Report: AMD Radeon Fury Specs and Photos Leaked

Subject: Graphics Cards | July 7, 2015 - 11:59 AM |
Tagged: Radeon Fury, radeon, HBM1, amd

As reported by the upcoming Radeon Fury card specs have been leaked (and confirmed, according to the report), and the air-cooled card is said to have 8 fewer compute units enabled and a slightly slower core clock.

View Full Size

The report pictures a pair of Sapphire cards, both using the Tri-X triple-fan air cooler. The first is a reference-clocked version which will be 1000 MHz (50 Hz slower than the Fury X), and an overclocked version at 1040 MHz. And what of the rest of the specs? VideoCardz has created this table:

View Full Size

The total number of compute units is 56 (8 fewer than the Fury X), which at 64 stream cores per unit results in 3584 for the non-X GPU. TMU count drops to 224, and HBM1 memory speed is unchanged at 1000 MHz effective. VideoCardz is listing the ROP count at an unchanged 64, but this (along with the rest of the report, of course) has not been officially announced.

View Full Size

The board will apparently be identical to the reference Fury X

Retail price on this card had been announced by AMD as $549, and with the modest reduction in specs (and hopefully some overclocking headroom) this could be an attractive option to compete with the GTX 980, though it will probably need to beat the 980's performance or at least match its $500 price to be relevant in the current market. With these specs it looks like it will only be slightly behind the Fury X so pricing shouldn't be much of an issue for AMD just yet.

Video News

July 7, 2015 | 12:07 PM - Posted by Keven Harvey (not verified)

With only 1/8 less cores than the fury X and the same memory bandwidth, it should beat even the highest clocked 980s no problem.

July 7, 2015 | 12:34 PM - Posted by Martin Trautvetter

Usually not a fan of cut-down cards, but AMD's second-in-line card has been quite close to the flagship for the last couple of generations.

Then again, $550 is really steep for an air-cooled card if and when a fully fixed Fury X is both quicker and presumably quieter under load for only $100 more.

July 7, 2015 | 02:21 PM - Posted by JohnGR

Wow. A week ago everyone was pointing fingers at the R9 Fury X for being a little noisy on idle ignoring the fact that it was quiet under load. Now that air cooled R9 Fury is going to be in the center of attention, suddenly everyone cares about the noise under load. LOL

July 7, 2015 | 02:35 PM - Posted by Martin Trautvetter

The broken-pump Fury X is not quiet under load, in fact, the pump doesn't seem regulated at all, providing constant whine regardless of GPU load.

So, tell me again, how's your Kepler card doing?

July 7, 2015 | 04:31 PM - Posted by Heavy (not verified)

while amd does have some blame for the pump noise, shouldnt most people be dissapointed in cooler master, it is their pump not amd.i know amd has to check their products but they have fixed it with newer pumps

July 7, 2015 | 05:21 PM - Posted by arbiter

Its more AMD's fault as they should seen the issue before they left the factory but they didn't. All they end up saying is their fixed it but some retail units had issue as well. Is cooler master to blame for this, a little maybe only 20%. 80% of blame falls on AMD for letting it slip in to the market.

July 7, 2015 | 05:28 PM - Posted by Martin Trautvetter

I don't agree with blaming the supplier. AMD speced and ordered the AIO cooler design.

AMD had their samples built and assembled, tested them and shipped them to reviewers, knowing full well how they sound. (Unless everyone in QA has a major hearing defect, which might go a long way explaining how anyone in their right mind would sign off on the 290X/290 blowers.)

AMD then claimed to have fixed this issue and was, again, suprised by already having shipped dozens?/hundreds? of units using the old, noisy design into the channel.

This one is squarely on AMD.

July 8, 2015 | 02:53 PM - Posted by ppi (not verified)

It is entirely possible the pre-production prototypes did not have that issue. And it might not be detectable in production phase Q&A, just e.g. because of other noise in production hall.

BUT, attempting to cover that was seriously lame.

July 8, 2015 | 04:39 PM - Posted by arbiter

Problem was AMD was asked about the issue, and AMD claimed retail cards were fixed and didn't have it. Pretty said they claimed it was fixed in retail cards when it clearly wasn't completely fixed as some had the issue.

