AMD Radeon R9 Fury X 4 GB

[yeeeeman]

So, everyone is taking a guess as what the causes of this situation might be. Really, the things are simple if you look carefully at the specs. Quoting Hexus on this:
1. Titan X has 6,611 GFLOPS of SP compared to 8,602 for the Fury. So here is the first offset.
2. Titan X has, ready for it, 207 GFLOPS of DP compared to 537 for the Fury. But both of them are under what the 290X is capable (739GFLOPS DP). So here is the second offset.
3. The ROPS are for sure the key element and here I will quote the folks at techreport:
4. HBM memory is wider but slower

"In other respects, including peak triangle throughput for rasterization and pixel fill rates, Fiji is simply no more capable in theory than Hawaii. As a result, Fiji offers a very different mix of resources than its predecessor. There's tons more shader and computing power on tap, and the Fury X can access memory via its texturing units and HBM interfaces at much higher rates than the R9 290X.

In situations where a game's performance is limited primarily by shader effects processing, texturing, or memory bandwidth, the Fury X should easily outpace the 290X. On the other hand, if gaming performance is gated by any sort of ROP throughput—including raw pixel-pushing power, blending rates for multisampled anti-aliasing, or effects based on depth and stencil like shadowing—the Fury X has little to offer beyond the R9 290X. The same is true for geometry throughput"

So, they increased the efficiency by removing some of the DP hardware compared to Hawaii. Still, they didn't do as much of a cut-down as NVIDIA did with Maxwell. Maxwell is so efficient because it's releaved of the DP hardware. Basically, that is how nvidia got this efficiency jump from Kepler to Maxwell, so that's not magic. The second part is indeed related to the ROPS and I guess they could've taked a bit from the shaders (3584 instead of 4096) and raise the number of ROPS to 96. Well, they could've but they didn't.
Now, the HBM part is tricky. This must be done at the driver level, and we know that AMD is a master guru when it comes to optimizing them. Only time will tell.

In conclusion, nvidia got rid of most of the unimportant bits and used those savings (which are huge, btw) to increase graphics horsepower. So, kudos to nvidia for doing this, they always pursuit what's the best even if that means losing something else. Still, considering all these things, I think AMD has done a very good job.
And also, stop looking at the maximum power consumption of this card. If you read the description of this test you'll see that it's just a furmark test, which does not mean anything. The real power consumption you should follow is the peak or the average.

Cheers!

 

[Petey Plane]

But it's not an Nvidia card. Its AMD. They said it was the best card in the world. Their pre-release benchmarks were evidently lies. This is like a cult whose leader says the coming comet will bring death and when nobody dies the cult leader says - "your faith has saved you".

There are assholes here saying it's fail. It's obviously not. It's a damn fine card - make no mistake. The main issue is it was touted and hyped, and hyped and hyped as a Titan X slayer (more by fans). The problem for the Fury X is......

...

...

the GTX 980ti is better. Nvidia is allowing custom variants with silly clocks and better coolers. Now, if AMD can (for technical reasons) allow AIB's to play with the Fury X PCB, then maybe this card will grow wings and kick ass properly. But we dont know for now. So yeah, the card has failed to meet the fan based perceived objective. Deal with it.

Let's hope they allow the board partners to work their magic on it.

I'm not sure they will be able to help. I haven't seen a review that was able to get more than a 100mhz overclock out of the Fury X. Most seem to be in the 50mhz range. And that's with a 500watt capable AIO cooler. I can't imagine a windforce or strix cooler will be able to do better. Also, memory overclocking is locked down on the hardware level.

I'm guessing that the standard, air cooled Fury will be the one that the AIBs get to put their spin on, and not the Fury X. Which shouldn't be an issue, because my understanding is the standard Fury has all the same silicon, it will just thermally throttle much sooner than the X. We'll see.

Also, you're right. All of AMD's benchmarks, and those "leaked" ones from Chinese forms were completely bogus.

I'm really looking to the Nano to restore some faith in AMD. I'd love to get ≈ 980 performance in a SFF case like the PC-Q30 (the little curved upright ITX case with the window)

 

[Nihilus]

Tahiti is 1024 Gflops in DP while Tonga is 207 Gflops in DP. We all saw a huge efficiency jump when they re-leaved Tahiti of this hardware (sarcasm).
This was not the time for cool experiments by AMD, it was a time to restore consumer confidence. They failed IMHO.

 

[mouacyk]

The second part is indeed related to the ROPS and I guess they could've taked a bit from the shaders (3584 instead of 4096) and raise the number of ROPS to 96. Well, they could've but they didn't.

According to AnandTech, AMD went with a 65nm interposer. Could they have chosen a 32nm interposer process that allowed more ROPS to be added? Would a simple (but more costly) upgrade in the process like that deliver the hype of this GPU? 65nm is what my Q6600 processor was 8 years ago!

 

[Petey Plane]

According to AnandTech, AMD went with a 65nm interposer. Could they have chosen a 32nm interposer process that allowed more ROPS to be added? Would a simple (but more costly) upgrade in the process like that deliver the hype of this GPU? 65nm is what my Q6600 processor was 8 years ago!

