Hardware AMD Radeon RX 480 Review

i have an overclocked ( factory overclock ) 750ti, and after testing i can confirm that during heavy gaming im averaging about 80~ watts all arriving through the PCIE slot so i really dont see what all the fuss is about...

i saw someone claim that a 480 ruined 3 high end motherboards ( if you think its the card destroying your motherboard, why install it into another suspiciously top end board )
perhaps when 3rd party cards arrive they will likely have 8 pin power delivery ( although the better power delivery could well make up for this and reduce the power draw through the slot )

but the Nvidia fanbois are already out in force over this...

 

Thanks for the reply...
Maybe unfair wasn't the correct term to use and although the review may be more balanced, some of my comments were directed more at others who have been signing the praises of the 480.

In response to AMD's press release...
This is really not sitting well with me, it seems like they are trying to justify technically dangerous profiles by the fact that they reach higher memory speeds. I can understand GPU's shutting down or underclocking due to issues but to actually cause PCI-e failures or even full MB failures... seems like really poor testing to me and they won't be 'fixing' that with future drivers, it certainly won't help the fence sitters choose a 480 over a 970 or 1060.

Can someone please explain to me why nVidia would revert to 192 memory bit rates for 3/6GB as they are rumoured to have done for the 1060 rather than use 256 for 4/8GB as AMD does? Is there really that much of a financial difference or a technical reason?

 

80W is technically within the PCI-E spec, since the spec is "75W within a margin of error", said margin is 10%, so a total of 82.5W can be pulled through the socket before a card can be determined to be "Out of spec", the 480s are allegedly pulling over 110w, far outside of spec error margins, and definitely out of specifications.

Motherboards aren't built to handle those sorts of voltage loads on the PCI-E sockets because they simply don't need to be, there's no point designing a PCI-E socket capable of delivering 150W because it should never be needed, so long as the parts being installed meet the specifications to which the socket is built.

If this is true and AMD have seriously been pulling that much power from the socket rather than the power connector; they dropped the ball hard, and it'll probably cost them a fair bit in the long run.

 

Pleasant audit. Appears a better than average exertion from AMD at an aggressive value point, that may well catch this end of the business sector.

 

AMD have released a statement they are now aware of the problem and will release a new driver on Tuesday to correct the power state problem

http://www.anandtech.com/show/10465/amd-releases-statement-on-radeon-rx-480-power-consumption

 

On a further note, isn't it more than a little backwards to have the device able to draw more power than specified? A device should be able to 'ask' the socket for more power with the socket responding with 'Sorry, but that's your lot!'.

 

margin of error?

erm no - 75w is the headline power draw - 80w is out of spec , and in fact the GTX 960 has been known to spike over 150w from the socket alone.

go read TPU and anand for proof (or watch the video from vault tech as links are provided)

its only because AMD has beenr ecorded doing it its suddenly `bad` - NV have been doing it for about 2 YEARs but thast ok

 

I don't claim to know much about this but I haven't heard about nVidia GPUs frying sockets and MBs due to excessive power draws. At least I have never seen the same reaction on websites as this instance so I find it hard to believe that it's that simple.

 

Of which; it can be split between the 3.3V and 12V lines, the balance of which may be essential to keeping boards from melting. I would not be surprised if Nvidia did a lot of testing on budget boards to see if it were possible to safely draw in excess of that voltage, which AMD likely did not have the funds to manage.

At the end of the day; This is seemingly happening. This soon after release; RX480s have been caught causing board failures, while I've heard very little about Nvidia parts causing the same, so they're doing something differently, to the effect of causing, or at least provoking, failures.

Question now is what's going to happen. Customers could viably have a case there, since the card purports to conform to standards which it does not, and could arguably be the cause for failures.

