AMD Radeon RX 480 Power Measurements Repeated And Clarified

[AjvarD]

Hmmm, does your hypotetical conslusions contradict a video with a multimeter which clearly shows that 3 VRMs go to PCI-E slot and 3VRMs go to 6-pin connector?
Watch since 53:40 https://www.twitch.tv/buildzoid/v/75850933

 

[FormatC]

@Onemoar:
I know this. But read my statement above. The small "shunts" were not on the right place. Not on my card, not on that retail I got for testing from a shop. We found 4 of this phases electrically connected to the PEG and two rails to other source. The PCB design is really flexible to use a variety of such bridges to connect all to each and back

AMD wrote, that there are some different connected boards on the market. So the guy on the video was lucky

 

[AjvarD]

I respect this product and it is one of the best deals for this price point...but what I absolutely hate: if a company doens't respect their own products and engineers but let the marketing guys make such totally unnecessary mistakes like this.

AMD also doesn't respect consumers misleading people claiming GPU is 110W holding whole GPU board in one hand while company's employees fit a 6-pin connector to hide AMD's lie in expense of our safety and money for buying new mobo.

Exact word is LIE. Because as soon as they started hiding in expense of failing specifications and at the same time even at expense of their payments for 8GB memory size in a 4GB cards!
Yeah, AMD decided to rush paying more for RAM and making us pay for burnt Pcs. Polaris isn't wrong? Well, AMD IS and AMD is building Polaris...

 

[FritzEiv]

I wonder if they could simply solve this by a driver update, so that the card could function as any other card (which means taking most of the power from the 6-pin connector rather then the PCI-E).
Anyway at this point it is clear to me that anybody wishing for an rx-480 should wait for aftermarket solution. Brands like Sapphire will definitely solve those issues

The protocoll, used by their PWM controller allows a re-balancing and it is in theory also possible to supply the whole card with only two of this oversized phases
I'm sure they will move a little bit more load to the 6-pin PCIe, beacuse he is completely connected. My own tests showed that they will survive 300 watts and more without damage.

Specifications covering just how much power is allowed to flow through the pins of a 16x PCIe motherboard slot were established a long time ago. However, saying that they include a maximum wattage isn’t just misleading, but flat out false.

This is exactly what I wrote on top of page Two. Only the current is interesting, the voltage is within the ATX specs and a completely other story. The translation from German is not clear enough - in original I wrote, that this wattage product is used from the masses to simplify the problem but it is totally unusual

I wrote this also to our editor-in-chief and hope that he can fix it asap. Thx for finding.

fixed

 

[blahsaysblah]

Hmmm, does your hypotetical conslusions contradict a video with a multimeter which clearly shows that 3 VRMs go to PCI-E slot and 3VRMs go to 6-pin connector?
Watch since 53:40 https://www.twitch.tv/buildzoid/v/75850933

Same curious about the discrepancy of TH analysis. He shows physically, there is 3/3 split.
Same video on his youtube channel, https://youtu.be/plC7tOYIqBw?t=3241

 

[FormatC]

BTW:
To figure this out we also were in contact to a community (no, not Reddit). It really seems, that existing different boards but I'm sure that all the board partners will do a better job.

All this things were the reason why I wrote especially "hypothetic" in this review (in the launch I also wrote 3:3). But after longer discussions and different findings it was more plausible to got a 4:2 card. This also might be the reason for such different measuring results between different sites, I don't know. But let me check this later. We are overloaded with some other cards

 

[FritzEiv]

"Summary
1. Our modified second round of measurements replicated the results of our first round. The amount of current flowing through the motherboard PCIe slot’s 12V rail exceeds the upper limit of the tolerance range described in the PCI-SIG specifications. Our repeated measurements of the reference card show currents 23 percent above the PCI-SIG norm."

So, the summary was the same. The new article just puts more emphasis on the minimal risks/dangers than the compliance failure itself. Seems like an attempt to please someone, or appease some source of revenue.

As I've pointed out in a couple of places, Igor sent his log files/test results to AMD several days before we published our initial review. AMD has STILL not contacted us directly about any of this. In other words, I can promise you there was/is absolutely zero pressure from AMD. Now I do not know if AMD is an advertiser (I'm the editor-in-chief, and while I'm sometimes aware of who advertises, it's only because I see the advertisements, just like you), and I hope they are, but nobody on the business/marketing/sales/executive side of our parent company has said a single word to me about our review, so I can also assure you there is no other pressure here. The pressure is self-imposed: Our data, our graphs, are all over the place, causing confusion because of the way that data is being used & quoted, and in some corners even being critiqued. We wanted to re-do the testing, we wanted to add some clarification, and we wanted to set the record straight. That simple. And several people gave up significant portions of their weekends (holiday weekend here in the US) to do this.

