GPU Now this could be big: R9 Nano!

[Phuncz]

Yes I would agree, though as a case buyer, not designer. I can understand Necere's concern about non-popular design choices though.

The length of a GPU is the one of the most variable of dimensions across the GPU spectrum when most are dual-slot and most are within an inch of the PCIe full-height limit. But with the length ranging from 150mm to 300mm, I personally would have the power sockets on the side. It also seems better/easier for cable management and aesthetics.

 

[veryrarium]

I watched the video Vittra linked in another thread that shows an R9 Nano held in hand in slowly varying angles, and I could be mistaken but there seems to be no heatpipes buried under or going through the heatsink fins, so it probably adopts a vapor chamber design I guess? Hard to imagine the heatsink would be both heatpipes and VC free for a card requiring a 8pin PEG connector.

 

[iFreilicht]

Yes I would agree, though as a case buyer, not designer. I can understand Necere's concern about non-popular design choices though.

The length of a GPU is the one of the most variable of dimensions across the GPU spectrum when most are dual-slot and most are within an inch of the PCIe full-height limit. But with the length ranging from 150mm to 300mm, I personally would have the power sockets on the side. It also seems better/easier for cable management and aesthetics.

I think I'll have to disagree on the last point. Having the connectors on the front allows vastly different and in some cases better options when it comes to cable management, as they won't ever run in front of the GPU you might want to show off. The difference is demonstrated here:

I personally think that power connectors in that location have something unique to offer. Not that I'd ever make use of it

 

[Phuncz]

I think I'll have to disagree on the last point. Having the connectors on the front allows vastly different and in some cases better options when it comes to cable management, as they won't ever run in front of the GPU you might want to show off.

I think you've misunderstood me. For me the sides of the card are the where the video connectors are and the side opposite to that.
Top is where the SLI connector is and where most cards have their power connectors, bottom is where the PCIe x16 connector is.

Since a GPU doesn't have a "direction" like a car, I look at it like most people do:

So left is 'side', right is 'side', up is 'top', down is 'bottom', heatsink is 'front', backplate is 'back'. Seems logical to me, otherwise why call it backplate ?

So I agree with you, if your point is that PCIe power connectors on the right side are prefered for cable management.

 

[iFreilicht]

I think you've misunderstood me. For me the sides of the card are the where the video connectors are and the side opposite to that.
Top is where the SLI connector is and where most cards have their power connectors, bottom is where the PCIe x16 connector is.

Since a GPU doesn't have a "direction" like a car, I look at it like most people do:

So left is 'side', right is 'side', up is 'top', down is 'bottom', heatsink is 'front', backplate is 'back'. Seems logical to me, otherwise why call it backplate ?

So I agree with you, if your point is that PCIe power connectors on the right side are prefered for cable management.

Oh I see, so were actually agreeing, but I didn't realise it I look at it like when it is installed in an ATX PC but it is quite confusing at times.

 

[Vittra]

I watched the video Vittra linked in another thread that shows an R9 Nano held in hand in slowly varying angles, and I could be mistaken but there seems to be no heatpipes buried under or going through the heatsink fins, so it probably adopts a vapor chamber design I guess? Hard to imagine the heatsink would be both heatpipes and VC free for a card requiring a 8pin PEG connector.

Some sites are reporting it as a vapour chamber as well, likely based on the same assumption, but I don't recall AMD ever stating it as such anywhere - I don't recall mention of it in E3, or any subsequent videos involving AMD/Richard Huddy.

 

[Phuncz]

Wasn't the 290(X) reference card also equiped with a vapor chamber cooler, albeit underspecced ? It seems reasonable to assume they have done so again, hopefully with a better performing and more quiet cooler.

 

[Vittra]

Yeah, the entire 290(x) assembly seems to have been taken from the 7970, which proved to be inadequate.

The 980 reference uses a heatpipe design, despite superficially looking like it's brethren that uses a vapour chamber - until you remove it from the PCB and look at it underneath. AMD could be using a similar design.

