AMD once again violating power specifications? (AMD RX-480)

[Majestic]

2 minutes ago, ace_cheaply said:

Quote, directly from the article:

 

"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. We just didn’t want to do that to our test platform. "  200w peak through the pci-e slot.  Not sure how much more simple i can make that. 

That is overclocking, meaning the average and peak went up from 82/155 to 100/200. Still don't understand why you fail to understand this.

[SteveGrabowski0]

12 minutes ago, Sakkura said:

The GTX 970 uses 170-175W, so 165W on the 480 is a little less, not more.

Not according to Guru3D's measurements:

 

original image:

Spoiler

 

Edited June 29, 2016 by Ryan_Vickers
that image was death on dark theme

[Trixanity]

4 minutes ago, Majestic said:

Nobody is talking about electricity bills. It's a very specificly defined issue, this strawman is utterly pointless.

 

@Trixanity I expect AMD to know how to fix it, but I'm pretty sure it will hurt performance as it has to throttle more. But ofcourse, that will be after all the reviews are out. 3,5/4 stars AMD *golfclap*

I just looked at a few different reviews. It really seems like Tom's has a dud in hand. 300W on the card itself seems like something has gone very wrong in either the card or the equipment. Anandtech says 300W total at peak (wall measurement I guess). TPU measured 166W maximum with a similar methodology from the card. TPU measured at both bus and 6-pin. PCPer says 150-160W from the card.

 

It's becoming clear that Tom's is so far the only source to measure the freakish 155W bus draw or 300W total card draw.

 

Something's fishy. And no I'm not saying Tom's being shady. I'm saying that there is something wrong with either the card or their equipment. Such an increase in power draw from the card is not normal. A card doesn't just spike from 160W to 300W even intermittently.

[TheRandomness]

8 minutes ago, Crazy Ginger said:

It draws that power through the power connectors, NOT THE PCIE SLOT.  there's a reason why you have to plug in those wires coming straight from the PSU to your gpu

And the PCIe slot*, 75W from it too.

4 minutes ago, Majestic said:

Not with 95% of your statements being logical fallacies, no.

My fantastic flawed logic showing through. I need to learn how to argue logically.

3 minutes ago, Ryan_Vickers said:

I know you're being sarcastic but it certainly isn't the first time.  We can all help try to make it the last time though by sending a message about them doing this.  Maybe companies lie because every time they do, the product sells anyway and people move on and just expect it to happen again?

Maybe companies lie because it makes their products sell better therefore they generate a larger profit. That's the whole point of a company, right? To make a product to generate as much money as possible, therefore sustain said company. They may do this is whatever way they want, be it lying or some other dodgy tactic that I can't think of.

[Sakkura]

2 minutes ago, SteveGrabowski0 said:

Not according to Guru3D's measurements:

Which are inaccurate. Hence the "calculated TDP".

 

Tom's Hardware has much better measuring equipment. As you can see in OPs link, the RX 480 draws 164W where the GTX 970 draws 171W.

[ace_cheaply]

3 minutes ago, Majestic said:

That is overclocking, meaning the average and peak went up from 82/155 to 100/200. Still don't understand why you fail to understand this.

I still don't understand how you fail to understand that I don't believe the numbers are accurate until they are reproduced...The numbers make no sense, i showed you why i think they make no sense.  The idea that the card is using over 345 watts...(that is, if the previous 300W was 155 from mainboard, 145 from 6 pin.  200W+145W=345W) is unbelievable to me. If it is true, I will be amazed, and will certainly not buy a 480 until it is resolved.

[SteveGrabowski0]

2 minutes ago, Sakkura said:

Which are inaccurate. Hence the "calculated TDP".

 

Tom's Hardware has much better measuring equipment. As you can see in OPs link, the RX 480 draws 164W where the GTX 970 draws 171W.

OK, so it's a slightly less power hungry GTX 970. Still pretty disappointing to have pretty much the same performance per watt as a 21 month old 28 nm card.

[Majestic]

2 minutes ago, Trixanity said:

I just looked at a few different reviews. It really seems like Tom's has a dud in hand. 300W on the card itself seems like something has gone very wrong in either the card or the equipment. Anandtech says 300W total at peak (wall measurement I guess). TPU measured 166W maximum with a similar methodology from the card. TPU measured at both bus and 6-pin. PCPer says 150-160W from the card.

 

It's becoming clear that Tom's is so far the only source to measure the freakish 155W bus draw or 300W total card draw.

 

Something's fishy. And no I'm not saying Tom's being shady. I'm saying that there is something wrong with either the card or their equipment. Such an increase in power draw from the card is not normal. A card doesn't just spike from 160W to 300W even intermittently.

They are measuring at very high polling rates (100khz -> 500mhz). Any equipment that is slower either misses the peaks or uses a moving average filter. Meaning it more resembles the average load.