The interposer is not the same as the GPU die. The ROPs are on the GPU die, which is 28nm. The interposer is more like a substrate that the GPU die and memory sit on (in a kind of inaccurate sort of way)

 

[yeeeeman]

According to AnandTech, AMD went with a 65nm interposer. Could they have chosen a 32nm interposer process that allowed more ROPS to be added? Would a simple (but more costly) upgrade in the process like that deliver the hype of this GPU? 65nm is what my Q6600 processor was 8 years ago!

As Petey before me said, the interposer is not an issue. This was solely a design choice and I guess it has its reasons, but we can only guess what those reasons were.

 

[DarkOCean]

has anyone found a review that shows vram usage on this card ?

 

[Nihilus]

Was expecting more, i hope Nano will do better for it`s price range.

Just where in the hell will Nano and Fury pro fit in?? Fury X costs $650 while the 390x 8 GB goes for around $400. Here is the problem-
Fury X is only 10-20% faster than a mildly clocked 390x (the Fury X doesnt overclock much at all). So then what? Nano will be 5% better than the
390x while the Fury pro is 15% better. Price-wise I am certain they will be closer to the Fury X due to HBM costs. Price/performance will suck with them
as well.

Then there are those that plan on setting up crossfire, trifire, or even quad fire. HAHA can you imagine 4 of these in a case - what nightmare!
Not to mention it will only be 4k of RAM. On the other hand, three or four 8 GB 390x or even a pair of 295x would be alot more manageable (and cheaper)
for those that want to game beyond 4k.

 

[Hood]

No point in being disappointed about Fury X, the real performance of any flagship video card has NEVER lived up to the pre-release hyperbole. If you are surprised you may be guilty of wishful thinking...what, you thought maybe they'd suddenly decide to underplay the hype just this once? Not likely!

 

[Yolokila]

I have been waiting a while to see FURY x do well but this is a disappointment. I have had the 290x, 780 ti and 980 ti. I am no fanboy but AMD you guys are terrible. False claims, poor performance and various issues such as pump noise.

I cannot even justify a FURY X over a 980 ti. There is no comparison lol.

Cons of FURY X;
-4gb Limit. Yeah its HBM but it is still a limit anyone who does not believe that is stupid.
-Pump noise/ coil whine (yeah NVidia has but as for 980 ti and titan x they addresses coil whine issues - it is much less than 970 and 980s)/ audible fan on idle - Overall these are completely ridiculous issues.
- POOR drivers as showed by TP review if you don't believe you are stupid. AMD I am sorry but you live in BETA driver world. Why can you not release real drivers. WTF are AMD programmers doing.
- 100mhz overclock is a "overclockers dream" SERIOUSLY AMD WTF-********* The fact that I have 980 ti reference currently 1450mhz core, 8.1ghz mem at 110% tdp I am speechless with AMD's claim. Second If I watercool the 980 ti with a EK block at 89 quid more FURY X has no chance. 1500mhz easy.
- watercooling - still 105c on VRM is terrible I think AMD have actually oc there fury x to compete with the 980 ti as the 980 ti release was unexpected. As some have said AMD I think was releasing this thinking there was only going to be a gtx 980 and titan x. AMD are too slow and that it self has killed off this product launch.

Sorry for gunning you AMD but they are FACTS. Too bad im not a reviewer so you cannot discipline me like you did kitguru lol muhahaha

LMAO OMG

 

[SonicZap]

Then there are those that plan on setting up crossfire, trifire, or even quad fire. HAHA can you imagine 4 of these in a case - what nightmare!

That's not really an issue because trifire and quadfire don't make sense to begin with.

Fury Pro and Nano might still prove interesting, but after this launch I'm not expecting much from them. The Pro version will likely have somewhat better performance for the price (Pro versions of AMD GPUs have almost always been price/perf winners), while Nano will be used to milk money from the gamers who want a powerful GPU in a small form-factor case (ITX).

 

[Yolokila]

Considering the shambles with the FURY X claims by AMD. I doubt the FURY nano will be any more false than there claims. You will be probably be better off with those smaller nvidia 960s

Is 4k @ 60Hz not possible via DP? HDMI sucks anyways. Goddamn HDCP.

How is the compute capability of this chip? Is it still capable or have they crippled/limited functionality, a la nVidia, in order to gain 'efficiency'?

If these were queried previously, disregard as I didn't read all 9/10 pages...

yeah but HDMI usually used if you are connecting to a 4k tv. FUry x does not support hdmi 2.0 so cannot do 60hz on hdmi.

 

[btarunr]

According to AnandTech, AMD went with a 65nm interposer. Could they have chosen a 32nm interposer process that allowed more ROPS to be added?

What!? Interposer lithography has nothing to do with rop count. Just reading that analysis gave me nausea.

What AMD did wrong, in my opinion, is lazily reused 1x the front-end of Tonga and 2x its back end, while merely doubling CU count. There's a bottleneck hidden somewhere there.

They thought memory will bail them out in the end. It did, to an extent. Just not enough.

