AMD Radeon RX 480 8GB Review

[SpAwNtoHell]

this is the only good thing about this card: "Finally, AMD gives us a reference card that doesn't sound like a hair dryer under heavy load. Well-done!"

otherwise in uk:

Sapphire Radeon RX480 AMD Polaris Graphics Card 8GB £233.40 Inc VAT in stock
Asus AMD Radeon RX 480 8GD5 Video Card £254.99 Inc VAT in stock
EVGA GeForce GTX 970 SC ACX2.0 GAMING Graphics Card - 4GB £218.60 Inc VAT
Zotac GeForce GTX 970 Gaming Graphics Card - 4GB £214.03 Inc VAT
MSI GeForce GTX 970 GAMING Twin Frozr 5 Graphics Card - 4GB £238.76 Inc VAT

Who would buy this compared to gtx970??? based on this review results... no one at least in UK ... unless in vr is at least double the performance of a GTX 970...

 

[jasonelmore]

same wattage as a 1080?!?! wtf you smoking.. it's a full 50w lower. see anandtech power numbers

 

[quilciri]

same wattage as a 1080?!?! wtf you smoking.. it's a full 50w lower. see anandtech power numbers

He probably meant the 1070.

Wow all the hype and it didn't deliver on any of it.

Exactly what claim did AMD make for the 480 that they didn't deliver on? As far as I can tell, the card is everything AMD said it would be.

About the only thing that worries me was mentioned in the article already - the 150w TDP, while a leap over the previous generation, is not only to be expected by the reduction in manufacturing process, but is matched by the much better-performing 1070.

 

[-Fran-]

Could somebody explain frame time and frame time variance to me? I do not understand those graphs.

It's just a fancy way of describing "stutter". When you see a frame taking 2x or 3x the time of the previous one (say, 10ms, 10ms, 40ms 40ms, 10ms, 10ms) they will represent what you describe in simple terms as "stuttering" on the motion perceived. Or in other words, the perceived motion of the sequence got a delay big enough that you noticed it.

When you aggregate them you use "variance" to determine how spread and different the overall sample is to the average. You have different ways to represent the data, so it is important to know that context as well. 99% or statistical distributions use Gaussian for it's simplicity, but you have fancier distributions of data.

Cheers!

 

[robinspi]

As always, we just need to wait for the 3rd party coolers, and the AIB with the 8/6+6 power connectors. There is where you will find the 1480-1600 MHz OC results.

But bear in mind, this is still GCN. You will not be able to get the same NVIDIA like OC results or performance/MHz

 

[kiniku]

But two in crossfire could beat a GTX1070! LOL

 

[kiniku]

I Polaris thought this was supposed to be a power efficiency improvement cycle? This thing sucks almost as much power as a GTX1080FE.

 

[quilciri]

terribe, i repeat rerrible perf / watt ratio.

jeez 480 is almost at the same numbers as 1080.

Nobody I know anywhere really cares are performance-to-watt ratios......only price-to-performance.

In absolute terms, 150w for the performance of the 480 is pretty good, and if you want to buy a 480, wattage isn't likely a big concern. However, I care about the wattage and you probably should too. Given that AMD is actually on a *smaller* node process than Nvidia, it's worrisome that NVidia was able to squeeze much more performance out of 150w in the 1070 than AMD has in the 480. It's doubly worrisome in that it indicates that the Fury successor cards may have some significant power draw and heat issues.

Personally, I'm worried because I bought a 1440 freesync 144hz for super cheap off slickdeals; I've been sitting on a GTX 970 waiting for the high end polaris cards and I may end up upgrading to a 1070 and foregoing freesync. Seriously..I want to buy an AMD card and I'm not sure I will.

