AMD Official Statement - Polaris is NOT delayed

[maddie]

How is my calculation wrong?

If 1070 and Polaris 10 are equal in performance, it means a 314mm2 chip (35% larger) is only as fast as a 232mm2 chip (Polaris 10). This means AMD would have pulled off a 35% increase in performance /mm2 over Polaris 10. I don't believe this is realistic.

> If we use die sizes, Polaris 10 loses
> If we use 2.5X perf/watt claims from AMD, Polaris 10 still loses

There is no realistic math that can get us to the point where a Polaris 10 chip is as fast as a 1070. In all my calculations, I also gave AMD the best possible chance and I low-balled 1070's performance by equating it to Titan X.

This is why I think you're wrong.

By the way, I'm not saying that P10 HAS to equal a 1070, just that it's not as ridiculous a situation as you make it out to be.

Die size:
The 1080 is 9.0 TFlops and the 1070 is 6.5 TFlops, suggesting that the 1070 is equal to 72% of the full die, or maybe 80% if lower clocks in the 1070. Remember that shaders are fused off and are non functional silicon, so can't be used as providing any use. Roughly a 227mm^2 to 251mm^2 GP104 full equivalent in functional silicon

This would suggest that a full Polaris 10 should be equivalent in perf/mm^2 to GP104 if it is equal in performance to the 1070, since Samsung is roughly 10% denser so 232 Samsung = 255 TSMC.

We get the working silicon for both P10 and the 1070 as being very close, so perf/mm^2 would be similar if they had equal performance.


2.5X perf/watt: [this is what I wrote]
I notice you took the peak gaming power as the value for 290X. Do you think this is equivalent to the 1070 150TDP value? I bolded the relative sentence.

From TPU charts:

• Average: Metro: Last Light at 1920x1080 because it is representative of a typical gaming power draw. The average of all readings (12 per second) while the benchmark was rendering (no title/loading screen) is used. In order to heat up the card, the benchmark is run once without measuring power consumption.
• Peak: Metro: We use Last Light at 1920x1080 as it produces power draw typical to gaming. The highest single reading during the test is used.

If the average gaming power of 258W is used, then P10 best case in your calcs become 103W and you end up with this value instead.

Polaris 10 = 103W x 1.39% => New 143W TDP

edit:
In that same TPU chart we see the peak gaming power of the 970 as 191W, and we all know this is way above the official Nvidia value, leading me to believe that the 150W for Nvidia 1070 is the average gaming value.

 

[DiogoDX]

Good and updated power usage test:

http://www.lesnumeriques.com/carte-...ectrique-plus-40-cartes-graphiques-a2733.html

 

[tential]

Good and updated power usage test:

http://www.lesnumeriques.com/carte-...ectrique-plus-40-cartes-graphiques-a2733.html

Am. I reading that right? Are some games requiring more power from. Nvidia than amd? What's going on there?

Is the amd card not being fully utilized or something?

 

[Schmide]

If 1070 and Polaris 10 are equal in performance, it means a 314mm2 chip (35% larger) is only as fast as a 232mm2 chip (Polaris 10). This means AMD would have pulled off a 35% increase in performance /mm2 over Polaris 10. I don't believe this is realistic.

Since it's area

314^2 = 98,596
232^2 = 53,824

diff = 44,772

44,772 / 53,824 = 83.2 % larger

That's without the 14-16 node diff. Which would equate to what 10-15%?

So you could almost say 2x the size?

 

[itsmydamnation]

Good and updated power usage test:

http://www.lesnumeriques.com/carte-...ectrique-plus-40-cartes-graphiques-a2733.html

that's actually really nice. be really interesting to see polaris/vega and GP10* once the are all out.

 

[Abwx]

Since it's area

314^2 = 98,596
232^2 = 53,824

314mm2 is already a product...

The ratio of the areas is simply 314/232 = 1.35, that is 35% more area, or more mm2 if you prefer...

Or are you implying that 314mm2 of Nvidia can be filled by 232 x 1.83 = 425mm2 from AMD.???..

I guess that you posted before drinking your coffee...

 

[SimianR]

People expecting Polaris 10 to compete with the 1070 are setting themselves up to be disappointed. Why are we setting the bar so high for what is widely expected to be an inexpensive, low power, mainstream GPU? If it does end up coming anywhere near the 1070 then great, but why not wait for reviews before hyping this thing to death? The only nice thing about NVIDIA launching first for AMD is that it lets them price Polaris 10 accordingly.

 

[Buttercream]

Since it's area

314^2 = 98,596
232^2 = 53,824

diff = 44,772

44,772 / 53,824 = 83.2 % larger

That's without the 14-16 node diff. Which would equate to what 10-15%?

So you could almost say 2x the size?

314mm² is the area, mm² is a unit of area.

So RussianSensation was correct, GP104 is 35% larger than Polaris 10.

 

[Schmide]

I get it. lol

 

[MrTeal]

314mm2 is already a product...

The ratio of the areas is simply 314/232 = 1.35, that is 35% more area, or more mm2 if you prefer...

