Cant say I am happy with the price but love the potential performance. Wonder if any makers will let it use a high TDP for less throttling? If that possible?
Well, you would have to cool this higher TDP. Maybe on really niche mini ITX watercooled system you would be able to gain from this, but I highly doubt it would be possible to cool more of it without increasing its footprint. And its footprint is its main argument as for higher TDPs you get other fidji cards...
They completely failed by not making this a blower. Most ITX cases, specifically single-fan cases like the FT03-mini, can not handle 175W of "recycled" hot air. Even if half the heat produced actually leaves the case with this split design (which physically won't happen) that is still 87.5w of heat in addition to all the other components like the CPU (65-88 watts) motherboard (20+ watts) and SSD/HDD (up to 5 watts each)
Why not make it a blower? I don't understand why these ITX cards are not coming with blowers? I had to ditch my Zotac 970-ITX for a cheap PNY 970 with a blower (9.5") that cools twice as good because 90% of the heat actually exhausts from the case.
Exactly what ITX case worthy of being useful for ITX applications (tiny amount of space) isn't going to overheat with a 175w videocard that effectively has a "top-down" cooler?
Some mITX cases have a sufficiently beefy case fan (or CPU/case fan) that they can handle the extra thermal load inside the case, but from experience with an mITX case that has a 92mm fan that can spin up to absurd speeds (so much so that sourcing a replacement fan is very difficult), you're going to get less noise with a blower GPU since it'll let that CPU/case fan spin a lot slower.
You answered your own question, didn't you? You replaced your SHORT card with a 9.5" blower card. That won't fit in some mITX case designs. You need to make sure your chassis is capable of shunting enough air before you consider a card like this - or mod it so it does. Otherwise go lower power.
I suspect the price is that high because they can't supply enough of them if the price was at or below the Fury. I believe the price will come down a bit after manufacturing improves. It likely should be priced between the Fury and Fury X.
A full 100 W of TDP less is a big chunk of performance less.
You also have to remember that these are binned high quality Fuji chips, so they are actually more expensive and rare than those used in Fury X. So the price is not surprise!
So you can take small, low energy and not so fast alternative or, big, energy hungry, fast variant... I actually like that you have that choice!
Honestly, I doubt it. This is a specialty product for small form factors. Like everything else SFF, you're paying a price premium *because* it's a smaller footprint even if performance-wise it's not as powerful. You're not buying one of these for fun, you're buying one of these specifically because the regular, cheaper one does not fit in the space you're trying to put it in.
Yeah, I'm not feeling this either. My mITX SUGO case can fit a full length card, so this is immediately in competition with the $500, 165 watt GTX980. If Nano is going to cost me another $150 and 10 watts, it better have some stellar performance.
Then again, 14nm is supposed to be out in 2016. The first new manufacturing node in almost four years probably makes those worth waiting for.
Yeah, all Fury line products should have debuted $50 lower MSRP, but with their inability to produce enough, they can price gouge until supply starts to exceed demand.
I was thinking the same thing. I believe it states the power delivery circuitry is cut down (especially compared to R9 fury) making it useless for record breaking type overclocking
Well, HTPC and small builds are not *that* common, even among "gaming" builds, so you could say this card is a niche of a niche.
I can't say I like the price either. If it were at the same price point as the Fury or between the Fury and the 390X, it would be a steal. But at that price point and expected performance, I can't say that I'm thrilled.
I really do not see it as an ideal solution for HTPC since, I believe, it doesn't support either HDMI 2.0 or HDCP 2.0. Do 4k Blu-ray playing isn't an option. Maybe a dedicated gaming build for the living room.
Agreed, this is MiniITX and not really HTPC. Where HTPC is the main focus and gaming is secondary the GTX 950/960 is ideal. I think the Fury Nano price will also weed those people out too.
according to doom9, HEVC main10 on the GTX960/950s is still hybrid and will struggle @ hi bitrates. I think I'll wait another generation before upgrading my HTPC - I'm not paying 200+ USD for a card that can't basically eat whatever video encode I throw @ it for breakfast
More like no sale, no money product. How many gamers willing to splash $650 on a graphics card is going to let themselves be this crippled by case size? I thought the Nano was going to be their mainstream Fury card, I'm not so concerned about this card but that it probably means the Rebrandeons will be all AMD has for the <$500 market for quite a while. This can not end well.
In the end, 20nm planar is what killed AMD. I'm certain both AMD and Nvidia have a gigantic backlog of designs just waiting for implementation when it's their turn at the 14-16nm trough. AMD still has a shot, but they need to kill it with the next generation.
NVidia has a 7 nm design since last year ready for production. 14 and 10 nm has been done long time ago.
I like AMD, but at this price they will sell at most 500 units in all. I don't think AMD will be able to recapture even its R&D cost for the new Fury architecture, unless they manage to win in next generation consoles. By that time NVidia and many others will have HBM. NVidia will have it by Feb-March next year, as far as I know. Sad, but I think AMD are financially done - no aggressive pricing, no market share, no enough time for exclusivity.
Fury will just end up in the next series as a slightly lower tier part I should imagine, unless it's AMD's Big Experiment which would naturally lead to HBM2 parts on 16nm next year with Fury on 28nm being a one-time thing.
The price is high, but let's put this into some sort of perspective - it's meant to fill a niche which is limited by cooling and size, both of which the Nano should be able to overcome. Turning up to a LAN event with a machine that performs within a few percent of the larger single GPU machines at a much reduced size would be amusing. There are more than seven billion people on this planet, so I imagine more than a handful see this as a good idea. Also, remember for some people, money is no object - the exact reason people buy Titan graphics cards and hexa/octocore Intel CPUs when a 980Ti and 4790k will do. Small usually means a price premium anyway.
