AMD Radeon RX 480 Power Measurements Repeated And Clarified

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.

tommy4242

Commendable
Jul 5, 2016
2
0
1,510
0
PCIe version 1.0, the spec version with 4x 12V pins, only allowed 60W (10W 3.3V, 50W 12V).

PCIe version 1.0a increased the number of 12V pins to 5 and increased the connector power to the current 75W.

More than 1.1A per connector pin was never allowed.
 

Calculatron

Honorable
Apr 3, 2013
2,240
0
12,460
384
I think that it's great that Tom's Hardware did a follow-up article on this issue; I saw how many other reviewers/influencers used the information of the original article, and I hated how they formed the information to fit their views. From my perspective, it was infuriating, and I can only imagine how the staff of Tom's felt.
 

drinkingcola86

Distinguished
Oct 10, 2008
546
0
19,160
61
I see that the rx 480 has issues with power over the pice slot and the 6 pin cable, what I've seen some other websites test is undervolting the last 3 power phases in watt man. It seems to fix the issue dropping the power at the wall socket by 10-30 watts.

I'm wondering if toms will test undervolting it?
 

AjvarD

Commendable
Jul 5, 2016
4
0
1,510
0

I don't want to be as harsh as this followup possibly deserves because I want to sleep so being short:
1. 4 PINS x 1.1A (max) x 12V = 52.8W + 9.9W (from 3.3V) = 62.7W >>> 60W Max - this is old ECR PCI-E Standard
2. 5 PINS x 1.1A (max) x 12V = 66W + 9.9W (from 3.3V) = 75.9W >>>> 75W Max after changed to CEM 1.0

So all that mass of text and photos about PCI-E pins called "Here's why we don't think this is such a big deal" contains nothing but meaningless info and you now have to Update article AGAIN:
- delete your false speculations about comparison older and newer graphics cards PCI-E pins
- delete words "less drama" and pasting clear "is not safe to use" or "there is a risk"
- and again rewrite all Summary taking this to account.

Next time spend more time reading how numerous fanboys and Electric Engineer experts on payroll try to defend failed manufacturer and how rare enthusiasts destroy their arguments instead of spending 2 days on the same article (and doing it way worse).
 

FormatC

Distinguished
Apr 4, 2011
981
1
18,990
1
You are wrong, the 5.5A-Limit was never changed (only the reserved pin is now also declared as 12V), it is existing since 2005 (version 1.1). And to be honest: 5 x 1.1 A are 5.5A and not more. The voltage (and their tolerance due the ATX specifications) is here totally uninteresting (as I wrote), the limit is set only to CURRENT without any tolerance and not to power draw as procuct of current and voltage.

If you read the Chapter 4 of the CEM carefully you will note, that the 75 Watt limit is only valid in the case of an maximum overvoltage on all rails, nothing else. This doesn't mean, that you can serve generally 75 watts, also if the voltage doens't exceed the 12 or/and 3.3 volts. This is totally wrong. The table in 4.2 is based at the limits of 4.1.

The short piece on page Two is from the current PCI SIG, provided in a guideline from Hewlett Packard. I especially asked two mainboard manufacturers and one OEM für plugs and sockets. Current, not wattage. Why? The limit was defined for cooling purposes, the contacts itself can survive 3A or more without any technical problems. But only the current is the reason for heating, the voltage is totally irrelevant for this. The approximate value for this are 55°C (PCI SIG), that must be cooled over the PCB :)

This is from my sight the only one thing that can be a problem, because the PCIe slot has inner temperatures of 70°C and more due the heat from the VRM section, directly above. The heat is wandering from VRM through the PCB and we see no cooling effects. But there is also a very large headroom. The most of OEMs specified their contacts for 3A or more and higher temps to work properly. Simply try to read their datasheets and certifications. ;)

This is from one of my German colleagues and we are in a closer contact since a while because he is figuring out this things on the practical way. I hope, his English is understandable:
https://www.youtube.com/watch?v=TF57OZoJ8M8
 

psiboy

Distinguished
Jun 8, 2007
180
1
18,695
3
I'll be waiting until one of their partners release a board with an 8 pin connector... not prepared to risk damage to my Gigabyte 970 board!
 

technoholic

Distinguished
Feb 27, 2008
800
0
19,160
69
Yeah that'd be wise to do so, i was trying to sell my r9 380x and hoping to buy a AIB rx480 but now i'm so irritated from all this sht probably will wait for Vega or will check Nvidia's mainstream offerings
 

maccorf

Commendable
Jul 6, 2016
1
0
1,510
0
People are the worst. You come out and clarify your original statement and people on here just talk about how you're either wrong (about something that you did yourself) or being leashed by AMD, because you're not completely validating the views they started with. I wonder how many of these people think we didn't land on the moon either.
 

problematiq

Reputable
Dec 8, 2015
443
0
4,810
16


I'm sorry I can't help myself..

