AMD Confirms Radeon R9 Nano Launching In August

[BiG StroOnZ]

 

AMD has spilled the beans on at least one aspect of the R9 Nano The product would be 19cm in length, offers 2x the perf of the 290X at half the power 2X the performance per watt and 2X the performance density of the 290X.

 

AMD confirmed this info during its quarter result conference call. 

 

AMD CEO Lisa Su -

 

“Fury just launched, actually this week, and we will be launching Nano in the August timeframe.”

 

It will be the 3rd product deriving from the Fiji range GPUs and follows a mITX form factor. The spec for this product have not been released just yet but expect, howwever the it'll be a cutdown Fiji GPU, the original design has  64 ROPs, 256 texture mapping units 64 GCN units with 4096 stream processors. 

 

We've speculated on how much lower clocks would need to be to meet the 175W target with full Fiji silicon, and it's going to be significant. The air coolers we've seen on the Fury (non-X) cards to date have extended well beyond the PCB, and the Nano is a mini-ITX form factor design.

 

Regardless of where the final GPU and memory clock numbers are I think it's safe to assume there won't be much (if any) overclocking headroom. Then again, of the card does have higher performance than the 290X in a mini ITX package at 175W, I don't think OC headroom will be a drawback. I guess we'll have to keep waiting for more information on the official specs before the end of August.

 

 

Source 1: http://www.guru3d.com/news-story/amd-confirms-radeon-r9-nano-launching-in-august.html

Source 2: http://www.pcper.com/news/Graphics-Cards/AMD-Confirms-August-Availability-Radeon-R9-Nano

 

I noticed although there has been a decent amount of interest in the Fury lineup, I think the card out of all of these that has taken the most interest is the Nano. While there haven't been any solid leaks or information on the specs of this card, we do have a concrete date now of release sometime in August. Which means we might get some specs closer to the end of July from leaks. Some have speculated it will be a full fledged Fiji chip, but of course if this is the case clockspeeds will have to be seriously reduced in order to meet the 175W power target. But I think as long as the card lives up to being more powerful than a 290X at a 175W TDP with an mITX design it could be one nice card for small form factor builds.

[Humbug]

It's not 2x the performance of the 290x. That would be crazy.

I believe what they claimed was 2x the perf / watt.

 

hopefully HBM supply issues will get sorted and more Fiji cards will hit the market. They are selling out too fast atm.

[AlwaysFSX]

It's not 2x the performance of the 290x. That would be crazy.

I believe what they claimed was 2x the perf / watt.

 

hopefully HBM supply issues will get sorted and more Fiji cards will hit the market. They are selling out too fast atm.

Yeah, it's not like AMD suddenly figured out energy efficiency overnight.

[BiG StroOnZ]

It's not 2x the performance of the 290x. That would be crazy.

I believe what they claimed was 2x the perf / watt.

 

hopefully HBM supply issues will get sorted and more Fiji cards will hit the market. They are selling out too fast atm.

 

I'm not really sure, Lisa Su herself said that it will be significantly faster than the 290X:

 

Lisa Su has said that the six-inch card will offer “significantly more performance than the Radeon R9 290X”

 

http://www.pcworld.com/article/2949623/components-graphics/amds-incredibly-small-yet-insanely-powerful-radeon-r9-nano-is-launching-in-august.html

 

While that may not add up to 2x the performance of a 290X, she did say significantly.

[iamdarkyoshi]

That's adorable!

[BiG StroOnZ]

Here's some maths:

 

Since we know that performance per watt is FP32/TDP, we can go ahead and extrapolate the power efficiency of the R9 290X.
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.

 

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

[79 Potatoes]

 

So in terms of FP Performance it's "close" to the Fury X?(8,6 TFLOPS)

Anyway I'll sit here till the review samples are given, I want to know the real power consumption, Including maximum peaks given that performance for a little case mod I've in my mind...

[IrshaadH]

man, this card sounds exciting.  I really hope it delivers

[BiG StroOnZ]

 

So in terms of FP Performance it's "close" to the Fury X?(8,6 TFLOPS)

Anyway I'll sit here till the review samples are given, I want to know the real power consumption, Including maximum peaks given that performance for a little case mod I've in my mind...

 

 

According to the math it should offer more compute performance than the Fury (non-X) since that is rated at 7.2TFLOPs, but be under the Fury X at 8.6TFLOPs so this should be somewhere in between. Which still, based on the Fury seems kind of hard to believe is possible considering the size of the Nano and the size of the cooler on the Nano. But hey, maybe they can pull it off.

[WereCat]

According to the math it should offer more compute performance than the Fury (non-X) since that is rated at 7.2TFLOPs, but be under the Fury X at 8.6TFLOPs so this should be somewhere in between. Which still, based on the Fury seems kind of hard to believe is possible considering the size of the Nano and the size of the cooler on the Nano. But hey, maybe they can pull it off.

Those will most likely be the best binned chips with the lowest leakage and it will be definitely clocked bellow 1000MHz on core. Hence the 175W TBP.

[Spoonfull]

Here's some maths:

 

Since we know that performance per watt is FP32/TDP, we can go ahead and extrapolate the power efficiency of the R9 290X.

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.

 

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

So its a tiny R9 Fury?

[BiG StroOnZ]

 

Those will most likely be the best binned chips with the lowest leakage and it will be definitely clocked bellow 1000MHz on core. Hence the 175W TBP.