As far as I remember, AMD have first crack at HBM2, so I wouldn't expect NVIDIA to beat them to it.
This is similiar to titan X, how many gamer will pay extra 100 - 350 USD for additional 3% performance?
I have a mini-ITX 970 in my secondary gaming machine, using xigmatek nebula case. This is the half show case, matching the design of my def tech studio monitor speakers. I am definitely in the market for this nano fury, just because it's the fastest small card. The same way titan x is the fastest flagship.
Neither nano nor titan are cost effective, but price/perf is not part of their marketing role.
It would be interesting in the review if you attempt to replicate Fury Nano's power usage via clockspeed, voltage, and power limiter adjustments particularly with the Fury X. Possibly the reverse as well.
Already been done. Just go into the AMD control panel and lower the power limit to -50. You lose about 5% performance, but power usage drops significantly.
The effectiveness of this is also based on the hardware power controls of the card too. The 290/390 have better power stepping than previous versions and the capability goes down hill as you go through each hardware generation. Now, with AMD selling cards whose silicon is not different for the last 3-4 years as a new model your mileage will vary.
Most mini-ITX cases can house up to 10" graphics cards. Even the smallest cube-shaped ones can take something like the Fury X because they're designed to support AiO coolers, and good-performing SFX PSUs rated up to 500W are rather easy/cheap-ish to find nowadays.
There's no other way to put it: price on this is ridiculous. It's like AMD doesn't even want to sell them.
It seems AMD can't produce too many of them. It's more of a technical demonstration than a real mainstream product. The price/performance will probably be way less ridiculous than your typical Titan product, though. All AMD needs is a R9 Femto Ti.
And then... If cases support it, it's because they need to support it. There are often quite difficult concessions made, and gming mini ITX cases are often much bigger than regular mini ITX cases. And good 400W SFX PSUs will be even easier/cheaper to find, or you will have more margin/less heating on your higher-rated SFX PSU.
I'm kind of surprised this uses a full fidji GPU. I knew highly restrictive binning would be involved, but I thought it would be easier to pull off with a few CU stream processors disabled.
Well I guess this somewhat confirms the impression that when a task is easily multithreaded, more low speed cores can do the same job as less high speed cores for less power. Or it could just be some marketing stuff. I don't know.
Running 4096 cores @ 750MHz and low voltage allows better performance per watt compared to running 3072 cores @ 1GHz and high voltage, at least in some cases.
Though this depends on the properties of the silicon and frequency drop required to reach the desired voltage drop.
No, that occurs because the Titan X in certain situations is bottle-necked by parts of the card that are not cut-down in the 980Ti OC. So the 980Ti OC can benefit by the increased clockspeeds to those parts.
A mini case for this GPU should have an opening (+dust filter) directly where the fan sucks cold air in and should exhaust the warm air from the front directly to the outside. That would be neat cooling without heating a case up.It would be very difficult to combine such an approach with a generic case, though.
What the hell is a "mini-ITX card"? Gamer ITX cases accept much longer cards. Non-gamer cases, if they support any cards at all, are typically limited to single width and half height, so this card fails on two counts (not to mention thermal problems).
For those of us who long for increasing FPS in decreasing case liters, having GPUs like this available is the first step. Hopefully a case maker will put out something like the TU-100b (but done right with respect to airflow, etc.) that takes advantage of these cards' short length to allow a smaller case.
This should be inside the "Ultimate" Steam box... :-). It really deserves a case/system build around it. And the size is not unlike a Gigabyte 970 I've seen, so I believe there are some mini ITX boxes smaller then the "gaming" ones that can fit this. But the price... oh, the price... :-(
Because Linux Gaming is a whopping "less than 1%" of the gaming market and because some guys just don't fucking get that an underdog has to spend its resources only on things that matter.
This is a niche product, yes. It is going to sell in small numbers, yes.
But this is HALF of the future Fury X2, the DUAL-GPU Fiji card AMD will launch next.
It has 175W TDP, so the dual GPU card will be around 300-350W TDP, like always. And its performance will be 170% of Fury X.
AMD is just testing the waters. How does this card perform? How many chips can it make. Seeing the binning get better so that it can make the flagship Fury X2.
My thought exactly, this is half the FuryX2, which will be the same size as the Fury X, and they've tuned the power usage with this card. And it will fit in most small cases, too. What about pricing... $1100-1200?
The Fury x2 has a chance to be great, but it is limited to the classic SLI issues. But even if this is the testing grounds, it's a somewhat disappointing product at the price.
Wow, all that separates it from the Fury X is 50MHz on the core clock? That's pretty tiny a difference for the form factor benefit, in usage that would be unnoticeable. I thought it would be closer to or even under the Fury non-X.
Nano won't be hitting it's max boost nearly as often as the Fury X. It's typical clocks under load will be significantly lower, and thus so will it's performance. Published benchmarks from AMD themselves show the Nano performing closer to the 980 than the Fury X, which makes me question its MSRP. I can understand a performance per $ penalty for the form factor and higher efficiency, but I think asking the same price as another card in your lineup that performs ~30% better (and $150 more than your competitors card that performs similarly) is a bit much.
