AMD Radeon RX 480 8 GB

[TRWOV]

Anyone knows how apart the mounting holes are? I want to know if any of my heatsinks will work on this.

The reference cooler simply won't cut it. I can buy an AIB card but I have a bunch of GPU coolers lying around gathering dust.

 

[Eroticus]

Anyone knows how apart the mounting holes are? I want to know if any of my heatsinks will work on this.

The reference cooler simply won't cut it. I can buy an AIB card but I have a bunch of GPU coolers lying around gathering dust.

http://videocardz.com/61697/how-fast-radeon-rx-480-can-be

4Chan >.< but RX480+H100

 

[TRWOV]

Thank you that'll help.

I really want to get a reference card from the first batchs to get Samsung/Hynx RAM chips. We know that manufacturers switch to Elpida RAM down the line.

 

[sergionography]

I'm just saying that the RX 480 dissapointed and that the 970 is a better buy if found cheaper. I'm not the one filling every reply with irrelevant bullshit benches

Old benches are not irrelevant at all, they are shown here to show the pattern of benches between day1 reviews and later reviews with mature drivers and newer relevant games. almost every launch review has the nvidia part beating its amd counterpart by up to like 10%, and then you skip about a year or more and check recent reviews and find out that the AMD part is beating its direct nvidia counterpart by a good 10%, if this doesn't ring a bell for you idk what will. And if you dont expect to see the same pattern for the rx480 down the road then you are in denial. Even with my previous point clarified, NO the gtx 970 is NOT a better card, as after 2 years of optimization it is losing to an rx480 on day one drivers by up to 10%. I mean compared to pascal yes amd is more power hungry, but to tell me get a maxwell card instead then thats just ridiculous fanboyism there lol.

So to summarise, AMD is a better futureproof investment especially for people who only buy a gpu once every 3-4 years like myself, and with dx12 becoming mainstream down the road will only make gcn shine even more, while nvidia by then will have a new architecture to better tackle dx12, those who invested in pascal and maxwell must realize that their architecture doesnt magically change or fix itself when nvidia releases a successor halo product, the only thing it fixes is their ego on forums about how nvidia is number one blah blah, which doesnt do sane people any good

 

[Casecutter]

What's not to understand?

That there are two different test both indicating "typical gaming power" and we really don't know for certain how either of those games are representative to either brand.

Just saying Metro: Last Light @1080p could load one brand more than the other skewing all the summary results. Now that we're in a time when dynamic clocks (boost) such operation from a particular brand may cause a card to work less in that particular a game, while the other brand might need to work harder. Using one game(one point) to average across all games will skew the results.

 

[rruff]

It's actually a very good indicator both in averages and peaks. That's why I picked it. No, it's not Furmark, that's why we have the Furmark test.

Any thoughts on whether the high power consumption is due to the card being volted and clocked way past its sweet spot? Thinking AMD went to extreme measures to hit a performance target with this card, and the 470 will be much more power efficient.

 

[Dethroy]

That there are two different test both indicating "typical gaming power" [...]

Same test, 2 different measured values: peak & average

 

[truth teller]

http://videocardz.com/61697/how-fast-radeon-rx-480-can-be

interesting

 

[Casecutter]

Same test, 2 different measured values: peak & average

so is the summary based on the 970-156W / 480-163W or the Peak of the 480-167W / 970-191W

What if the title was Hitman and the 970-166W / 480-154W, while the Peak of the 480-153W / 970-198W? Wouldn't that change the summary averages?

 

[W1zzard]

so is the summary based on

average of course

Wouldn't that change the summary averages?

yes, and that's something we all have to live with, it's just too much work to bench all those games for power consumption a separate time

 

[efikkan]

When the news of vega came, it clearly said is was going to have better performance x watt than polaris.
And also, there are two vegas, 10 & 11. V10 is supposed to be in the 1080 territory and V11 in 1080Ti or whatever is it called.

- AMD claims Vega will be significantly more power efficient than Polaris, we'll see if this is true. They claimed Polaris would improve efficiency by 2.5×, but it's more like half of that. (even less if you compare to the R9 Nano). Pascal is about twice as efficient as Polaris, so AMD needs to get their act together and needs to close that gap to compete.
- Vega 10 will be the big chip, Vega 11 will be smaller. Just like with Polaris 10/11.
- Vega is scheduled for end of Q1/ early Q2 2017. The development of the chip is soon finished, tape out is next (proably July), then ~3 months for the first test batch...

