[EDIT]hmmm.....whats going on at AMD? (and history of 3Dfx Rampage)

[SickBeast]

Yeah that's really bad. I have those drivers installed as well. I guess I'm stuck with them now.

[RolloTheGreat]

Yeah that's really bad.  I have those drivers installed as well.  I guess I'm stuck with them now.

Doubt it. One (or ten guys) with an issue does not mean it's a real issue.

[SteelCrysis]

RX 480 clock speeds leak: http://techreport.com/news/30303/rumor-a...peeds-leak

[SteelCrysis]

Zen is having problems with its integrated chipset: http://www.techpowerup.com/223583/amd-ze...late-costs

[ocre]

This is embarrassing: http://www.techpowerup.com/forums/thread...le.223568/

It took no time for red team troopers to come in with full force. Attack and discredit the op.

They literally have twisted it in an attempt to confuse, they are claiming his SSD is aged and wore out. They have his smart report, which clearly "ok" across the board, yet they are claiming it is failing. It's crazy, the normalized values are showing everything in great shape, it counts down. 100. Means 100% good. 87 me
would be 87% percent used. His values are fine.

It is amassing. Seems that the attribute types that smart uses as categories for each test is an angle that is being used to confuse, I guess in hopes the OP won't notice or that he may not know.
PreFail is not the result of any test, the results look good, 100 percent the best then they go down as the drive is used up. PreFail tests are showing good results.

[RolloTheGreat]

Sadly unsurprising.  There is no low that the MAD defense farce won't stoop to when they perceive their beloved MAD to be under assault from critics.

This is the difference in them and me GStan. (the main one anyway)

The stuff NVIDIA was giving me was always either the best stuff you could get, or you could reasonably make an argument that it was the best stuff you get. Crossfire was a mess back then, and ATi/AMDs flagship cards were not ever beating NV in significant ways. (and what they did better, I admitted)

These people are paid for disinformation, they just slander NV and pimp AMD no matter what the situation is.

Which brings up the second difference- I never had less than $100K household income while I was doing the Focus Group. I could always afford my own video cards. I was doing stuff like buying new cars, recreational property, and power boats. I didn't need NVIDIA to give me a card to get it.

I think a lot of these people are young, less affluent. They probably see what they're doing as a part time job to supplement their income. If AMD doesn't give them the card, they may not be getting the card.

No one seems to get this, but there's a big difference in perspective when you have the cash. I'm currently buying my new house before my old one is sold, and I could have done that the last time I saw NVIDIA too. (GTX680 launch) Do you honestly think I went scampering home from CA thinking, "Have to start pimping these video cards, the master might give me a free game!"?

Uh huh, sure. I earn my living in the software industry, doing tech support, training, QA, light design work, implementation. I don't need to lie about products to get them, unless the products are yachts.

[SteelCrysis]

More info on RX 480:
http://videocardz.com/61434/amd-radeon-r...k-analysis
http://videocardz.com/61396/new-amd-rx-4...cking-tool

[ocre]

So....390 to 480 a side grade

https://m.reddit.com/r/Amd/comments/4pzk...pact=false

Let's see where Polaris lands can't wait to see the massive spin efforts...this is gonna be entertaining

[happy medium]

Yea its pretty much confirmed...480 ~ 390/X.

VR performance is on par with a stock gtx970.

Clocks are 1080 base ,1266 boost.
32 rop's

http://www.techpowerup.com/forums/thread...st-3478890





This guy gasoline is testing the card.

Starting from page 18 I think.
http://www.techpowerup.com/forums/thread...51/page-17

He is using 16.6.2 drivers...
http://www.mediafire.com/download/h13f6s...+16.20.rar

[SickBeast]

If it does overclock to 1500mhz though it could come quite close to a GTX 980 Ti. This is going to be an interesting card. It's certainly priced right.

[happy medium]

If it does overclock to 1500mhz though it could come quite close to a GTX 980 Ti.  This is going to be an interesting card.  It's certainly priced right.

It don't, seems its hitting about 1330 and he said it gave him another 2 to 4 fps.
A gtx980 with a 1490MHZ overclock will be only 9% slower than a gtx980ti @ 1080p.

480 seems about as fast as a 220$, 150 watt, gtx970 @ 1450mhz.

480 cards are on newegg.

http://www.newegg.com/Product/ProductLis...12219&SID=

[SickBeast]

It sucks that they are reference cards though and the rumor is that they run quite hot. You would think that the reference cooler would be one of the best. They should really come up with better reference coolers. I don't know why they insist on making those blower versions that no one wants. They are hot and loud and they just plain suck unless you want to run Crossfire/SLI.

