AMD Fury X Reviews

[Elixer]

When will AMD finally learn that with their tiny market share they can't make the rules, and they have to beat Intel and Nvidia at their own game?

Isn't that rather presumptuous?

Everyone was stung by the lack of both TSMC and GloFlo having anything available that was under 28nm tech. 14nm tech is just not available, and it will be that way until both TSMC and GloFlo get their crap in gear sometime late '16 for mass production.
Both camps had new tech coming, AMD with Artic Island, and nvidia with Volta.
They both read the tea leaves and saw 14nm wasn't going to happen in '15, and most likely not '16 either.

That left AMD in a predicament, they had 28nm parts working, but, doing a complete redesign of 14nm back to 28nm would have been incredibly expensive so, they instead chose to rework hawaii + HBM hoping that will get them by until 14nm tech is ready and affordable.
Heck, I am sure they would have rather used HBM2 as well, but, that is also delayed.

AMD did get HBM right, it seems to be working as advertised.
In this respect, they are the first out of the gate that beat both intel & nvidia to the punch. However, guess what, at least for current games, memory bandwidth is only part of the problem, they need more pipes (ROPs) to take full advantage.
But, in things where memory bandwidth matters the most, looking at AT's SystemCompute C++AMP test (http://anandtech.com/bench/GPU15/1232), it is at the poll position. This could mean that for specific tasks, Fury is looking very, very good.

Intel took their own sweet time getting to 14nm, and, as a leader, and as the 500lb gorilla in the room could afford to take their time until it made financial sense for them to get it out.

Nvidia is hoping to stem the tides with Pascal--if they can get 14nm tech *and* HBM2 sometime in '16, and AFAIK, they have no fallback, so, the current crop of cards is what you will see for the remainder of the year, and for most of '16 (and maybe into '17) as well.

I know most of us would have loved to see a completely new design to go with HBM, and AMD just didn't have the tech available to do it.
Yeah, it sucks, but, they are trying their best with what they have, and they did lead with HBM, you got to give them that.

HBM2 in '16 won't change the picture that much, a respin of Fiji won't really help either, they just need more ROPs. Will DX12 help? Possible, but, DX12 games are far into the future.

What I find the most puzzling is, in AMD's livestream, one of the designers said that the Fury can "O/C like no tomorrow, it is a O/C'er's dream" so, what is holding back the fury, is it BIOS, or drivers, or both? It came from the horses mouth, and I really doubt that Joe lied, so... Hmm.
refs: https://youtu.be/7MEgJLvoP2U?t=3760

 

[BrentJ]

Tell me why Dying Light is justified as being popular that a lot of people play it:

http://steamcharts.com/top

Witcher 3, yes no problem its an awesome game with lots of players and indeed deserving of "GOTY". But why run it with HairWorks enabled? Heck, did [H] even realize AMD users can force tessellation to 16x and get the SAME image quality but at significant performance improvements?

If any FPS needs to be included in tests, its still ARMA 3. That game is very popular for years and still crushes GPUs.

Just a point of correction, perhaps you should re-read our Witcher 3 page in regards to HairWorks being on, and off, what we found playable, and what we found by enabling it as well for tessellation performance.

 

[Elfear]

Why get a waterblock (unless you are a water jumkie). The EVGA SC+ is $680, 35db under load, and out-of-the-box is 12% faster than a reference 980 TI / 21% faster than a reference Fury X!

My current GPUs and CPU are under water so it seems silly to add an air-cooled card to the system (despite some aftermarket 980Ti coolers being very good).

Decisions, decisions.

 

[ocre]

2 reasons:

1) Steam/GOG/UPlay/Origin sales prove that most gamers don't buy AAA games on day 1.

2) When assessing the card's overall performance, we want to try to get as much data as possible. Are there are only 4-5 most demanding titles? If not, throw 10-15 of them in. TPU does, or are you suggesting it's only worth buying a 980Ti for the 5 games HardOCP tested? ^_^



TPU, Sweclockers, Computerbase, Hardware.fr, PCgameshardware, AT are all better than HardOCP by miles when it comes to assessing the card's overall performance.



Whatever you say. Have fun playing broken Batman AK and Project CARS and trying to suggest these games fairly represent the performance of a modern flagship card for future games for the next 2-3 years.

Not to mention you still didn't address the point how 980 is apparently ~ Fury but every other site besides TR and HardOCP has it crushing the 980 by 24-35% at 1440P-4K. I would totally buy an after-market 980TI and OC it but only a totally clueless person or a brand biased gamer would state things like 980 and Fury X are competitors at 1440P-4K.