Is there any game I will show you that you will not call AMD biased?

There are no good Direct3D 12 games yet. Games like AofS was written for Mantle and then ported, of course it's going to be biased. There is still no evidence that Nvidia's hardware is inferior in Direct3D 12. But a lot of people are confused since Nvidia get "lower gains from Direct3D 12", but that's because they've brought the Direct3D 12 driver optimizations to all APIs(1)(2).

And can you guarantee that the extra ram will not be useful in the near future? Not all people buy cards for 6 months. Some people buy cards and keep them for years.

A card in the class of RX 480 is not fast enough to utilize more than ~4 GB. If you are trying to use more in a single frame the performance will tank because of bandwidth limitations.

 

[Eroticus]

- AMD claims Vega will be significantly more power efficient than Polaris, we'll see if this is true. They claimed Polaris would improve efficiency by 2.5×, but it's more like half of that. (even less if you compare to the R9 Nano). Pascal is about twice as efficient as Polaris, so AMD needs to get their act together and needs to close that gap to compete.
- Vega 10 will be the big chip, Vega 11 will be smaller. Just like with Polaris 10/11.
- Vega is scheduled for end of Q1/ early Q2 2017. The development of the chip is soon finished, tape out is next (proably July), then ~3 months for the first test batch...


There are no good Direct3D 12 games yet. Games like AofS was written for Mantle and then ported, of course it's going to be biased. There is still no evidence that Nvidia's hardware is inferior in Direct3D 12. But a lot of people are confused since Nvidia get "lower gains from Direct3D 12", but that's because they've brought the Direct3D 12 driver optimizations to all APIs(1)(2).


A card in the class of RX 480 is not fast enough to utilize more than ~4 GB. If you are trying to use more in a single frame the performance will tank because of bandwidth limitations.

4GB vs 8GB - 90% + Developers release their products downgraded.

btw my 4GB 290x and Fury X has same problem as i know.


GTX 970 has 224GB/s
1070 has 256GB/s
RX480 has 256GB/s memory bandwidth, so it's pretty OP.



DX11 vs 12

My 290x gains 25 fps over DX11. [1080p/Ultra]

DX12 +CPU Usage is Higher 6C/6T
http://i.imgur.com/YF5fBgG.jpg

DX11 CPU Usage Lower 6C/6T + 25 fps less.
http://i.imgur.com/mgmozPw.jpg

 

[john_]

Because within those 15 days Nvidia urged card manufacturers to cut prices on 970, 980 and 980Ti. My feeling is Nvidia sees it as competition to overpriced 1070 and 1080, and want them and the 980's gone from all stock.

Plus, 970 prices have fallen bit by bit since their production halted 2 months ago.

I think you got it wrong here. You don't make the competition better(Maxwell cards) if you feel they are a threat to your new products(Pascal). Nvidia saw that AMD managed to have big quantities of RX480 ready for the market. That was the only thing they didn't knew 15-30 days ago. They knew how RX480 was performing, but if AMD didn't had enough cards for the market, prices would have gone up quickly and RX480 wouldn't be a threat. It would have been a lost chance. But AMD flooded the market, so Nvidia needed to counter that fast. So they lowered the prices on Maxwell cards, something that gives people the argument that GTX 970 is not much more expensive than RX 480 today, and also moved GTX 1060 closer. Instead of August, July.

A card in the class of RX 480 is not fast enough to utilize more than ~4 GB. If you are trying to use more in a single frame the performance will tank because of bandwidth limitations.

And you are an expert in GPUs? Do you design GPUs for a company? Can you tell me what is the bandwidth of GTX 1070? The same maybe? Will GTX 1070s performance TANK because of bandwidth limitations? Or are you going to tell me that GP104 is a marvelous chip that also bypass bandwidth limitations?

 

[Legacy-ZA]

I think you got it wrong here. You don't make the competition better(Maxwell cards) if you feel they are a threat to your new products(Pascal). Nvidia saw that AMD managed to have big quantities of RX480 ready for the market. That was the only thing they didn't knew 15-30 days ago. They knew how RX480 was performing, but if AMD didn't had enough cards for the market, prices would have gone up quickly and RX480 wouldn't be a threat. It would have been a lost chance. But AMD flooded the market, so Nvidia needed to counter that fast. So they lowered the prices on Maxwell cards, something that gives people the argument that GTX 970 is not much more expensive than RX 480 today, and also moved GTX 1060 closer. Instead of August, July.