[SteelCrysis]

I saw a video of a guy running BF4 at 1080p on an RX 480. He indicated that with the reference cooler, it was hitting 85C.

[SickBeast]

Yes, it looks like Polaris runs hot which is very bad news. AMD GPUs have always run hot, for the past several generations anyhow. I would have thought that the new node would have helped with that, but apparently not.

[RolloTheGreat]

If it does overclock to 1500mhz though it could come quite close to a GTX 980 Ti.  This is going to be an interesting card.  It's certainly priced right.

It don't, seems its hitting about 1330 and he said it gave him another 2 to 4 fps.
A gtx980 with a 1490MHZ overclock will be only 9% slower than a gtx980ti @ 1080p.

480 seems about as fast as a 220$, 150 watt, gtx970 @ 1450mhz.

480 cards are on newegg.

http://www.newegg.com/Product/ProductLis...12219&SID=

The mighty AMD Defense Crew, who just got done howling about how the "Flounders Edition" 1080s have a "horrible" aluminum shroud and vapor chamber cooler will now be spanking over that plastic, rear blower bargain bin cheapie that face plants on every AMD launch.

"The plastic is so futuristic looking, they look badass! "

"I've heard plastic has better thermals than that cheesy looking aluminum!"

Etc.

[RolloTheGreat]

Yes, it looks like Polaris runs hot which is very bad news.  AMD GPUs have always run hot, for the past several generations anyhow.  I would have thought that the new node would have helped with that, but apparently not.

For who?

I just bought a card with somewhere between 1070 and 1080 performance for $390 shipped that has excellent cooling and 14/3phase power delivery. Sounds like you got similar for $360..

We don't have to buy AMDs spare parts.

Which 980 Ti did you get?

[RolloTheGreat]

Yes, it looks like Polaris runs hot which is very bad news.  AMD GPUs have always run hot, for the past several generations anyhow.  I would have thought that the new node would have helped with that, but apparently not.

For who?

I just bought a card with somewhere between 1070 and 1080 performance for $390 shipped that has excellent cooling and 14/3phase power delivery. Sounds like you got similar for $360..

We don't have to buy AMDs spare parts.

Which 980 Ti did you get?

[SickBeast]

Yes, it looks like Polaris runs hot which is very bad news.  AMD GPUs have always run hot, for the past several generations anyhow.  I would have thought that the new node would have helped with that, but apparently not.

For who?

I just bought a card with somewhere between 1070 and 1080 performance for $390 shipped that has excellent cooling and 14/3phase power delivery. Sounds like you got similar for $360..

We don't have to buy AMDs spare parts.

Which 980 Ti did you get?

I never got a 980 Ti. Where did you hear that? You read my posts at BTR? I almost bought one for $440CAD the other day which I think is $340USD. I didn't bite though. I have ordered a DisplayPort to HDMI 2.0 adapter for my 280x and I am waiting for a good deal on a new GPU. I have thought about the GTX 980 Ti and I'm not going to go that route unless they really clear them out for a blowout price. The GTX 1070 has more VRAM, it runs cooler, it consumes less power, and it overclocks better. For the sake of $40USD difference I'm willing to pay that much extra for all those advantages. Now, if they undercut the GTX 1070 price by $100USD then now we're talking.

I'm not in any rush. I can continue to game at 1080p on my 4k TV, it still looks quite nice. Plus I actually have a ton of older games that I think my 280x can run at 4k resolution. Most of the games my son likes are PS3 ports. I'm sure they will run at 4k resolution.

[RolloTheGreat]

Yes, it looks like Polaris runs hot which is very bad news.  AMD GPUs have always run hot, for the past several generations anyhow.  I would have thought that the new node would have helped with that, but apparently not.

For who?

I just bought a card with somewhere between 1070 and 1080 performance for $390 shipped that has excellent cooling and 14/3phase power delivery. Sounds like you got similar for $360..

We don't have to buy AMDs spare parts.

Which 980 Ti did you get?

I never got a 980 Ti.  Where did you hear that?  You read my posts at BTR?  I almost bought one for $440CAD the other day which I think is $340USD.  I didn't bite though.  I have ordered a DisplayPort to HDMI 2.0 adapter for my 280x and I am waiting for a good deal on a new GPU.  I have thought about the GTX 980 Ti and I'm not going to go that route unless they really clear them out for a blowout price.  The GTX 1070 has more VRAM, it runs cooler, it consumes less power, and it overclocks better.  For the sake of $40USD difference I'm willing to pay that much extra for all those advantages.  Now, if they undercut the GTX 1070 price by $100USD then now we're talking.