I stopped believing anything AMD claims. AMD is better at marketing than making GPUs. Every year they make promises they can't keep... I resent the fact they are essentially putting it at the same price as the Fury X. Nobody in their right, educated mind, will buy something just because it is expensive for the sake of spending money. Nvidia is pricier because, by most the benchmarks, they are always ahead of AMD. Give AMD credit for trying, but seriously they really need to quit the hardware business and become a software company of some sort. Its sad that they are the first with HBM but are barely able to beat the GTX GDDR5 after 8 years of "careful" development. Honestly why not just sell the card, lay down some guidelines and let others choose their own cooling options.
Yeah. LIke FreeSync. What a freaking failure. No 100$ overhead, no "you only get single port on a monitor" no perfomance hit, no "exclusively ours, fuck competitors". They surely can't keep promises, eh.
"AMD tells us that the typical gaming clock will be around the 900MHz range, with the precise value depending on the power requirements of the workload being run. As to why AMD is shipping the card at 1000MHz even when they don’t expect it to be able to sustain the clockspeed under most games, AMD tells us that the higher boost clock essentially ensures that the R9 Nano is only ever power limited, and isn’t unnecessarily held back in light workloads where it could support higher clockspeeds."
I'm sure it has nothing at all to do with being able to advertise the card as running at 1000MHz... Marketing spin is so annoying. It's also disappointing to see them continue to market this as a 4K card. No single card solution is good for 4K yet, even the best cards can't hit a consistent 60FPS in modern games.
On a positive note, it does seem to fill a niche and open up the possibility of even smaller gaming PCs. It's too bad they didn't partner with someone to demonstrate how small a case could be using this card.
This pretty much confirms the theory about the Nano's unusually high advertised perf per W around the time Fury X launched. It looks like Fiji is in fact operating well outside its optimal frequency range in the Fury X. AMD really had to push Fiji to its limits to make it performance competitive with GM100. Now that GM100 is no longer the objective, Fiji can show it's true efficiency, albeit at a significantly lower performance profile. While binning should account for some of the efficiency gains in the Nano, the vast majority of it probably comes from the lower clock speeds and power targets, as performance benchmarks have already shown.
I'm currently not looking to upgrade, but I would be more likely to upgrade if AMD comes out with something good, just to encourage competition. But with prices like this, with obviously less performance, no one will buy.
Also, isn't the nano the same size as the fury X? Doesn't that make the nano irrelevant, as the fury X is just a faster nano, with free water cooling? The only advantage is perfect/watt, but I bet the 980 will beat it in that regard, and taking a nano and underclocking it should have a similar result.
The voltage is going to make the biggest difference. I would be interested to see how close an under volted Fury X can get. It would help us to better understand Fiji's optimal frequency range, and how much of the nano's efficiency gains actually result from binning.
While I won't be buying this card due to price and needed performance for my quickly approaching HTPC build, I am very happy that AMD (and NV to a degree) are pursuing these small cards.
It's not all about cramming a small card into a small case. I want a small card so that the (barely) ATX case I have has sufficient internal airflow, and not a huge card blocking the airflow. These mini cards are a perfect solution for a low power system with low RPM fans. Now they just need to make a $300 one for 1080p like the mini 970s and I might bite.
Let's wait for actual games tests and not Ashes of the Singularity benchmarks before we praise or slate anybody for their DX12 performance. Sorry, I just think it's premature.
I think this is a great concept, but I wish there was also a R7 Nano for the SFF Pre-built desktop and future Thunderbolt III market (laptops and UCFF desktops) as well.
40W TDP (using a large GPU die running at lower clocks and voltage) would be great to have for these two purposes.
Actually, I have read some comments telling that such card would not fit an mini-ITX case. The problem is that such a card occupies two slots...so it doesn't match the expectation. Although we must admit that such a card is an achievement in any case. But AMD fans will have to wait until next year to see a really new middle-range video cards from AMD. For now, we have to buy something like this: http://www.newegg.com/Product/Product.aspx?Item=N8... http://hardware.nl/sapphire/11230-00-20g.html The price is just under $200, and the performance is great. Nvidia would cost you more. It has a lower power consumption, though. So, we just have to wait until the next year to see a real battle.
Chip binning for a difference of 100 W of power? Something is not right here, or rather something is very wrong with the Fury X and Fury. It almost gives the impression that the explanation for this is entirely the other way around. There is no "chip binning" going on with Fury Nano. AMD just wanted to sell off its faulty chips in the form of Fury X without the competition from Nano, and then start selling the good chips with Nano. What kind of fault can raise the TDP by 100 W and still allow the chip to work is very interesting though.
Throttling does not explain it. It is a 1000 MHZ card. If it keeps throttling downwards all the time and is at home only at 900 MHZ why sell it as a 1000 MHZ card? I am not deeply knowledgeable about chips, but if the MHZ is linearly proportional to the TDP then a 100 MHZ decrease does not explain a 100 W decline in TDP from 275 W. Even if it is not linear then the same technology must work differently on different chips, as in CPUs where an overclock of a 100 MHZ does not send the TDP higher by 100 W. Again chip binning for power draw is a first in any chip, be it CPU of GPU. Supposedly, there are Intel or AMD CPU or GPU chips that can work at half of third of the advertised TDP, yet they have never been found and reported anywhere before.
Hell yes, *the* card I've been waiting for. Finally a decent card that fits in a small case and can be run off a single 8pin PCIe connector. Can't wait to see Nvidias response or I'll buy one of those expensive suckers...
