Official AMD Polaris Review Thread: Radeon RX 480, RX 470, and RX 460

[obidamnkenobi]

just watch the part of paul's video I pointed out it only gets loud at 75% or higher fan speed which it never gets on stock/reference. it will answer all your questions.

fyi, I would not trust techpowerup for anything for your own good. I am 100% sure 390 is 30% + better than a 380x, there is no doubt on my end. by that logic, 480 will be 30% or better than 380x. if you want to spend only 150$ go for the 470 when it comes out

OK, I did watch the video. But frankly the "noise level" part was pointless. It will depend on the sensitivity of his mic, the input settings, the ambient levels in his space and the level of my headphones, compared to the noise level inside my case from where I'm sitting; there's so many layers of abstraction it's not worth anything.

And apart from said noise section, everything in his video could have been done better and more organized as a write-up (lol at studying graphs in a video), so unless you're dyslectic I don't quite see the point of video reviews either.

If I was looking at the ~$250 range the 480 does look good, except the noise/TDP issues. In the $150-200 range I prefer, I hope the 470 will be as good, with a more quite custom cooler

 

[obidamnkenobi]

Just going by the performance summary of computerbase.de RX480 is 46% faster than 380X. In conclusion 150$ for 380X performance is the worse deal.

Why? The 480 cost 60% more. So how is paying 60% more for 46% more performance a good deal?

 

[guachi]

I was too harsh on the RX 480. Today I've been playing DX12 Total War Warhammer and I am really liking locked 60 FPS even in huge battles, something that wasn't possible with DX11. It reminded me how important strong DX12/Vulkan performance matters already and it will only be more important later this year.

Someone... admitted they were wrong.... on the internet?

I'm not much of a gamer anymore, but I do love me some Total War games ever since the first (though I've skipped some titles)

Thanks for the first hand user account.

 

[Thala]

Why? The 480 cost 60% more. So how is paying 60% more for 46% more performance a good deal?

Not sure what you calculated, but 200/150 = 1.33 -> 33% higher price for 46% higher performance

 

[sirmo]

The often understated thing about DX12 / Vulkan and async compute is that the games developed using this paradigm remove the common CPU bottlenecks we've been living with now for years.

The marginal IPC improvements in the CPU tech can't keep up with the growing need for Tflops on the GPU side.

DX12/Vulkan and async compute is absolutely the way forward.

 

[William Gaatjes]

The often understated thing about DX12 / Vulkan and async compute is that the games developed using this paradigm remove the common CPU bottlenecks we've been living with now for years.

The marginal IPC improvements in the CPU tech can't keep up with the growing need for Tflops on the GPU side.

DX12/Vulkan and async compute is absolutely the way forward.

I think we have a real winner with the RX 480 for the 1080p market and it will show in the next 6 months.

I also have some thoughts about those leakage fuses and Boot Time Calibration. I keep them for myself for now before i sound very negative. I think it is best i try to do some reading up on what the real deal is with the Samsung 14nm process GF made their own. That is, if i can find any.
Looking at it as an outsider(with hardly any knowledge about it), i see a method to bin as much dies as possible which is necessary because there are going to be a lot of polaris customers (consolemarket, linux, Apple and windows pc-hardware. And a short time to market.

 

[boozzer]

OK, I did watch the video. But frankly the "noise level" part was pointless. It will depend on the sensitivity of his mic, the input settings, the ambient levels in his space and the level of my headphones, compared to the noise level inside my case from where I'm sitting; there's so many layers of abstraction it's not worth anything.

And apart from said noise section, everything in his video could have been done better and more organized as a write-up (lol at studying graphs in a video), so unless you're dyslectic I don't quite see the point of video reviews either.

If I was looking at the ~$250 range the 480 does look good, except the noise/TDP issues. In the $150-200 range I prefer, I hope the 470 will be as good, with a more quite custom cooler

I don't understand you now.

I never stated whether video or written reviews are better, I was just telling you that paul's video has the noise part. as that was what you were concerned about. what are you getting upset over? like what?

but 480 has no noise issues, and why is the TDP an issue unless you have a 400 watt oem level psu? that is just fud.