I'm not in any rush.  I can continue to game at 1080p on my 4k TV, it still looks quite nice.  Plus I actually have a ton of older games that I think my 280x can run at 4k resolution.  Most of the games my son likes are PS3 ports.  I'm sure they will run at 4k resolution.

I saw your post on Babble Tech.

(shrugs)

Nothing wrong with 1070s, but they don't offer much over the good 980Tis. If you're willing to run them far over spec, that's about it.

[SteelCrysis]

First Polaris gaming benchmarks have leaked, a little less performance than the 970: http://videocardz.com/61512/first-radeon...it-the-web

[SickBeast]

Those benchmarks look completely fake. Look at the 380x scores.

[RolloTheGreat]

Those benchmarks look completely fake. Look at the 380x scores.

I actually feel bad for AMD and I feel bad for America.

There "should" be a place in the world for AMD but the competition intel and NVIDIA is too well funded.

They do their jobs the second best on the planet, how many industries are there where second best on the planet gets you kicked in the balls, knocked into the gutter, and spat upon?

If you build the second best fishing reel, shotgun, boat, truck, TV (to name some things I buy) you sell it for big bucks and your buyers are happy.

The PC industry is the only market I can think of where it is "number one or go bankrupt".

If they wouldn't have sicced their attack mongrels on me, I'd probably be trying to help them as I always did up to the FX-60.

[happy medium]

Just watched this review . the rx480 sucks!
sower than a 390! no overclocking and runs hot.
https://www.youtube.com/watch?v=5CJBaZ9V2Eo

[ocre]

Talk about no overclocking!! What the heck? How are the amd troopers gonna justify all the gp104 crap they talked. They trashed and trashed the gp104. Another interesting thing is how they are saying that a 480 is worthy for 390 (x) owners to upgrade to but they claim that getting the 1080 or 1070 is dumb if you own a 980ti

[ocre]

As for performance, it is exactly my projection.
I predicted:
-would not be faster than the 390x consistently or on average.
-that it would be Hawaii class performance.

The real shocker is the overclocking...geez

Let's wait some to see more reviews...but this is really looking like a phenom (original) like blunder...
Haven't seen power consumption yet, but it may end up a bulldozer like failure.

But just think about what their scientist raja was saying...something like pascal is just a higher clocks but Polaris is so much better because it's a vastly improved arch that will be higher clocked too.

Geez, they can't even pass maxwell clocks

[happy medium]

https://hardforum.com/threads/first-gami...1042385515
Well Kyle is saying AIB partners could have cards that clock up to 1600mhz.

I wouldn't hold my breath.
I still think it would be slower than a gtx980 and with a 980's overclock, it would get destroyed.

[RolloTheGreat]

Talk about no overclocking!! What the heck?  How are the amd troopers gonna justify all the gp104 crap they talked.  They trashed and trashed the gp104.  Another interesting thing is how they are saying that a 480 is worthy for 390 (x) owners to upgrade to but they claim that getting the 1080 or 1070 is dumb if you own a 980ti

Do not be foolish!

RedTeamLeader @AMD says the 480 is an overclocker's dream!


Like I said, I feel bad for AMD. My guess is they can't get the best chip engineers and don't have the resources to give the ones they have.

They're becoming a tech shell game of misdirection.

[SteelCrysis]

Those benchmarks look completely fake. Look at the 380x scores.

I actually feel bad for AMD and I feel bad for America.

There "should" be a place in the world for AMD but the competition intel and NVIDIA is too well funded.

They do their jobs the second best on the planet, how many industries are there where second best on the planet gets you kicked in the balls, knocked into the gutter, and spat upon?

If you build the second best fishing reel, shotgun, boat, truck, TV (to name some things I buy) you sell it for big bucks and your buyers are happy.

The PC industry is the only market I can think of where it is "number one or go bankrupt".

If they wouldn't have sicced their attack mongrels on me, I'd probably be trying to help them as I always did up to the FX-60.

You do have a good point.

[SteelCrysis]

http://www.techpowerup.com/reviews/AMD/RX_480/

Quoting from the AMD reviewer's guide "A lot of work has gone into reducing noise levels for the Radeon RX 480." Sorry, but no, gaming noise levels are bad. The reference card is noisier than every single card released in recent times, and it runs at well above 80°C too. We confirmed the temperature and noise levels with other reviewers, so it's not only our sample.
...
Maximum overclock of our sample is 2250 MHz on the memory (13% overclock) and 1335 MHz on the GPU (5% overclock).

