R9 Nano benchmarks surface

[Bloodyvalley]

It has been almost 2 months since we first heard about the existence of the R9 Nano in a rumour. Very little was known about the card then except the its name. Now however, we have pretty much all the relevant info on the Nano sized Fiji offering. We have another nano sized update today from DGLee over at IYD.KR. The pictures of the surprisingly tiny but very powerful GPU are here for everyone to see – as well as performance numbers (official).

A slide from the AMD Press Deck. @AMD Public Domain

Pictures of the AMD R9 Nano surface – official performance and efficiency numbers revealed

The R9  Fury X is tiny, at around 19 centimetres but the R9 Nano is well, nano, measuring in at only 15 centimetres. Its small form factor design would enable it to be fitted in the tiniest mini-iTX casings out there and its impressive efficiency would enable it to be remarkably powerful at the same time. The card rocks an allegedly full sized Fiji core with 4096 SPs. It is touted as being an efficiency monster with the same performance as the Radeon R9 290X but half the power draw. This is a remarkable card in any way you consider it and here is a good luck at the the shape of the card itself.

Pictures courtesy of DGLee @IYD.KR

The card features a tiny heatsink (by GPU standards) and a lone single axial cooler. If the chip is indeed the full Fiji die then I would assume it is going to be clocked low for it to be efficient enough (and not overwhelm the modest heat dissipation equipment). Now that I have mentioned that, I cant help but wonder what would happen if someone were to mod onto it an after-market heatsink with double or tipple the heat dissipation capacity. Ofcourse, AMD has probably used cherry picked chips in the Fury Nano, so I wouldn’t be surprised if its actually priced higher or near the Fury X (caution: opinion).

Its time for the benchmark and efficiency numbers provided by AMD. The benchmarking by AMD was conducted at rig rocking an i7-5960X (no CPU bottlenecking of any sort) and X99 setup with ample ram. The program used was the Unigine Heaven Benchmark at its 4K settings (3840×2160 resolution, Extreme present, 0xAA). Now we know the Fury Nano has a TDP of 175W and AMD has obligingly given the efficiency numbers for both cards.

Pictures courtesy of DGLee @IYD.KR

The R9 Fury Nano has an efficiency of 0.152 fps/Watt at the above settings in this 4K benchmark, whileas the R9 290X has an efficiency of 0.076 fps. Simply multiplying the efficiency by the TDP we arrive at the actual score of the R9 Nano: 26.6 frames per second at a 4K benchmark – not bad at all. Ofcourse its worth noting that running a benchmark in 4K fully exploits the HBM capability of the Nano, so the efficiency difference might not be as high at lower resolutions where the HBM factor does not come into play. These are after all the official numbers, so it would be exceedingly unwise to take them without a grain of salt. 

It is also worth noting that while the bandwidth of the R9 290X is 320GB/s it only has 2816 stream processors so the Bandwidth Per SP of the R9 290x is 0.1136 GB/s. On the other hand the Nano (if the rumors are true) has 4096 SPs with 512 GB/s bandwidth – the effective Bandwidth Per SP becomes 0.1250 GB/s. So when you take a look at the full picture, it isn’t actually that huge of a difference as one might think at first glance. In a way the Fiji Core is not benefiting as much as it could from the HBM due to the fact that it has such a large number of cores. Needless to say, HBM2 should be quite a step up from the HBM1 used here.

 

Perfect for LAN rigs

 

Source

[themaniac]

so its pretty much a fury x in a smaller form the is also more power efficient?

[Bloodyvalley]

so its pretty much a fury x in a smaller form the is also more power efficient?

pretty much the same as a 290X but with better efficiency and less power consumption

[NinjaTitan]

I think it has HBM2 memory

Nothing has HBM2 memory at this point in time...

[miagisan]

pretty much the same as a 290X but with better efficiency and less power consumption

HBM2 is not in production yet

[Bouzoo]

pretty much the same as a 290X but with better efficiency and less power consumption

 

also has HBM2

HBM is not out yet.

[alexyy]

It's a downclocked fury x but I'd love of see a waterblock for it and see what it could achieve.

[Bloodyvalley]

Nothing has HBM2 memory at this point in time...