Not to worry. Now that Binning for TDP and lovely Throttling is out of the closet, NVIDIA will hand you a card half the size of Nano in no time. GTX 980 Ti Nano or even Titan Nano (imagine that), is the card to wait for.
Guys, what happens to "throttling" if one slaps water cooling on it. Aren't cards naturally consuming less power at lower temps? Wouldn't it then be able to run at 1000Mhz withing given power envelope?
Usually when thing get physically smaller, and functionally amost the same, the price would get higher due to many factor. R9 Nano is the same case.... Just wonder why people expect it to be cheaper than says R9 Fury or Fury X.... is it because it's physically smaller ?
It is interesting to see so much power in a such a small card. This is brilliant for people like me that favors an ITX rig. Just that the price is a little off for me currently.
What a awesome card. I cannot wait to see what type of water block someone will come up with for this card. What could -10C and to this chip ? Fury X or 980Ti cost more then just the purchase price 175w is huge!!! When you are running 2,3,4 cards after all they only make PSU's so big and it is possible for video cards to suck up 1200+ watts. Running 2X1200W PSU's in 1 comp can be a real pain. Imagine a 4 way CrossFire and only need 1200W to run a machine that can do things only dreamed of even 5 years ago. Just look at how much power we all could have that will fit on a desktop. RDY for some 8k gaming anyone ?
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
95 Comments
Back to Article
jragonsoul - Thursday, August 27, 2015 - link
Cant say I am happy with the price but love the potential performance. Wonder if any makers will let it use a high TDP for less throttling? If that possible?nightbringer57 - Thursday, August 27, 2015 - link
Well, you would have to cool this higher TDP. Maybe on really niche mini ITX watercooled system you would be able to gain from this, but I highly doubt it would be possible to cool more of it without increasing its footprint. And its footprint is its main argument as for higher TDPs you get other fidji cards...Samus - Friday, August 28, 2015 - link
They completely failed by not making this a blower. Most ITX cases, specifically single-fan cases like the FT03-mini, can not handle 175W of "recycled" hot air. Even if half the heat produced actually leaves the case with this split design (which physically won't happen) that is still 87.5w of heat in addition to all the other components like the CPU (65-88 watts) motherboard (20+ watts) and SSD/HDD (up to 5 watts each)Why not make it a blower? I don't understand why these ITX cards are not coming with blowers? I had to ditch my Zotac 970-ITX for a cheap PNY 970 with a blower (9.5") that cools twice as good because 90% of the heat actually exhausts from the case.
Exactly what ITX case worthy of being useful for ITX applications (tiny amount of space) isn't going to overheat with a 175w videocard that effectively has a "top-down" cooler?
Guspaz - Friday, August 28, 2015 - link
Some mITX cases have a sufficiently beefy case fan (or CPU/case fan) that they can handle the extra thermal load inside the case, but from experience with an mITX case that has a 92mm fan that can spin up to absurd speeds (so much so that sourcing a replacement fan is very difficult), you're going to get less noise with a blower GPU since it'll let that CPU/case fan spin a lot slower.Alexvrb - Monday, August 31, 2015 - link
You answered your own question, didn't you? You replaced your SHORT card with a 9.5" blower card. That won't fit in some mITX case designs. You need to make sure your chassis is capable of shunting enough air before you consider a card like this - or mod it so it does. Otherwise go lower power.eanazag - Thursday, August 27, 2015 - link
I suspect the price is that high because they can't supply enough of them if the price was at or below the Fury. I believe the price will come down a bit after manufacturing improves. It likely should be priced between the Fury and Fury X.A full 100 W of TDP less is a big chunk of performance less.
haukionkannel - Thursday, August 27, 2015 - link
You also have to remember that these are binned high quality Fuji chips, so they are actually more expensive and rare than those used in Fury X. So the price is not surprise!So you can take small, low energy and not so fast alternative or, big, energy hungry, fast variant...
I actually like that you have that choice!
zodiacfml - Thursday, August 27, 2015 - link
Right. People would still pay for performance than size. I'm still surprised with the price though.Mushkins - Friday, August 28, 2015 - link
Honestly, I doubt it. This is a specialty product for small form factors. Like everything else SFF, you're paying a price premium *because* it's a smaller footprint even if performance-wise it's not as powerful. You're not buying one of these for fun, you're buying one of these specifically because the regular, cheaper one does not fit in the space you're trying to put it in.Mr Perfect - Thursday, August 27, 2015 - link
Yeah, I'm not feeling this either. My mITX SUGO case can fit a full length card, so this is immediately in competition with the $500, 165 watt GTX980. If Nano is going to cost me another $150 and 10 watts, it better have some stellar performance.Then again, 14nm is supposed to be out in 2016. The first new manufacturing node in almost four years probably makes those worth waiting for.
beginner99 - Friday, August 28, 2015 - link
TDP is not actual power usage. It's a marketing number an NV has been cheating with this for years.D. Lister - Tuesday, September 1, 2015 - link
TDP is not actual power usage, but it isn't just a "marketing number" either.https://en.wikipedia.org/wiki/Thermal_design_power
Also kindly indulge us in how NV (Nvidia?) has been cheating with it for years.
