Official OcUK RX480 4GB and 8GB review thread

[james.miller]

Melmac I dont disagree, other than AMD stating it was a 380x replacement - i don't remember that but maybe they did. I don't think anyone with the cards you listed would have been looking at the 480 as a replacement at all had the 1070 delivered on all fronts, but you know how it goes.

I'm not disappointed that it's placed where it is; i'm disappointed known how they've managed it given where nVidia are now with pascal. What was it, 1070 is 80% more efficient? that's a massive massive gap, lots of ground to gain with vega and what does that mean for AMDs next big core? Will there be one any time soon?

 

[andybird123]

The hype wasnt just built by forum goers. When AMD threw up slides saying things like "crossfire 480's faster than a 1080 and at 51% usage", they were the ones building the hype

 

[eddyr]

Most of the statements are true though. I've seen far more lies about Nvidia products in the last days than about the RX480. Nonsense like the 1080 is only 10% faster than the 980Ti.

Even if the ratio is as you say, which I don't think it is, the mindshare Nvidia already have make it insignificant. Like a fly in the face of a high speed train.
AMD have a less receptive audience, and lets face it, an uphill struggle to gain customers even when they are competitive. The decision for a casual gamer polling for opinion or browsing forums (who wants to confirm their provisional purchase decision after reading reviews) could be easily swayed by a few disingenuous or false posts and shifted to an alternative brand with proportionally more positive 'community press'. In reality thay have been turned away from what is a good value product. I have come across plenty of people who have purchased an inferior product this way.


In all cases extreme/false statement should be countered with neutrality and fact rather than posters taking it upon them selves to try to manually counter balance on some form of positive-negative point scoring basis.

 

[Mr Latte]

That's fine, you can play that game. These points are all real. The power draw, the poor clocking, the fact that decreasing the voltage improves the clocks on the cards tested, it all points towards the card being pushed further than it was ever intended. It's less than AMD had hoped and less then we hoped. You can Ignore the out-of-spec power draw and uncompetitive performance per watt because you're not interested, that's fine. You can ignore what could be possible with nVidias competing card because you're not interested in it, as well, and just praise the 480. That's up to you of course, but it doesn't change reality. As said, great card for the money, but its obvious AMD have had to cut corners to get there and for the rest of us enthusiasts it doesnt do much that we couldnt already do and it has us questioning where AMD will go from here. I cant ignore that, no matter how good the 4gb card is at that price point.

This has been an odd couple of weeks from both camps to be honest. Hype, disappointment, hype, disappointment, hype, disappointment. About the only thing we could rely on was floppers unwavering stratospheric enthusiasm for Polaris.

I think at £199 or less indeed some corners are cut. Its mainly plastic shell, basic alu heatsink is certainly not like $500 cards they said it was similar too at the presentation.

At no point have I said the card has no issues or is perfect. Yet some people like yourself focus more on picking at the issues and not give credit to other factors.

Indeed you present issues some people may of had with it which is fine and is in the tech news. I'm not a technician or can say if drivers can help or if the issue is with all the reference based designs or only when overclocking etc. Some might be concerned about it others might not be but from what I seen mentioned it may be an issue with budget motherboards. I guess if a lot of them have to be returned then yeah big fail on their part.

The power draw image was shown by me as it lets people decide for themselves if it is decent poor or not an issue for them. I think "Paul" is one of the more balanced and fair reviewers around so having bench-marked the card and been able to compare it his own comments in the video count more than mine.

If it matters to you the DX11 to me was a bit below my expectations but believe AIB will bring it more inline with 980 performance which for me personally was the goal I sought from it to warrant a purchase.

Would I rule out a 1060, not at all the last Nvidia card I bought was the Titan at £749 from here. If signs for the 1060 indeed become the better card for the masses to the RX480 then well done to Nvidia but credit to AMD for improving, making a card that has received good reviews, having plenty of stock available for day of release and enabling vendors to hit MSRP or close to it.

To deny the excellent things about it, the interest it has gathered or the sales of it already is ignorance...

 

[melmac]

The RX480 is trading with the 970 GTX and the GFLOP figures indicated it would trade with the 980 GTX. Reality has fallen short of that for stock cards. I've not seen any third party stuff so I can't comment on that..

Those figures are never an out and out guide to performance, there are cards with less performing better than cards with more.

My point was simply that AMD stated that they were aiming at the mainstream market. If they had cards that were going to be faster than the 980 they would have marketed them as 390 replacements.

 

[RavenXXX2]

I was put off the 480 by the power consumption. I reckon we could see the 1060 hit 980 performance at 100-120 watt, I'll be all over it.

 

[The Running Man]

How does it consume more power than a GTX 1070 though?

The card would be decent if launched 6 months ago..but waiting so long for such a lacklustre product. I don't know but it doesn't have a feel good factor.

 

[FoxEye]

I was put off the 480 by the power consumption. I reckon we could see the 1060 hit 980 performance at 100-120 watt, I'll be all over it.

Ditto. Could go either way, tho, couldn't it.

A 1060 that's 980+ perf if they really want to go for the jugular.

or

A 1060 that's worse perf than the 480, whilst being the same price or dearer, just to milk those with strong brand loyalty to nVidia.

I know which I'm hoping for, but wouldn't be surprised to see them go the milking route.

 

[Matteysan]

I think a AIB card will be replacing my 280x. But I'm still unsure.