 

I do admit that something is abit off with people claiming the card is killing Motherboards... but i wonder how many of those claims are
A) Nvidia fanbois trying to urinate all over AMD's success
B) Sensationalist headlines from 'Tech' Forums fishing for views
C) people noticing the cards push beyond its specification, so purchase the cards to cause them to fail so they get a refund / replacement Motherboard

Now im not saying that there isnt an issue, but i wonder how out of proportion its getting ( like the GTX 970 memory issue that was a legitimate issue, but gained WAY to much hate )



750ti pulling 92 watts from the PCI-E socket alone ( average, not accounting for Spikes in power ) then they overclocked the card +10%~ but they dont show overclocked power consumption ( something that most review sites dont do that really irks me )

nice to see AMD coming out and saying that there is an issue, and taking immediate steps to try to solve the problem

 

The 950 is the same too if not worse. They made one with no six pin, so every one who has overclocked one of those has broken spec.

AMD have done this before with the 6990 and god knows how many other cards. Nvidia have done it too. It's just Nvidiots desperately trying to stop people staying with AMD.

I've only ever seen pins burn out back in the X58 days and people were running three GTX 480s under water for that to happen. And even then the 24 pin socket would burn, not the pins in the PCIE socket. They've since fixed this by adding a 6 or 8 pin PCIE plug onto the motherboard itself and the lanes get extra juice from that. But as I said, this was a problem specific to 3/4 way SLI and CFX not 2 way or even a single card. It took absolutely extreme circumstances to make anything burn.

 

Calculated TDP is at best inaccurate, At worse its downright dangerous.


Ive saw people base PSU assumptions off it and the PSU goes POP under any load as those calculations do not include alot of stuff.

5990 and 6990 where out of PCIE spec and above 6/8pin spec once overclocked. Did not do anyone any harm. I have ran both under water and with heavy clocks and not noticed any issues. They both draw a heck of alot more power than anything else on todays market.


A worrying trend of low end PSUs seem to be developing where they can not take any sustained load.

I wonder how much of this is the 480 out of spec and how much is the PSUs people are using causing the issues.

Any story is a story in tech, Apple gets hated on for even the most basic of issues, The clock stopped peoples Alarms going off one morning and they got press for a week based on that issue lol.

AMD have took the step to say we may have a issue, It could effect sales which would not be good for them. This should of all off been picked up on in internal testing though.

If AMD are replacing the dead boards then good for them.

 

FWIW, there is a difference between what a card does at stock and what it does overclocked. Overclocking is the user's choice, and remember all of those disclaimers about "doing so at your own risk"? Stock settings shouldn't break spec, and I say that regardless of whether it's the green or red team doing it.

 

Thank you for this information . Been a CPC subscriber since issue 1 and luv the reviews in there and Bit Tech .

Just for my own info I noted only in Hitman did RX 480 gain over Fury X, I think a driver optimisation issue with Fury X possibly? The rest for 1080P Fury X on min FPS compare is faster than RX 480 by :-

AOTS 18% , Fallout 4 36%, Division 34%, Warhammer 11%, Witcher 3 26%

So average for those set of games = ~25%, would have been interesting to see power usage stats per game due to how Polaris's main aspect is perf.per watt. Not too sure about TPU perf.per watt chart is accurate as it says at top:-

So I would think they used RX 480 @ 163W and Fury X @ 246W from page 22 of review, I think time to ask on TPU.

A little disappointed with power usage TBH on RX 480, at Computex the slide where they showed CF 480 is better performance+efficiency than GTX 1080 I was expecting lower power usage but then they did have a slide with "Power" as 150W for 1 card.

In TPU RX 480 review for "Typical gaming" chart where average W is shown, a single RX 480 is on a par with GTX 1080 and no way CF RX 480 gonna be same W so the AOTS slide was "over-hype" IMO on "efficiency".

In case as such where I was thinking of buying or bought a card with no PCI-E plugs I would accept that, but after seeing how past gen AMD cards with PCI-E plugs (the few I collated in post 34) I'm disappointed. If I had an RX 480 and was OC'ing like I do with all cards it would be risky on the PCI-E slot power usage IMO, on average RX 480 under load is using 50/50 spilt of power over PCI-E slot/plugs, where as past cards is way way lower on PCI-E slot (data in post 34).

