The 480: power consumption, PCI-E powerdraw

[IEC]

I've said it multiple times already in this thread: you'll burn out the 24-pin ATX motherboard power connector @ the +12V pins long before you'll fry a PCI-e slot. Doing so requires a pathological load such as multi-card mining without powered risers and/or too much resistance at the connection point, e.g. with oxidized +12V pins (i.e. out of spec cable/connector). Too much resistance = too much heat with current draw = melt.

Two cards run at 110% mining doesn't seem to be causing any issues. I'll update if that changes.

 

[Azix]

more analysis.

https://www.reddit.com/r/Amd/comments/4rbw8p/facts_about_pcie_connectors/ - fits well with why the german test video showed no problem at 5+ A

I think its safe to say this isn't going to fry pci-e slots unless something is really wrong with the slot.

I'll try to dispel some myths generated by the RX 480 power draw controversy.
I am an electrical engineer and read spec sheets and design components on a day to day basis. Here is the product page of a PCIe connector manufacturer. There we can find a qualification report. Within are diagrams concerning ampacity (the current carrying capability of electrical conductors).

Connector Setup

Temperature Rise

3.3V Curve

12V Curve

Those test show what the max. ampacity of one contact over a ambient temperature curve. So another thing we need to look at is what kind of temperature those contacts are exposed to. Intel requires that cases are designed so that ambient inlet temperatures do not exceed 45°C. I can't really imagine a scenario where one has such an awful setup with such bad airflow that the case temp is above 50°C. But Lets do some calculations. Here is a table so you know what kind of temps we are talking about.

Safe max continuous power draw from a PCIe connector at not realistic scenario with 65°C ambient temperature

3.3V * 4 contacts * 6.1A= 80W
12V * 5 contacts * 4.5A = 270W
Total 350W

Safe max continuous power draw PCIe connector at a realistic worst case scenario with 50°C ambient temperature

3.3V * 4contacts * 7A = 92W
12V * 5contacts * 6.4A = 384W
Total 476W

Of note is that these values are with a derated 20% safety margin. Plus the way they tested it was at a stupid high temperature with 6 contacts adjacent i.e. worst case and not how the PCIe connectors are actually setup. (max of 3 contacts adjacent) If we disregard the safety margin, temp and the unrealistic setup of adjacent pins we could probably pull almost 800W - 900W from a PCIe connector without burning out any contact.
I don't have access to the PCIe 3.0 electrical specs so I can't say anything on that side except those are probably whit a ton of safety margins. My personal opinions as someone who does this professionally on a day to day basis is that interfaces are designed for a reason and should be adhered to. But the amount of overdraw worries me not one bit because we go over spec all the time to make things work. If we would design things with such a small safety margin basically everything would break down immediately all the time. As a matter of fact I wonder why it is brought up at all?! IMHO this whole situation is laughable and I would not be surprised if AMD engineers would have been completely been caught of guard as this is not even remotely an issue to them. What they are doing with the driver is satisfying politics not physical necessities. (As we so often do *sigh...)

I also don't like the test done by laymen such as pcper for example. They have an air of seriousness to them but for somebody in the know it's kind of like watching an Adam Sandler movie (that Malventano guy... dear lord). Also you can't just draw a line, on a sampled current measurement, guesstimate where the median is and announce that this is the average power draw. What you'd need is a program that must obtain the average by integration. Placing some random cables and connectors between them and using an oscilloscope for power measurement has it's own issues. All in all very unprofessional. Pure cringe material. They should have asked an electrical engineer specialized in equipment testing and another one in in circuit/component design to come on to the show.

 

[Shivansps]

Again, its highly unlikely that its gona fry the slot, because the ATX plug its the failure point, and one card is not enoght to damage it.

Then again, you can't guarantee a thing either way.

 

[MrTeal]

more analysis.

https://www.reddit.com/r/Amd/comments/4rbw8p/facts_about_pcie_connectors/ - fits well with why the german test video showed no problem at 5+ A

I think its safe to say this isn't going to fry pci-e slots unless something is really wrong with the slot.

No, they likely aren't. Really, the idea of the average user frying a PCIe slot is pretty far out.

The reddit guy is using the wrong qualification report though. He used the qualification report for this PCIe jumper He should be using this one for an actual PCIe slot with a card edge connector, which is what a GPU is.

You can see the results on page 15; for a 30°C temperature rise the current rating is 2.4A with 2 contacts energized, 2.0A for 4 contacts, 1.7A for 6 contacts, and 1.6A for 8 contacts. Now, there's 5 energized 12V lines on PCIe connector essentially all right beside each other, and the current has to return through the grounds. There's 3 grounds by the 12V lines before the notch, 6 more up to the 1X width, and a bunch more further down for 16x cards. For heating you're primarily interested in the grounds close to the 12V lines, but the average current in those 3 will be much smaller than the 12V current due to all the parallel paths and so will have a small effect. You can probably easily used the 20% derated 1.7A current without any worries. You can look at the graphs on page 27 to see the curves for maximum current under the 125°C limit at various different ambients.

