AMD Polaris Thread: Radeon RX 480, RX 470 & RX 460 launching June 29th

[Dribble]

Almost no one who took the time to carefully study AMD's statements and Polaris 10 specs expected Fury X and especially not Fury X+ levels of performance out of a <240mm2 die, 2304 shaders with a 256-bit bus, back then rumoured 1.05Ghz-1.1Ghz clocks, and regular GDDR5.

But I get the feeling if Polaris had been able to use TSMC and had therefore got 25% higher clocks it would have been close. I can imagine the frustration in the radeon tech group right now as they release this - the AMD cpu group firstly blows all the money on Zen, but even with this we produce a decent card only to have it torpedo'd by that stupid wafer agreement the cpu group negotiated meaning we have to use GloFo and compete with one arm tied behind our backs.

 

[boozzer]

base on all the info so far, polaris could be king in every market segment under 300$.

when will 1060/Ti have their nda lifted? once we have single card benchmarks of polaris and 1060, everything will be clear.

 

[Gikaseixas]

Seeing a lot of disappointment in different forums

You should stop looking for flaws and appreciate the tech. How about this GTX 970/980 performance for GTX 960 prices?

 

[Burpo]

4 minutes into the video, Raja drops more than price.. lol

https://www.youtube.com/watch?v=eqCo0LTUhsE#t=1.309083

 

[boozzer]

You should stop looking for flaws and appreciate the tech. How about this GTX 970/980 performance for GTX 960 prices?

960 prices were higher than 200$ :/ that piece of crap were pushed by a few posters on this very forum.

 

[Gikaseixas]

960 prices were higher than 200$ :/ that piece of crap were pushed by a few posters on this very forum.

Yes i know that and fully agree with you

 

[monstercameron]

You should stop looking for flaws and appreciate the tech. How about this GTX 970/980 performance for GTX 960 prices?

Lol dont buy into his confirmation bias...

 

[FatherMurphy]

Am I the only one who feels the TPD for the RX480 is a bit high or, at least, unexpectedly high for this card and performance level? If the performance in DX11 and DX12 games is roughly between 970 and 980, as has been reported, then AMD is only just catching up to or barely surpassing Maxwell's perf/w with the jump to 14nm FinFet. Moreover, add 15-30W, and you have a stock 1080, which is much much faster than a 980. What am I missing, because AMD is touting enormous performance/watt gains (presumably against Hawaii, I suppose)?

 

[Bryf50]

Seeing a lot of disappointment in different forums. That has to do with the huge expectations around this card, many expecting Fury X+ all-around performance. Of course we have yet to see numbers, but Geforce GTX 970/980 performance (told to the press in the Macau event according to Videocardz) is not out of reach for GP106, especially if rumours of a 192-bit 6GB VGA replacing GM204 are correct.

Seeing a lot of excitement in different forums. That has to do with the aggressive launch pricing of the RX 480 and meeting most reasonable expectations, many expecting $299+.

See I can make vague unsubstantiated statements also.

 

[swilli89]

Am I the only one who feels the TPD for the RX480 is a bit high or, at least, unexpectedly high for this card and performance level? If the performance in DX11 and DX12 games is roughly between 970 and 980, as has been reported, then AMD is only just catching up to or barely surpassing Maxwell's perf/w with the jump to 14nm FinFet. Moreover, add 15-30W, and you have a stock 1080, which is much much faster than a 980. What am I missing, because AMD is touting enormous performance/watt gains (presumably against Hawaii, I suppose)?

That is a concern of mine as well, but like others have said, AMD has a history of over rating their TDP by as much as 50% so for wattage we will need to wait for independent testing. This card may very well draw 80-90 watts, and since a single slot is 75W they had to go to 6-pin.

 

[Sweepr]

See I can make vague unsubstantiated statements also.

You don't expect me to provide links to specific forum posts, do you? And Geforce GTX 970/980 performance is not meeting expectations for most people, just look at pre-unveiling posts here. Currently it looks like a good deal for the price, but if you throw used Hawaii/GM204 into the mix (should be even cheaper from now on), along with rumoured GP106 pricing/specs, it's just 'ok'.