While memory overclocks really well, reaching the Overdrive adjustment limit, GPU overclock is extremely limited and one of the smallest I've seen in my reviews so far. AMD does offer the option of voltage control, but its effect seems extremely limited, with no noteworthy frequency gains in my testing.
...
Using these clock frequencies, we ran a quick test of Battlefield 3 to evaluate the gains from overclocking.
Actual 3D performance gained from overclocking is 5.0%.
...
Idle temperatures are good, because the fan does not stop in idle. During gaming temperatures are quite high, reaching 84°C - which isn't alarming, but doesn't exactly boost confidence in the reference design cooler, especially since it's kinda noisy, too.
...
As mentioned before, power efficiency is improved significantly, mostly thanks to the 14 nanometer FinFET process. These improvements bring the RX 480 roughly on the same performance per Watt level as NVIDIA's last-generation Maxwell architecture. NVIDIA's current Pascal architecture, is still over 70% more power-efficient. This makes me wonder a bit, how NVIDIA managed to gain such huge improvements, while bumping up GPU clock to 2 GHz, and AMD is still running in the 1200 MHz range, with worse power consumption.
...

The weakest point of AMD's reference design is certainly the thermal solution. It doesn't use any heatpipes or other high-tech means to keep the card cool. Rather, there is a big slab of metal, with a copper core that has the blower fan sending air across its fins. As a result we are seeing temperatures of up to 84°C, which has the effect that the card will clock down further to keep cool. On average our card ran 1239 MHz, which is in the upper range of AMD's rated 1120 - 1266 MHz clock window. What's even worse than the heat is the terrible fan noise. While idle noise is fine with 29 dBA (an idle-fan-off feature would still have been nice), in gaming the fan ramps up a lot, emitting 41 dBA noise during gaming (not Furmark). This makes the RX 480 the loudest card launched in recent history, much noisier than for example GTX 1080 (which is almost twice as fast). AMD has mentioned to us that the reference design is deliberately weak to leave room for partners to improve on their custom designs. To me this sounds a bit like "let the partners deal with the problem".

With Polaris AMD is introducing a new overclocking control panel called "WattMan" which has tons of options, including voltage control and several ways to adjust the thermal profile. However, overclocking potential on our sample was very slim. All we managed to do stable was increase the GPU clock from 1266 MHz up to 1335 MHz - a lousy 5% increase, again the worst I've seen for years on a reference board. This is further complicated by the fact that the card will often clock down during OC because it a) exceeds the board power limit or b) runs too hot. If you increase the power limit using WattMan, you'll run into the thermal limit quicker. It does seem that there is a huge spread between GPUs on review samples. I've heard of reviewers who see stock temperatures well below 80°C, while others reach up to 89°C. Assuming that AMD has selected the best cards for press review, retail cards might be even worse, which means higher temps, more noise, lower performance.

The real highlight of the Radeon RX 480 is its pricing though. It starts as low as $199 for the 4 GB version, which also comes with slightly lower memory clocks. In this review we have tested the 8 GB version, which is priced at $239. This ensures both cards claim the price/performance throne of all cards we've tested so far. Especially the 4 GB version seems really affordable, I seriously doubt the extra 4 GB on the 8GB variant will be able to provide an extra 20% performance to make up for the 20% increase in cost. So, AMD has successfully captured the market segment for 1080p cards with the RX 480. This effectively obsoletes all previous AMD cards except for Fiji maybe, which itself has been obsoleted by GTX 1070. Everything faster is owned by NVIDIA's iron-grip on the high-end, where huge margins are possible. Also I expect that NVIDIA will be dropping the price of GTX 970 further to get rid of its inventory, which could entice potential shoppers to go for NVIDIA's alternative, and then there's GTX 1060 coming soon, too.

http://www.techpowerup.com/reviews/AMD/R...re/21.html

At just $398, about the same price as the cheapest GeForce GTX 1070 you can find, or $478 for a pair of 8 GB cards like we have, the Radeon RX 480 CrossFire is not a viable solution, if you plan to buy two cards upfront. When averaged over all our games it is consistently slower than a single GeForce GTX 1070 at all the resolutions that matter - 1080p, 1440p, and 4K. Instead of buying two cards upfront, you're much better off putting your monies into a single GTX 1070, not just for better performance but also to dodge the spectre of application multi-GPU support, which continues to haunt both SLI and CrossFire.

We averaged games that do take advantage of CrossFire in a separate relative performance data-point than the overall relative-performance. The findings are interesting, when averaged among games that do scale, the RX 480 CrossFire is about 5-10% faster than a GTX 1070. However, only 6 out of 16 tests are taking advantage of the second card.