I'm a licensed electrician and you should totally use 14 gauge wire to your receptacles and a 40 amp breaker, because moar power right? (Please for the love of god do not actually try this.)
 

todd evo

Commendable
Jul 6, 2016
3
0
1,510
0
The 75W limit of the 6-pin connector is laughably under specified. Look at the datasheets of the components the mfg would use and the sustainable amperage is almost certainly much higher, especially if they use quality plating. In your test the draw from the power connector doesn't seem to go past the 78W which makes me believe the 75W is a "hard" limit set inside the PSU when only 6-pins is detected. Using a 6+2 and shorting the unused +2 works around that, the PSU should be able to supply 150W, more for the conn. to use, see how the load distributes then?
If I'm right, anyone with an RX480 and a 6+2 pin capable PSU can 'fix' their card without any power concern for the cost of a paperclip and about 5 minutes of their time.
I don't have an RX480, I wish I did, so I can't try this myself. I want to know what happens if you plug a 6+2 style connector into the card instead, and short the +2 pins to force the PSU into 'enhanced power mode', the same thing a real 8-pin card would do. Also, I'd like to know if the middle power pin of the 6-pin connector is actually connected to the other power pins or is it just there for cosmetics. Not a huge problem if it isn't but it would be better if it was.
Shorting the "+2" would drive the sense pin that is required to unlock the additional power (for "8-pin" cards) and should offer additional card power.

I build motherboards for a living, and their statement about "Standards exist for a reason" is right. You should follow them or else the standard is meaningless but breaking it isn't illegal (unless you say something is compliant when it isn't) and almost any device worth buying from a reputable vendor is going to exceed it the specs. The build quality would have to be exceptionally poor and if it were you would most likely have further problems.
 

turkey3_scratch

Polypheme
Ambassador


There is no "hard limit" in the PSU, but you are correct aboiut it being underrated. A 6-pin PCIe cable should be able to handle 250W under normal conditions. People have hooked them up to load testers and stressed them past 200W for hours without issues.

What also matters, though, is the joining of wires in the power supply. Sometimes they connect two or even three of those 12V wires into one lead of solder on the PCB of the PSU, which can then be a limiting factor.
 

todd evo

Commendable
Jul 6, 2016
3
0
1,510
0


To be completely accurate, that's the PCIe spec, NOT the spec of the crimps or connector cabling which is typically way higher.
 

todd evo

Commendable
Jul 6, 2016
3
0
1,510
0


All I'm suspecting is, the card is choosing to get more power from the slot than the PSU, that could be the way the card is designed or a side effect of the 6-pin sense wire telling the PSU "By spec you can only give me 75W." However, drive pin 8 and 4 (on a 6+2 connector) to ground and now you're telling the PSU "Hey, I can have up to 150W if you have it." and then see what happens.

I think the 75W is a hard limit because there are sense wires (not all of them are ground) there for a reason and I imagine the PSU is listening to it. Again, a wild guess. A 5 minute test with a 6+2 connector, multimeter, and an RX480 (the one thing I don't have) would tell me.

Connect multi to 12v of PCIe, run, check voltage. Change to my 6+2 idea, repeat, the voltage drop should be a bit lower on the connector. That's the quick & dirty but an amp probe on the 6+2 wires would confirm it. I'm simply looking if the card suddenly can draw more than 78W from there (as I think it can) or not.
 

TJ Hooker

Glorious
Ambassador

Did you read the article? The fact that the card is only drawing around 75 W from the 6 pin is based on the design/distribution of VRMs, as well as firmware and/or drivers. As turkey3_scratch said, there's no "hard limit" that the PSU is imposing.
 

chazcon

Distinguished
Mar 21, 2011
3
0
18,510
0
I feel like Tom's got pressured into writing this second article, which to me is a complete sell-out. All these years we value Tom's objective opinion. No longer, how can we trust them now? Very bad choice by Tom's to cave and write this article.
 

turkey3_scratch

Polypheme
Ambassador
envy14tpe never said anything about there not being a problem, but I agree that it makes zero sense Tomshardware would have to avoid a lawsuit when literally AMD could not sue Tomshardware for anything. The only thing they could do is threaten to not give Tomshardware any hardware in the future.
 
Driver version 16.7.1 lowers Motherboard slot by 8 watts and ups power plug by 10 watts. Performance increases by a good number of most tests. AMD added a compatible mode that lowers the draw more but does effect performance a bit.
 
Status
Not open for further replies.

ASK THE COMMUNITY