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

[AlwaysFSX]

7 hours ago, djdwosk97 said:

I wonder if that affects/affected my z97s SLI plus since it's basically the same

board. Do you have a link to the problem? 

No but I can find one.

[DevilishBooster]

55 minutes ago, Valentyn said:

Also 75W is nothing, the PCIE standard allows for a maximum of up to 300W. 75W is the default minimum.


http://composter.com.ua/documents/PCI_Express_Base_Specification_Revision_3.0.pdf

 

[Prysin]

15 minutes ago, raphidy said:

I agree that it can be fix. I don't know if it's going to be easy. 

Your graph of the 960, found the source, it's more about strix fucking up than Nvidia. There's even an example of the Gainward under it.

 

http://www.tomshardware.com/reviews/nvidia-geforce-gtx-960,4038-8.html

Don't worry. This is a GTX card doing it. So there is no issue here.

 

right? @Majestic

[Briggsy]

Seems like this thread keeps going in circles, where someone new comes in and doesn't understand that the 480 is pulling more than spec specifically from the pcie slot which is bad, and no other graphic card tested is doing this. 

 

In a couple more pages there will be another "herp derp, guys xyz card is pulling more power than 480 from pci-e cables!" and the rest of us facepalm again.

[Prysin]

59 minutes ago, Valentyn said:

Also 75W is nothing, the PCIE standard allows for a maximum of up to 300W. 75W is the default minimum.


http://composter.com.ua/documents/PCI_Express_Base_Specification_Revision_3.0.pdf

 

Saving this for later. Thanks

 

@Majestic Perhaps you, AND Tom's should review your knowledge of the rules?

[patrickjp93]

4 minutes ago, Prysin said:

I think you should try review your statistics there. You may find yourself coming up a few hundred links short.

Nope, I used a Java web crawler that we had to write back in sophomore year. I've also posted more articles than both of you combined.

[zMeul]

5 minutes ago, DevilishBooster said:

-

 

[patrickjp93]

2 minutes ago, Prysin said:

Saving this for later. Thanks

 

@Majestic Perhaps you, AND Tom's should review your knowledge of the rules?

You do realize that's for the entire PCIe bus, not a single slot, right? (it's in the same link provided btw).

[laminutederire]

5 minutes ago, patrickjp93 said:

You do realize that's for the entire PCIe bus, not a single slot, right? (it's in the same link provided btw).

That says "slot power limit value", why wouldn't it be for a single slot?

[raphidy]

12 minutes ago, Prysin said:

Snip.

Sure, it's a GTX, not reference. Did you find a GTX over 75W on average? I did not follow all the posts.

 

Anyways, if AIB version of 480 does not have that problem,it's fine. The reference still needs a proper fix though.

I personally waiting on the AIB version, i'll pick the grater probably.

[patrickjp93]

5 minutes ago, laminutederire said:

That says "slot power limit value", why wouldn't it be for a single slot?

Go ahead and try to pull 150W sustained and see what happens. The 24-pin connector can't pull 1200W, and thus the 6+ slots on an mATX can't all be taking up 200+W each, capiche?

[Briggsy]

from Asus pci-e FAQ https://www.asus.com/support/faq/104406/

 

What is the maximum power consumption of PCIE slots?

 

... A standard height x16 add-in card intended for graphics applications must, at initial power-up, not exceed 25 W of power dissipation, until configured as a high power device, at which time it must not exceed 75 W of power dissipation.

...x16 graphics card is limited to 75 W. The 75 W maximum can be drawn via the combination of +12V and +3.3V rails, and the sum of the draw on the two rails cannot exceed 75 W.

[KeltonDSMer]

58 minutes ago, mariushm said:

Snip

Thank you for taking the time to write out your posts in the way you have. You are clear, concise, and provide sources along with real world examples that are shockingly relevant to the points you make. For every person that posts in threads like this, there are hundreds if not thousands of people that read them for information and do not participate in the conversation. I read OCN and AT news sections every day but do not participate; filtering through all the back and forth BS looking for informative posts, like yours have been today, is exhausting. Thanks

 

Edited June 30, 2016 by KeltonDSMer
Added punctuation

[patrickjp93]

18 minutes ago, DevilishBooster said:

 

Except that information is misleading. The entire bus combined can handle up to 300W. The 24-pin MOBO connector can't supply 1200W, yet you really think any one us on the motherboard can draw that much (4x 300W for quad configs)? See below. 75W is the limit per slot, period.

 

4 minutes ago, Briggsy said:

from Asus pci-e FAQ https://www.asus.com/support/faq/104406/

 

What is the maximum power consumption of PCIE slots?

 

... A standard height x16 add-in card intended for graphics applications must, at initial power-up, not exceed 25 W of power dissipation, until configured as a high power device, at which time it must not exceed 75 W of power dissipation.

...x16 graphics card is limited to 75 W. The 75 W maximum can be drawn via the combination of +12V and +3.3V rails, and the sum of the draw on the two rails cannot exceed 75 W.

 

[mariushm]

It's maximum 75w in total for each pci-e slot. If I remember correctly, the recommended limits in the standard are 10w for pci-e x1 slots, 25w for pci-e x4 and 60w for pci-e x8 or higher on 12v. There's also some note about half height cards should use only 25w.

The rest of the 15w is for 3.3v and there's also a 3.3v aux capable of about 0.4A so about 1.5 watts max (it's like 5Vsb on power supplies, a stand by rail).

 

It's totally possible to carry even 100w through 5 tiny traces, how much current can be carried through traces is directly proportional to surface .. so thin but short traces can carry lots of current, thin and long traces not so much.