If you only have money for a single card now, and you want to buy a second card later, then you're in for more than playable frame-rates at 2560x1440 resolution, in games that do scale, and even 4K performance with eye-candy watered down a little. The RX 480 CrossFire loses out its overall relative performance big time, due to the number of games that don't scale well (6 out of 16).

In games that do scale, you're treated with upwards of 85% performance uplift. From our Radeon R9 Nano CrossFire review till now, we see that AMD hasn't really spent a lot of time optimizing CrossFire for more games. We can't fault AMD too much, though, because since then, a lot of games that came out don't support multi-GPU at all, due to engine limitations. There's not much AMD or NVIDIA can do about such games, and this is what scares us about multi-GPU solutions going forward.

We have to give AMD credit where due, for including support for 3-way or 4-way CrossFire, unlike NVIDIA which has dropped support for 3-way and 4-way SLI. The RX 480 supports both 3-way and 4-way configurations, and in any application that can take advantage of CrossFire, not just synthetic benchmarks.

With the advent of DirectX 12, we are promised new multi-GPU rendering modes thanks to Microsoft giving developers more control over per-GPU resource allocation, but we doubt that we'll see widespread use of those techniques. Nowadays, games are developed for consoles first (which are single-GPU), and publishers have little interest in spending a lot of developer time (= money) on adding support for exotic multi-GPU configurations that are used by only a small percentage of their customers.

Some of the latest DirectX 11 games (eg: Just Case 3) and the early DirectX 12 ones, are struggling with multi-GPU support. NVIDIA too, has cut out 3-way and 4-way SLI support (but for slightly different reasons), making us wonder if it's worth the trouble following a multi-GPU upgrade path with mid-thru-performance segment cards such as the RX 480. We rather recommend you spend $398-458 on a single GPU solution.

[ocre]

The card is far worse that I thought, it's not performance specifically but the power consumption is astronomical when compared to pascal. This blows my mind. It uses more power than the 6 pin can provide, consistently!!


Wholly smokes. Amd manages to just about reach maxwell perf per watt!! Maxwell!!!
That is insane. How the fuck did they come up with 2.5x more efficient? These are huge massive lies they told. Bold face lies. Yet Nvidia gets railed for their gp104 launch which has claims that are variable, even if some are special or best case scenario...amd flat out lied.

The power consumption is horrible for a 14nm card running at 1266mhz. I am truly blown away

[BoFox]

Yes, it looks like Polaris runs hot which is very bad news.  AMD GPUs have always run hot, for the past several generations anyhow.  I would have thought that the new node would have helped with that, but apparently not.

Hell, with that aluminum heatsink, it's like having a stock Intel heatsink and trying to overclock the thing to 4.5GHz. 

Those who really want more performance from overclocking should just avoid the stock cooler like the plague and wait for custom coolers by AIB's (especially if there's a version that allows for increased voltage).  Temperatures would drop by maybe 25-30C, reducing power leakage by quite a bit (perhaps 20W or more), and allow for much higher overclocking (perhaps as high as 1500MHz, but I think more than 1450MHz would be hard to accomplish with just 1 6-pin PCIE power connector.  By the time MSI or Gigabyte designs a board with an 8-pin PCIE power plug, AMD would probably have launched Vega by then, heh.

[BoFox]

The card is far worse that I thought, it's not performance specifically but the power consumption is astronomical when compared to pascal.  This blows my mind.  It uses more power than the 6 pin can provide, consistently!!


Wholly smokes.  Amd manages to just about reach maxwell perf per watt!! Maxwell!!!
That is insane. How the fuck did they come up with 2.5x more efficient?  These are huge massive lies they told. Bold face lies. Yet Nvidia gets railed for their gp104 launch which has claims that are variable, even if some are special or best case scenario...amd flat out lied.

The power consumption is horrible for a 14nm card running at 1266mhz.  I am truly blown away

Spot on!  Consistently around 165W, or higher.  AMD rated the card at 150W TDP.  It's a lie, AMD.  I like how TPU includes both average and peak gaming power consumption.

What's worse is that AMD's ***STILL*** lying about it being not 2.5x but 2.8x more efficient than 28nm cards:


Moreover, the card runs at around 1200MHz in most games, rather than 1266MHz.  Some games make it drop down to 1140MHz - see:
https://www.computerbase.de/2016-06/rade...80-test/5/

[RolloTheGreat]

Somebody good with Photoshop should put together a picture of homeless guy sleeping by dumpster with a 480 sticking out of his pocket and another one sticking out of his change cup.

Posted on ATVF, it would be pandemonium while it lasted.

They'd run around in circles, rip up the grass and throw it in the air, beat their chests and roar at the sky!