 

[obidamnkenobi]

Not sure what you calculated, but 200/150 = 1.33 -> 33% higher price for 46% higher performance

I used $240 for the 480. As that'll be the price of custom parts with better cooler.

 

[coercitiv]

I used $240 for the 480. As that'll be the price of custom parts with better cooler.

and better performance

 

[obidamnkenobi]

I don't understand you now.

I never stated whether video or written reviews are better, I was just telling you that paul's video has the noise part. as that was what you were concerned about. what are you getting upset over? like what?

but 480 has no noise issues, and why is the TDP an issue unless you have a 400 watt oem level psu? that is just fud.

No, the video review tangent was my bad. I just find this trend strange.

But I do think someone recording with his mic with unknown settings is rather pointless to evaluate noise levels. Oh it's too low for his mic to pick up, woho, so what. Give me dBA measurements

Can we ban the word fud? It's only used about information people don't like.. FWIW I own an AMD card now, I have nothings against them. I want it to be good. I just want a quite card

Several others say (and measure) the 480 fan is loud. This is not "fud"

http://www.legitreviews.com/amd-radeon-rx-480-video-card-review-polaris-arrives_183428/11

The AMD Radeon RX 480 also is the loudest graphics card that we have tested recently as it topped out chart at 51.5dB.

http://www.guru3d.com/articles_pages/amd_radeon_r9_rx_480_8gb_review,8.html
42 dBA, only beaten by SLI cards.

and Techpowerup at 41 dBA.

Maybe not terrible, but to me 41 dBA is pretty loud. When I can get way quieter cards with 0 dB fans at idle that's what I want.

 

[jpiniero]

I used $240 for the 480. As that'll be the price of custom parts with better cooler.

Probably more like $275, esp for the ones that can clock to 1.4+. Which is completely fair regardless of what nVidia does.

 

[Bacon1]

Probably more like $275, esp for the ones that can clock to 1.4+. Which is completely fair regardless of what nVidia does.

Why would you be comparing it against the 8gb model not the 4gb one?

4GB custom models won't go for more than 8GB ref.

 

[Hitman928]

Can you get a 380x for $150?

 

[Thala]

I used $240 for the 480. As that'll be the price of custom parts with better cooler.

But the 46% performance gain we are talking about is with stock cooler, stock setting...so apparently we are talking about the 199$ model.

 

[boozzer]

No, the video review tangent was my bad. I just find this trend strange.

But I do think someone recording with his mic with unknown settings is rather pointless to evaluate noise levels. Oh it's too low for his mic to pick up, woho, so what. Give me dBA measurements

Can we ban the word fud? It's only used about information people don't like.. WWIW I own an AMD card now, I have nothings against them. I want it to be good. I just want a quite card

Several others say (and measure) the 480 fan is loud. This is not "fud"

http://www.legitreviews.com/amd-radeon-rx-480-video-card-review-polaris-arrives_183428/11


http://www.guru3d.com/articles_pages/amd_radeon_r9_rx_480_8gb_review,8.html
42 dBA, only beaten by SLI cards.

and Techpowerup at 41 dBA.

Maybe not terrible, but to me 41 dBA is pretty loud. When I can get way quieter cards with 0 dB fans at idle that's what I want.

I call posts for what they are. if they consistently contain false info, I call it fud.

http://www.gamersnexus.net/guides/2497-amd-rx-480-fan-noise-comparison super in depth. 480 is just normal, not good not bad. calling it noisy is fud. if 40 dba is loud for you, that just means you need to get a AIB card that will get you 38 dba or less.

 

[obidamnkenobi]

Can you get a 380x for $150?

Not that I've found. $150 for 380. $200 for 380X. With the 960 between. I'm curious what happens when the 470 comes out, if it's the rumored $150. If the 3xx drops significantly I think they might be good value.

 

[obidamnkenobi]

I call posts for what they are. if they consistently contain false info, I call it fud.

http://www.gamersnexus.net/guides/2497-amd-rx-480-fan-noise-comparison super in depth. 480 is just normal, not good not bad. calling it noisy is fud. if 40 dba is loud for you, that just means you need to get a AIB card that will get you 38 dba or less.

Anandtech measured 50 dBA, others measured 40 dBA. I'm just repeating what they said. How is that "false information" in my posts? Unless you say these tests are wrong?