July 8, 2015 | 07:57 AM - Posted by obababoy

Would you just stop! You are generalizing EVERYONE like the guy who commented was the exact same person who posted about the Fury X pump noise last week/s. For all you know this guy may have never posted about Fury X. You sound like an idiot.

July 7, 2015 | 12:46 PM - Posted by funandjam

the picture of that beefy cooler on that really small board for the fury makes me wonder just how much slower the nano will be, since what they showed off is not only a really small board but also a really small cooler

July 7, 2015 | 04:35 PM - Posted by Heavy (not verified)

yea i laughed a little when i saw the heatsink longer then the pcb.nano might compete with the 970 or something between it and the 980 but thermal throttling might keep it down from reaching its full pontential man someone should do the math on this.

July 7, 2015 | 12:56 PM - Posted by BBMan (not verified)

ONLY $549 and uses 100W more than 980 GTX.


Why might I not be striking a <$500 980 GTX from the Christmas list yet?

July 7, 2015 | 01:48 PM - Posted by Anonymous (not verified)

Not to mention any 980 overclocks well beyond 1500+ on the core and will blow this thing out the water garanteed!

July 8, 2015 | 08:07 AM - Posted by obababoy

I mean unfortunately for me as an AMD fan you are probably right, but guaranteed you are not :) And enlighten me, how much more "well beyond" 1500 core does the 980 OC? Don't tell me about the stupid LN2 crap. No one really cares what a computer can do under a completely NON functional liquid nitrogen.

July 8, 2015 | 11:48 AM - Posted by Anonymous (not verified)

Some are reference and some are custom pcb from MSI, Gigabyte and Asus. Check on the guru3d forums in the benchmark mayhem section, also and hardforums for proof.

Any enthusiast knows this already since they were released and no point doubting it when its been out there for over 8 months now.

July 7, 2015 | 01:02 PM - Posted by Anonymous (not verified)

The snapshot says "Stream Processors: Up to 3584 unit."

Does that mean some cards will have less than stream processors?

July 7, 2015 | 01:16 PM - Posted by Anonymous (not verified)

great question. ive always wondered how they get exact specs every time from a manufacturing process.
sure its mature. but every chip?
I wonder how exactly close 4 gallon jugs of water really are.
what I mean by that is there is some variance.

July 7, 2015 | 04:27 PM - Posted by Heavy (not verified)

they'll just run some off every gpu and cpu company has done this,

July 8, 2015 | 12:27 AM - Posted by Anonymous (not verified)

They don't manufacture a separate chip. They get parts that don't make Fury X bin, maybe there's an error in one of the shader units, maybe it has too much leakage/uses too much power to fit the FuryX TDP, so a chunk of the GPU is disabled. This has been widespread across the GPUs and CPUs for the last decade. "Unlocking" can sometimes be done to restore those parts of the GPU, if they aren't physically disabled (laser cut)

July 7, 2015 | 05:38 PM - Posted by arbiter

that was my thought to, they list is as "Up to" is 2 worst words ever used. same for clocks

July 7, 2015 | 02:23 PM - Posted by JohnGR

Many "anonymous" Nvidia fanboys in here. Probably R9 Fury will be a good option.

July 7, 2015 | 02:53 PM - Posted by Anonymous (not verified)

Yeah, okay fanboy. The fanboy here, it seems, is you. Look, its not anyones fault but AMDs that they've been on the losing end of GPU and CPU business for a while. Not Nvidia, not Intel but AMD. Not fans or fanboys, but AMD. Someone needs to sit you down and explain that, this is business, not some game. They suck as a company, making products that people actually want and will buy. Name of the game, if they actually were truly competitive and not "just" there, they'd do better. But they haven't, and they aren't.

You'll scream fanboy, I'm sure, go ahead. I won't take it personally, but just remember, I don't have a Nvidia or Intel picture, while you have what?

July 7, 2015 | 03:21 PM - Posted by Anonymous (not verified)

A stupid Sempron avatar.