And you are an expert in GPUs? Do you design GPUs for a company? Can you tell me what is the bandwidth of GTX 1070? The same maybe? Will GTX 1070s performance TANK because of bandwidth limitations? Or are you going to tell me that GP104 is a marvelous chip that also bypass bandwidth limitations?

Both the new GTX1000 series and RX480 from nVidia and AMD come with new texture compression technology, that allows for the cards to do more while still running slower memory bandwidth than we are used to.

 

[rtwjunkie]

I think you got it wrong here. You don't make the competition better(Maxwell cards) if you feel they are a threat to your new products(Pascal). Nvidia saw that AMD managed to have big quantities of RX480 ready for the market. That was the only thing they didn't knew 15-30 days ago. They knew how RX480 was performing, but if AMD didn't had enough cards for the market, prices would have gone up quickly and RX480 wouldn't be a threat. It would have been a lost chance. But AMD flooded the market, so Nvidia needed to counter that fast. So they lowered the prices on Maxwell cards, something that gives people the argument that GTX 970 is not much more expensive than RX 480 today, and also moved GTX 1060 closer. Instead of August, July.

No sir, I think you got it wrong. If they still hat shitloads of the older Maxwells in stocks, and which are all cheaper than their monstrously priced new cards, and which perform at a level that is all people need unless they want 4K, and if the new cards are all obscenely priced, it stands to reason Nvidia and their distributors must find a way to unload the old stock quickly. That way, as they see it, people will be more willing to buy their overpriced, overhyped cards.

As long as they have a cheaper, perfectly adequate option in the pipeline, they don't have much hope of actually selling the new stuff. Nvidia stopped worrying about what AMD "might" do long ago. and the new Pascals are not in the 480 performance or price tier at all, so not a consideration. The only reason they started the sales was to liquidate stock of their 1070/1080 competitors.

 

[john_]

Both the new GTX1000 series and RX480 from nVidia and AMD come with new texture compression technology, that allows for the cards to do more while still running slower memory bandwidth than we are used to.

Yes I know that. Nvidia offers color compression from Maxwell if I am not mistaken and AMD introduced it with Tonga.

No sir, I think you got it wrong. If they still hat shitloads of the older Maxwells in stocks, and which are all cheaper than their monstrously priced new cards, and which perform at a level that is all people need unless they want 4K, and if the new cards are all obscenely priced, it stands to reason Nvidia and their distributors must find a way to unload the old stock quickly. That way, as they see it, people will be more willing to buy their overpriced, overhyped cards.

As long as they have a cheaper, perfectly adequate option in the pipeline, they don't have much hope of actually selling the new stuff. Nvidia stopped worrying about what AMD "might" do long ago. and the new Pascals are not in the 480 performance or price tier at all, so not a consideration. The only reason they started the sales was to liquidate stock of their 1070/1080 competitors.

No you did.

Making Maxwell cheaper you only drive customers to Maxwell. Every customer gone to Maxwell, will not buy a Pascal card, but more important, will still be an Nvidia customer. And why would someone who just bought a Maxwell card "be more willing to buy their overpriced, overhyped cards"? It doesn't make sense. He just got a new shiny Maxwell card.

If you have "shitloads of the older Maxwells in stocks" with prices that will interest no one, you will probably still have "shitloads of the older Maxwells in stocks" 12 months latter. That doesn't really affect how Pascal cards sell. Having "shitloads of the older Maxwells in stocks" at high prices only affect AIBs and shops having difficulties selling those cards.

Giving those "shitloads of the older Maxwells in stocks" at lower prices, you drive people to those cards which means they are not going to buy Pascal cards because they just got a new card. Especially when those Pascal cards have high prices.

I don't know in how many ways I have to explain it. I thought it was obvious. No. IT IS obvious.

What changed was AMD's RX480. Every customer going to RX480 is a lost customer to Nvidia, more market share for AMD. With no GTX 1060 in the market, Nvidia had to do something fast to not lose customers to AMD, to not have "shitloads of the older Maxwells in stocks" 12 months latter. And the only thing they could do was to lower Maxwell prices.

If the rumors about Polaris having problems where true, if Polaris was pushed back in October, we wouldn't have seen any significant price drops in Maxwel cards, except 980Ti that was way too expensive compared to GTX 1070. Prices would have stayed relatively high until GTX 1060 was released and Maxwell cards would still be selling nicely. With RX480 in the market, who would have gone and bought a GTX 970 if it was $50-$100 more expensive than an RX480? Who would have bought an even more expensive GTX 980? Too few. Not to mention that RX480 would look so much better today with NO competition at it's price range.