[BoFox]

http://www.techpowerup.com/reviews/AMD/R...re/21.html

At just $398, about the same price as the cheapest GeForce GTX 1070 you can find, or $478 for a pair of 8 GB cards like we have, the Radeon RX 480 CrossFire is not a viable solution, if you plan to buy two cards upfront. When averaged over all our games it is consistently slower than a single GeForce GTX 1070 at all the resolutions that matter - 1080p, 1440p, and 4K. Instead of buying two cards upfront, you're much better off putting your monies into a single GTX 1070, not just for better performance but also to dodge the spectre of application multi-GPU support, which continues to haunt both SLI and CrossFire.

We averaged games that do take advantage of CrossFire in a separate relative performance data-point than the overall relative-performance. The findings are interesting, when averaged among games that do scale, the RX 480 CrossFire is about 5-10% faster than a GTX 1070. However, only 6 out of 16 tests are taking advantage of the second card.
.....

Some of the latest DirectX 11 games (eg: Just Case 3) and the early DirectX 12 ones, are struggling with multi-GPU support. NVIDIA too, has cut out 3-way and 4-way SLI support (but for slightly different reasons), making us wonder if it's worth the trouble following a multi-GPU upgrade path with mid-thru-performance segment cards such as the RX 480. We rather recommend you spend $398-458 on a single GPU solution.

OUCH!  6 out of 16!!!  It's never been that bad since AMD first started doing Crossfire with the X850XT cards, I think.  Even the generation after that when I had SLI cards, I never went back to a multi-GPU configuration because it's just not worth it.  I noticed microstuttering way before review sites ever started talking about it (where 40 fps felt like 22 fps for example). 

Then there's this horrible hitching in GTAV with all AMD cards, not just with Crossfire but with single GPUs:



Is AMD ever going to fix this hitch-fest, for one of the by far most popular games ever? 

One artifact that these charts don’t entirely capture is an intermittent stutter suffered by AMD’s cards. Big frame time spikes on the Radeon R9 390 are our only data-based indicator of this, but every card exhibits the behavior.
...
This time the spikes are observable on the 480, 390X and 290, though they appear to be outliers when, in reality, they’re far more persistent.

http://www.tomshardware.com/reviews/amd-...616-5.html

I'd gladly lose 40% of the frame rate just to see the game-breaking hitching go away, but thankfully we have Nvidia.

If AMD's the 2nd-best GPU designer in the world, with brilliant engineers, they could definitely resolve the hitching, but I think they deliberately left that there just to maintain 10-15% higher frame rate for benchmarks.  And, hardly any vocal AMD fanboy really wants to complain about it, while NV fans don't bother with AMD cards to even care.

[ocre]

Thanks for the Tom's reference Bofox.
I haven't went there for reviews in a long time, but they have come a long long way. Probably one of the most comprehensive power testing I have ever seen done on graphics cards.

You know, all the hype and AMD troopers really had me grinning, I simply could not wait for Polaris launch. There were a lot of signs and a few whispers, I knew the card was not gonna be nearly as good as all the claims. I was eagerly waiting for the is one to blow up. But man....I wasn't ever expecting things to be this bad. The atrocious power hog it turned out to be, it blows NY mind. This was the one metric AMD harped on the most. It's not like I expected it to beat pascal, I mean everywhere on every site people had taken it as a given that AMD was gonna be far ahead of Nvidia in perf per watt. But, it was real obvious to me that pascal efficiency was extremely great. Most everyone overlooked it due to AMD and their claim that Polaris was 2.8x times as efficient. So, PASCAL'S phenomenal perf per watt was completely disregard. As a matter of fact, I remember so many railing on Nvidia saying that their gains from moving to 16nm was pathetic.
But geeZ... there are mountains of crow that needs to be eaten.

I knew that Polaris wouldn't surpass pascal perf per watt, but I never ever expected such utter trash as this-

AMD’s Radeon RX 480 draws an average of 164W, which exceeds the company's target TDP. And it gets worse. The load distribution works out in a way that has the card draw 86W through the motherboard’s PCIe slot. Not only does this exceed the 75W ceiling we typically associate with a 16-lane slot, but that 75W limit covers several rails combined and not just this one interface.

With peaks of up to 155W, we have to be thankful they're brief, and not putting the motherboard in any immediate danger. However, the audio subsystems on cheaper platforms will have a hard time dealing with them. This means that the "you can hear what you see" effect will be in full force during load changes; activities like scrolling may very well result in audible artifacts.

We’re also left to wonder what we'd see from a CrossFire configuration. Two graphics cards would draw 160W via the motherboard’s 24-pin connector; that's a tall order. Switching from the bars back to a more detailed curve makes this even more evident.

