Some math on Fury Nano performance

[Tiber1337]

So ! WCCFTECH ! did some math on the performance of the Fury Nano, its supposed to be darn strong by their calculations.

 

My take is that the Nano is going to be somewhere between r9 380 and r9 390 , since there's a big gap in price there and a card would fit nicely. BUT these calculations show a different story.

 

Here ya go :

 

http://wccftech.com/fast-amd-radeon-r9-nano-find/

[NeatSquidYT]

Anyone else think that this is a pretty cool reference card?

[Unlikelyjoker]

WCCFTECH did get some things right so can we stop the bashing guys?

[CarlosRex]

So ! WCCFTECH ! did some math on the performance of the Fury Nano, its supposed to be darn strong by their calculations.

 

My take is that the Nano is going to be somewhere between r9 380 and r9 390 , since there's a big gap in price there and a card would fit nicely. BUT these calculations show a different story.

 

Here ya go :

 

http://wccftech.com/fast-amd-radeon-r9-nano-find/

AMD already said that this card will have "significantly more performance" than the R9 290X. So its going to target the GTX 980 and up.

The math is pretty sound, and puts the Nano at a competing level with the GTX 980, maybe a bit faster.

 

We know AMD’s defining performance per watt as the peak FP32 TFLOPS divided by the “typical board power”, which for the Nano we know is 175W

R9 290X’s peak FP32 = 5.6 TFLOPs, in other words 5600 GFLOPs, and its TDP is 250W.

Perf/W = 5600 GFLOPs/250W = 22.4 GFLOPs/W

We also know that the R9 Nano has 2X the perf/watt of the R9 290X.

Which means it’s 2X (5.6TFLOP/250W)

= 2X 22.4 GFLOPs/W

= 44.8 GFLOPs/W.

Thus the perf/watt rating of the R9 Nano is 44.8GFLOPs/W.

Incidentally we also have the TDP for the Nano, and that’s the last missing piece in the puzzle.

R9 Nano

Perf/watt = FP32 in GFLOPs (unknown) / TDP (175)

44.8 = FP32 (unknown) / 175

44.8 x 175 = FP32 (unknown)

44.8 x 175 = 7840 GFLOPs or 7.84 TFLOPs.

I should point out that it’s still debatable whether AMD can actually reach this 7.84 TFLOPs figure with the R9 Nano. It would still be an impressive feat if the Nano ends up close enough to the 7 TFLOPs mark. After considering additional data points that AMD shared with us about the Nano with regards to its performance density compared to the 290X, we reckon it will probably end up around he 7.3TFLOPs mark.

 

 

 

Remember that you can't compare AMD and Nvidia cards based on FP32. Nvidia makes up for compute with a stronger geometry engine, so Nvidia GPUs with lower tflops can still keep up with AMD cards with more tflops in games.

[Bsmith]

we will see once these cards get released,until then i will do my best to remain neutral, but damn.. if this is correct.

[ZetZet]

Why would nano be more efficient than Fury?

[Aytex]

titan x divided by 2

[stconquest]

Just, wow.

[Misanthrope]

So it will compete with the 970 itx versions? Not a bad place to be in fact.

[ImBleu]

That thing is so cute.

[stconquest]

So it will compete with the 970 itx versions? Not a bad place to be in fact.

 

It has the potential to destroy the 970 in any form factor... as far as I understand it.

 

Imagine performance slotted between the 980 and the 980Ti, priced at $500.... that is my "wow".

[Kinda Bottlenecked]

Why would nano be more efficient than Fury?

 

I think its mentioned that the nano will be more efficient than the 290x, therefore that assumption.

[BookerDeWitt]

Why would nano be more efficient than Fury?

Lower clock speeds, lower voltages, specially selected low leakage chips.

We already see this with mobile vs desktop graphics cards. Often having the exact same configuration but lower clock speeds and much better efficiency.

[Misanthrope]

Why would nano be more efficient than Fury?