 

[johnnyfiive]

That power consumption though... jeez.

 

[HumanSmoke]

The claim was "2.5x improved performance-per-watt compared with its 28nm-class hardware".

The quantifier here has always been what constitutes its 28nm-class hardware and what Perf/W matrix? Is "class" based on die size or level of performance today? In other words is it compared to a Pitcairn like the 270X, or say the 290X/390.

Been in a coma?
AMD's own slide from where the claim was made clearly shows the 28nm GPU baseline as late 2014. The ONLY AMD card launched in that timeframe is the perf/watt dog...R9 285. It's called "setting a low bar"

 

[Eroticus]

Been in a coma?
AMD's own slide from where the claim was made clearly shows the 28nm GPU baseline as late 2014. The ONLY AMD card launched in that timeframe is the perf/watt dog...R9 285. It's called "setting a low bar"

700 series are 28nm to.

280x =51% = 2.0p/m

51% = Overall = Could reach 2.5 in some games or apps.

 

[HD64G]

- AMD claims Vega will be significantly more power efficient than Polaris, we'll see if this is true. They claimed Polaris would improve efficiency by 2.5×, but it's more like half of that. (even less if you compare to the R9 Nano). Pascal is about twice as efficient as Polaris, so AMD needs to get their act together and needs to close that gap to compete.
- Vega 10 will be the big chip, Vega 11 will be smaller. Just like with Polaris 10/11.
- Vega is scheduled for end of Q1/ early Q2 2017. The development of the chip is soon finished, tape out is next (proably July), then ~3 months for the first test batch...

- Polaris has the #11 also which could be the best one in efficiency, so wait 1-2 weeks more to be sure judging this aspect for Polaris.
- Vega 10 is the smaller of the Vegas, as Vega 11 is the beasty one.
- Vega 10 could be out for sale in the late 2016 with the development already done (that's what Raja told about the celebration last week after all)

 

[Eroticus]

- Polaris has the #11 also which could be the best one in efficiency, so wait 1-2 weeks more to be sure judging this aspect for Polaris.
- Vega 10 is the smaller of the Vegas, as Vega 11 is the beasty one.
- Vega 10 could be out for sale in the late 2016 with the development already done (that's what Raja told about the celebration last week after all)

I'm not strong in Astronomy.

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

 

[HumanSmoke]

700 series are 28nm to.

ffs why don't you give the trolling and bad logic a rest.
The slide is an AMD slide of AMD products.
The slide was part of an AMD presentation comparing AMD architectures

 

[Eroticus]

ffs why don't you give the trolling and bad logic a rest.
The slide is an AMD slide of AMD products.
The slide was part of an AMD presentation comparing AMD architectures

Simple English ..

28nm GPUs 1x p/w
Polaris GPUs 2.5x p/w


---------------

Awww sorry ..

I mean "7000" - Radeon HD 7000 Series.

i need to learn to write better

Any way you are nvidia fan so 280x = Re branded 7970 Ghz.

RX 480 has 2.0p/w over 280x overall and 7970 GHz Editions were even more hungry.

so this could reach 2.5p/w over 28nm.

 

[HumanSmoke]

I mean "7000" - Radeon HD 7000 Series.
i need to learn to write better
Any way you are nvidia fan so 280x = Re branded 7970 Ghz.

Remember what I said about bad logic?
The 280X launched in October 2013....now check the timeline on the AMD slide that positions the 28nm GPU being compared. It shows late 2014
Even if you had used the last HD 7000 rebrand, the R9 280, it launched at the beginning of 2014 and still doesn't correlate to either the slide timeline, nor the architectural presentation that accompanied it. I'm surprised you needed an "nvidia fan" to supply the actual facts on this matter.

 

[Eroticus]

Remember what I said about bad logic?
The 280X launched in October 2013....now check the timeline on the AMD slide that positions the 28nm GPU being compared. It shows late 2014
Even if you had used the last HD 7000 rebrand, the R9 280, it launched at the beginning of 2014 and still doesn't correlate to either the slide timeline, nor the architectural presentation that accompanied it. I'm surprised you needed an "nvidia fan" to supply the actual facts on this matter.

Ohhh so you want to talk about last 2years.

Did AMD released all Polaris cards ? did you know there APU wiith Polaris to notebooks ? that is using only 35w ?

Why are you comparing whole graph to the first and most powerful Polaris card ?

Great logic dude!!!!!