In the PC perspective RX 480 power article they do explain that THG's data the spikes to (x)W on RX 480/other cards doesn't represent typical power being used.

From what I made from this article and others I've read average is what we as users should be looking at in this case not instantaneous. Average PCI-E slot power draw on 6990/7990/295X2 was well within spec, THG data. Due to comment on the PC Per article from viewer stating THG had a 960 Strix which was spiking heavily they did the same new testing treatment on one.

Next as someone who has been delving into Hawaii/Fiji bios mod a lot and as RX 480 uses IR3567B and is PowerPlay 7 card like Tonga/Fiji I think they won't be able to curb PCI-E slot power usage to bring it in line with how Hawaii/Grenada/Fiji was.

Some info I posted on HUK regarding my thoughts on RX 480 / BIOS / IR3567B Post 250, Post 187, 247.

 

Can I ask if Bit-Tech will be re-testing the 480 once the software update has been released to see how it impacts on performance or if it does?

 

AMD Radeon RX 480 Undervolting Performance

It seems to work better by Undervolting the card

http://www.legitreviews.com/amd-radeon-rx-480-undervolting-performance_183699

 

Like in CPUs how much voltage each card needs will vary widely to get a stock stable setting AMD has set 1100mv as its stable level 100%. How many cards will even do 1050 will vary widely.

The fact they saw a power spike of 1.31v is nuts and does explain why they are struggling to stay in spec. Does not help that the card is not well wired judging by the tear downs.

What I can see happening is AMD banning all overclocking easily done in drivers, and forcing it to 1075 which will be within the margin for error in built.

Custom cards will still ignore these guidelines but they will come with much better power delivery systems.

 

The spike to 1.31V is not biggie IMO.

For example on Hawaii/Fiji you'll have say card setting VID per DPM at x level but what we see in monitoring is drooped VDDC. This droop is down to how LLC is set on AMD EVV cards (Electronic Variable Voltage). My current Fury X DPM 7 (highest state 1050MHz) VID on stock ROM is 1.212V when monitoring it you will never see 1.212V but every so often you may see a spike close to it.

The other aspect of EVV/PowerTune is even when card is in like DPM = same VID you will see varying VDDC depending on app load card. For example you can load card with 3DM FS and get different VDDC to Crysis 3 but on both occasions card was giving flat 1050MHz GPU clock.

Besides how the whole PCI-E slot / plug distribution is currently, I don't know how the card passed PCI-SIG. View this video

 

Just as update to members here, The Stilt who is a pro overclocker and has good AMD product knowledge has released a method to reprogram IR3567B to take more of the loading to PCI-E plugs but due to how the VRM is spilit 50/50 between PCI-E slot/plugs the redistribution yields only a 10W decrease in PCI-E slot usage.

1. The post explaining the ref PCB.

2. The thread with i2c command fix and soon bios fix release.

3. Anyone interested in photo showing phase distribution and has a link to Buildzoid's video testing RX 480 PCI-E slot/plug setup view this post by McSteel on TPU.

Due to only 10W lowering on PCI-E slot and how PCB design is I would assume this would easily be used up when OC'ing. Hopefully when AMD release a fix it will not hamper performance to gain lower PCI-E slot power usage. If there was another controller that could change PCI-E slot/plug power distribution or PowerPlay in ROM could I know The Stilt would have done this.

 

I don't get the line of thinking that AMD can reduce power usage and retain performance... If they could achieve that performance with lower power why the hell did they not go with the from the start or increase the power for further performance? Is AMD really being that shoddy with their product designs and implementation now? It just doesn't add up.

I can only see that any power reduction will lead to performance drops, the question is, how much performance will they lose once they settle on a fix? Obviously some performance will be saved due to improved drivers but I doubt then that there will be much of a future gain to come.

Anyway, whatever happens I'll be waiting longer than I thought... I just don't trust the market right now.