 

Not saying it would be of course but if it ends up being the case there is an explanation: diminishing returns. You can push the power and clock speeds and everything to get to the 980ti level but if you scale back a bit the performance doesn't necessarily scales back in the same proportion because it took a lot of basically wasted efficiency to get to max performance. This is why technically and according to prices the 970 is actually more efficient for the money than the 980 yet they're based on the same chip.

 

Think of it as reverse SLI/Crossfire: If you remove one card from the equation you don't get 50% the speed actually quite a bit more than that due to the fact that a lot of performance is lost in the process of going up in dual card more which is never 100% scaleable.

[Lethal Seraph]

Reminded me of GoT 'Even small men can cast a bug shadow.'

[ZetZet]

Lower clock speeds, lower voltages.

so it will use less power, efficiency is not that though.

If it's the same card scaled down it will be as efficient as fury x is, just weaker.

 

Not saying it would be of course but if it ends up being the case there is an explanation: diminishing returns. You can push the power and clock speeds and everything to get to the 980ti level but if you scale back a bit the performance doesn't necessarily scales back in the same proportion because it took a lot of basically wasted efficiency to get to max performance. This is why technically and according to prices the 970 is actually more efficient for the money than the 980 yet they're based on the same chip.

 

Think of it as reverse SLI/Crossfire: If you remove one card from the equation you don't get 50% the speed actually quite a bit more than that due to the fact that a lot of performance is lost in the process of going up in dual card more which is never 100% scaleable.

that might be it, but speculating that it somehow becomes 25% more efficient is still strange to me.

[TERAFLOP]

Would you look at that, someone's talking about teraflops....

[BookerDeWitt]

 

so it will use less power, efficiency is not that though.

If it's the same card scaled down it will be as efficient as fury x is, just weaker.

 

that might be it, but speculating that it somehow becomes 25% more efficient is still strange to me.

 

It's not speculation AMD said Nano has 2X perf/watt as the 290X, fury X only has 1.5X.

 

Just like CPUs, when you clock the chips down and reduce the voltages the efficiency improves. The R9 285 for example is a 190W card, the mobile version with even more cores 2048 vs 1792 and only an 850mhz clock speed vs 918mhz is rated 125W.

http://www.notebookcheck.net/AMD-Radeon-R9-M295X.129043.0.html

 

[Grrizz]

WCCFTECH did get some things right so can we stop the bashing guys?

Throw enough darts at a board and eventually you'll hit the bull's-eye

[Misanthrope]

 

so it will use less power, efficiency is not that though.

If it's the same card scaled down it will be as efficient as fury x is, just weaker.

 

that might be it, but speculating that it somehow becomes 25% more efficient is still strange to me.

 

 

Well I do know that the nano will come later so they might also be optimistically taking in to account eventual driver optimization gains post release by the time the nano hits the streets. Or they might be full of shit either or.

[Alurit]

WCCFTECH did get some things right so can we stop the bashing guys?

wccftech is like firing a shotgun onto a pack of ducks, you're going to hit a lot of thing and miss a lot of thing

[BookerDeWitt]

wccftech is like firing a shotgun onto a pack of ducks, you're going to hit a lot of thing and miss a lot of thing

Well looks like they nailed every single duck with the fury specs.... gotta give it to em, they've got a mighty shotgun.

http://wccftech.com/amd-radeon-fury-x-specs-fiji/

[Mo5]

There's a lot of 'if's and marketing speech in this calculation

[BookerDeWitt]

There's a lot of 'if's and marketing speech in this calculation

You should read the fine print, no really.

http://i.imgur.com/xuWUyS7.png

[Tiber1337]

 

snip

 

I remember that around a month ago they said 3X performance/watt at a conference with HBM, but they only delivered 1.5x . CEOs exaggerate most of the time, perhaps what she meant was " significantly more performance density" than the 290x. 

 

I'm just speculating and I'm in no way saying that I am right. But considering prices , 390x is starting 430 (perphaps much more for other models) and Fury aircooled is 550$ , there's little  room left for a card in between, it would just make the 390x completely useless if u could get half the tdp and more performance for 50 bucks more