AMD Radeon RX 480 8 GB

[HD64G]

I'm not strong in Astronomy.

But isn't Vega 10 is much bigger than Vega 11 ?

In this gen of AMD GPUs, numbers show when design was started, so V10 is the 1st and the smaller of the 2.

 

[BiggieShady]

80W actually doesn't tell you much because you need to look at both voltage and amperage. In the test the RX480 draws 12V @ 7V = 84W on average.

Actually it does, normal working PSU keeps voltage constant as atx spec commands ~12 V and the current (or as you say amperage) is what changes and consequently the power spent (current multiplied with voltage) ... so yeah, nitpicking

7A is 27% over the absolute maximum current limit of 5.5A.

That current limit sets pcie power limit at 66 W, but we colloquially take it as 75 W ... I just don't know, we'll have to wait for it to kill couple of cheap motherboards that use uncooled cheap mosfets in cases with bad airflow

 

[Hellraiser1981]

If you check the pinout for the card edge connector you can see that 5 pins are allocated to the +12V rail. This rail is sourced by the ATX header and any Auxillary pcie plugs that apply power closer to the pcie bus for sli, xfire. The original molex standard states 1.1A per adjacent pin. Some on ncix and digikey state up to 2.2A per pin. 66W to 132W through the slot connector 12V pins. Its going to depend on the quality and design of the power distribution systems on your motherboard. A board with a 20 pin ATX header would probably burn.
Out of compliance with pcie spec at 75W. If you look on the pcb front, above the small section of card edge before the notch, you will see the regulator circuit thats pulling all that power. This regulator should be configurable in the VBIOS, however it probably supplies power for the core functions of the gpu.
We've all seen the "Your power connecter is not connected" on a black screen when you forget to hook up the 6pin. That's bare minimum Gpu running off power from the slot.
Overclocking is supported but not recommended, once you pop something overclocking, you can't send it back for the free magic smoke re-injection.

 

[TRWOV]

I just hope AMD can fix this with a bios update. I guess motherboards with supplemental power connectors won't have a problem but those with budget boards could have issues.

Considering the price of the 480 many with budget rigs will likely buy them so it's a potential time bomb for AMD if it's not corrected right away

 

[nem..]

AMD RX 480 PCB & VRM Design Beefier Than GTX 1080 Founder’s Edition – PCIe Power Issue Detailed

Read more: http://wccftech.com/amd-rx-480-pcie-power-issue-detailed-overclocking-investigated/#ixzz4DDxltUVA

Article 01/07/2016

 

[HD64G]

An official AMD statement about the PCI-E topic:

As you know, we continuously tune our GPUs in order to maximize their performance within their given power envelopes and the speed of the memory interface, which in this case is an unprecedented 8Gbps for GDDR5. Recently, we identified select scenarios where the tuning of some RX 480 boards was not optimal. Fortunately, we can adjust the GPU’s tuning via software in order to resolve this issue. We are already testing a driver that implements a fix, and we will provide an update to the community on our progress on Tuesday (July 5, 2016).

https://www.techpowerup.com/223833/official-statement-from-amd-on-the-pci-express-overcurrent-issue

 

[john_]

So, I checked a review from TPU about a low power GTX 950, those with no extra power connector

ASUS GTX 950 2 GB (no power connector) Review | techPowerUp

79W from the PCIe bus.

And that's with NO overclocking. What will happen when you overclock that card? 80W? 85W? 90W? Throttling?

 

[okidna]

So, I checked a review from TPU about a low power GTX 950, those with no extra power connector

ASUS GTX 950 2 GB (no power connector) Review | techPowerUp

79W from the PCIe bus.

And that's with NO overclocking. What will happen when you overclock that card? 80W? 85W? 90W? Throttling?

I think you might want to read the testing methodology description :

Peak: Metro: We use Last Light at 1920x1080 as it produces power draw typical to gaming. The highest single reading during the test is used.