40 dBA is loud to me. Especially when there are cards with 75%+ of the performance with significantly lower noise (and lower $$). Even in the link you posted it's at 39-43 dBA and the only cards louder than the 480 is the 980Ti, 1080 and 290x! I'd prefer to see noise comparison against cards with ~ same performance.

 

[boozzer]

Anandtech measured 50 dBA, others measured 40 dBA. I'm just repeating what they said. How is that "false information" in my posts? Unless you say these tests are wrong?

40 dBA is loud to me. Especially when there are cards with 75%+ of the performance with significantly lower noise (and lower $$). Even in the link you posted it's at 39-43 dBA and the only cards louder than the 480 is the 980Ti, 1080 and 290x! I'd prefer to see noise comparison against cards with ~ same performance.

the link I posted has a 13 min video that goes along with the article and charts.

I trust the one that allows me to see the entire testing process, provides more details than one that does not.

and if 40 dba is loud, then all blower cards are not what you want. you need to buy AIB cards. but they are only 2-3 dba less. that makes a difference?

you realize all the ones that are quieter are all AIB cards right? in the link I provided? when they are all on stock/auto fan speed, 480 beats every other blower and 1 AIB card.

 

[Erenhardt]

seems like people are desperate for another hawaii cooler.

 

[Bacon1]

JayTwoCents did CFX review: https://youtu.be/cVVJPbFRDEc?t=223

Benchmarks start @ 3:30ish

vs 1080:

Ties: FC4, Metro Last Light

Slightly slower: Crysis 3, GTA V, Witcher 3

Price: $699 vs $478

Scales fine all the way to 4k

Also faster, but more expensive, than 1070, obviously more power too.

Tested DX12 ROTTR, so obviously no scaling (why do sites still do DX12 in ROTTR???)

 

[Silverforce11]

I call posts for what they are. if they consistently contain false info, I call it fud.

http://www.gamersnexus.net/guides/2497-amd-rx-480-fan-noise-comparison super in depth. 480 is just normal, not good not bad. calling it noisy is fud. if 40 dba is loud for you, that just means you need to get a AIB card that will get you 38 dba or less.

Yeah, any review that says the card is noisy is utter bull. It's the same as the GTX 980, actually identical in many reviews on temps & noise. Almost the same as the 1080. Yet did reviews say those cards were noisy?

Also this: https://www.reddit.com/r/Amd/comments/4qtuu5/rx_480_undervolt_oc_1340mhz/

Under-volted and OC. Instead of a 150W card that struggles against a 390X, the above under-volted and overclocked RX 480 is now a 125W card that beats the 390X.

This goes back to my point earlier, AMD needs a better system or feature to regulate voltages & clocks automatically. They need a tech like NV's boost because manual setting each card to 1.15v (I've seen some with faulty bioses defaulting to 1.25v) is utter stupid when there's just so much potential left on the chip.

I'll be testing this when my cards arrive, too many orders the etailer got hammered and there's delays in processing... -_-

 

[topmounter]

Have any of the other reviews been able to replicate the results in the techaltar.com review?

 

[Silverforce11]

Have any of the other reviews been able to replicate the results in the techaltar.com review?

Nope, Techaltar is like an AMD fansite. The likes of ABT & PClab.PL as an NV fansite.

I don't put much weight in outliers.

 

[zinfamous]

Nope, Techaltar is like an AMD fansite. The likes of ABT & PClab.PL as an NV fansite.

I don't put much weight in outliers.

that sounds about right. Those results were ridiculous.

 

[garagisti]

JayTwoCents did CFX review: https://youtu.be/cVVJPbFRDEc?t=223

Benchmarks start @ 3:30ish

vs 1080:

Ties: FC4, Metro Last Light

Slightly slower: Crysis 3, GTA V, Witcher 3

Price: $699 vs $478

Scales fine all the way to 4k

Also faster, but more expensive, than 1070, obviously more power too.

Tested DX12 ROTTR, so obviously no scaling (why do sites still do DX12 in ROTTR???)

It's a Gameworks title, and that may have something to do with it. As i don't see many sites testing other DX12 titles, including the free Forza 6: Apex