amd commented on this whole power issue beyond "we are working on it"? It just kind of seems like they're:

Raja himself commented in the Reddit AMA/RX480 that he was aware of some reviewers showing abnormal power readings and that they had engineers working with them to try to duplicate.

 

[Shivansps]

The 75W limit comes from pci-e 1.0 rev, and as devices must be retro-compatible, they must meet this, when pci-e was introduced extra pins where added to the ATX plug in order to provide the power for pci-e slots, rev 2.0 raised the limit to 150W and rev 3.0 to 300W, but since no extra reinforcements was added, there is just no way to make a MB up to PCIE power specs, so they all settle on 75W as is the 1.0 requeriments, and is optional to them to provide extra power.

Anyway, there is a reason of why +75W devices must use external power even on a 3.0 slot.

This issue with the PCI-E is just a footnote on reference RX480 problems, anyone with 1 card should be OK, CF could be a long time problem as this could damage motherboard/PSU ATX connector.

You may be able to kill a PCI-E 1.0 motherboard thought, but who gona put a RX480 on a 1.0 mb?

On 2.0 and 3.0 you should be OK on passing this limit by a few W, still there is more chance to melt the ATX connector before killing the mb.

 

[tonyfreak215]

The PCI Express slot specifies +/-8% tolerance. Also a lot of multimeter is not perfectly accurate. Are these calculated in the results?

I have a Picoscope (Oscilloscope). When I eventually get the card, I could do some tests. That's going to be a while though.

 

[LTC8K6]

I'm comfortable with the 750ti 64W average from my PCI-E slot

I'm not comfortable with the RX480 if it is averaging 84 watts from my PCI-E slot.

I don't think that's an unreasonable line of thought.

So, let's get to the bottom of the issue and clear it up?

 

[Abwx]

Tom's description of why they didn't overclock.
And we see an average of 80 watts during the Metro run for the 480?

That looks considerably worse than the 750ti? Average of 64 watts?

Average means nothing since spikes are way higher, you could drain 20W on average with spikes of 200W, that would be worse than an average of 70W with 100W spikes...

 

[Stuka87]

The 75W limit comes from pci-e 1.0 rev, and as devices must be retro-compatible, they must meet this, when pci-e was introduced extra pins where added to the ATX plug in order to provide the power for pci-e slots, rev 2.0 raised the limit to 150W and rev 3.0 to 300W, but since no extra reinforcements was added, there is just no way to make a MB up to PCIE power specs, so they all settle on 75W as is the 1.0 requeriments, and is optional to them to provide extra power.

Anyway, there is a reason of why +75W devices must use external power even on a 3.0 slot.

This issue with the PCI-E is just a footnote on reference RX480 problems, anyone with 1 card should be OK, CF could be a long time problem as this could damage motherboard/PSU ATX connector.

You may be able to kill a PCI-E 1.0 motherboard thought, but who gona put a RX480 on a 1.0 mb?

On 2.0 and 3.0 you should be OK on passing this limit by a few W, still there is more chance to melt the ATX connector before killing the mb.

The connection is not going to be damaged by going 10W over the limit. This is misinformation, and you should not be posting it. A connector is not going to "melt" because of an extra 10W of average power usage. Going from 6.25A to 7.08A @12VDC is not enough of a difference to cause any issue to the connector, and certainly not enough to cause it to melt as you state.

And then there is the fact that we don't even know if this is a widespread issue, or an issue with the reviewers flashing the bios to test the 4GB and 8GB speeds on one card.

 

[Bacon1]

Really good testing in this article here:

http://www.pcper.com/reviews/Graphics-Cards/Power-Consumption-Concerns-Radeon-RX-480

Has anyone other than computerbase used the newer drivers?

 

[Shivansps]

The connection is not going to be damaged by going 10W over the limit. This is misinformation, and you should not be posting it. A connector is not going to "melt" because of an extra 10W of average power usage. Going from 6.25A to 7.08A @12VDC is not enough of a difference to cause any issue to the connector, and certainly not enough to cause it to melt as you state.

And then there is the fact that we don't even know if this is a widespread issue, or an issue with the reviewers flashing the bios to test the 4GB and 8GB speeds on one card.

It may do so in CF...nothing gona happen with a single card at all.

These are not strange events

Its more common that you may think

 

[LTC8K6]

Average means nothing since spikes are way higher, you could drain 20W on average with spikes of 200W, that would be worse than an average of 70W with 100W spikes...

We don't know what the component ratings are.

Peak ratings are often quite high for components.

It may be that 200W peaks are okay, but the 70W average isn't, for a given component.

Overall, I would think the average number is the important one.

 

[IEC]

It may do so in CF...nothing gona happen with a single card at all.

These are not strange events

Its more common that you may think

IIRC, that second image is from a rig running SLI video cards, and doesn't have anything to do with Crossfire...

Edit: Yep. Here's the original thread from 2010:
http://forums.evga.com/tm.aspx?m=500697&high=raw2dogmeat

 

[MrTeal]

You're not going to melt your PCIe slot with a single RX 480. AMD screwed up here, but it's not going to be a real danger for anyone not running a bunch of them. Spec on the connector is 1.1A per pin and there's 5 12V pins which is where the 66W comes from, but running 20% over that PCI-SIG spec won't melt things. Many of the connectors just rate things by the PCI-SIG rating, but others actually give a contact rating for the individual pins.
Samtec - 2.2A/pin

AFAIK, on pretty much any motherboard I've seen the PCI-E 12V power is just directly connected to the MB 12V plane, so it's not like you have a dedicated supply feeding the card that could by overloaded. The bigger danger (IMO) with high draw cards like this is the possibility of overloading your MB 24 pin connector which only has two 12V wires coming into it if you're running multi-GPU. That's not a unique to the RX 480 though, it's just that AMD pitches multi-GPU systems with the RX480 and usually those configurations are reserved for cards that draw much less power from the socket and get all the core power from the PCI-e power plugs. Running GTX960 SLI using those Strix cards would put more strain on a motherboard than a single RX480, even if those are in spec and the RX480 is out of spec.

 

[tonyfreak215]

Really good testing in this article here:

http://www.pcper.com/reviews/Graphics-Cards/Power-Consumption-Concerns-Radeon-RX-480

Fantastic article. Thanks for sharing!

 

[Azix]

I asked around our friends in the motherboard business for some feedback on this issue - is it something that users should be concerned about or are modern day motherboards built to handle this type of variance? One vendor told me directly that while spikes as high as 95 watts of power draw through the PCIE connection are tolerated without issue, sustained power draw at that kind of level would likely cause damage. The pins and connectors are the most likely failure points - he didn’t seem concerned about the traces on the board as they had enough copper in the power plane to withstand the current.

from the pcper link. seems he's worried about the pins the GPU connects to and not the traces on the motherboard. he can't be worried about the 24 pin connector either, that has to be able to handle it.

wonder how much power the physical slot can handle then

 

[Kenmitch]

Weird. Reading thru the thread it looks like it's overblown too me.

Looks like the 960 and 750ti would have killed lots of boards by now if the tolerance/spec is so tight.

Not saying ignore it....Just don't use it as an excuse for purchasing/justifying an alternative product choice.