 

The 300W is a very minute peak, averages are on par with other outlets. Don't mistake different results as an assured error on the person who measures them. Unless all outlets were using the same equipment, that is a grossly erronenous assumption.

[vanished]

3 minutes ago, TheRandomness said:

Maybe companies lie because it makes their products sell better therefore they generate a larger profit. That's the whole point of a company, right? To make a product to generate as much money as possible, therefore sustain said company. They may do this is whatever way they want, be it lying or some other dodgy tactic that I can't think of.

I think if a company really cares about long term sales and image, they will choose consumer trust that leads to lasting, repeat customers over high initial sales of a product.  That is, unless they can count on us to all just "take it", in which case they can do whatever they want because we'll just buy their stuff anyway.

[Sakkura]

2 minutes ago, SteveGrabowski0 said:

OK, so it's a slightly less power hungry GTX 970. Still pretty disappointing to have pretty much the same performance per watt as a 21 month old 28 nm card.

Arguably, yes. But at least it's a major improvement over AMD's previous generation.

[Crazy Ginger]

1 minute ago, TheRandomness said:

I think I'm the only one here who has the mentality of 'who cares?'.

Oh no, a company lied. That's a first.

If you have a PCIe SSD and an RX 480, you put your money in the wrong places. Also, with the SMCs on the motherboard they have their own specs, meaning that it's possible for certain components to run even with the increase power draw and whatnot. Kinda like how VRMs can go to 125C whilst the GPU itself has a separate thermal limit.

 

I should stop arguing, I'm never going to win.

I only use the NVME drive as an example for argument's sake, and because I have one so it was one of the first things to come to mind (it was on sale and I couldn't resist, don't tell me I wasted money I could have used on getting a gpu) and sure, you could get away with running slightly over spec, but that doesn't change the fact that because it is running over the specs of pcie, it adds unnecessary risk.  That's why people are uneasy about this.  

[dlink377]

I believe the PCIe connector and the molex connector will able to provide a bit extra without problem of the maximum amp for the connector, although, it is the best if they just follow the power limit of the connector by using 8 pin or dual 6 pin connectors, just for overhead.

 

Just wait for AIB cards, I am quite sure they will include those extra connectors.

[Majestic]

3 minutes ago, ace_cheaply said:

I still don't understand how you fail to understand that I don't believe the numbers are accurate until they are reproduced...The numbers make no sense, i showed you why i think they make no sense.  The idea that the card is using over 345 watts...(that is, if the previous 300W was 155 from mainboard, 145 from 6 pin.  200W+145W=345W) is unbelievable to me. If it is true, I will be amazed, and will certainly not buy a 480 until it is resolved.

You find it hard that I don't understand your belief in something not being true? U wot mate. They are minute peaks, not averages. It's not drawing 345W continuously. Whatever....

[Majestic]

1 minute ago, dlink377 said:

I believe the PCIe connector and the molex connector will able to provide a bit extra without problem of the maximum amp for the connector, although, it is the best if they just follow the power limit of the connector by using 8 pin or dual 6 pin connectors, just for overhead.

 

Just wait for AIB cards, I am quite sure they will include those extra connectors.

Maybe, but that's why the post is only to serve as a warning against the reference design. Just like Club3D started issueing out better manuals for power supply selection with the 295X.

[ace_cheaply]

1 minute ago, Majestic said:

You find it hard that I don't understand your belief in something not being true? U wot mate. They are minute peaks, not averages. It's not drawing 345W continuously. Whatever....

Yes, i find it hard that you don't understand that I don't believe the results until they are reproduced. It should never, ever be drawing 345W, peak or otherwise. 

[TheRandomness]

1 minute ago, Crazy Ginger said:

I only use the NVME drive as an example for argument's sake, and because I have one so it was one of the first things to come to mind (it was on sale and I couldn't resist, don't tell me I wasted money I could have used on getting a gpu) and sure, you could get away with running slightly over spec, but that doesn't change the fact that because it is running over the specs of pcie, it adds unnecessary risk.  That's why people are uneasy about this.  

Offtopic: I really should give more shits about my computer then, currently have a radiator sat onto of the CPU heatsink because I had nowhere else to put it. Also, threw 1.46V at my 290 because 1200MHz pls.

Ontopic: The R9 295x2 ran over the specs of PCIe by quite a lot. People didn't exactly bitch so hard about that though, did they? They just got the message that they should have a very good PSU system for it. Might give some people wanting the RX 480 a similar message, if you're running on a $30 mobo maybe you should prioritise something else to upgrade instead.