Cellar Door - Thursday, August 27, 2015 - link
ITS WAY TO EXPENSIVE - good luck AMD.jay401 - Thursday, August 27, 2015 - link
Yeah, all Fury line products should have debuted $50 lower MSRP, but with their inability to produce enough, they can price gouge until supply starts to exceed demand.hodakaracer96 - Friday, August 28, 2015 - link
I was thinking the same thing. I believe it states the power delivery circuitry is cut down (especially compared to R9 fury) making it useless for record breaking type overclockingD. Lister - Tuesday, September 1, 2015 - link
None of the "Fury" branded GPUs overclock very well unfortunately.YoloPascual - Thursday, August 27, 2015 - link
So a low volume, high margin product?YukaKun - Thursday, August 27, 2015 - link
Well, HTPC and small builds are not *that* common, even among "gaming" builds, so you could say this card is a niche of a niche.I can't say I like the price either. If it were at the same price point as the Fury or between the Fury and the 390X, it would be a steal. But at that price point and expected performance, I can't say that I'm thrilled.
Cheers!
danjw - Thursday, August 27, 2015 - link
I really do not see it as an ideal solution for HTPC since, I believe, it doesn't support either HDMI 2.0 or HDCP 2.0. Do 4k Blu-ray playing isn't an option. Maybe a dedicated gaming build for the living room.eanazag - Thursday, August 27, 2015 - link
Agreed, this is MiniITX and not really HTPC. Where HTPC is the main focus and gaming is secondary the GTX 950/960 is ideal. I think the Fury Nano price will also weed those people out too.wintermute000 - Thursday, August 27, 2015 - link
according to doom9, HEVC main10 on the GTX960/950s is still hybrid and will struggle @ hi bitrates. I think I'll wait another generation before upgrading my HTPC - I'm not paying 200+ USD for a card that can't basically eat whatever video encode I throw @ it for breakfastKjella - Thursday, August 27, 2015 - link
More like no sale, no money product. How many gamers willing to splash $650 on a graphics card is going to let themselves be this crippled by case size? I thought the Nano was going to be their mainstream Fury card, I'm not so concerned about this card but that it probably means the Rebrandeons will be all AMD has for the <$500 market for quite a while. This can not end well.dagnamit - Thursday, August 27, 2015 - link
In the end, 20nm planar is what killed AMD. I'm certain both AMD and Nvidia have a gigantic backlog of designs just waiting for implementation when it's their turn at the 14-16nm trough. AMD still has a shot, but they need to kill it with the next generation.Ananke - Thursday, August 27, 2015 - link
NVidia has a 7 nm design since last year ready for production. 14 and 10 nm has been done long time ago.I like AMD, but at this price they will sell at most 500 units in all. I don't think AMD will be able to recapture even its R&D cost for the new Fury architecture, unless they manage to win in next generation consoles. By that time NVidia and many others will have HBM. NVidia will have it by Feb-March next year, as far as I know. Sad, but I think AMD are financially done - no aggressive pricing, no market share, no enough time for exclusivity.
silverblue - Thursday, August 27, 2015 - link
Fury will just end up in the next series as a slightly lower tier part I should imagine, unless it's AMD's Big Experiment which would naturally lead to HBM2 parts on 16nm next year with Fury on 28nm being a one-time thing.The price is high, but let's put this into some sort of perspective - it's meant to fill a niche which is limited by cooling and size, both of which the Nano should be able to overcome. Turning up to a LAN event with a machine that performs within a few percent of the larger single GPU machines at a much reduced size would be amusing. There are more than seven billion people on this planet, so I imagine more than a handful see this as a good idea. Also, remember for some people, money is no object - the exact reason people buy Titan graphics cards and hexa/octocore Intel CPUs when a 980Ti and 4790k will do. Small usually means a price premium anyway.
As far as I remember, AMD have first crack at HBM2, so I wouldn't expect NVIDIA to beat them to it.
testbug00 - Thursday, August 27, 2015 - link
How do they have a 7nm design when there are no design rules for ANY 7nm processes? You're full of it.Never mind that most 10nm process likely only finalized design rules under a year ago, at best.
piiman - Saturday, August 29, 2015 - link
"I like AMD, but at this price they will sell at most 500 units in all."LOL Want to bet?
meacupla - Thursday, August 27, 2015 - link
How about I want a gaming PC that isn't too heavy and fits in my luggage case?RVZ02/ML08, SG13, etc. that sort of deal.
PEJUman - Thursday, August 27, 2015 - link
This is similiar to titan X, how many gamer will pay extra 100 - 350 USD for additional 3% performance?I have a mini-ITX 970 in my secondary gaming machine, using xigmatek nebula case. This is the half show case, matching the design of my def tech studio monitor speakers. I am definitely in the market for this nano fury, just because it's the fastest small card. The same way titan x is the fastest flagship.
Neither nano nor titan are cost effective, but price/perf is not part of their marketing role.
limitedaccess - Thursday, August 27, 2015 - link
It would be interesting in the review if you attempt to replicate Fury Nano's power usage via clockspeed, voltage, and power limiter adjustments particularly with the Fury X. Possibly the reverse as well.nightbringer57 - Thursday, August 27, 2015 - link
I'd assume the Nano's regulators are tightly designed as to fit just right. I'd be quite surprised if it had a good TDP margin.Stuka87 - Thursday, August 27, 2015 - link
Already been done. Just go into the AMD control panel and lower the power limit to -50. You lose about 5% performance, but power usage drops significantly.jragonsoul - Thursday, August 27, 2015 - link
Wow, really? Do you know if this works on other cards?testbug00 - Thursday, August 27, 2015 - link
http://techreport.com/forums/viewtopic.php?f=3&...I'll leave that there. With the OP 290x, they dropped the power limit ~25% for lowering ~60W AND kept the same performance.