 

[17seconds]

In TechPowerUp's 480 Crossfire review, they found that only 6 of the 16 games tested took advantage of the second card. (In the GTX 1080 SLI review, they found that 11 of the 16 games supported SLI.)
http://www.techpowerup.com/reviews/AMD/RX_480_CrossFire/21.html

 

[ern88]

The card is drawing more power from the PCI-E slot the the slot is designed for. Plus the RX 480 is using more power then the TDP was listed. If you have a cheap oem motherboard. I bet that the card will burn out the PCI-E slot on it. Old classic AMD making promises that they can't keep when it comes to power usage.

 

[quilciri]

In TechPowerUp's 480 Crossfire review, they found that only 6 of the 16 games tested took advantage of the second card. (In the GTX 1080 SLI review, they found that 11 of the 16 games supported SLI.)
http://www.techpowerup.com/reviews/AMD/RX_480_CrossFire/21.html

Hopefully this will be less and less of an issue as DX12 makes inroads. I don't think that moment will happen until the next generation of video cards, though (i.e., when windows 10 has the largest % of desktop usage)

 

[bignastyid]

So another AMD release where you should avoid the reference card. Maybe they will fix it next year when they refresh the series.

 

[cangelini]

Exclusively DX11 review of a DX12 card? Why would anyone write that?

To be fair, DX12 games currently account for less than 1% of games currently available. The rest are DX11 or lower, or OpenGL/Vulkan.

Moreover, we do have numbers for DX12, which are lower than DX11. The issue is that Fraps is not DX12-capable. So, while I can run the in-game test in DX11 vs DX12 to see performance drop, I cannot generate more granular charts in DX12 using anything but Ashes at this time. There are other tools we can use for this (FCAT is one), but the workflow is not in place yet.

 

[king3pj]

Wow all the hype and it didn't deliver on any of it.

Exactly what claim did AMD make for the 480 that they didn't deliver on? As far as I can tell, the card is everything AMD said it would be.

Here is the claim that many people are hung up on.
http://www.pcworld.com/article/3077432/components-graphics/polaris-confirmed-amds-200-radeon-card-will-bring-high-end-graphics-to-the-masses.html

In a livestream from Computex in Taipei, AMD announced that the Radeon RX 480 will be the first graphics card based on its forthcoming Polaris graphics processors. And get this: The Radeon RX 480 stands ready to deliver performance equivalent to what today’s $500 graphics cards offer, as first reported in the Wall Street Journal earlier today. That’s roughly in line with the Radeon R9 390X, GeForce GTX 980, or air-cooled Radeon Fury.

During the announcement they said that the RX 480 would have performance on par with $500 graphics cards. People came out of that announcement event expecting RX 480 to match a GTX 980 and R9 Fury for $200. Because of that overly optimistic claim you have lots of people talking about how they are disappointed instead of raving at what a good price to performance card this is.

I think the RX 480 will sell much better than the 380 and 280 did. I just think they made a mistake by setting the expectations higher than what can reasonably be expected from a $200-$240 card.

Edit: I'm not sure why the part of the article I quoted double posted itself. I have looked at my post in the editor and can't seem to see what I did wrong. I guess I don't know how to use the quote feature properly.

 

[17seconds]

Exclusively DX11 review of a DX12 card? Why would anyone write that?

To be fair, DX12 games currently account for less than 1% of games currently available. The rest are DX11 or lower, or OpenGL/Vulkan.

Moreover, we do have numbers for DX12, which are lower than DX11. The issue is that Fraps is not DX12-capable. So, while I can run the in-game test in DX11 vs DX12 to see performance drop, I cannot generate more granular charts in DX12 using anything but Ashes at this time. There are other tools we can use for this (FCAT is one), but the workflow is not in place yet.

Furthermore the DirectX 12 games we do have are either poorly coded, or straight out of the AMD Gaming Evolved program, or both.

 

[robinspi]

I Polaris thought this was supposed to be a power efficiency improvement cycle? This thing sucks almost as much power as a GTX1080FE.

As I think I said, this is still GCN. AMD's GCN will always be less power efficient than NVIDIA's Maxwell/Pascal. Look at the last gen. This is a massive jump, and more of a gap-closer in terms of performance/watt, and price/performance.