July 8, 2015 | 08:13 AM - Posted by obababoy

Ok, that JohnGR guy is lame but AMD makes plenty of products that people buy. I could easily justify to you why I have an R9 290 well beyond the fact that it still remains as the best performance/price ratio on the market at $240..They make great products but they cut corners being the underdog.

Where ill agree with you is this latest iteration of fiji cards. The Fury X would have taken an easy win priced at $550-575, or had it performed or overclocked better than the 980ti. Same with the upcoming Fury cards here...They are going to be great cards but if they are not priced to beat the 980 by at least $75 they will lose yet again.

The moral is to not use one idiot as a generalization for a populace.

July 7, 2015 | 03:46 PM - Posted by BBMan (not verified)

In fairness, I was an ATI fanboy when they started selling GPUs. However, they've been having to play leap-frog and catch-up since the the 9800 Pro. More recently I've been turned off mostly by the power comparisons.

However, this engine could represent a turning point as they have obviously found a viable memory solution.

Now if they can tweak rather than twerk it ....

July 7, 2015 | 04:34 PM - Posted by Anonymous (not verified)

The problem with that, is that Nvidia decided to ignore Gen 1, and instead invest their R&D into Gen 2. So when Gen 2 comes around, chances are that Nvidia will be the first out the door while AMD will be lagging behind once again. And remember, the memory isn't really doing them too much good performance wise.

July 7, 2015 | 05:18 PM - Posted by arbiter

if you look last 2-3 generations of AMD cards, memory bandwidth has been what kept them in the game. Nvidia cards have been memory slowed a bit. Look at Kepler cards most performance you would get outta is when did memory overclock. Put both current chips on same memory it would easily be a good 20-30% advantage for nvidia.

cue the fanboyz to argue that.

July 8, 2015 | 08:18 AM - Posted by obababoy

Your abuse of the word "fanboyz" makes me nauseous. Look, NV makes a faster card right now and that is fine, but don't bring up pointless what ifs right now. NV doesn't have HBM so they don't have your imaginary 20-30% advantage. Instead they have a more efficient and better computing GPU chip to give it the edge. That is all.

I could say the same crap about AMD. "Oh by next year they will have HBM2 and a more powerful/efficient chip"...blah blah.

July 8, 2015 | 01:19 PM - Posted by arbiter

AMD fans started with the fanboyz bs first in this story so can suck it. As for bandwidth, you know its true hence why you argued it. Reason say they would be faster is cause nvidia cards respond performance wise to memory overclocks then gpu.

As for saying same crap, well Nvidia will have HBM2 on pascal so lead will be gone.

July 9, 2015 | 11:10 AM - Posted by obababoy

Yep and they still have the lead on bandwidth. Their struggle is price/performance right now. They failed there.

July 8, 2015 | 12:20 PM - Posted by NvidiotBumLover (not verified)

You make us NvidiaFanBoyz look bad, stop posting comments publicly.

July 7, 2015 | 06:31 PM - Posted by Anonymous (not verified)

There is very little difference between HBM gen 1, and HBM gen 2, and the HBM2 stacks will probably be drop in compatible with AMDs GPUs, so while Nvidia may still be in the process of engineering and certifying HBM2, AMD will probably be able to push out some revised Fury SKUs with HBM2 long before Pascal gets to market. AMD already has the interposer work done, and any internal changes to HBM will be handled by the 5th chip die at the bottom of the HBM stack which handles the HBM control/logic. So HBM2 gets some taller stacks and more memory, and some faster clock speeds, and AMD is ahead of the game with HBM integration! The hard work is already done for AMD and HBM. As its stands, I do not think AMD is using all of the available bandwidth that the HBM standard provides, but HBM2's 8GB will provide plenty of 4k texture/other storage and the extra clock speed will probably help some, even if to get things into the GPU a little faster to the units that need them. That 5th logic DIE can probably completely abstract any internal HBM to HBM2 differences from the GPU/memory subsystem so not very much extra added work needs to be done by AMD in the matter. AMD is more than one step ahead for getting HBM2 inside revised Fury SKUs sooner rather than later.

I see that the FUD-Tastic 4 are pretty much all over any AMD bad news, as if the end times of the world are at hand and its AMD's fault, and AMD's imminent demise has been happening for how many years now!