D. Lister - Tuesday, September 1, 2015 - link
Wow, AMD should hire these guys, and fire those awful manatees* that set up the clocks and wattage on their GPUs.*Southpark reference.
eanazag - Thursday, August 27, 2015 - link
The effectiveness of this is also based on the hardware power controls of the card too. The 290/390 have better power stepping than previous versions and the capability goes down hill as you go through each hardware generation. Now, with AMD selling cards whose silicon is not different for the last 3-4 years as a new model your mileage will vary.ToTTenTranz - Thursday, August 27, 2015 - link
Most mini-ITX cases can house up to 10" graphics cards. Even the smallest cube-shaped ones can take something like the Fury X because they're designed to support AiO coolers, and good-performing SFX PSUs rated up to 500W are rather easy/cheap-ish to find nowadays.There's no other way to put it: price on this is ridiculous. It's like AMD doesn't even want to sell them.
nightbringer57 - Thursday, August 27, 2015 - link
It seems AMD can't produce too many of them. It's more of a technical demonstration than a real mainstream product. The price/performance will probably be way less ridiculous than your typical Titan product, though. All AMD needs is a R9 Femto Ti.And then... If cases support it, it's because they need to support it. There are often quite difficult concessions made, and gming mini ITX cases are often much bigger than regular mini ITX cases.
And good 400W SFX PSUs will be even easier/cheaper to find, or you will have more margin/less heating on your higher-rated SFX PSU.
ant6n - Thursday, August 27, 2015 - link
next up: half-height mini-itx cardjimjamjamie - Thursday, August 27, 2015 - link
That would be super cool - I'd want one just to play around with itmedi03 - Thursday, August 27, 2015 - link
Shouldn't its performance be in Fury X area?testbug00 - Thursday, August 27, 2015 - link
no, it will not hit 1000Mhz in most workloads. As the article says, AMD reps stated they expect around 900Mhz in most cases.nightbringer57 - Thursday, August 27, 2015 - link
I'm kind of surprised this uses a full fidji GPU. I knew highly restrictive binning would be involved, but I thought it would be easier to pull off with a few CU stream processors disabled.Well I guess this somewhat confirms the impression that when a task is easily multithreaded, more low speed cores can do the same job as less high speed cores for less power. Or it could just be some marketing stuff. I don't know.
zepi - Thursday, August 27, 2015 - link
Running 4096 cores @ 750MHz and low voltage allows better performance per watt compared to running 3072 cores @ 1GHz and high voltage, at least in some cases.Though this depends on the properties of the silicon and frequency drop required to reach the desired voltage drop.
testbug00 - Thursday, August 27, 2015 - link
Especially considering GCN seems really to shine in the 850-925Mhz range. Over that it's pref/watt seems to die.nikaldro - Thursday, August 27, 2015 - link
Doesn't a too high core count reduce performance scaling though?Isn't that why an OC 980ti can push even past a Titan X at the same power envelope?
praeses - Thursday, August 27, 2015 - link
No, that occurs because the Titan X in certain situations is bottle-necked by parts of the card that are not cut-down in the 980Ti OC. So the 980Ti OC can benefit by the increased clockspeeds to those parts.MrSpadge - Thursday, August 27, 2015 - link
A mini case for this GPU should have an opening (+dust filter) directly where the fan sucks cold air in and should exhaust the warm air from the front directly to the outside. That would be neat cooling without heating a case up.It would be very difficult to combine such an approach with a generic case, though.cbm80 - Thursday, August 27, 2015 - link
What the hell is a "mini-ITX card"? Gamer ITX cases accept much longer cards. Non-gamer cases, if they support any cards at all, are typically limited to single width and half height, so this card fails on two counts (not to mention thermal problems).keninct - Thursday, August 27, 2015 - link
For those of us who long for increasing FPS in decreasing case liters, having GPUs like this available is the first step. Hopefully a case maker will put out something like the TU-100b (but done right with respect to airflow, etc.) that takes advantage of these cards' short length to allow a smaller case.Mugur - Thursday, August 27, 2015 - link
This should be inside the "Ultimate" Steam box... :-). It really deserves a case/system build around it. And the size is not unlike a Gigabyte 970 I've seen, so I believe there are some mini ITX boxes smaller then the "gaming" ones that can fit this. But the price... oh, the price... :-(bug77 - Thursday, August 27, 2015 - link
Nothing from AMD belongs in an Ultimate Steam box. Mostly because their abysmal performance on Linux.medi03 - Thursday, August 27, 2015 - link
Because Linux Gaming is a whopping "less than 1%" of the gaming market and because some guys just don't fucking get that an underdog has to spend its resources only on things that matter.Yojimbo - Friday, August 28, 2015 - link
Steam boxes are Linux based, are they not?Michael Bay - Monday, August 31, 2015 - link
Yes, promptly wiped and enchanced with Windows.stanleyipkiss - Thursday, August 27, 2015 - link
You people are missing the point. Ryan as well.This is a niche product, yes. It is going to sell in small numbers, yes.
But this is HALF of the future Fury X2, the DUAL-GPU Fiji card AMD will launch next.
It has 175W TDP, so the dual GPU card will be around 300-350W TDP, like always. And its performance will be 170% of Fury X.
AMD is just testing the waters. How does this card perform? How many chips can it make. Seeing the binning get better so that it can make the flagship Fury X2.
You read it here first.