[Majestic]

6 minutes ago, ace_cheaply said:

Yes, i find it hard that you don't understand that I don't believe the results until they are reproduced. It should never, ever be drawing 345W, peak or otherwise. 

http://www.tomshardware.com/reviews/nvidia-geforce-gtx-1080-pascal,4572-10.html

 

GTX 1080 also can draw 311W. Power consumption is always this erratic, it's just never brought up. The difference is, out of that 311, only 62W is out of the PCI-E, well within limits. Not 155W.

[Majestic]

4 minutes ago, TheRandomness said:

Ontopic: The R9 295x2 ran over the specs of PCIe by quite a lot. People didn't exactly bitch so hard about that though, did they? They just got the message that they should have a very good PSU system for it. Might give some people wanting the RX 480 a similar message, if you're running on a $30 mobo maybe you should prioritise something else to upgrade instead.

But you can't optimize your PSU for something that is drawn through the damn mainboard. Please just practice what you preach and stop arguing. You're defending it just to defend it. Never making any valid points.

[TheRandomness]

1 minute ago, Majestic said:

But you can't optimize your PSU for something that is drawn through the damn mainboard. Please just practice what you preach and stop arguing. 

Seriously, you're defending it just to defend it. Never making any valid points.

I don't- what?

The little connectors on the top of the GPU. You see them? They also have power drawn through them. Yes, I know you explained what you mean on the previous page, but still...

[ace_cheaply]

5 minutes ago, Majestic said:

http://www.tomshardware.com/reviews/nvidia-geforce-gtx-1080-pascal,4572-10.html

 

GTX 1080 also can draw 311W. Power consumption is always this erratic, that's why we look at averages, because for those we can build. The difference is, out of that 311, only 62W is out of the PCI-E, well within limits. Not 155W.

While I'll admit the peaks on the 1080 are suprisingly high, I still am waiting to see another reviewer have similar results before any conclusion can be reached. It's possible, maybe even probable that nobody goes to the lengths tom's does, so they found what nobody else would have and now everyone will find the same problem. It's also possible it was a testing error.  
 

[xAcid9]

6 minutes ago, ace_cheaply said:

While I'll admit the peaks on the 1080 are suprisingly high, I still am waiting to see another reviewer have similar results before any conclusion can be reached. It's possible, maybe even probable that nobody goes to the lengths tom's does, so they found what nobody else would have and now everyone will find the same problem. It's also possible it was a testing error.  
 

Not everyone have equipment like Tomshardware and do power consumption test extensively. Many reviews use crappy Watt meter.

[Majestic]

13 minutes ago, TheRandomness said:

I don't- what?

The little connectors on the top of the GPU. You see them? They also have power drawn through them. Yes, I know you explained what you mean on the previous page, but still...

 

The bars on the graph, do you see them?  PCI-E 12V is the connector, Mainboard 12V is the pci-E slot.

 

8 minutes ago, ace_cheaply said:

While I'll admit the peaks on the 1080 are suprisingly high, I still am waiting to see another reviewer have similar results before any conclusion can be reached. It's possible, maybe even probable that nobody goes to the lengths tom's does, so they found what nobody else would have and now everyone will find the same problem. It's also possible it was a testing error.  
 

That testing error is then a systemic error, because it's present on multiple cards and reviews. Also, that requires all outlets to ditch their Kill-A-Watt and get professional kit. Which not all of them will. I assume AMD will fix it by making the card throttle. But it's unsure if tests will be redone afterwards. That said though, their averages are not really any different. They just have more extensive datasets.

 

I'm also nowhere claiming it to be infallible proof. Only arguing against misrepresentations or flatout denials of the issue.

 

 

[Misanthrope]

24 minutes ago, ace_cheaply said:

Yes, i find it hard that you don't understand that I don't believe the results until they are reproduced. It should never, ever be drawing 345W, peak or otherwise. 

No offense to Luke and well most other reviewers out there, but apart from the 2 sites that tested this how many you honestly think even test this?

 

As they said themselves most durable motherboards will run into no issues and most reviewers have good motherboards and fairly clean in terms of other pci-e cards and such since well, it's a test bench. It could have honestly gone by them.

 

Last but not least, let's not forget how a very slight overclock on Tech of Tomorrow's review shot up GPU temps to 94c and such. To me this would indicate his card is probably also drawing too much power as well maybe we should contact him and ask him to test wattage again he could replicate.

 

Edit: Also @Slick I'd be great if you could try to replicate this

 

[juretrn]

1 hour ago, Majestic said:

Because it peaks 155W over a bus designed for 75W. Meaning it will stress the mainboard momentarily for double it was designed for.

They note it won't catch fire, but it will cause perhiperal issues like sound or add-in cards.

 

But who knows when someone is using a cheap board and is overclocking it. Meaning it will average 100W over something designed for 75W. It could cause damage.

But it's AMD, so they are forgiven in advance!

[Majestic]

3 minutes ago, juretrn said:

But it's AMD, so they are forgiven in advance!

Forever the benefit of the doubt, indeed.