Polaris: 2x Perf/Watt of Current AMD mainstream GPUs

[maddie]

It's on AMD's investor website and it was published within the last couple of days.

Thanks. Yeah, missed the investor May 2016 at bottom left.

 

[Glo.]

It's on AMD's investor website and it was published within the last couple of days.

Lisa Su always claimed that Polaris will be 2 times perf/Watt. RTG on the other hand always claimed it will be 2.5 times perf/watt.

 

[lilltesaito]

I wonder if they changed the clock speeds after seeing what the 1080 is doing. Lets hope this change is for the better.

 

[maddie]

I'm watching their February 2016 Investor Presentation slides and they have the same 2X perf/W value. No change in 3 months.

Seeking clarity here.

Is it normal to be very conservative during these presentations? I know some companies understate forward guidance in order to claim, "we did better than expected". This seems better for stock values, always beating estimates.

 

[Glo.]

I'm watching their February 2016 Investor Presentation slides and they have the same 2X perf/W value. No change in 3 months.

Seeking clarity here.

Is it normal to be very conservative during these presentations? I know some companies understate forward guidance in order to claim, "we did better than expected". This seems better for stock values, always beating estimates.

Well making expectations lower than end result, is better than overhyping the product and not delivering. No matter how I try to calculate perf/watt using 2x mark - it does not make any sense.

Only 2.5 times makes, with understanding thermals, and performance of AMD GPUs.

 

[Arachnotronic]

I'm watching their February 2016 Investor Presentation slides and they have the same 2X perf/W value. No change in 3 months.

Seeking clarity here.

Is it normal to be very conservative during these presentations? I know some companies understate forward guidance in order to claim, "we did better than expected". This seems better for stock values, always beating estimates.

In investor presentations companies need to be careful because if they knowingly lie to investors, they can be in deep trouble.

 

[ZGR]

I'm guessing the 2.5x perf/watt improvement is from the conservatively clocked mobile GPUs.

 

[JoeRambo]

I'm guessing the 2.5x perf/watt improvement is from the conservatively clocked mobile GPUs.

And sadly from the other thread it seems that AMD could be well behind clock game. Fiji when downclocked/undervolted actually has decent perf/watt. Problem was that GM200 forced AMD to clock them rather high.

Same seems to be happening with GP104 - 2Ghz clock with minor overclock is like 1.4Ghz Maxwells all over again.

If AMD has 30-40% clock deficit - that is gonna be real bad. Shaders and teraflops are nice, but 2Ghz chip chewing pixels, textures and geometry has its own advantage.


P.S. just speculation, Polaris could well be running 3Ghz or maybe not.

 

[Head1985]

2x perf/watt vs 380/380x is good.
980TI is now 88% faster than 380x.

 

[IntelUser2000]

Lisa Su always claimed that Polaris will be 2 times perf/Watt. RTG on the other hand always claimed it will be 2.5 times perf/watt.

Lisa Su: Card
RTG: Chip?

I mean, power usage is more than the chip. You need to take board power into consideration, and that includes memory chips.

It seems that despite AMD's first showing of Polaris and rumors that Nvidia wasn't doing well in yields with Pascal, so far Pascal to Polaris looks just like last generation.

 

[Azix]

I hope it doesn't become 1.5x performance/watt by the time Polaris actually launches. Knowing AMD's history, it is actually quite possible.

but if they do that to offer a 1070 and 1080 competitor at lower price, would be fine.

 

[Qwertilot]

More likely something like 2.5 times vs some madly overclocked 290x, 2 times vs their more efficient chips? Marketing departments do of course love taking the most rabidly optimistic values possible

Utter speculation, but maybe the confidence of AMD of being ahead when they showed Polaris was maybe more down to plans of having it out before now? Certainly one big motivation of going tiny die first would have been to strike first.

Its a lot easier to believe that (say) the timetables at GF slipped a little somehow than AMD really believed that NV/TSMC were out to lunch.

They definitely want it out soon, to at least get it out before the 1060 and friends, and have a chunk of market to themselves for as long as possible.

Not that it matters in some ways. Polaris is already a success PS4++ contract, some apple stuff etc. Very handy all of that.

 

[Qwertilot]

but if they do that to offer a 1070 and 1080 competitor at lower price, would be fine.