 

[Udgnim]

960 prices were higher than 200$ :/ that piece of crap were pushed by a few posters on this very forum.

and it sold like gangbusters alongside the GTX 970

 

[Daedalus685]

I don't understand the negativity, but I suppose I rarely do.

There are usually 4 or so chips (sometimes more if notebooks get their own dedicated parts) in a full generational lineup, we have known for some time that this would be the case for AMD with polaris and vega (a big and small of each). GP100(2), GP104, GP106, and GP108 have been expected for as long. Low end, mainstream, high end, and enthusiast (or whatever names we colloquially designate to the tiers)

Why then, is everyone losing their minds when these companies went with differing strategies and released products that targeted different markets? One released their mainstream, the other their high end. I don't know which strategy is right (halo effect vs volume targeting) but we will find out by next year at this time.

Can you imagine if AMD's mainstream part (chip 2 of 4) competed with GP104, leaving 2 full chips (perhaps 4 SKUs) ABOVE it in performance? A world where that is true is a world where NVidia is as far behind AMD as AMD was during the 8800GTX days... To expect that with any semblance of rationality is a joke.

I want to upgrade to a 1440p ultrawide, I want to have adaptive frame rate tech with it and I don't want to pay the g-sync premium. I would have much preferred if AMD released Vega first so I know how it stacks up with GP104 and I can get on with upgrading my 970, but that has absolutely no bearing on if the RX 480 is a good card or not. Mainstream is not 'for' us (folks on a tech forum).

To add, folks assuming the clock speed is more determined by process tech than design targets is dumbfounding.... There are people assuming that pascal not clocking as high as GP104 means Zen will have bad clock scaling compared to intel for crying out loud! Maybe GloFlo is awful, and I look forward to MONTHS from now when we can compare the entire product stack of AMD to Nvidia to see if that is the case (but even then it could be that GCN1.4 just isn't as good as pascal...). Most of the 900 series clocks a good 20-30% higher (boost) than its competing products and that was on the exact same process...

 

[kondziowy]

Am I the only one who feels the TPD for the RX480 is a bit high or, at least, unexpectedly high for this card and performance level? ... ...What am I missing, because AMD is touting enormous performance/watt gains (presumably against Hawaii, I suppose)?

R9 Fury has TDP of 275W. My entire rig with Asus Fury is consuming 250-280W at the wall on full load with stock clocks and voltage. Conclusion: never trust the TDP.

 

[Riek]

Am I the only one who feels the TPD for the RX480 is a bit high or, at least, unexpectedly high for this card and performance level? If the performance in DX11 and DX12 games is roughly between 970 and 980, as has been reported, then AMD is only just catching up to or barely surpassing Maxwell's perf/w with the jump to 14nm FinFet. Moreover, add 15-30W, and you have a stock 1080, which is much much faster than a 980. What am I missing, because AMD is touting enormous performance/watt gains (presumably against Hawaii, I suppose)?

Currently their are two 480 chip in the rumours (C4 and C7 ) where the 3dmark (rumoured) score of C7 is 7% above the 980. (+- FURY)
We also only know 150W design card with a 6pin. (1070 is a 150W TDP card with 8pin).

They also showed a 480CF which had similar performance in ashes as 1080 with higher efficiency. So for this these cards should be running <<90W.

 

[Despoiler]

R9 Fury has TDP of 275W. My entire rig with Asus Fury is consuming 250-280W at the wall on full load with stock clocks and voltage. Conclusion: never trust the TDP.

More like TDP does not equal power draw.

 

[oussama-tn]

R9 Fury has TDP of 275W. My entire rig with Asus Fury is consuming 250-280W at the wall on full load with stock clocks and voltage. Conclusion: never trust the TDP.

The r9 390 has a Tdp of 275W
http://www.kitguru.net/wp-content/uploads/2015/07/power-consumption.png

Kitguru's rig shows a use of 328W

 

[exar333]

R9 Fury has TDP of 275W. My entire rig with Asus Fury is consuming 250-280W at the wall on full load with stock clocks and voltage. Conclusion: never trust the TDP.

The TDP is helpful, but a full review/power analysis is needed to understand TDP in different scenarios. Also, it can differ from card to card slightly too, so looking at a few different reviews is helpful.