On average the GTX 950 is fine, 74 W, so still below the specification.

 

[Grings]

I was really hoping they would have sorted out that blu ray power draw by now

I'm after a new card for my 2nd pc at some point this year (primarily for kodi, netflix etc on a 4k tv) and was hoping this would fix these issues

I'm also not a fan of both companies pushing the power limit of the pci-e slots just to have a card with 1 pci-e power connector, every psu i have bought in the last 8 years or so has had at least 2 pci-e connectors ffs

 

[john_]

I think you might want to read the testing methodology description :

On average the GTX 950 is fine, 74 W, so still below the specification.

Yes I see that. But those 74-79W are NON overclocked results. It doesn't talk about what happens when you overclock the card and the memory like in the overclocking page of that review. It says that most of the time the card stayes at 1200MHz. In the overclocking page it talks about "1447 MHz on the GPU and 2060 MHz on the memory". 24-25% overclock. And I bet the card doesn't throttle to maintain those 74W, because "Actual 3D performance gained from overclocking is 19.1%." Can you get 19% more performance while having the same power consumption?

@W1zzard Can you tell me if I am thinking it wrong? Could this be interesting enough for you to do a quick test?

 

[okidna]

Can you get 19% more performance while having the same power consumption?

I'm most certain that you can't, it will increase power consumption, especially if you're playing with power target and voltage.

But why does it matter anyway? Why do you feel the need to add overclocking in the equation?
The card working as intended in stock setting, within regulation and doesn't violate PCI-E specification, that's the main point.

 

[john_]

I'm most certain that you can't, it will increase power consumption, especially if you're playing with power target and voltage.

But why does it matter anyway? Why do you feel the need to add overclocking in the equation?
The card working as intended in stock setting, within regulation and doesn't violate PCI-E specification, that's the main point.

Many people and the press make too much fuss about this and they should. But at the same time you see many reacting like this is the ONLY card in the world, overclocked or not, that passes that 75W limit. Well it's not. While this is AMD's mess, it shouldn't remain as a "AMD messed up" story, but also become an opportunity to educate people. Take models that work at their defaults at the limits of the power they can get from the PCIe bus and probably one or two extra power connectors, and show people that when overclocking a graphics card, it is not just the GPU going over some specs. It's not just temps and furmark/stability testing that you have to keep in mind. I am afraid the main point here, the big picture, is completely bypassed from almost everyone. Even those who point at GTX 960 Strix at RX480's defence.

 

[RejZoR]

I think you might want to read the testing methodology description :



On average the GTX 950 is fine, 74 W, so still below the specification.

Who says the spike only happened ONCE and never ever again, but it's somehow doing it injustice because the highest reading was noted in the review? If it spiked once, you can be pretty sure it would spike several times repeatedly during normal gaming session.

 

[okidna]

Who says the spike only happened ONCE and never ever again, but it's somehow doing it injustice because the highest reading was noted in the review? If it spiked once, you can be pretty sure it would spike several times repeatedly during normal gaming session.

The average/typical gaming power draw says it all, at 74W there's little possibility that the 79W spikes happened very frequently or during an extended period of testing.

 

[RejZoR]

You can have 150W spikes every 10 seconds for 1 second and the average would show you 74W if the rest of 9 seconds is 60-70W... You know, that's how averaging works. That's like showing average framerate of nice 60fps, but if you look at the actual data, it can fluctuate from 5fps to 350fps. Which can be pretty unplayable, but the average looks awesome...

 

[okidna]

You can have 150W spikes every 10 seconds for 1 second and the average would show you 74W if the rest of 9 seconds is 60-70W... You know, that's how averaging works. That's like showing average framerate of nice 60fps, but if you look at the actual data, it can fluctuate from 5fps to 350fps. Which can be pretty unplayable, but the average looks awesome...