I'm marginally in the camp of AMD is pushing it a little far with their design, but even then that is more out of concern for the ATX power connector rather than the PCIe slot itself.
Luckily the RX480 isn't powerful enough that people would be looking at tri- and quad-fire setups with them for gaming, and while miners will blow up boards if they plug six of them into a MB, they should know better than to do so without using powered risers anyway.

 

[Element115]

No, they likely aren't. Really, the idea of the average user frying a PCIe slot is pretty far out.

The reddit guy is using the wrong qualification report though. He used the qualification report for this PCIe jumper He should be using this one for an actual PCIe slot with a card edge connector, which is what a GPU is.

You can see the results on page 15; for a 30°C temperature rise the current rating is 2.4A with 2 contacts energized, 2.0A for 4 contacts, 1.7A for 6 contacts, and 1.6A for 8 contacts. Now, there's 5 energized 12V lines on PCIe connector essentially all right beside each other, and the current has to return through the grounds. There's 3 grounds by the 12V lines before the notch, 6 more up to the 1X width, and a bunch more further down for 16x cards. For heating you're primarily interested in the grounds close to the 12V lines, but the average current in those 3 will be much smaller than the 12V current due to all the parallel paths and so will have a small effect. You can probably easily used the 20% derated 1.7A current without any worries. You can look at the graphs on page 27 to see the curves for maximum current under the 125°C limit at various different ambients.

I'm marginally in the camp of AMD is pushing it a little far with their design, but even then that is more out of concern for the ATX power connector rather than the PCIe slot itself.
Luckily the RX480 isn't powerful enough that people would be looking at tri- and quad-fire setups with them for gaming, and while miners will blow up boards if they plug six of them into a MB, they should know better than to do so without using powered risers anyway.

Guy here. I used this report as it has 3.3V and 12V tests. The one you linked to were done with 500VDC. This is important as current throughput is lower the higher the voltage as seen with the 3.3V and 12V tests. Insofar it does not matter that I used another report as the contacts touching the GPU are the weakest point.

 

[MrTeal]

Guy here. I used this report as it has 3.3V and 12V tests. The one you linked to were done with 500VDC. This is important as current throughput is lower the higher the voltage as seen with the 3.3V and 12V tests. Insofar it does not matter that I used another report as the contacts touching the GPU are the weakest point.

That doesn't even make sense. The dielectric withstand voltage and insulation resistance tests were done with 500VDC, not the current tests. The current tests were done at whatever voltage caused the test current to flow down one set of contacts and up the other set.

 

[Stuka87]

Guy here. I used this report as it has 3.3V and 12V tests. The one you linked to were done with 500VDC. This is important as current throughput is lower the higher the voltage as seen with the 3.3V and 12V tests. Insofar it does not matter that I used another report as the contacts touching the GPU are the weakest point.

The current only goes down as voltage increases if you use wattage as your measurement. But it states he is using current, so voltage is irrelevant as long as insulation and spacing is large enough to prevent any possible arcing.

 

[Flapdrol1337]

Guy here. I used this report as it has 3.3V and 12V tests. The one you linked to were done with 500VDC. This is important as current throughput is lower the higher the voltage as seen with the 3.3V and 12V tests. Insofar it does not matter that I used another report as the contacts touching the GPU are the weakest point.

Current is what matters.

The contact has resistance and the heat generated is it's resistance*current^2

 

[Element115]

That doesn't even make sense. The dielectric withstand voltage and insulation resistance tests were done with 500VDC, not the current tests. The current tests were done at whatever voltage caused the test current to flow down one set of contacts and up the other set.

You are right. I just did a cursory overview of that spec sheet.

 

[MrTeal]

You are right. I just did a cursory overview of that spec sheet.

No worries. At the end of the day your conclusion isn't incorrect; the connector itself is able to handle much more current than even an overclocked RX480 would draw with acceptable heating.

 

[Element115]

No worries. At the end of the day your conclusion isn't incorrect; the connector itself is able to handle much more current than even an overclocked RX480 would draw with acceptable heating.

Yeah I wanted to give a nice visual overview and the 12V and 3.3V graphs were quite handy for that. But you are right. Doesn't change much.

 

[Yakk]

Good article today over at Tomshardware http://www.tomshardware.com/reviews/amd-radeon-rx-480-power-measurements,4622.html

 

[MrTeal]

Good article today over at Tomshardware http://www.tomshardware.com/reviews/amd-radeon-rx-480-power-measurements,4622.html

Weird. They've got a whole suite of nice test equipment, why wouldn't they grab a multimeter and just check which source the 12V on the high side switch is connected to in order to actually verify their hunches. It'd take all of 2 minutes.