 

[Stuka87]

The r9 390 has a Tdp of 275W
http://www.kitguru.net/wp-content/uploads/2015/07/power-consumption.png

Kitguru's rig shows a use of 328W

That is ENTIRE SYSTEM power consumption, NOT just the card on its own.

 

[Headfoot]

But I get the feeling if Polaris had been able to use TSMC and had therefore got 25% higher clocks it would have been close. I can imagine the frustration in the radeon tech group right now as they release this - the AMD cpu group firstly blows all the money on Zen, but even with this we produce a decent card only to have it torpedo'd by that stupid wafer agreement the cpu group negotiated meaning we have to use GloFo and compete with one arm tied behind our backs.

I dont think its fair to assume the 25% clock speed advantage is due to TSMC process. Perhaps in part. But Maxwell clocked better than GCN on identical processes, and they made a big hulabaloo about focusing on the frequency critical path for Pascal. IMO Pascal as an architecture targets high clock speeds as its primary feature.

Plus it remains to be seen what sort of overclocking headroom there is. Pascal seems to have pushed out of the box clocks quite hard. If you recall the 7950 launch, they left tons of headroom around. Could be they've left lots of headroom on Polaris in order to hit lower TDP targets. Plus I would anticipate AMD / AIBs will still let us manually increase voltage like GCNs past

 

[tviceman]

The price is definitely fantastic. The architecture vs. Pascal is still up in the air. Is the rx 480 full die P10? If so 390x performance is not particularly close to 1070 - a part lambasted for being so cut down. If it's not full die, wtf is wrong with AMD choosing to hold back on such a small die when they need all the help they can give themselves?

If the underlying P10 is inferior in perf/mm2 and perf/w, then Nvidia can eventually match Polaris in performance/$ if they lose sales and still be forever better off going forward with lower costs and higher margins at similar perf/$. And, as everyone likes to think around ATF, Nvidia gets the benefit of the doubt 100% of the time when it comes to choosing a new card at similar prices and performance, so I don't see how this plays out well long term unless AMD is holding full die P10 back and waiting on GP106 to release full P10 (which is still a dumb thing to do in their situation).

 

[boozzer]

I don't understand the negativity, but I suppose I rarely do.

There are usually 4 or so chips (sometimes more if notebooks get their own dedicated parts) in a full generational lineup, we have known for some time that this would be the case for AMD with polaris and vega (a big and small of each). GP100(2), GP104, GP106, and GP108 have been expected for as long. Low end, mainstream, high end, and enthusiast (or whatever names we colloquially designate to the tiers)

Why then, is everyone losing their minds when these companies went with differing strategies and released products that targeted different markets? One released their mainstream, the other their high end. I don't know which strategy is right (halo effect vs volume targeting) but we will find out by next year at this time.

Can you imagine if AMD's mainstream part (chip 2 of 4) competed with GP104, leaving 2 full chips (perhaps 4 SKUs) ABOVE it in performance? A world where that is true is a world where NVidia is as far behind AMD as AMD was during the 8800GTX days... To expect that with any semblance of rationality is a joke.

I want to upgrade to a 1440p ultrawide, I want to have adaptive frame rate tech with it and I don't want to pay the g-sync premium. I would have much preferred if AMD released Vega first so I know how it stacks up with GP104 and I can get on with upgrading my 970, but that has absolutely no bearing on if the RX 480 is a good card or not. Mainstream is not 'for' us (folks on a tech forum).

To add, folks assuming the clock speed is more determined by process tech than design targets is dumbfounding.... There are people assuming that pascal not clocking as high as GP104 means Zen will have bad clock scaling compared to intel for crying out loud! Maybe GloFlo is awful, and I look forward to MONTHS from now when we can compare the entire product stack of AMD to Nvidia to see if that is the case (but even then it could be that GCN1.4 just isn't as good as pascal...). Most of the 900 series clocks a good 20-30% higher (boost) than its competing products and that was on the exact same process...

just a glance at their post history would tell you why, pretty easy to understand, and something they can never hide.