 

[robinspi]

Furthermore the DirectX 12 games we do have are either poorly coded, or straight out of the AMD Gaming Evolved program, or both.

Yep. See RotTR for #1 and Hitman for #2.

 

[bramahon]

About the only thing that worries me was mentioned in the article already - the 150w TDP, while a leap over the previous generation, is not only to be expected by the reduction in manufacturing process, but is matched by the much better-performing 1070.

This is what concerns me most. It feels like AMD is still trailing the green camp by roughly one generation in overall efficiency. Is it truly a leap over the previous gen? Let's see, first of all none of the cards compared to RX480 were exactly entry/mid level back when they're introduced which makes it difficult to probe the real efficiency gain made by Polaris. I mean I could compare my HD6870 to the HD7870/270x when it launched to figure out exactly how much improvement it made - new architecture and smaller node combined. But this time it's not that simple. Correct me if I'm wrong, the most efficient (performance/watt) SKU from the last gen would be Nano or even Fury(non-X) to some extent; so price or value aside, it would be only fair to pit RX480 against those. OTOH if we compare perf/watt gain of GTX1070 over GTX970, it seems to me that the parity is far from restored!

 

[cangelini]

The last time Chris used stock-clocked cards in a comparison, he got grilled by people saying that nobody runs their high-end GPUs at plain stock clocks. Use factory OC, then you have people complaining that it isn't a fair stock-to-stock comparison or that he should be comparing between factory overclocks or between manual overclocks.

There is just no pleasing everyone at the same time. At least not without including a hundred different GPUs and GPU variants. No matter which way you slice it, it becomes a nightmare.

I think its also a bit unfair to have the field stacked with overclocked cards, sure factory overclocks from board partners, but overclocks none the less.

In this case, I would have run all cards at stock clock rates, if some of the partner boards weren't throwing odd results or outright crashing when I tried to force reference-class frequencies. With my hand forced, the best I could do was full transparency and a partial accounting of the higher clock rates in the benchmark analysis.

 

[Kahless01]

maybe im retarded but wouldn't it have made better sense to put the 380 or 380x in the charts and not the 290 that hasn't been available for awhile and doesn't matter much anymore?

 

[Onus]

I posted this in another thread, and thought it bears putting here too:

I agree that for a card that touted low power usage, the RX 480 is a FAIL. Price / performance is excellent, but that wasn't the big selling point. Not to sound like a nVidiot, but I'm inclined to believe the GTX1060 will render this card irrelevant.

 

[Ags1]

This review is missing VR benches. It seems the RX480 really performs well on specifically VR tasks, as shown on this Dutch site: https://tweakers.net/reviews/4709/10/amd-radeon-rx-480-crossfire-vr-dx12-en-overklok-getest-vr-games-op-de-oculus-rift.html.

 

[LORD_ORION]

Why did you not bench a 380 4GB :\

 

[tomc100]

$200 is a very competitive price to beat. I'm looking forward to see AMD's flagship gpu though with hbm2 technology hopefully soon.

 

[quilciri]

Wow all the hype and it didn't deliver on any of it.

Exactly what claim did AMD make for the 480 that they didn't deliver on? As far as I can tell, the card is everything AMD said it would be.

Here is the claim that many people are hung up on.
http://www.pcworld.com/article/3077432/components-graphics/polaris-confirmed-amds-200-radeon-card-will-bring-high-end-graphics-to-the-masses.html

I don't see why people would be hung up on that, it's somewhat accurate. Both the 980 and the 390X debuted at $500. In this review, the 480 beat or matched (within 1 fps) at least one of those cards in half the tests (GTA, Hitman, CARS, Tomb Raider, Division), and came close in at least one other (Witcher - 2 fps off 390X). I wouldn't say it performs quite as well overall as the 980 or 390X, but it's close enough to make complaints over it's relative performance seem nitpicky.