 

[WYNIR0]

Hmmm, seems they are 'power throttling'

 

[Overlag]

Cant believe how hot and noisy this card is. I suppose this is following the trend in CPU's, smaller process MORE heat these days?

 

[Mr Latte]

Hmmm, seems they are 'power throttling'

Good video explains quite a bit...
Roll on the following and anything else to come from AIB:

• Asus Strix
• Sapphire Nitro
• Devil Card

 

[PhoenixUK]

http://www.gamersnexus.net/guides/2498-complete-disassembly-of-rx-480-and-road-to-diy-hybrid

No surprise that the card runs hot and gets held by the temps, LOOK AT THAT COOLER!! It's almost non existent. The heatsink is so tiny it looks like something from a £50 GPU from 10 years ago.

Also, powercolor has a giveaway announced for 11th July - looks like AIB cards could come out sooner than we thought?

And:

 

[Beren]

So, having read this thoroughly entertaining thread all the way from beginning to end, I have one question...

Are any of you going to buy one?

The reference one? No. The noise and motherboard power requirement bothers me.

An AIB one with decent power supply and cooler? Almost definitely.

I am not going to be building until little Vega is out so it is possible I will go that way. However for 1080p where I am going to game an AIB 8gb card looks like almost the perfect choice.

:edit: Also as others have said the performance of this card is all over the shop at the moment. Overall card performance is looking far worse than it should if they can eradicate the shoddy performance if the outliers. Looking at the graphs I think that by the time the AIB boards are out we are going to see driver revisions massively improve the overall performance if this card.

 

[Angilion]

If the PCI power draw is indeed in breach of the specs. I can see a whole heap of small claims cases for new motherboards.

I think it would be worse than that. Any specification organisation has to enforce those specifications. If the PCI-E power draw is in breach of the specs (and multiple independent testing of different cards has shown that it is), then if the specifications are enforced the card can't be sold as a PCI-E card. If the specifications aren't enforced then they're no longer specifications and not only are the specs useless but the specification organisation is useless as well.

But I'm having trouble with the idea that AMD could be so bogglingly incompetent that they could release a card that didn't comply with the specs. It does look like they have done just that, but it would be a ludicrous level of incompetence.

My guess at the moment is that AMD intended the card to be throttled down to meet the specs and the throttling isn't entirely working.

 

[madspunk]

WAT. It matches a heavily OC'd custom cooler and heatsink 970 at stock in DX11, for less money.

Matches 1070 / 980Ti in at least half the DX12 games available.

This is a new architecture on first lot of drivers ... Pascal / Maxwell have had 2 years of driver revision.

I'd expect it to significantly exceed a 980 in DX11 with a couple of months worth of driver releases.

Imagine what it will look like in 2 years ...

Also, for those expecting some kind of miracle re: performance ... this is a 212mm2 die. It's:

74% the size of GP104
58% the size of GM204

New drivers, and non-reference coolers plus OC = NVIDIA have absolutely nothing anywhere near it.

Sums everything up perfectly for me, nice to see at least one other person with similar views.

 

[mark-san]

Some interesting findings from our German friends at Computerbase:

https://www.computerbase.de/2016-06...el_potenzial_fuer_undervoltage_bei_der_rx_480

Basically, with some undervolting, the card starts showing it true potential. So ~390X performance and using only around 120W.

I just wonder what's up with AMD and this constant factory overvoltage on it's parts.

 

[madspunk]

Undervolting using Wattman? or whatever the program is called?

 

[Angilion]

Hmmm, seems they are 'power throttling'

I suggested that back in post 188, asking if anyone had seen any testing for it.

I'm not surprised that someone has done at least superficial testing for it and found it. I'd like to see someone do a proper test, though. That video was very much lacking in information (and not well presented). I am surprised that none of the serious hardware sites have done some proper testing to evaluate how much power throttling is going on. Even free monitoring software will give you a decent idea of the numbers.

It bodes well for performance of non-reference cards. On my 7950 I was able to increase performance by 36% in the benchmark I was using (Unigine Heaven) without overclocking and it was still throttled by power consumption (but much less). It also ran cooler (because I had reduced the voltages).

I made a post about it with some figures:

http://forums.overclockers.co.uk/showpost.php?p=26333269&postcount=5

So a properly made RX 480 that isn't artificially limited to meet a marketing TDP that the hardware isn't actually capable of could give significantly better performance than a reference one. Give it an 8-pin connector (or 2 6-pin ones), a decent cooler and a higher imposed TDP and I wouldn't be surprised to see a 15% increase in performance at stock speeds, no overclocking at all. Maybe more. After all, I got a 36% increase. It's not just some quirk of that one benchmark either - my 7950 is well above normal in everything I've run.

 

[humbug]

Some interesting findings from our German friends at Computerbase:

https://www.computerbase.de/2016-06...el_potenzial_fuer_undervoltage_bei_der_rx_480

Basically, with some undervolting, the card starts showing it true potential. So ~390X performance and using only around 120W.

I just wonder what's up with AMD and this constant factory overvoltage on it's parts.

Yeah several people are finding some sort of power to performance related issue.

OC3D found by upping the power limit (not the Mhz and the MHZ remain the same during testing) the performance increases.

Polaris seems to have some design where the core is hitting TDP limits and a reduction in performance occurs without a change in Mhz, by lowering the volts power levels are lower.