Gadgety - Thursday, August 27, 2015 - link
My thought exactly, this is half the FuryX2, which will be the same size as the Fury X, and they've tuned the power usage with this card. And it will fit in most small cases, too. What about pricing... $1100-1200?TallestJon96 - Thursday, August 27, 2015 - link
The Fury x2 has a chance to be great, but it is limited to the classic SLI issues. But even if this is the testing grounds, it's a somewhat disappointing product at the price.extide - Thursday, August 27, 2015 - link
Exactly, this was the first thing I thought too!tipoo - Thursday, August 27, 2015 - link
Wow, all that separates it from the Fury X is 50MHz on the core clock? That's pretty tiny a difference for the form factor benefit, in usage that would be unnoticeable. I thought it would be closer to or even under the Fury non-X.nikaldro - Thursday, August 27, 2015 - link
It runs around 900MHz in game.So much just to claim that low TDP...
dragonsqrrl - Thursday, August 27, 2015 - link
Nano won't be hitting it's max boost nearly as often as the Fury X. It's typical clocks under load will be significantly lower, and thus so will it's performance. Published benchmarks from AMD themselves show the Nano performing closer to the 980 than the Fury X, which makes me question its MSRP. I can understand a performance per $ penalty for the form factor and higher efficiency, but I think asking the same price as another card in your lineup that performs ~30% better (and $150 more than your competitors card that performs similarly) is a bit much.Laxaa - Thursday, August 27, 2015 - link
Next gen with HBM2 and I'll bite. I need a small and efficent(but still powerfull) card next time around.Mitrovah - Thursday, August 27, 2015 - link
I stopped believing anything AMD claims. AMD is better at marketing than making GPUs. Every year they make promises they can't keep... I resent the fact they are essentially putting it at the same price as the Fury X. Nobody in their right, educated mind, will buy something just because it is expensive for the sake of spending money. Nvidia is pricier because, by most the benchmarks, they are always ahead of AMD. Give AMD credit for trying, but seriously they really need to quit the hardware business and become a software company of some sort. Its sad that they are the first with HBM but are barely able to beat the GTX GDDR5 after 8 years of "careful" development. Honestly why not just sell the card, lay down some guidelines and let others choose their own cooling options.medi03 - Thursday, August 27, 2015 - link
Yeah. LIke FreeSync. What a freaking failure. No 100$ overhead, no "you only get single port on a monitor" no perfomance hit, no "exclusively ours, fuck competitors". They surely can't keep promises, eh.silverblue - Thursday, August 27, 2015 - link
AMD aren't Sega, you know.Doobs - Tuesday, September 1, 2015 - link
"Nobody in their right, educated mind, will buy something just because it is expensive for the sake of spending money."What planet are you living on?
cfenton - Thursday, August 27, 2015 - link
"AMD tells us that the typical gaming clock will be around the 900MHz range, with the precise value depending on the power requirements of the workload being run. As to why AMD is shipping the card at 1000MHz even when they don’t expect it to be able to sustain the clockspeed under most games, AMD tells us that the higher boost clock essentially ensures that the R9 Nano is only ever power limited, and isn’t unnecessarily held back in light workloads where it could support higher clockspeeds."I'm sure it has nothing at all to do with being able to advertise the card as running at 1000MHz... Marketing spin is so annoying. It's also disappointing to see them continue to market this as a 4K card. No single card solution is good for 4K yet, even the best cards can't hit a consistent 60FPS in modern games.
On a positive note, it does seem to fill a niche and open up the possibility of even smaller gaming PCs. It's too bad they didn't partner with someone to demonstrate how small a case could be using this card.
dragonsqrrl - Thursday, August 27, 2015 - link
This pretty much confirms the theory about the Nano's unusually high advertised perf per W around the time Fury X launched. It looks like Fiji is in fact operating well outside its optimal frequency range in the Fury X. AMD really had to push Fiji to its limits to make it performance competitive with GM100. Now that GM100 is no longer the objective, Fiji can show it's true efficiency, albeit at a significantly lower performance profile. While binning should account for some of the efficiency gains in the Nano, the vast majority of it probably comes from the lower clock speeds and power targets, as performance benchmarks have already shown.medi03 - Thursday, August 27, 2015 - link
A good point.TallestJon96 - Thursday, August 27, 2015 - link
I'm currently not looking to upgrade, but I would be more likely to upgrade if AMD comes out with something good, just to encourage competition. But with prices like this, with obviously less performance, no one will buy.Also, isn't the nano the same size as the fury X? Doesn't that make the nano irrelevant, as the fury X is just a faster nano, with free water cooling? The only advantage is perfect/watt, but I bet the 980 will beat it in that regard, and taking a nano and underclocking it should have a similar result.
TallestJon96 - Thursday, August 27, 2015 - link
An under locked fury X should have a similar result*dragonsqrrl - Thursday, August 27, 2015 - link
The voltage is going to make the biggest difference. I would be interested to see how close an under volted Fury X can get. It would help us to better understand Fiji's optimal frequency range, and how much of the nano's efficiency gains actually result from binning.Ryan Smith - Thursday, August 27, 2015 - link
The Fury X is 7.64" long, and you still have to account for the CLLC. R9 Nano is 6" long.Kakti - Thursday, August 27, 2015 - link
While I won't be buying this card due to price and needed performance for my quickly approaching HTPC build, I am very happy that AMD (and NV to a degree) are pursuing these small cards.It's not all about cramming a small card into a small case. I want a small card so that the (barely) ATX case I have has sufficient internal airflow, and not a huge card blocking the airflow. These mini cards are a perfect solution for a low power system with low RPM fans. Now they just need to make a $300 one for 1080p like the mini 970s and I might bite.
extide - Thursday, August 27, 2015 - link
Well, now we know what the dual GPU car will look like! Two of these!extide - Thursday, August 27, 2015 - link
card* of course...jeffrey - Thursday, August 27, 2015 - link
Chart lists R9 390X as 4GB card instead of 8GB.Ryan Smith - Thursday, August 27, 2015 - link
Got it. Thanks!KenLuskin - Thursday, August 27, 2015 - link
AMD's GPU will continue to post far superior performance to NVDA GPUs!So, there will be NO more comparison, and price will fade as a concern.
gamer demand for top of the line AMD GPUs will explode.