 

[veryrarium]

I like Fury X and Fury X2 for their almost complete absence of electrolytic caps. The only two e-caps present on X2 are very short and placed along the bottom edge. AMD hasn't revealed the bare PCB of R9 Nano yet, but if Fury X and Fury X2 are any indication Nano as well will likely share a similar characteristic, which I believe greatly benefits the 2-slot card heatsink design over such a small PCB, especially if it contains a horizonal vapor chamber across its base. On the other hand, the PCB is too short for a centrifugal fan and a heatsink to be side by side without extending the length like the GTX 670/760/970 reference, and though I personally like AMD's decision to keep the total card length short by going with an axial fan, I believe a heatsink with a horizontal vapor chamber is better combined with a centrifugal fan than an axial fan for a 2-slot card (and a heatsink with heatpipes better with an axial fan), so I'm kind of torn.

 

[jeshikat]

It could by launching in just a few days:

http://www.techpowerup.com/215507/amd-radeon-r9-nano-launch-date-revealed.html

 

[Phuncz]

I'm curious how it performs, this is going to be good times for SFF people. Suddenly a lot of mITX cases could become a lot more potent. But I hope the cooling is sorted properly this time.

 

[veryrarium]

My speculation on the heatsink was wrong, it's a heatpipe design (I don't know if a vapor chamber lies underneath HPs though). The axial fan partially makes sense now with air path under the heatsink on the front side of the card. Edit: I guess it's cooling capacity resembles that of Sapphire's 285 Compact which are equipped with four heatpipes.

Source: http://iyd.kr/717

 

[Vittra]

Paper launch. Full Fiji XT core. $650 USD. Availability Sept. 10:

http://www.hardwarecanucks.com/forum/hardware-canucks-reviews/70148-amd-r9-nano-preview.html
http://www.hardocp.com/article/2015/08/27/amd_radeon_r9_nano_video_card_paper_launch

 

[EdZ]

I'm going to reserve the confetti until benchmarks are available. The Fury XT achieved it's perf/watt by keeping junction temperatures low with the 'overspecced' liquid cooler. With the compact air cooler, the R9 Nano is going to have much higher junction temperatures, so clock-for-clock will be less efficient than the Fury XT.
Either the Nano is going to throttle very aggressively and rarely hit its rated 1000MHz core boost clock, or AMD decided to keep a batch of unparalleled quality Fiji chips for the Nano and sell them at the same price as the Fury XT, rather than using those chips to squeeze even more performance from the Fury XT design in a higher margin product.

 

[Phuncz]

It could also be that the Nano has better binned chips that allow a lower voltage for the 5% lower clock. It might explain the staggered launch: Fury X - Fiji mid-end chip quality
Fury - Fiji low-end chip quality
Fury Nano - Fiji high-end chip quality

No doubt Anandtech is going to look at the performance-over-time but that will probably be for the end of September. Given the Fury X availability, that doesn't seem like an issue.

 

[Vittra]

This card appears as if it will be a decent showcase of what the future holds for Pascal/Arctic Islands, and as a result the implications for SFF.

In other words, things are looking pretty good.

Right now though, $650 USD is going to be a really tough sell for a majority of people, regardless of what they are attempting to accomplish.

 

[iFreilicht]

I was extremely skeptical when I heard about the price tag, but I think if the card can beat the 970 by a long margin, some people might find it to be worth it.
We'll see how it performs in benchmarks, but right now I'm imagining something close to 980 performance, maybe even better, and that's not possible at the moment in such a small form factor. Please AMD, give us the power our minuscule PCs deserve!

 

[craigbru]

Although the price is a bit of a disappointment, I've already started planning a build around the Nano. After looking at the revised OSIDIAS plans I've been working on, I thought to myself, "I must go smaller!"

 

[esplin2966]

I also think this opens up a lot of possibilities for case manufacturers. For the first time, there is a SFF GPU that can be considered "top of the line". Others will follow.

 

[MJVR1]

Due to the fact that the chips in these cards are basically AMDs best Fiji chips, I wouldn't be surprised that these cards will be hard to find. Though honestly, I would go Fury X over the Nano almost every day of the week. Unless you want to fit a Fiji chip into a case like the S3/S4 or some lian li cases that REQUIRE short cards, the Fury X will perform better in nearly all aspects. That doesn't mean this card doesn't have its place until he market. It still does, but it's a very small corner imo.