No, not really. If there's a truly big perf/watt gap then NV get all the premium laptops, an almost perpetual lock on the fastest card at a given time etc etc.

Hopefully they'll get it plausibly close. Better for everyone that way

 

[Azix]

No, not really. If there's a truly big perf/watt gap then NV get all the premium laptops, an almost perpetual lock on the fastest card at a given time etc etc.

Hopefully they'll get it plausibly close. Better for everyone that way

who cares about premium laptops? those are going to be just like high end desktop GPUs, small percentage of the market. Lower end will want something like polaris 11 (sub $1000 laptops)

 

[3DVagabond]

Says 2x of current "mainstream AMD GPUs." Fiji is not a mainstream GPU.

So, since you are asking the question and answering it, what is it?

Also you are saying GPU but it might be whole cards? Or it might not? Is the chip 2X as efficient? Or is the entire card? Because the entire card might be 2x while improving the GPU by 2.5?

 

[geoxile]

Lisa Su: Card
RTG: Chip?

I mean, power usage is more than the chip. You need to take board power into consideration, and that includes memory chips.

It seems that despite AMD's first showing of Polaris and rumors that Nvidia wasn't doing well in yields with Pascal, so far Pascal to Polaris looks just like last generation.

It might be the other way around. Remember Fury Nano got significant power savings from its optimized PCB design. RTG's 2.5x efficiency number may be reference card vs reference card.

 

[Glo.]

Lisa Su: Card
RTG: Chip?

I mean, power usage is more than the chip. You need to take board power into consideration, and that includes memory chips.

It seems that despite AMD's first showing of Polaris and rumors that Nvidia wasn't doing well in yields with Pascal, so far Pascal to Polaris looks just like last generation.

I think AMD already answered that question with their demo of small Polaris.

They did not demoed only chip but whole card.

 

[Qwertilot]

who cares about premium laptops? those are going to be just like high end desktop GPUs, small percentage of the market. Lower end will want something like polaris 11 (sub $1000 laptops)

But a very profitable end of the market, and AMD do need profit

It matters a lot for mainstream desktop GPUs too. A 10% gap say is fine, but the gap that Maxwell opened up was >> than that and basically forced AMD to compete a whole tier up in terms of TDP to get rough performance equality.

Non ideal for margins but NV preferred extra profits and AMD at least had the 290/x etc going 'spare' to do that with so it was an option of sorts.

AMD don't even have that option for this generation. They're starting small, and going bigger. If they're as far behind as with Maxwell they'd a chip running well over 200w to match the 1080 - the biggest Vega basically - at which point there would honestly be little point either developing or releasing it.

Thankfully there is every reason to think they probably have closed the gap at least a bit. It'll have been AMD's absolute top priority for this generation. If nothing else it basically life and death for their custom sillicon stuff like games consoles, various embedded APU stuff etc etc.

Conversly it doesn't(I think?) look like NV nudged things on so much more with Pascal, beyond the gains from the process. Their tock into Volta probably will move performance efficiency on again, but that's not for a while

 

[Slaughterem]

Mainstream is always 199-249$ bracket.

So R9 380 and 380X.

We are looking at 50 GFLOPs/Watt for Polaris.

The picture in the slide is either Nano or Fury x. I know that you may consider mainstream 380 380x due to price but remember AMD now considers fury as mainstream from their latest road map.

 

[antihelten]

At 200W you would get 10 TFLOPs of compute power from Polaris GPU. It would be faster than GTX 1080...

No it would not be faster than GTX 1080.

First of all a 2x improvement in perf/W would not be 50 GFLOPS/W, it would only be about 42 GFLOPS/W, since a 380X only does 21 GFLOPS/W (3973.1 GFLOPS, 190W TDP).

Secondly GCN generally has worse utilization than Maxwell. For instance the 390X and 980 are more or less similar in gaming performance, but the 390 has almost 20% higher processing power (5914 GFLOPS vs. 4981 GFLOPS).

So a 200W Polaris with a 2x improvement in perf/W over 380X, would be 8400 GFLOPS, which gaming performance wise would be equal to about 7100 GFLOPS for a Maxwell (and presumably the same for a Pascal card). The 1080 is roughly 9100 GFLOPS or about 28% more.