Or are you implying that 314mm2 of Nvidia can be filled by 232 x 1.83 = 425mm2 from AMD.???..

I guess that you posted before drinking your coffee...

As sad as that is. If nVidia or AMD would launch a 232mm x 232mm GPU, I would buy one even though it would take a whole wafer to produce and would likely cost as much as a midsized car.

JHH could even talk about it being an irresponsible amount of power without sounding like a preening tool.

 

[Azix]

People expecting Polaris 10 to compete with the 1070 are setting themselves up to be disappointed. Why are we setting the bar so high for what is widely expected to be an inexpensive, low power, mainstream GPU? If it does end up coming anywhere near the 1070 then great, but why not wait for reviews before hyping this thing to death? The only nice thing about NVIDIA launching first for AMD is that it lets them price Polaris 10 accordingly.

Its not out of reach at all. Its a 6.5Tflop part. AMD was already near 6 with hawaii. If full polaris has 2560 cores then its almost guaranteed to pass 1070 without the dx11 bottlenecks and with the dx12 advantages. If the clocks arent too bad it could reach close to 1080.

1070 is very much at risk depending on how good it is and what AMD can do.

But this is AMD, maybe they have their own weird plans and difficulties preventing that.

 

[railven]

That's two people now saying Polaris 10 is going to be within reach of GTX 1080.

Good luck with those predictions.

 

[boozzer]

That's two people now saying Polaris 10 is going to be within reach of GTX 1080.

Good luck with those predictions.

that is just irresponsible hype.

 

[airfathaaaaa]

we dont know nothing about them for all we know p10 might be faster than 1080
or p10 might be slower than even 1070

 

[krumme]

Look at the rabadged 400m series. I think it shows where p10 and p11 slots in. Whats your take on this perspective and what does it show?

 

[airfathaaaaa]

Look at the rabadged 400m series. I think it shows where p10 and p11 slots in. Whats your take on this perspective and what does it show?

you are comparing a rebranded chip from 2012 to p10?

 

[krumme]

I think we will even see vega on mobile. Looks to me like they cut the die size just fine for this portfolio fitting a leaner and more mobile market.

 

[krumme]

you are comparing a rebranded chip from 2012 to p10?

http://www.anandtech.com/show/10322/amd-unveils-radeon-m400-series-for-notebooks

P10m is goin to slot in as m490 and 495 above the 2048 shader 485.
P11m is 475 and 480.
Imo show how powerfull even the slowest p11 is on mobile. Its a gigantic jump from prior gen.
Nv 1080 will surely be much faster. But p10 is surely also much cheaper and can run on standard gddr5. Damn fine positioning imho as its probably difficult for gf to produce bigger dies now.

Vega will be new line perhaps new naming. I am pretty sure they will use hbm2 higher efficiency to introduce an expensive 150w mobile part that will doninate mobile as the fastest card.

 

[xpea]

we dont know nothing about them for all we know p10 might be faster than 1080 or p10 might be slower than even 1070

seriously?
what are the odds that it will happen? 0.1% ? 1% ? 10% ?



















IMHO 0%

 

[Krteq]

Yeah, It's impossible to achieve similar performance of GTX 1080 with aprox. 35% smaller die.

 

[Vaporizer]

As far as I know AMD will have at last roughly 4 additional chips above Polaris 10 (2 for vega 10 and 2 for vega 11) whereas NV will only have two chips for GP102.
If P10 already levels the 1070 than vega 10 cut should easily match 1080 and full vega 10 would compete with cut GP102.
Then there would be the two Vega 11 chips without competition. That is not very likley.

 

[coercitiv]

As far as I know AMD will have at last roughly 4 additional chips above Polaris 10 (2 for vega 10 and 2 for vega 11) whereas NV will only have two chips for GP102.
If P10 already levels the 1070 than vega 10 cut should easily match 1080 and full vega 10 would compete with cut GP102.
Then there would be the two Vega 11 chips without competition. That is not very likley.

Full Polaris 10 matching 1070 simply puts AMD and Nvidia on the same terms as far as perf/mm2 goes. Nothing more. How you managed to come up with Vega 11 having no competition is beyond me.

 

[airfathaaaaa]

seriously?
what are the odds that it will happen? 0.1% ? 1% ? 10% ?



















IMHO 0%

do you know anything we dont?i personally dont know for all i know it might be graphene based and run on 1 watt

 

[james1701]

It is well known that Polaris is going after mainstream. You are probably looking at 390 performance at around a 299.99 price tag.

 

[seitur]

It is well known that Polaris is going after mainstream. You are probably looking at 390 performance at around a 299.99 price tag.

+1

It advantage will be some architectural changes that may result in better performance in newer titles, low power consumption and thus heat generation&noise, support for 10-bit/newest codec hardware encoding/decoding, better FreeSync, ports, etc but overally it will be around 390 performance. Maybe 390X if we're lucky.

btw. anyone know what happened to TrueAudio? Is that chip thing still present in newer AMD GPUs or it was quickly axed and forgotten?