 

 

HBM2 is not in production yet

yeah I missread

[Humbug]

Hmm so it's just an underclocked fury x.

I guess the underclock allows them to effectively cool it with just one fan.

[azurezeed]

I don't think that cooler will do a good job

[Deletive]

DGLee calculated that to be 24FPS for R9 Nano and 18.4FPS for R9 290X in Unigine Heaven 4K. hmmm, like 30% is pretty good

[The Benjamins]

Full FIJI, that is interesting. that could be a great card to water cool and OC.

[Virgule]

Full FIJI, that is interesting. that could be a great card to water cool and OC.

Exactly what I was thinking. AMD better release a good driver for HBM that unlocks overclocking potential and I just might get two of these with waterblocks.

[bcredeur97]

I think it's worth getting this over the fury... Waterblock and/or custom cooler may allow you to get up to Fury X clocks

[FakezZ]

I think they should have used the Fury's GPU for the Nano and have the Fury be the aircooled Fury X version. But oh well... logic I guess.

[Colonel_Gerdauf]

Nice card, but this is confusing me.

 

Assuming that it has the same specs as the Fury X;

 

How does the nano keep cool whereas the Fury X needed an AIO? To make a big deal out of AIOs and then suddenly trying to kill off the idea is not the smartest PR move I have seen.

If the nano can do the cooling well, then what exactly is the point of the Fury X? Not exactly a stopgap solution as the Titan X was, as they were announced on the same day.

How does it maintain those specs at an even smaller form factor? I am guessing that the VRMs have reduced in size.

[Notional]

Nice card, but this is confusing me.

 

Assuming that it has the same specs as the Fury X;

 

How does the nano keep cool whereas the Fury X needed an AIO? To make a big deal out of AIOs and then suddenly trying to kill off the idea is not the smartest PR move I have seen.

If the nano can do the cooling well, then what exactly is the point of the Fury X? Not exactly a stopgap solution as the Titan X was, as they were announced on the same day.

How does it maintain those specs at an even smaller form factor? I am guessing that the VRMs have reduced in size.

 

It runs at 850 MHz apparently (Fury X is like 1050 MHz or so). Sadly you cannot put a water block and OC it, as it only has 1 8 pin power connecter.

[The Benjamins]

Exactly what I was thinking. AMD better release a good driver for HBM that unlocks overclocking potential and I just might get two of these with waterblocks.

 

its not locked, just no one has figured out how to change the voltage. you can already change the GPU/MEM frequencies.

 

just wait for MSI afterburner to get support.

[carolkarine]

Exactly what I was thinking. AMD better release a good driver for HBM that unlocks overclocking potential and I just might get two of these with waterblocks.

 

that's not how it works. there is no driver unlocking of voltage control, the maker of msi afterburner needs to physically have a fury x sample that he can then work with to unlock it. unlike nvidia, AMD has never "locked the voltage" and they've never released an API that allows voltage control.

[ManWithBeard1990]

Even though it's clocked lower and the voltage probably went down with it, I imagine at 175 watts this might still get quite toasty and loud.

[carolkarine]

It runs at 850 MHz apparently (Fury X is like 1050 MHz or so). Sadly you cannot put a water block and OC it, as it only has 1 8 pin power connecter.

 

you could rape the connector like the 295x2

[Notional]

you could rape the connector like the 295x2

 

I guess it depends on the VRM of the card. Maybe beefing it up would help keep it cooler, but not sure. It's going to be very interesting to see though.

[Ellis_D]

I'm really curious about the price of this thing.  If it costs anywhere close to the Fury X, it's just not going to be worth it.  Most ITX cases are built to accommodate large cards so unless OEMs like Origin or Digital Storm make an ultra-tiny case specifically for the Nano, it makes no sense.

 

Now if it's priced lower between $450-$500, it might prove to be the overclocking card everyone will want to buy.

[wpirobotbuilder]

It's such a cute little card with the one fan.

[LucidMew]

I don't trust these numbers, even moreso than I normally don't trust numbers from AMD/NVidia. AMD skewed the numbers for FuryX, what's to say they're not doing the same thing here? I'm going to wait until it gets released and benchmarks from reviewers come out.