So all in all this hypothetical 200W Polaris GPU would be about equal to a Titan X, which isn't too bad honestly, especially if the price is good. By extension, a 150W Polaris would be about equal to a 980, and a 100W polaris about equal to a 380X.

If AMD manages to hit a 2.5x improvement instead of 2x, then they will basically be neck and neck with Pascal, efficiency wise. So fingers crossed I guess.

 

[tential]

This whole discussion is based on whether amd has their act together in marketing. Which no one has evidence to support.

Reading into marketing claims is a hilarious waste of time.

 

[maddie]

No it would not be faster than GTX 1080.

First of all a 2x improvement in perf/W would not be 50 GFLOPS/W, it would only be about 42 GFLOPS/W, since a 380X only does 21 GFLOPS/W (3973.1 GFLOPS, 190W TDP).

Secondly GCN generally has worse utilization than Maxwell. For instance the 390X and 980 are more or less similar in gaming performance, but the 390 has almost 20% higher processing power (5914 GFLOPS vs. 4981 GFLOPS).

So a 200W Polaris with a 2x improvement in perf/W over 380X, would be 8400 GFLOPS, which gaming performance wise would be equal to about 7100 GFLOPS for a Maxwell (and presumably the same for a Pascal card). The 1080 is roughly 9100 GFLOPS or about 28% more.

So all in all this hypothetical 200W Polaris GPU would be about equal to a Titan X, which isn't too bad honestly, especially if the price is good. By extension, a 150W Polaris would be about equal to a 980, and a 100W polaris about equal to a 380X.

If AMD manages to hit a 2.5x improvement instead of 2x, then they will basically be neck and neck with Pascal, efficiency wise. So fingers crossed I guess.

I'm thinking that is one of the targeted improvements with all of the new Polaris IP blocks. New command processor, shader cores, primitive discard unit, etc.

We can't use the past except to use it as the absolute performance floor. Your perf/W comparison will be the lowest. A 150W P10 could be around 6500 GFOPS in the old AMD tradition but with better utilization than in the past, equal, or dare I say, exceed the same 6500 GFLOPS 1070. it depends on how much the efficiency has improved.

Having said that, I can't see a 1080 match in any realistic scenario. 1070 however, could be fair game.

I'm also wondering if AMD is waiting out Nvidia for the 1070 tests to go live before displaying P10 details. This will be the short term battleground between the two of them. No other overlap should exist for now.

 

[selni]

It's just a marketing bullet point without context. Compared to what? At what power target? How does it scale? Looking like it won't be long for actual details now so just wait for that - you can't read much into this sort of thing.

 

[maddie]

This while discussion is based on whether amd has their act together in marketing. Which no one has evidence to support

I agree with your assessment and can understand why, seeing that out of 9000 employees, 7000 are engineers. A survival measure not without it's costs.

Hopefully they can add more staff to round out the company if sales and profits improve.

By the way, what's happening to the polished Nvidia marketing machine. Did you see this video? https://www.youtube.com/watch?v=bNCfn4y8dBw


Derailing threads is not allowed here. This is about AMD, not Nvidia
Markfw900

 

[tential]

I agree with your assessment and can understand why, seeing that out of 9000 employees, 7000 are engineers. A survival measure not without it's costs.

Hopefully they can add more staff to round out the company if sales and profits improve.

By the way, what's happening to the polished Nvidia marketing machine. Did you see this video? https://www.youtube.com/watch?v=bNCfn4y8dBw

18k views meh. Nvidia gonna be fine. You think people are going to stop founding the 1080 because of some small confusion? No, they'll defend it. I'd give you evidence but it's so abundantly clear and I don't want to put any members in an awkward spot. For supporting the founders edition.
Everyone wants to found something, and if you want to found the 1080 God willing found away. Without founders how would a card ever even be produced?

Like I've said, expect Nvidia to dynamically price better. Did you think Uber, airbnb, etc. Pricing is coming out of thin air. Dynamic pricing is the most efficient way of capturing value all along the demand curve instead of at a fixed price point.

Amd, they need a marketing department founders card edition for vega. "please help us fund a real marketing department!" might work with some fans.....

Again, stay on topic please, this is not about marketing. -Shmee