July 7, 2015 | 05:15 PM - Posted by arbiter

as you say that there are just as many AMD fanboyz if not MORE in there.

July 7, 2015 | 05:47 PM - Posted by JohnGR

Yeap, I was right. Many Nvidia fanboys in here.

July 7, 2015 | 03:23 PM - Posted by amdfanboy (not verified)

i'm a amd fanboy look it says so in my name deal with it.

July 7, 2015 | 03:42 PM - Posted by Anonymous (not verified)

Wonder if this will be one of those situations where it can be bios flashed to a fury x.

July 7, 2015 | 06:59 PM - Posted by StephanS

AMD pricing is so wrong...

The Fury wont sell in volume above $459,
same goes for the Fury X.. its a $559 card.

The only way AMD can charge that extra 100$ is if they get state of the art Dx11 drivers released, now, and beat nvidia across the board.

Until this happen, lower your prices or continue to get DECIMATED.

AMD just dont grasp the value of good marketing...
So its sad to see a company digging its own grave just because of a few bad apples in management.

Also AMD has been destroyed by the press for their power usage,
yet we know hawaii perform amazingly at 900mhz..
Why cant AMD release a 'powertuned' card to put a cork in all this negative press, and render the green army troller inactive.

Well, now that nvidia is pricing the 980 ti at $599 (its coming)
AMD is toast, even so they created technically the most advanced GPU to date with equivalent power efficiency than nvidia maxwell ...

July 7, 2015 | 09:42 PM - Posted by mil (not verified)

Even with the price tag of $649,it is still not easy to find a fury x these days.

July 7, 2015 | 10:48 PM - Posted by Joe (not verified)

Dude let ask you a question about that. If the fury non x beats the gtx980 across the board or hell 80%, why can't they charge $549 for it if it's pretty much better than a 980. Do you think the gtx980 is that much better than gtx970 to be charging still $500 for that card? The answer to that is no. If you ask me they are all overpriced from both sides. This is coming from a guy that owns a gtx970.

I have owned both AMD and Nvidia so I can speak from experience from both sides of the coin. If it's on par with a 980 why can't they at least charge $500 like the 980? Why does it have to be cheaper because it's AMD? They need to make money, bottom line and going super cheap all the time doesn't always cut it. Is that their own miscue? Yes it is and I agree with you there.

July 8, 2015 | 02:28 AM - Posted by JohnGR

If AMD drops the prices tomorrow, not $100 but $200, do you know what will happen? Nvidia will do the same. Fury X's price was dictated by Nvidia when they put that $650 price tag on 980Ti. Most people where NOT expecting a price under $700-$750 for a card that uses a full GM200.

It's not a matter of marketing. Nvidia is controlling the hi end market. AMD doesn't have the pockets, neither the GPU to play freely in the market. So they play along and just hope Nvidia not to do anything crazy, like 980 at $400 and 980Ti at $550.

July 8, 2015 | 02:11 PM - Posted by Giant Oyster (not verified)

GTX 980, at stock, already provides a much better experience than does a Fury X.

That's without even taking into account that a GTX 980 can be overclocked to in excess of 1500MHz. The Fury card is already easily pre-fragged by the 980.

July 9, 2015 | 11:07 AM - Posted by obababoy

Not a better experience. The Fury is a decent amount more powerful. Your link is for perf/dollar which you are correct that the 980 is better at. R9 290 and 290x still crush that scale though. I love my 290.

July 9, 2015 | 08:49 AM - Posted by Anonymous (not verified)

If one want high end I guess NV has the upper hand right now.

290/290X is still the best in mid performance, especially if one plans to go dual graphics which is not a good idea with gtx970 due to its memory config.

The energy efficiency of the 970 is fine and all but it does not really change anything in considering actual of economic realities.

July 10, 2015 | 09:03 AM - Posted by obababoy

I actually like that there is no extended backplate. Think, now that last fan can push air straight through the heatsink with very low resistance.

With that said, the price is not great and AMD better lower it.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

This question is for testing whether you are a human visitor and to prevent automated spam submissions.