 

[Beholder88]

Cause for this second article aside, this whole thing will be moot once the partner companies start making their own versions. I can see a swap from the 6 pin power connector to the 8 pin, or even two 6 pin connectors. I think the power regulation will be changed as well. I only hope their designs aren't already too far along to make such an adjustment, unless it was already in the works. I just bought a GTX 1070, but hope AMD can get this right for the good of the overall industry.

 

[L the otaku]

This may put some idiots at bay, others will make reasons why it's still bad. For how many more years do we have to have backwards compatibility for boards released in 2004? That's the question I have. Since I upgrade board and all every 2 years, this is a problem that is foreign to me.

 

[tommy4242]

The PCIe spec is quite clear on current capability per pin (PCI EXPRESS CARD ELECTROMECHANICAL SPECIFICATION, REV. 1.1):
"Power pins (+3.3V, +3.3Vaux, and +12V) are defined based on the PCI Express power delivery requirements specified in Chapter 4, with the connector contact carrying capability being 1.1 A per pin."

There is no headroom. If you look at PCIe connector manufacturer specifications, they are rated at 1.1A.

The 23% are misleading. Power dissipation in the connector / mainboard traces is quadratic with current and increased by 50%.

 

[FormatC]

I will disassemble a RX 480 and measure the resistance between the Drain of high-side FETs and the different rails again. Just bringing my digital multimeter back from the other lab.

 

[hannibal]

Finnish overclocking spesialist did say in gpu Forum that you can change the power balance between the pci slot and 6 PIN extra power by 30.6 % so it is actually quite easy to reduce the current draw From the pci slot.
The interesting thing is why this happened at all... Because the system does allow that change... But all in all this can be remedied, so not as bad as it first seemed to be.

 

[hannibal]

One other interesting thing is that

GPU-Z- 0.8.9. ASIC Quality for rx 480

Muropaketti: 87,1 %
techPowerUp!: 83,0 %
Nordichardware: 75,6 %
Hexus.net: 75,2 %
Sweclockers: 74,3 %

The sweet spot seems to be 75%. Those 480 runs below 80 celcius.
If the Asic is near 90% the temperature can go ower 90 degree celcius in 3d. So the production system is not quite mature yet.

 

[FormatC]

Ok, interesting findings...

I measured that PEG slot is connected to three of the left six phases (0.049 Ω) and also to the single phase right on top next to the PCIe connector (0.042 Ω). The other three phases are going to the PCIe connector. One of the bridges has no function, it seems it's a dummy or something else. The single phase on the top left is also connected to the PCIe. At the end we have 8 VRM sections, four of them are connected to the PEG, the other four to the PCIe.

GPU-Z is monitoring only the six phases on the left, that means 110 Watts were going to this part (+ losses). The rest of power consumption is splitted between this two phases.

 

[Shankovich]

I think Igor (just as an aside) should have noted AMD tends to have a habit of shipping new cards generously over-volted to avoid issues. A lot of people are successfully undervolting their cards and solving this issue.

Could this test be done by OC'ing a reference 970? I'm sure overclocking most reference cards would pull above spec on PCI-E slots.

 

[technoholic]

As a potential consumer, i do not like whatsoever to hear "blurry" statements. IS THIS PRODUCT SAFE TO USE OR NOT? The statement "quality motherboards can handle this board good" just does not cut it. Which motherboards and which PSUs are those?I like to hear definitive statements if i'm going to shed my hardly earned money on a product and if my motherboard (or some slots/components) is under risk of being damaged

 

[FormatC]

I think Igor (just as an aside) should have noted AMD tends to have a habit of shipping new cards generously over-volted to avoid issues. A lot of people are successfully undervolting their cards and solving this issue.
Could this test be done by OC'ing a reference 970? I'm sure overclocking most reference cards would pull above spec on PCI-E slots.

It doesn't exist not one authorized GTX 970 Reference from Nvidia, they launched only board partner cards. Only Manli (a daughter of PC Partner) has produzed a similar card, based on Nvidias reference cooler and specs.