We skipped long-term overclocking and overvolting tests, since the Radeon RX 480’s power consumption through the PCIe slot jumped to an average of 100W, peaking at 200W. We just didn’t want to do that to our test platform.

Stress Test Power Consumption

Believe it or not, the situation gets even worse. AMD's Radeon RX 480 draws 90W through the motherboard’s PCIe slot during our stress test. This is a full 20 percent above the limit.

To be clear, your motherboard isn't going to catch fire. But standards exist for a reason. All of the components around the PCIe slot and along the path from the slot to the 24-pin ATX connector will suffer from the peaks. And depending on your platform's design, audio problems may also materialize. 

http://www.tomshardware.com/reviews/amd-...616-9.html

You know, this is a very scary thing. TOM'S and every other reviewer out there have robust enthusiasts boards. The market AMD is aiming at with their card, it's bare minimum budget guys. You know, this is off the charts. Tom's discovery was jaw dropping, and as mean as AMD fans might accuse it.....Tom's went very very light on the issue.

On their super dooper high class top of the line motherboard, they were afraid of overclocking due to the extreme violation of the PCIe power specs observed using the 480 at stock. There will be pure people buying these cards, with and troopers persuasion, then straight up popping them in crapware budget computers. Using Crap PSUs with mole to 6 pin adapter on a tiny case with bare minimum El cheapo motherboards. It is completely unfathomable. Tom's went so so light on the issue. Sure, his top of the line psu and mobo may be able to withstand such abuse but people with that kind of HW most likely we buy a comparable nice TOP of the line GPU.

AMD'S 480 is worse worse than the Fermi 480, as the Fermi was used and aimed at the top of the line enthusiasts market. Not a card that blast past PCIe specs (mobo and psu) aimed at the budget market. If Fermi was ever bad, this is far worse.

[ocre]

Somebody good with Photoshop should put together a picture of homeless guy sleeping by  dumpster with a 480 sticking out of his pocket and another one sticking out of his change cup.

Posted on ATVF, it would be pandemonium while it lasted.

They'd run around in circles, rip up the grass and throw it in the air, beat their chests and roar at the sky!

Please do!!!

That would be the most awesome thing I ever saw. I would literally die, nothing could ever top it.

It's so freaking funny just thinking about a picture like that

[ocre]

Yes, it looks like Polaris runs hot which is very bad news.  AMD GPUs have always run hot, for the past several generations anyhow.  I would have thought that the new node would have helped with that, but apparently not.

Hell, with that aluminum heatsink, it's like having a stock Intel heatsink and trying to overclock the thing to 4.5GHz. 

Those who really want more performance from overclocking should just avoid the stock cooler like the plague and wait for custom coolers by AIB's (especially if there's a version that allows for increased voltage).  Temperatures would drop by maybe 25-30C, reducing power leakage by quite a bit (perhaps 20W or more), and allow for much higher overclocking (perhaps as high as 1500MHz, but I think more than 1450MHz would be hard to accomplish with just 1 6-pin PCIE power connector.  By the time MSI or Gigabyte designs a board with an 8-pin PCIE power plug, AMD would probably have launched Vega by then, heh.

The power consumption would not drastically change. Amd took a page from Nvidia and tried their best to keep the Temps in the low to mid 80s. This is the same spot Nvidia has been capping their cards at. It's not running 95 degrees.

Look at Nvidia cards reference cards like the 1070, 970, 980, etc. Their target was low 80s and their performance per watt was optimal. Custom cards that run cooler don't show radically improved efficiency. Actually, the custom cards tend to not be better at all when it comes to efficiently. I don't ever remember seeing that, not even once. custom cards with better cooling do drop Temps for maxwell and pascal, but the overall consumption tends to be higher than Nvidia reference designs running in the low 80s.

See, custom cards have more components and are more robust but those things come with added power draw. The improved cooling doesn't seem to lower power consumption and with pascal at least, they haven't brought major leaps in overclocking capabilities.

I think the 1600mhz Polaris over clocks were just bologna, just as the 2500mhz pascal rumors. Just inflated bs hype.

So, better cooling on the 480 wouldn't be bad but I seriously doubt there would or could be such a radical drop in power consumption. They seem to stay around 80c already, Nvidia has the same temp targets and I trust they have aimed their cards in the ideal range. It's not like their blowers are the issue, not so much, this is just their fan target speeds. If a few degrees lower was so much more efficient, they would be targeting below 81c. Their cards hold 70 degree Temps with just a minor change in fan curve. So i am pretty sure that the low 80s is the ideal zone for temps, power, and efficiency...at least this is what Nvidia seems to think.

