Now this could be big: R9 Nano

theGryphon

[H]ard|Gawd
Joined
Nov 21, 2011
Messages
1,295
I don't think anyone saw this coming. :eek:
Half the power of 290X, but more powerful? Very unlike AMD, lol. Could put Nvidia in trouble if this becomes the trend. Real competition will be welcome! :)

amdradeonrinano-1000x667.jpg
 
Oooooo, I'm looking this a card A LOT! I'm thinking new HTPC here!
 
According to recent news about upcoming AMD HBM memory architecture for gpu's it'll soon be possible to get such powerful cards in such small form factors. If you put memory directly inside the same package with the gpu and put IHS on it the whole pcb gets simplier and also it needs less power to transmit data between memory and gpu.

It's also possible that AMD will scale down the power consumption depending on available production process( at TSMC/Global Foundries?).

Other stuff worth noting is Framerate Target Control technology:

Framerate Target Control, which abbreviates to FRTC, is a feature that is useful for dealing with games that run at very high framerates. Rather than rendering a super-high framerate, but only displaying a handful of them, the GPU's clockspeeds are adjusted to deliver the performance that is necessary, and no more. This reduces power consumption, and therefore leads to a cooler and quieter gaming experience.

Which is something that might be a game changing feature for the AMD. Till now both companies solved this problem at the end of the rendering process with vsync while constantly rendering more than enough frames and wasting energy. If it really works, that might mean getting low temperatures on notebooks and long battery life while playing 30fps on the go and having constant 60fps with cool and quiet high end pc.
 
The Fury X isn't all that longer either although it does have the AIO. It's all due to HBM not taking a bunch of pcb space.
 
The Fury X isn't all that longer either although it does have the AIO. It's all due to HBM not taking a bunch of pcb space.

Required PCB space is not just about how much space the components take. There is the cooling consideration due to power output. Full Fiji may fit on this board, but no way you can maintain reasonable temps on the GPU, let alone the VRM.

So, this card is bound to be a cut-down version of the Fiji die. From some reports, it will still have the 4GB ram with full HBM bandwidth, but it will most likely have a smaller GPU. I surmise it's half the full Fiji.
 
Half the power of 290X, but more powerful?
Is there anymore info on this? IMO this is more important for SSF then the actual size of the card.

At the same time I don't understand why a power efficient card would need a AIO cooler.
 
According to recent news about upcoming AMD HBM memory architecture for gpu's it'll soon be possible to get such powerful cards in such small form factors. If you put memory directly inside the same package with the gpu and put IHS on it the whole pcb gets simplier and also it needs less power to transmit data between memory and gpu.

It's also possible that AMD will scale down the power consumption depending on available production process( at TSMC/Global Foundries?).

Other stuff worth noting is Framerate Target Control technology:



Which is something that might be a game changing feature for the AMD. Till now both companies solved this problem at the end of the rendering process with vsync while constantly rendering more than enough frames and wasting energy. If it really works, that might mean getting low temperatures on notebooks and long battery life while playing 30fps on the go and having constant 60fps with cool and quiet high end pc.

This sounds very interesting, but I'm quite sceptical. Having memory inside the same package is very expensive if it is directly integrated with the rest of the processor, that's why it was not done before. Everything you say is true, but I wonder how AMD solved the cost issue of this. Regular Caches on CPUs, even the most ridiculously expensive ones, are still only a few Megabytes large, I'd really like to know how they can fit 1-4GB on a chip and still get reasonable pricing.

Is there anymore info on this? IMO this is more important for SSF then the actual size of the card.

At the same time I don't understand why a power efficient card would need a AIO cooler.

I'd like to know that, too. This is the next step in SFF GPUs. Especially interesting if AMD can get back in the game.

This one doesn't have an AIO cooler, it uses a single fan, at least on the pictures it does.
 
They showed the Fury X with a watercooler, anyone know if third party manufacturers will make models that are all air? (Want to put one in an A4-SFX)
I'm hoping MSI and Gigabyte (something that's not so extremely long) make air cooled ones.
 
They showed the Fury X with a watercooler, anyone know if third party manufacturers will make models that are all air? (Want to put one in an A4-SFX)
I'm hoping MSI and Gigabyte (something that's not so extremely long) make air cooled ones.

The regular Fury is air-cooled and comes out July 14th.
 