On motherboards, you don't have thin traces from the 24 pin connector all the way to the pci-e slot, you have a large slab of copper inside the pcb in a layer you don't see and right by the connector you have a few vias bringing the 12v to the surface to a polymer capacitor and maybe an inductor and then the 12v goes right into the slot. So those few mm of thin traces can indeed carry even 100 watts to one pci-e slot if needed, for short periods of time.

The surface of the layer of copper inside the pcb most likely is larger than the surface of a few awg18 cables going from the power supply to the motherboard.. not to mention the surface is easier to cool as well, in comparison to cables.

[Valentyn]

10 minutes ago, Briggsy said:

from Asus pci-e FAQ https://www.asus.com/support/faq/104406/

 

What is the maximum power consumption of PCIE slots?

 

... A standard height x16 add-in card intended for graphics applications must, at initial power-up, not exceed 25 W of power dissipation, until configured as a high power device, at which time it must not exceed 75 W of power dissipation.

...x16 graphics card is limited to 75 W. The 75 W maximum can be drawn via the combination of +12V and +3.3V rails, and the sum of the draw on the two rails cannot exceed 75 W.

Funny then how ASUS's own Strix GTX 960 exceeded that on average during use.

 

[patrickjp93]

2 minutes ago, Valentyn said:

Funny then how ASUS's own Strix GTX 960 exceeded that on average during use.

 

Not on average. In spikes, but still to nowhere near the degree the 480 does. The average was 50 wish watts.

[laminutederire]

16 minutes ago, patrickjp93 said:

Except that information is misleading. The entire bus combined can handle up to 300W. The 24-pin MOBO connector can't supply 1200W, yet you really think any one us on the motherboard can draw that much (4x 300W for quad configs)? See below. 75W is the limit per slot, period.

 

 

It doesn't say 300W per slot,it just says that you can allow up to 300W per slot. It's a matter of being able to give everything you can on one slot, or not allowing it.

To me they can perfectly change that value and it won't draw above that, and if it needs those extra watts, the card will get those on the 6 pin which won't be limited by that.

 

 

By the way mate, have some respect when you answer to people.

What the document says is that you can put the LIMIT value to 300W on the four slots, and then every slot can draw as much as it can while while still remaining under what the MOBO can give.

[Prysin]

@Majestic

PCPerspective made an article. Please add to OP
http://www.pcper.com/reviews/Graphics-Cards/Power-Consumption-Concerns-Radeon-RX-480/Overclocking-Current-Testing

 

Quote

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.

 

As we all know with hardware failures in PCs, this is something that could in theory happen during a single gaming session, or it might instead take months and months of gaming to wear down componentry. Or it might never affect your system at all. I tend to lean on the side of worrying less about this power draw concern than many in the community are pushing me to. Following the likes of ASUS, MSI and Gigabyte over the last 5-10 years, I have seen all the major motherboard vendors focus on quality of componentry, and specifically on power delivery, in a wide range of product families. That doesn’t mean that any one particular motherboard is not prepared for the overclocked power draw on a Radeon RX 480, or that lower cost or older motherboards that some gamers are still using in their PCs might not be built to withstand it. We just don’t know yet; not enough hardware is in enough hands.

 

For our part, we are going to be plugging the Radeon RX 480 into a couple of older platforms and running it in some “bad case” scenarios…just to see what happens.

 

The easy fix to this whole ordeal is for AMD to have used an 8-pin power connector on the RX 480. The PCI Express spec allows an 8-pin connection to draw 150 watts on its own, leaving the power from the PCI Express connection on the motherboard for overhead. This is how NVIDIA designed the GTX 970 (two 6-pin connections) and how AMD designed the R9 380 (one 8-pin connection): both cards have 150+ watt TDPs with power supply overhead available to them. The new GeForce GTX 1070 has an identical TDP to the RX 480 (150 watts) but uses an 8-pin power connection to relieve any concerns at stock performance or while overclocking.

 

[3DOSH]

10 minutes ago, Prysin said:

@Majestic

PCPerspective made an article. Please add to OP
http://www.pcper.com/reviews/Graphics-Cards/Power-Consumption-Concerns-Radeon-RX-480/Overclocking-Current-Testing

 

 

In nut she'll,  its fine since they are spikes and MB vendors confirm that.  

AMD knows about the issue and may require software patch only. 

[AlwaysFSX]

5 minutes ago, 3DOSH said:

In nut she'll,  its fine since they are spikes and MB vendors confirm that.  

AMD knows about the issue and may require software patch only. 

Nowhere in the article was that said ^

[laminutederire]

2 minutes ago, AlwaysFSX said:

Nowhere in the article was that said ^

It's said spikes under 95W is okay.

 

[Prysin]

1 minute ago, 3DOSH said:

In nut she'll,  its fine since they are spikes and MB vendors confirm that.  

AMD knows about the issue and may require software patch only. 

chip draws 110w, memory, leakage and other things add up another 40w.

 

i bet they can just lower memory speed to 7GB/s and atleast for stock speeds. Stay within the spec.

[Briggsy]

2 minutes ago, laminutederire said:

It's said spikes under 95W is okay.

 

 

"sustained power draw at that kind of level would likely cause damage"

[zMeul]

18 minutes ago, laminutederire said:

 

you do realize that you base your argument on PCI Express® Base Specification Revision 3.0 - 2010

while the rest of us counter it with PCI Express Card Electromechanical Specification Revision 3.0 - 2013