[BoFox]

Hell, with that aluminum heatsink, it's like having a stock Intel heatsink and trying to overclock the thing to 4.5GHz. 

Those who really want more performance from overclocking should just avoid the stock cooler like the plague and wait for custom coolers by AIB's (especially if there's a version that allows for increased voltage).  Temperatures would drop by maybe 25-30C, reducing power leakage by quite a bit (perhaps 20W or more), and allow for much higher overclocking (perhaps as high as 1500MHz, but I think more than 1450MHz would be hard to accomplish with just 1 6-pin PCIE power connector.  By the time MSI or Gigabyte designs a board with an 8-pin PCIE power plug, AMD would probably have launched Vega by then, heh.

The power consumption would not drastically change.  Amd took a page from Nvidia and tried their best to keep the Temps in the low to mid 80s.  This is the same spot Nvidia has been capping their cards at.  It's not running 95 degrees.

Look at Nvidia cards reference cards like the 1070, 970, 980, etc.  Their target was low 80s and their performance per watt was optimal.  Custom cards that run cooler don't show radically improved efficiency.  Actually, the custom cards tend to not be better at all when it comes to efficiently.  I don't ever remember seeing that, not even once.  custom cards with better cooling do drop Temps for maxwell and pascal, but the overall consumption tends to be higher than Nvidia  reference designs running in the low 80s.  

See, custom cards have more components and are more robust but those things come with added power draw.  The improved cooling doesn't seem to lower power consumption and with pascal at least, they haven't brought major leaps in overclocking capabilities.

I think the 1600mhz Polaris over clocks were just bologna, just as the 2500mhz pascal rumors.  Just inflated bs hype.

So, better cooling on the 480 wouldn't be bad but I seriously doubt there would or could be such a radical drop in power consumption.  They seem to stay around 80c already, Nvidia has the same temp targets and I trust they have aimed their cards in the ideal range.  It's not like their blowers are the issue, not so much, this is just their fan target speeds.  If a few degrees lower was so much more efficient, they would be targeting below 81c. Their cards hold 70 degree Temps with just a minor change in fan curve.  So i am pretty sure that the low 80s is the ideal zone for temps, power, and efficiency...at least this is what Nvidia seems to think.

You might be right - it seems that power leakage isn't as bad as it was at 40-45nm.  Some chips made at 45nm would generally leak as much as 40W extra per 20C increase (when over-volted really high in order to overclock as high as possible).  Now, the voltage is far lower, so perhaps leakage is hardly noticeable at all. 

I'd beg to differ with you, though - 28nm cards with custom cooling seemed to show lower power consumption figures than those with stock cooling.  Just look at the HD 280X or 7970 GHz Edition compared to the 7970.  Most review sites received custom-cooled cards for review, which ran at far lower temperatures along with lower noise, and the power consumption was also somewhat lower in order to compare better against GTX 680/770. 

I hope a hardware site would give a real good look at power leakage vs temperature with 14/16nm cards compared to 28nm (and also 40nm GPUs).  Anandtech did a real great analysis on 45nm Intel cpus (yeah, so many years ago, when power leakage was getting way worse than 65nm and 90nm, but Intel promised that it would get better after 45nm).

[ocre]

I'd beg to differ with you, though - 28nm cards with custom cooling seemed to show lower power consumption figures than those with stock cooling.  Just look at the HD 280X or 7970 GHz Edition compared to the 7970.  Most review sites received custom-cooled cards for review, which ran at far lower temperatures along with lower noise, and the power consumption was also somewhat lower in order to compare better against GTX 680/770.  

While it is true that the 7970ghz used more power than the original 7970 and the ghz ran a few degrees hotter but you seem to have completely ignored (or dont know) the other critical factors which are well known to directly impact power consumption.

1) voltage
2) clock speed.

Amd pushed up both the voltage and clock speed on the ghz. It's well documented that simply undervolting GCN reduces consumption and some claim to have dramatic results. While all chips have respond the same way, more volts equals more power consumption. It appears reasonable to assume GCN is sensitive to small increases in voltage.

As for the other gpu you bring up, it's also a common bet that as time goes by, the process, chips, and nodes mature. The later gtx 480s for example, they were greatly improved from the first batches.

All these things make it impossible to conclude, it's flawed bad. The only accurate way is to have a gpu, run it at 80c while measuring the wattage then take the same setup, same clocks, same voltages, same card and cool it to 70 and measure the difference. I don't think it will be that big

You also must remember, these things are not linear...it's on a curve

[ocre]

Not to mention, there is variation from chip to chip when it comes to power consumption