People who are currently making price comparisons, are similar to those that compare
a VW to a Porsche.
They are CLUELESS about DX12 performance!
silverblue - Thursday, August 27, 2015 - link
Let's wait for actual games tests and not Ashes of the Singularity benchmarks before we praise or slate anybody for their DX12 performance. Sorry, I just think it's premature.Michael Bay - Monday, August 31, 2015 - link
What shit are you smoking? I want some.Computer Bottleneck - Thursday, August 27, 2015 - link
I think this is a great concept, but I wish there was also a R7 Nano for the SFF Pre-built desktop and future Thunderbolt III market (laptops and UCFF desktops) as well.40W TDP (using a large GPU die running at lower clocks and voltage) would be great to have for these two purposes.
Mark_gb - Friday, August 28, 2015 - link
I was hoping for a $475 price... $650 is a no deal price on this for me.SviatA - Friday, August 28, 2015 - link
Actually, I have read some comments telling that such card would not fit an mini-ITX case. The problem is that such a card occupies two slots...so it doesn't match the expectation. Although we must admit that such a card is an achievement in any case. But AMD fans will have to wait until next year to see a really new middle-range video cards from AMD. For now, we have to buy something like this:http://www.newegg.com/Product/Product.aspx?Item=N8...
http://hardware.nl/sapphire/11230-00-20g.html
The price is just under $200, and the performance is great. Nvidia would cost you more. It has a lower power consumption, though. So, we just have to wait until the next year to see a real battle.
meacupla - Saturday, August 29, 2015 - link
No, you got it backwards.The problem is most modern mITX cases will fit 10.5", or longer, double slot cards no problem.
So this card has to compete with those.
versesuvius - Friday, August 28, 2015 - link
Chip binning for a difference of 100 W of power? Something is not right here, or rather something is very wrong with the Fury X and Fury. It almost gives the impression that the explanation for this is entirely the other way around. There is no "chip binning" going on with Fury Nano. AMD just wanted to sell off its faulty chips in the form of Fury X without the competition from Nano, and then start selling the good chips with Nano. What kind of fault can raise the TDP by 100 W and still allow the chip to work is very interesting though.nikaldro - Friday, August 28, 2015 - link
Chip binning and THROTTLING. AMD said that it runs around 900MHz in game.versesuvius - Friday, August 28, 2015 - link
Throttling does not explain it. It is a 1000 MHZ card. If it keeps throttling downwards all the time and is at home only at 900 MHZ why sell it as a 1000 MHZ card? I am not deeply knowledgeable about chips, but if the MHZ is linearly proportional to the TDP then a 100 MHZ decrease does not explain a 100 W decline in TDP from 275 W. Even if it is not linear then the same technology must work differently on different chips, as in CPUs where an overclock of a 100 MHZ does not send the TDP higher by 100 W. Again chip binning for power draw is a first in any chip, be it CPU of GPU. Supposedly, there are Intel or AMD CPU or GPU chips that can work at half of third of the advertised TDP, yet they have never been found and reported anywhere before.Daniel Egger - Saturday, August 29, 2015 - link
Hell yes, *the* card I've been waiting for. Finally a decent card that fits in a small case and can be run off a single 8pin PCIe connector. Can't wait to see Nvidias response or I'll buy one of those expensive suckers...versesuvius - Sunday, August 30, 2015 - link
Not to worry. Now that Binning for TDP and lovely Throttling is out of the closet, NVIDIA will hand you a card half the size of Nano in no time. GTX 980 Ti Nano or even Titan Nano (imagine that), is the card to wait for.medi03 - Sunday, August 30, 2015 - link
Guys, what happens to "throttling" if one slaps water cooling on it.Aren't cards naturally consuming less power at lower temps? Wouldn't it then be able to run at 1000Mhz withing given power envelope?
mr_tawan - Monday, August 31, 2015 - link
Usually when thing get physically smaller, and functionally amost the same, the price would get higher due to many factor. R9 Nano is the same case.... Just wonder why people expect it to be cheaper than says R9 Fury or Fury X.... is it because it's physically smaller ?watzupken - Tuesday, September 1, 2015 - link
It is interesting to see so much power in a such a small card. This is brilliant for people like me that favors an ITX rig. Just that the price is a little off for me currently.LeadSled - Tuesday, October 6, 2015 - link
What a awesome card. I cannot wait to see what type of water block someone will come up with for this card. What could -10C and to this chip ? Fury X or 980Ti cost more then just the purchase price 175w is huge!!! When you are running 2,3,4 cards after all they only make PSU's so big and it is possible for video cards to suck up 1200+ watts. Running 2X1200W PSU's in 1 comp can be a real pain. Imagine a 4 way CrossFire and only need 1200W to run a machine that can do things only dreamed of even 5 years ago. Just look at how much power we all could have that will fit on a desktop. RDY for some 8k gaming anyone ?