I know how averaging FPS or power consumption works, my assumption is based on how the game benchmark (Metro LL) power usage characteristic, something like this :

Pretty much stable for the whole run and no extended period of power spiking.

 

[NDown]

how is it that's fast enough to run VR while maintaining 970-like performance??? run all DX11.2 powered games at Medium-High for >60fps?? pfft. The 480's big older bros even butchered it across the deck...

what? if "butchered" simply means 1-3 fps difference then i have nothing more to tell you lmao

 

[gupsterg]

@W1zzard
On the performance per watt page of RX 480 review it says at the top:-

We used the relative performance scores and the typical gaming power consumption result.

So am I correct in thinking you have used RX 480 @ 163W and Fury X @ 246W from page 22 of review and applied that wattage to page 24 results to express perf.per watt?

Cheers .

 

[ikeke]

I'll just leave this here.

http://www.overclock.net/t/1604798/...express-overcurrent-problems/10#post_25315251


The Stilt@7/3/16 at 7:22am

Luckily the power distribution balance can be altered

Working on it and I'll should have further info available after w1zzard from TPU has tested my fix.

The Stilt@7/3/16 at 7:33am

It can done through the drivers too, however I'm not sure if AMD has even looked into this method. I've heard that they are looking to reduce the total power draw by other means. It is a VRM controller feature, so I'm not sure AMD is even aware of such possibility. But we'll see.

Looks to be a feature of IRF IR3567B, perhaps from 50/50 split to 30/70.

 

[RejZoR]

Well, I said just this and yet everyone continued with the drama about PCIe melting itself under the power of thousand suns. Polaris seems clever enough power delivery wise, meaning something like this was to be expected.

 

[Mike89]

Biggest issue for me would be the noise. I am one that simply cannot tolerate a noisy vid card.

 

[okidna]

 

[gupsterg]

I'll just leave this here.

http://www.overclock.net/t/1604798/...express-overcurrent-problems/10#post_25315251


The Stilt@7/3/16 at 7:22am


The Stilt@7/3/16 at 7:33am


Looks to be a feature of IRF IR3567B, perhaps from 50/50 split to 30/70.

All kudos to The Stilt for his effort in aiding community and @W1zzard but I was hoping for more than 10W drop, link to post by The Stilt. As then it would place RX 480 in line with how Hawaii/Fiji was on PCI-E slot power draw. Seems to me IR3567B can't redistribute the way power is drawn from slot/PCI-E plug. May well be due to ref PCB design viewing @buildzoid video on RX 480 and @McSteel 's post on TPU.

From meddling with bios mod on Hawaii/Fiji, PowerLimit values in PowerPlay don't differiate on where power is drawn from, they just limit GPU to (x) TDP W, (x) TDC A and (x) MPDL W.

All in all gutted with ref PCB as had been hoping to buy one for tinkering.

 

[W1zzard]

All kudos to The Stilt for his effort in aiding community and @W1zzard

I've completed the testing. You can shift power draw off the PCIe slot just fine.

the number at the bottom is how much to shift the power draw, 0 is default, 16 is maximum

However, this also means that the 6-pin will go further out of spec...

 

[gupsterg]

Thank you for your update .

When I say it can't redistribute the power I mean the fix shifts some load from 3 phases connected to PCI-E slot to 3 phase on PCI-E plug. It's not changing supply source to the 3 phases connected to PCI-E slot. Which originally I thought maybe possible, as more info (buildzoid's video) was published it become somewhat clearer the PCB design is the issue. I still thought perhaps The Stilt knows of some register within IR3567B which would change supply to the phases supplied by PCI-E slot.

This shift is reducing ~10W from PCI-E slot, which with some OC'ing will easily get consumed IMO.

As much as I welcome this fix from The Stilt's and your efforts it's still not placing RX 480's draw on PCI-E slot similar average W as 390X/Nano/Fury X.

I'll be waiting for either updated ref PCB or AIB card to purchase.

Any chance of information I was interested in post 343? thanks .