So they're selling a "new" engine with intercooled variable turbine turbocharger, but with an oldschool 2VperC/SOHC cast-iron block underneath it.

 

[BiggieShady]

So they're selling a "new" engine with intercooled variable turbine turbocharger, but with an oldschool 2VperC/SOHC cast-iron block underneath it.

It had to come eventually to a car engine analogy, didn't it it's like Godwin's law for tech forums

 

[TRWOV]

The price seems to imply that Fury X availability will be somewhat low. If AMD had priced it $50 less or so these would be flying off shelves so I guess they're trying to recoup as much cash as posible from sales.

Nano seems to be the GPU to watch,

 

[RadFX]

What!? Interposer lithography has nothing to do with rop count. Just reading that analysis gave me nausea.

What AMD did wrong, in my opinion, is lazily reused 1x the front-end of Tonga and 2x its back end, while merely doubling CU count. There's a bottleneck hidden somewhere there.

They thought memory will bail them out in the end. It did, to an extent. Just not enough.

So they're selling a "new" engine with intercooled variable turbine turbocharger, but with an oldschool 2VperC/SOHC cast-iron block underneath it.

Bad anology. My old school non-variable Turbo, SOHC, cast iron block Daytona regularly put the boots to higher HP/Torque cars with DOHC, VVT and other new tech.


I don't see any reason to purchase AMD's new GPU's, The prices are outrageous (Nvidia too) and all I'd have to do it pick up another R9 290 to get R9 295 x2 like performance...which is cream of the crop right now.

 

[hero1]

The price seems to imply that Fury X availability will be somewhat low. If AMD had priced it $50 less or so these would be flying off shelves so I guess they're trying to recoup as much cash as posible from sales.

Nano seems to be the GPU to watch,

With the way things are, I wouldn't hold my breath for Fury Nano. I'd more likely look to non X version with custom PCB designs and even then, you're better off with whatever you got unless they somehow, magically, improve the card before launch, i.e., tweak the ROPs to 96 or more. Dreamland.

 

[BiggieShady]

Speaking of pcb-s

 

[hero1]

Speaking of pcb-s
View attachment 66021

ROPs ROPs

That's one sexy dual GPU PCB for sure. Let's hope they go back to drawing board and make some changes in ROPs count, lower the shader units, and let the Fury out of the dungeon lol

 

[Casecutter]

They said it was the best card in the world. Their pre-release benchmarks were evidently lies.

Agree they need to learn to muzzle the executives' are rouge and say that crap. While I think Lisa SU said something to that effect. PR should watch such talking points from her.

As to their 4k B-M those were more I believe part of the launch kit for "accepted typical" performance, they where not meant circulate in the rumor mill as an AMD consumer proclamation as what most presented them as. That said I would absolutely hope someone independently verify and re-created their results. AMD provided all the detail and settings, I think it would be fair for a reviewer to revel them as either truth or lies.

It may be that so many reviewer just run max/ultra to provide reusability of the data down the road and call it day. It looks as though AMD is showing what presentable today in term of real or usable FpS with various adjustments. Honestly, if those charts are considerable different from other identically configured set-ups I'd really like to see the deviation. Sure you might never perfectly match their charts, but the deviation should not be that great while consistency of a trend should.

Honestly, I think AMD might get right in there with the 980Ti @ 4K, especially if more playable setting are presented, and not aided with proprietary algorithm (or yes more copasetic to AMD). Some of the reviews are showing only a couple of FpS difference and seem to spare back and forth.

 

[Nihilus]

So time to speculate on the Fury pro now? We can make some educated guesses. I say slightly better than
GTX 980 performance at $550. Most likely worse $/performance than the GTX 980

If you must go Team Red, the 295x2 is now $660 - a card that beats the Titan X in every benchmark.

Hell, one of the board partners could cook up an updated 7990 based on the 380x with more vram and AIO cooling that would match this for less money.

It feels like we are going back in time with these guys.

 

[btarunr]

So time to speculate on the Fury pro now? We can make some educated guesses. I say slightly better than
GTX 980 performance at $550. Most likely worse $/performance than the GTX 980

Yuo'll hear more about Fiji Pro only in July.

 

[N3M3515]

Just no.

By your own argument, the similarly priced 980ti beats the Fury X.

BEATS???
2% is a beating?? 50fps vs 51fps for example LOL, it is the same performance.

I think this card should be priced at $600 just because of the poor oc potential and medium 1440p perf.

 

[semantics]

Sure NVIDIA could do improvements too but Fiji is new silicon. Are the first drivers out (ones that aren't even publically downloadable) really going to be the best? As others have said, DX12 is irrelevant "now" because there are no games that use it.

You're talking about a few percent improvement for most end users. The only large driver improvements i've ever seen are dual-gpu single card. Else you're hoping that it will be finally on par with the 980ti. Which still makes it underwhelming.

 

[Fluffmeister]

BEATS???
2% is a beating?? 50fps vs 51fps for example LOL, it is the same performance.

I think this card should be priced at $600 just because of the poor oc potential and medium 1440p perf.

Hey, it turns out stock reference 980 Ti's are deadly, especially at the same price point.

But you're right, Fury X is overpriced.