 

Yes, but realize that the Fury is cut down and has lower clocks but still has the same 275W TDP as the Fury X. Therefore, as many current leaks suggest, that the Nano will be a full Fiji core like the Fury X. In order for them to reach a 175W TDP the clock speeds have to be reduced greatly. Which brings back the question, based on the math of 2x perf per watt of the 290X equaling 7.84TFLOPs how is that possible, when the Fury is only at 7.2TFLOPs but is cut down, still has reduced clock speeds, and maintains the same 275W TDP as the Fury X. 

 

So its a tiny R9 Fury?

 

Basically, significantly more powerful than a 290X but how much that actually equates to is currently unknown.

[Misanthrope]

Seeing the temps and the mammoth coolers on the Fury cards I think that this will have to be significantly cut down and not just a little bit, so I think this will sit between the 380 and the 390 in terms of performance. It's still about as powerful as you can get on this size yet nothing that really revolutionary since we already have 970s on this size and this will only match it or slightly edge it.

[Spoonfull]

Seeing the temps and the mammoth coolers on the Fury cards I think that this will have to be significantly cut down and not just a little bit, so I think this will sit between the 380 and the 390 in terms of performance. It's still about as powerful as you can get on this size yet nothing that really revolutionary since we already have 970s on this size and this will only match it or slightly edge it.

So, a 370 dollar price point perhaps?

[Misanthrope]

So, a 370 dollar price point perhaps?

 

That'd be the smart move not to cannibalize the 390 too much, but hopefully they'll see reason and drop the 390 and 390x a little bit but we'll have to wait and see, I doubt a price drop will come this early in the cycle.

[BiG StroOnZ]

Seeing the temps and the mammoth coolers on the Fury cards I think that this will have to be significantly cut down and not just a little bit, so I think this will sit between the 380 and the 390 in terms of performance. It's still about as powerful as you can get on this size yet nothing that really revolutionary since we already have 970s on this size and this will only match it or slightly edge it.

 

Only problem with it sitting between a 380 and a 390 or matching or slightly edging a 970 is that would mean:

 

A.) It doesn't have 2x performance per watt of a 290X

B.) It is not significantly faster than a 290X (which is what Lisa Su stated)

[WereCat]

 

 

 

Yes, but realize that the Fury is cut down and has lower clocks but still has the same 275W TDP as the Fury X. Therefore, as many current leaks suggest, that the Nano will be a full Fiji core like the Fury X. In order for them to reach a 175W TDP the clock speeds have to be reduced greatly. Which brings back the question, based on the math of 2x perf per watt of the 290X equaling 7.84TFLOPs how is that possible, when the Fury is only at 7.2TFLOPs but is cut down, still has reduced clock speeds, and maintains the same 275W TDP as the Fury X. 

 

When you lower/increase clocks the performance goes down/up lineary while power consumption goes down/up exponentialy.

[BiG StroOnZ]

When you lower/increase clocks the performance goes down/up lineary while power consumption goes down/up exponentialy.

 

Yes, but in order to achieve 2x the perf of a 290X at 7.84TFLOPs it seems like quite a stretch based on the fact that they have to lower clock speeds enough to achieve a 175W power consumption. Meaning, either what has been said so far about the Nano is exaggerated or they are achieving some magic here with the Nano.

[marldorthegreat]

Yeah, it's not like AMD suddenly figured out energy efficiency overnight.

Yeah, but the fury x is on par with the efficiency of maxwell. the fury x (At stock) consumes 275 watts, the 980ti consumes 250 watts, they are both about the same in terms of raw performance, so AMD is not that far behind nvidia at all. 

[Deletive]

Yeah, but the fury x is on par with the efficiency of maxwell. the fury x (At stock) consumes 275 watts, the 980ti consumes 250 watts, they are both about the same in terms of raw performance, so AMD is not that far behind nvidia at all. 

What I want to know is if it scales as well as Maxwell does, which would be the real question. 

[marldorthegreat]

What I want to know is if it scales as well as Maxwell does, which would be the real question. 

If the rumours are correct, toms hardware made the fury x consume 175 watts with only a 3fps dip in performance. 

[Dabombinable]

If the rumours are correct, toms hardware made the fury x consume 175 watts with only a 3fps dip in performance. 

That's pretty damn good. Maybe AMD should have started pushing Tonga sooner (Fiji is Tonga all beefed up). I just realised too that the IPC of AMD's last 2 generations of GPU (aka Hawaii and Tonga) is higher than that of Maxwell.

[TorqueS]

That's pretty damn good. Maybe AMD should have started pushing Tonga sooner (Fiji is Tonga all beefed up). I just realised too that the IPC of AMD's last 2 generations of GPU (aka Hawaii and Tonga) is higher than that of Maxwell.

I think that was kind of obvious, since we know that OCing AMD GPUs with like 50Mhz, is similar to 100 on Maxwell (so better scaling there).

Still...R9 Nano being 175W and 3FPS lower than Fury X???

Highly unlikely...what would be the purpose of the R9 Fury then(yeah, I know that it is a cut-down version of the Fiji core, but still)?

[Popingheads]

Yes, but in order to achieve 2x the perf of a 290X at 7.84TFLOPs it seems like quite a stretch based on the fact that they have to lower clock speeds enough to achieve a 175W power consumption. Meaning, either what has been said so far about the Nano is exaggerated or they are achieving some magic here with the Nano.

 

I have heard rumors that the Nano actually won't be a cut down chip, which would significantly help improve its performance per watt, since they could lower clock rates (and voltages) even further while still keeping good performance.

[Dabombinable]

I have heard rumors that the Nano actually won't be a cut down chip, which would significantly help improve its performance per watt, since they could lower clock rates (and voltages) even further while still keeping good performance.

So a reference mini ITX card?