As a potential consumer, i do not like whatsoever to hear "blurry" statements. IS THIS PRODUCT SAFE TO USE OR NOT? The statement "quality motherboards can handle this board good" just does not cut it. Which motherboards and which PSUs are those?I like to hear definitive statements if i'm going to shed my hardly earned money on a product and if my motherboard (or some slots/components) is under risk of being damaged

We can't give you any warranty, please ask AMD. The life is a bitch and coincidence their sister

 

[hannibal]

Wait for that power balance driver fix. After that you will be fine. They did try this with afterburner and did get well below the speks so this can and will be remedied! I did not understand a bit of those setting that They did, but it works, so it proves that this can be fixed by software upgrade.

 

[hannibal]

This is the setting that Stilt did use. On your own risk, but this proofs the point.

"MSIAfterburner /wi06,08,1E,BB" and "MSIAfterburner /wi06,08,1F,B0".

 

[JackNaylorPE]

Wait for that power balance driver fix. After that you will be fine. They did try this with afterburner and did get well below the speks so this can and will be remedied! I did not understand a bit of those setting that They did, but it works, so it proves that this can be fixed by software upgrade.

The "software Upgrade" is like buying a 6600k and then turning off Intel Turbo Boost. There is a simple solution. Resist the urge to tell your friends that you will be the 1st one on the block to have the "latest and greatest" and what for the AIB partners to release the 8 pin cards. Buying a "reference card" has never been a good idea....

Is sitting at your PC checking e-tailer stocks worth the T & E to get the 1st release products:

a) First stepping boards / cards represent the bleeding edge; wait for the later steppings and you get a better product w/ all the big fixes, fabrication fixes, updated fan profiles, tweaked BIOSs that you don't get on the 1st stepping cards

b) Other then the EVGA SC series, you almost invariable get a PCB from on the nonf=reference cards (Gaming, Strix, G1, etc) with improved components, extra heat sinks, bigger VRM, thermal pads, improved coolers, etc.

c) Non-reference cards almost always overclock better than reference ones.

Yes, AMD does over-volt ... or better said "aggressively clocks" their cards in the box... have done so ever since the R series. That's why we have historically seen only single digit fps increases on test sites while when nVidia cards are OC'd, its well into double digits. versus double digit (up to 31+%) . But clearly overclocking the reference 480 is not an option. You can overclock the nVidia 10xx reference cards, not that it actually does anything for you as the card throttles even at stock speeds.

The 480 exceeds specs, AMD admits this and will fix it. As with any other "power cap" scenario, performance will be impacted... simple fix, don't get a reference card.... or do the update and take the performance hit and have a card you can't OC.

The 10xx FE (reference) cards throttle ... nothing new here ... simple fix, don't get a reference card.

In either case, we are 2 days from the rumored release of the 1060.... or take the performance hit from the throttling (assuming that 1060 suffer from same deficiency of 1070 / 1080 ... which is not guaranteed) and again have a card you can't OC.

 

[FormatC]

Just for interest - I've updated the German version after the closer look to the board with my digital multimeter and we will wait for a short update here too. It's coming soon.

At work....

 

[uglyduckling81]

It's exactly the same story from Team Green. If you have any brain you don't buy reference. GTX 1080 runs too hot and throttles with reference cooler. Reference RX480 is poorly engineered to achieve some marketing goal and has power problem.
Both problems are easily fixed with aftermarket solutions.

 

[George Phillips]

If a motherboard can't even handle that 23% increase in amperage, it might as well as be tossed out of the window. One of the reasons this issue even showed up might be that people who buy the cards want to save money and have older systems/motherboards. All newer motherboards designed and made within 4 years must be handling this easily.

 

[the_countess]

On top of this, the motherboard connectors don't stay pristine forever,

if you look at the specs for from the manufacturers of the PCIe slots themselves you'll see that each pin is rated for up to 2.8a at a ambiant temperature of 65 degree's C.
with a minimum of 4 pin's that's 11.2 amps total.

at a much more realistic, but still very bad case of 50 degree's C that's up to 3.3 Amp's or 13.2 amps total. basically double what AMD puts on them.

 

[FormatC]

The VRM section is directly above. I measured up to 80°C (and more) inside the slots after long-term runs in a closed case. The heat is wandering through the PCB, also downside to the slot.