 

[Sweepr]

The AMD Radeon RX 480’s power supply configuration exceeds the limit defined by the PCI-SIG’s specifications. It doesn’t exceed them by a massive amount, but it does do so reliably. Norms should be respected, especially if they already have a very generous built-in tolerance range. We never implied in our launch article that a system made up of solid components might be directly damaged by an AMD Radeon RX 480 graphics card running at stock clock frequencies.

There shouldn’t be any problems unless cheap, dirty or outdated components are used, the card isn’t installed correctly, or the amount of power drawn is increased by overclocking the card.

Well cheap and dirty is kind of what you expect from (some) systems running $199-239 VGAs.


Threadcrapping, baiting, flaming and trolling are not allowed
Markfw900

 

[boozzer]

Well cheap and dirty is kind of what you expect from (some) systems running $199-239 VGAs.

hahahahahaah wow. this is almost as bad as the competition is bad for innovation bit.

 

[Sweepr]

hahahahahaah wow. this is almost as bad as the competition is bad for innovation bit.

In conclusion, they fail PCI-E specifications:

The AMD Radeon RX 480’s power supply configuration exceeds the limit defined by the PCI-SIG’s specifications. It doesn’t exceed them by a massive amount, but it does do so reliably.

Wasn't AMD going to come up with a software solution/statement today?

 

[Madpacket]

As I wait for my Sapphire reference 480 to arrive (shipping from California to Southern Ontario!) I came across this video test of a modified 480 running at high clock rates.

https://www.youtube.com/watch?v=Jq47qmwcus8

What's interesting is how the presenter states aftermarket cards will most likely not reach any higher than 1400Mhz, maybe 1450 on a really good card.

Given most reference models can be undervolted and overclocked to 1300+ speeds I'll likely just keep the card when it arrives rather than refusing shipment. An extra 100Mhz for more money and power usage won't really make much of a difference. I care more for power efficiency (for mining / gaming) than a few more FPS. Sure the blower fan isn't great but I prefer them for the longevity and heat exhaust properties (this card will be going into a mini ITX case).

My only regret is not ordering the 4GB model (although I had like 10 minutes before they sold out!) as they're identical cards simply in need of a BIOS flash. Would have been a good way to save $60.00 CAD.

Congrats to all the people who snagged 4GB models, hopefully the BIOS flash utility provided to reviewers is leaked soon.

 

[boozzer]

In conclusion, they fail PCI-E specifications:

Wasn't AMD going to come up with a software solution/statement today?

I don't know, I don't follow fud much.

 

[LTC8K6]

So Tom's says that the card is out of spec.

Will board mfgs chime in about using PCI-E cards that are out of spec?

I mean, assuming AMD's solution isn't satisfactory.

 

[exar333]

The whole issue seems (to me) to be a result of the GPU not quite hitting the expected power envelope. It appears that the phases and power delivery were designed based on the assumption this would be a 110w total card and that didn't happen. Aftermarket cards should address the issue, but it is interesting the reference card misses the mark here. At some point, it is a bit puzzling that this wasn't fixed prior to launch. Understanding this issue isn't a 'huge' issue or a 'deal breaker' but definitely damaging for the card at launch.

As a AMD stock owner and GPU enthusiast, it is (again) disappointing that AMD stumbles out of the gate here a bit. I think the 480 will be a fine card, but the fly in the ointment here isn't ideal.

As others have stated here and in other threads, GF always seems to be the X Factor for AMD recently, GPU or CPU. They just don't have the same expertise and/or track record of TSMC. I am not saying I think they are not executing, but they will need to really prove themselves for a few consecutive launches to gain some confidence IMHO.

 

[amenx]

The whole issue seems (to me) to be a result of the GPU not quite hitting the expected power envelope. It appears that the phases and power delivery were designed based on the assumption this would be a 110w total card and that didn't happen. Aftermarket cards should address the issue, but it is interesting the reference card misses the mark here. At some point, it is a bit puzzling that this wasn't fixed prior to launch. Understanding this issue isn't a 'huge' issue or a 'deal breaker' but definitely damaging for the card at launch.

Have a feeling they took a last minute risk by upping the core clocks to gain a performance edge (vs 970) without sufficient re-testing of the power draw. Some spec leaks in weeks before release had the clocks lower than final release.

 

[techne]

Facts about PCIe connectors?

https://www.reddit.com/r/Amd/comments/4rbw8p/facts_about_pcie_connectors/

 

[3DVagabond]

Well cheap and dirty is kind of what you expect from (some) systems running $199-239 VGAs.

The systems they are describing might just as likely be damaged supplying less than 75W too.

 

[Zstream]

Well cheap and dirty is kind of what you expect from (some) systems running $199-239 VGAs.

Uhh just no...

 

[96Firebird]

I'm curious as to if AMD passed the PCI-SIG Compliance Workshop with the cards that reviewers reviewed, or if something has changed. Would they be required to re-test if a fix is not viable?