I am not a pc enthusiast who is willing to spend 600-700$ without batting an eye. my budget is 300$ for a desktop gpu. nothing more. I read this forum so I can buy the best perf/dollar gpu with my budget, that is all.

and I love the way you post :thumbsup:

 

[Headfoot]

Bottom line cant come until we see real performance figures. If this thing truly does perform a-la 980 / Fury air at $200 it will be a smash hit. Consider the last new chip released at $200 was the 960 2gb which was garbage. Provided no screw ups like the 290 cooler fiasco.

I'd bet dollars to donuts that they're building stock on the full chip to either sell for premium laptops where it could command significant margin or to use as a counterplay to the final gp106 release.

 

[Azix]

To add, folks assuming the clock speed is more determined by process tech than design targets is dumbfounding.... There are people assuming that pascal not clocking as high as GP104 means Zen will have bad clock scaling compared to intel for crying out loud! Maybe GloFlo is awful, and I look forward to MONTHS from now when we can compare the entire product stack of AMD to Nvidia to see if that is the case (but even then it could be that GCN1.4 just isn't as good as pascal...). Most of the 900 series clocks a good 20-30% higher (boost) than its competing products and that was on the exact same process...

AMDs clocks are probably lower due to density. They have more cores in less space. But they do reach higher teraflops so its w.e. I would expect some decent overclocks on polaris though since 1266Mhz is just close to the upper limit of older GCN clocks. Should reach over 1300, maybe over 1400. If not the card would not be 150W max. Doesn't seem like they have had to push the clocks as much as hawaii et al.

But I get the feeling if Polaris had been able to use TSMC and had therefore got 25% higher clocks it would have been close. I can imagine the frustration in the radeon tech group right now as they release this - the AMD cpu group firstly blows all the money on Zen, but even with this we produce a decent card only to have it torpedo'd by that stupid wafer agreement the cpu group negotiated meaning we have to use GloFo and compete with one arm tied behind our backs.

Or TSMC might have made it worse. I doubt they have a problem with using 14nm. Its supposedly superior to 16nm if just slightly.

Kyle needs a doctor, he seems to have adhd which does strange things to your head and perception.
I worked with a few.

it seems obvious amd have covered everything up and beyond 500$ with the 480.

Some folks will have a hard time grasping launching mainstream first. He'll probably need nvidia to do it before he can come to terms with the concept.

The price is definitely fantastic. The architecture vs. Pascal is still up in the air. Is the rx 480 full die P10? If so 390x performance is not particularly close to 1070 - a part lambasted for being so cut down. If it's not full die, wtf is wrong with AMD choosing to hold back on such a small die when they need all the help they can give themselves?

If the underlying P10 is inferior in perf/mm2 and perf/w, then Nvidia can eventually match Polaris in performance/$ if they lose sales and still be forever better off going forward with lower costs and higher margins at similar perf/$. And, as everyone likes to think around ATF, Nvidia gets the benefit of the doubt 100% of the time when it comes to choosing a new card at similar prices and performance, so I don't see how this plays out well long term unless AMD is holding full die P10 back and waiting on GP106 to release full P10 (which is still a dumb thing to do in their situation).

The only thing they could have worse than pascal is that VR multiprojection feature. GCN was already ahead of maxwell, polaris has to be more so. We can compare new architectures when volta is here. Polaris should be even more complete dx12 GPU. Pascal on the other hand is still not.

 

[Qwertilot]

Am I the only one who feels the TPD for the RX480 is a bit high or, at least, unexpectedly high for this card and performance level? If the performance in DX11 and DX12 games is roughly between 970 and 980, as has been reported, then AMD is only just catching up to or barely surpassing Maxwell's perf/w with the jump to 14nm FinFet. Moreover, add 15-30W, and you have a stock 1080, which is much much faster than a 980. What am I missing, because AMD is touting enormous performance/watt gains (presumably against Hawaii, I suppose)?

Well, first up they started a really rather long way behind Maxwell in terms of Perf/watt, so - with NV not staying still - they could have genuinely made a lot of improvements and still be behind.

With this being a desktop SKU from the chip, they might also have driven it well past the point of sensible efficiency gains to get a touch more performance.

The other thing is what people said about not being sure what the power draw really is until it gets tested. If it is a genuine 150w though? Not a good place to be. Nil laptop sales, won't be near competitive at the top end etc.

We'll find out