Tuesday, August 25th 2015
AMD Radeon R9 Nano Core Configuration Detailed
AMD's upcoming mini-ITX friendly graphics card, the Radeon R9 Nano, which boasts of a typical board power of just 175W, is not a heavily stripped-down R9 Fury X, as was expected. The card will feature the full complement of GCN compute units physically present on the "Fiji" silicon, and in terms of specifications, is better loaded than even the R9 Fury. Specifications sheet of the R9 Nano leaked to the web, revealing that the card will feature all 4,096 stream processors physically present on the chip, along with 256 TMUs, and 64 ROPs. It will feature 4 GB of memory across the chip's 4096-bit HBM interface.
In terms of clock speeds, the R9 Nano isn't too far behind the R9 Fury X on paper - its core is clocked up to 1000 MHz, with its memory ticking at 500 MHz (512 GB/s). So how does it get down to 175W typical board power, from the 275W of the R9 Fury X? It's theorized that AMD could be using an aggressive power/temperature based clock-speed throttle. The resulting performance is 5-10% higher than the Radeon R9 290X, while never breaching a power target. Korean tech blog DGLee posted pictures of an R9 Nano taken apart. Its PCB is smaller than even that of the R9 Fury X, and makes do with a slimmer 4+2 phase VRM, than the 6+2 phase VRM found on the R9 Fury X.
Sources:
VideoCardz, IYD.kr
In terms of clock speeds, the R9 Nano isn't too far behind the R9 Fury X on paper - its core is clocked up to 1000 MHz, with its memory ticking at 500 MHz (512 GB/s). So how does it get down to 175W typical board power, from the 275W of the R9 Fury X? It's theorized that AMD could be using an aggressive power/temperature based clock-speed throttle. The resulting performance is 5-10% higher than the Radeon R9 290X, while never breaching a power target. Korean tech blog DGLee posted pictures of an R9 Nano taken apart. Its PCB is smaller than even that of the R9 Fury X, and makes do with a slimmer 4+2 phase VRM, than the 6+2 phase VRM found on the R9 Fury X.
101 Comments on AMD Radeon R9 Nano Core Configuration Detailed
We don't know the price yet. At this point, I wouldn't be surprised if the price is close to Fury X's. Nano effectively is Fury X for small form factor computers (Steam Machines comes to mind).
And please, stop trotting out the 3.5GB BS. The GTX 970 can address all 4 gigabytes of the graphics memory it has; that makes it a 4GB graphics card. If you claim anything else, you're either ignorant, stupid, or a combination of both.
Keep crying those fanboy tears though. They taste delicious.
Pricing though...$500?
It still performs well and I'll defend that but Nvidia CLEARLY mislead consumers about 4gb, when the last 0.5gb can hinder performance, though as stated, very few scenarios get to that point.
I doubt the R9 Nano will have much overclocking at all. I just want to see this cooler and the card in action so we can understand how it works because having this full core seems a bit weird over just binning out some less than stellar example chips of the bunch.
Nvidia itself basically said that they sold a 3 GB card, with a memory structure whose last bits were only designed such that "...GTX 970 is a 4GB card. However, the upper 512MB of the additional 1GB is segmented and has reduced bandwidth. This is a good design because we were able to add an additional 1GB for GTX 970 and our software engineers can keep less frequently used data in the 512MB segment..." That article that quote comes from can be found here: www.gamespot.com/articles/nvidia-boss-responds-to-gtx-970-false-advertising-/1100-6425510/
If you are to be truly honest, that's the Nvidia team peddling BS. You could bend over backwards and accept their logic, but if you do so then the Fury Xis the absolute best card currently on the market. You just have to reduce your sample size to a few hand picked titles at 4K, where the 980ti is beaten.
What we are arguing is to have a consistent standard for judgement. Right now, you've got two options. Either both companies peddle whatever BS will move cards, or both companies are 100% honest because they can find at least one instance where their claims are true.
Nvidia, AMD, and Intel all say whatever they need to to move hardware. This is why being an early adopter sucks so hard. If you don't wait for reviews, you'll always be disappointed. Stating that one manufacturer, or another, is uniformly better is stupid. Neither is better than the other, only their currently offered products are better or worse when measured to one another.
Edit:
If the 3.5 GB memory thing is still an impasse, maybe you should review an article from a year ago, that is surprisingly still accurate today streamcomputing.eu/blog/2014-08-05/7-things-nvidia-doesnt-want-know/.
That's right, people doing actual coding work with OpenCL and CUDA are calling Nvidia on BS. Kinda seems like the people actually using GPUs for stuff other than gaming recognize that AMD may not be doing well, but it's because of their marketing and not actual performance. If the AMD marketing was half as slimy as Nvidia they'd be claiming Fury X cured cancer, because it can be used for BOINC and the like.
Perhaps you're buying pre-clocked cards because 30% doesn't seem rare to me at all. Now 50% that's rare.