That's sad, even AMD admits that Intel is the way to go on the CPU for SFF:

http://i.imgur.com/fAS9j2f.jpg[/IM]
[IMG]http://i.imgur.com/pr1k08r.jpg[/MG]

Props to djotter for noting that the board is an ASRock Z97E-ITX/AC
[URL]http://www.pcper.com/news/General-Tech/AMDs-Small-Form-Factor-Project-Quantum-PC-Dual-Fiji-Powerhouse[/URL][/QUOTE]

Ugh, who cares for the purposes of this thread?

On topic, any mention of price for the the nano? I know they gave prices for the fury x and fury.
 
Why do I keep hearing that Project Quantum will be powered by two R9 GPUs? Will it use the dual-GPU card? Because there is clearly only one GPU inside if you look at the renderings in the video.
 
Why do I keep hearing that Project Quantum will be powered by two R9 GPUs? Will it use the dual-GPU card? Because there is clearly only one GPU inside if you look at the renderings in the video.

It uses a dual fiji card. This is why the image of the dual fiji card is floating around. AMD also more or less confirmed that the dual fiji card in Project Quantum is 9" long. Hopefully the retail version is the same.
 
Last edited:
This sounds very interesting, but I'm quite sceptical. Having memory inside the same package is very expensive if it is directly integrated with the rest of the processor, that's why it was not done before. Everything you say is true, but I wonder how AMD solved the cost issue of this. Regular Caches on CPUs, even the most ridiculously expensive ones, are still only a few Megabytes large, I'd really like to know how they can fit 1-4GB on a chip and still get reasonable pricing.

I didn't say the memory will be integrated into the chip/die of the gpu. I said PACKAGE, as here:

hynix_hbm.jpg


Also random access memory is not the same thing as on-die fast cache/registers.

There are some SoC units that have the memory inside the package.

AMD already has the experience with this kind of thing with Radeon E8860:

2550.jpg

zSAE886-2GD5-MA-2.jpg
 
Last edited:
Why is the 8 pin custom angled on that Intel board?

To save vertical space. There seems to be a full-cover waterblock for the board inside Quantum, so the connectors need to be as low profile as possible.

I didn't say the memory will be integrated into the chip/die of the gpu. I said PACKAGE, as here:

http://www.purepc.pl/files/Image/news/2015/02/hynix_hbm.jpg

Also random access memory is not the same thing as on-die fast cache/registers.

There are some SoC units that have the memory inside the package.

AMD already has the experience with this kind of thing with Radeon E8860:

http://www.techpowerup.com/gpudb/images/2550.jpg
http://www.sparkleglobal.com/mw/cuf...ae886-2gd5-ma/thumb_big/zSAE886-2GD5-MA-2.jpg

I see, I seem to have misunderstood that. I thought that this may not have been possible with 28nm technology, but apparently, it damn well was. Quite curious why it took them so long if that approach was something they've already had experience with.
 
Quite curious why it took them so long if that approach was something they've already had experience with.

I think it wasn't good idea to bring the new ideology while still using old type of the memory - that wasn't efficient to change production process for old/current tech for the mainstream
 
Why do I keep hearing that Project Quantum will be powered by two R9 GPUs? Will it use the dual-GPU card? Because there is clearly only one GPU inside if you look at the renderings in the video.

THIS;
small_fiji-x2-2.jpg

Dual Fury X in one PCB = CF
 
Other stuff worth noting is Framerate Target Control technology:
Framerate Target Control, which abbreviates to FRTC, is a feature that is useful for dealing with games that run at very high framerates. Rather than rendering a super-high framerate, but only displaying a handful of them, the GPU's clockspeeds are adjusted to deliver the performance that is necessary, and no more. This reduces power consumption, and therefore leads to a cooler and quieter gaming experience.
Which is something that might be a game changing feature for the AMD. Till now both companies solved this problem at the end of the rendering process with vsync while constantly rendering more than enough frames and wasting energy. If it really works, that might mean getting low temperatures on notebooks and long battery life while playing 30fps on the go and having constant 60fps with cool and quiet high end pc.
This sounds very much like the Battery Boost technique Nvidia rolled out last year for laptops. Moving this to the entire desktop line as well as laptops would definitely save on power and noise for low/mid workloads more effectively than the normal FPS limiter.
 
Back
Top