In fact the only cards that haven't at least hit 20% were my 9800 pro 256MB, an X1950XT, a 9600GT, and a GTX 295 FTW edition. Aside from the 9800 pro all the rest were preclocked by the manufacturer and you can't expect a 20% overclock on top of an existing overclock.
This could go on and on but we could be here forever, making a list of all the times Nvidia has been dirty. I just hope the time comes for Nvidia to pay for its business practices; hell, Intel waded through a load of shit for one bit of controversy surrounding OEMs.
Whoever said earlier that the R9 390 is a bullshit product has left me scratching my head. In what way does it not live up to the product it's marketed as? It sure as hell gives the GTX 970 a run for its money, even without any mention of the VRAM. What, do you not have a proper PSU?
Also, someone doesn't seem to understand the importance of binning. Binning is not for OCing alone. In a diminutive card like the Nano, where every bit of the heatsink matters, you don't want some garbage quality chip that takes absurd amounts of voltage to hit the boost clocks you want. 30% overclocks are not that rare in the light of big Maxwell, but suggesting that Nano should've been built around that goal is insanely ridiculous.
I don't know how you achieve that - must be something intentional, no? Like cherry picking cards plus water ?
I just don't think a Nano that costs the same as a 980 Ti is going to do very well because the performance, with the throttling, will probably hit more around the Fury non-X...
This product seems so niche it's silly. Why can't AMD just release a Fury X without the water cooler and with custom boards?
- R9 Nano will be faster product than R9 290X but not that fast to threaten any Fury;
- R9 Nano won't cost more than 630$, actually we expect price tag in line of around 450$;
- It won't be niche - it should be the new standard or paving the way for new generations of small cards;
- Fury X doesn't need custom boards - you won't achieve anything if you are seeking for guinness record clock heights. It's the best with water and let it please stay with WATER !
The price should be announced tomorrow.
The voltage has been tested by W1zzard and he got the result that it doesn't respond very well to over voltage, so perhaps custom cards are unnecessary. But it seems Nano may make use of a backwards approach, limit voltage input to keep tdp down tight, allowing the compact cooler.
I think Fiji in general is too immature to be tested properly. It's like when the initial Tahiti (7970) came out with conservative clocks. AMD may have been cagey to prevent chip problems but with hindsight, released the 'GHz Editions' that rivalled the GTX680.
Fiji came close to nailing it (but in PR terms, still so far away) so I think the next chip or a respin might make a huge difference.
We know DX12 will make a difference in AMD's favour so it'll be interesting next year with AMD's HBM experience and Nvidia getting their effort out.
As for Nano, we'll see tomorrow.
1) Faster than a product that is one generation old, and functionally a twice baked 7970. Kinda depressing really.
2) Price tag pulled straight from your backside. We don't expect anything, you have expectations. My only expectation is that the price tag will be north of what I would spend on the third time around for this process node. Of course, that's not a definitive number. You seem to be measuring only by "less than a fury," but "more than a 390." Seems like you've got a real winner there, with an almost $300 wide window.
3) Do you understand what niche means? The niche application for this is either an HTPC or other SFF computer, where money seems to be no object. Most people have a budget, which means the same performance could be had cheaper at the cost of space. Given that gamers generally have spacious cases, budgetary restrictions, the desire for raw performance, or some combination thereof you've got a niche product. Being king of a niche isn't bad, but believing reigning over a niche makes you a success is stupid. Blackberry ruled the niche of work smart phones, but died because their niche was too shallow.
4) Derp. Just plain derp. I'm sorry, but if you're spending that kind of money on a card, you should be able to cool it however you want. I'd be happy with Asus, MSI, Gigabyte, or another partner to come forward with a 10% increased cost card ($63 at $630 is exactly that), that supported excellent overclocking on air but required three slots. I'd be happy with someone releasing an underclocked "efficiency" version of the card that was cheaper because they cut some corners. What makes some people angry is AMD putting their foot down, especially for a product that might not be fantastic (coil whine on the pumps?) for the huge price. Cards sell better when there are options. Options are predicated on being able to choose performance targets and design to them. Saying "Fury is meant to be under water" is like saying Fury isn't meant to run, so break its knees. It already has a perfectly fine wheel chair it can get around in.
Seriously, stop trying to talk for everyone. Every time you do it makes me angry, because you assume we agree with your points. If you think something, so be it. If you tell me I think something, you'd better be prepared to retract your points when they are demonstrably unfounded.
But these things are mysteries for you.
Probably you are typing just to argue with someone who has better points than you.
I got used to your points and honestly - I am sick of them and want something better.
You also have to think a bit more. If it saves a couple of watts (which is being generous, actually) The difference between, say, 1.2v @ 85C vs. 1.2v @ 65C which is where water would take it, there really isn't much savings in power, is there ((NO))? With that said, wouldn't the pump, which most use more watts than a fan, negate those negligible gains for power savings ((YES))?
They are plenty of negatives to liquid cooling too, so it is far from silly to prefer air. I water cooled my systems for years, but went back to air to avoid the hassle. Now I run an AIO because they largely eliminate most of the hassle, but they still aren't as nice as easy as air cooling.