I find it extremely funny how you make post like these in the same thread where you yourself cherry pick a small handful of games as some kind of proof that H is so wrong and Fiji is 35% faster than the 980. Your consistent double standards are obvious to many as well as your general distaste of the gtx980 itself
.
I can't help but be disappointed with Fiji x performance but come one dude, lets be real. You say that a hand full of games is so wrong yet when you talk about the 980 vs Fiji, you are all for it.

But using your personal guidelines, that you insist must be used (except when talking about a 980), what do we have?

Sites like TPU and anand graphs show clearly that Fiji X on average is 5% faster at 1080. It is the same speed in lower resolutions. And at 4k, it pulls a 20% lead. So your 35% faster claim is rubbish. Total and complete rubbish. Fiji is no where near 35% faster on average. If you want to cherry pick, take a look at the anandtech graph and you will find cases where fury x is zero or negative.

So, H is so wrong huh? About as wrong as you are in your 35% counter claim.........actually, you being the one preaching these things........I would say that is even worse.

Here is the big elephant in the room, one everybody seems to completely have forgot..........

The 980 overclocks like mad. Its easy and even reference cards reach 1400mhz and beyond. My card runs 1450mhz without touching the fan, leaving it in auto. This is in every single game I have. From DAge to gta5, unity, crysis3, bf4, to batman AK.

I wish Fiji was faster, I really wanted it to be faster. But don't be so all over the place and take notice of your double standards. You talk about needing lots of games but want to focus on a handful when comparing Fiji to a 980. You talk about max overclock vs max overclock, but totally seem ignorant to the 980 and its high maxwell 2 overclocks. You need to crunch the numbers here. At max overclock, as it sits today, Fiji has no chance against the 980ti or titanX.

Now Fiji OC against the 980 OC, max OC to max OC........
Fiji might have a chance. It can win, depending on the game and depending on the resolution. The thing is, the 980 ram and bandwidth overclock like mad too. Which might even make that 4k advantage shrink.

You can say what you want, hate on the 980 all you like. But I am asking you to take off them blinders for a second, and crunch the numbers. A max overclocked 980 is a much better fight for fury x to be in.

At one time, my plan was to sell my 980 and buy Fiji. I was hoping for better than titanX performance and OC headroom to boot. The Fiji that launched, its not much faster than my overclocked 980, if it is at all.

As time passed, I started to realize that fiji wasnt gonna be what I hoped. When we had no performance numbers at the AMD press unleashing, I knew it wouldn't be faster than the titanX. I spoke about this here.

It is not the Fiji I wanted to see. There is little reason for anyone with a 980 or 290x to buy these new cards. And that is the last thing I wanted to see.

 

[moonbogg]

Yes, using those few increments it has. You can pass something like 3dmark or heaven at pretty unstable clocks, but once you start gaming in lots of games you expose weak overclocks fast. 1400 is pretty solid for a core clock that never crashes on you.

That's my personal goal and I like that number. I'm going after a solid 1400 once I get my loop together.

 

[Abwx]

And at 4k, it pulls a 20% lead. So your 35% faster claim is rubbish.

Less rubbish than the 20%...

How many games did TPU use..?.

http://www.computerbase.de/2015-06/amd-radeon-r9-fury-x-test/5/

http://www.computerbase.de/2015-06/...t/6/#diagramm-rating-1920-1080-hohe-qualitaet

 

[Enigmoid]

Less rubbish than the 20%...

How many games did TPU use..?.

http://www.computerbase.de/2015-06/amd-radeon-r9-fury-x-test/5/

http://www.computerbase.de/2015-06/...t/6/#diagramm-rating-1920-1080-hohe-qualitaet

He is using the OC numbers for the 980 from the computerbase review which are easily available from an aftermarket 980. Since Fury X is all reference its a pretty valid comparison.

 

[guskline]

That's my personal goal and I like that number. I'm going after a solid 1400 once I get my loop together.

With the loop you are going to build that will be very achievable and a BRUTE!:biggrin:

 

[beginner99]

Here you go.. this will help ease the pain of no review to read, but instead, look at the bar charts.

http://www.anandtech.com/bench/product/1442?vs=1513 Fury vs 980
http://www.anandtech.com/bench/product/1496?vs=1513 Fury vs 980 Ti

Hm, there seems to be an error in Far Cry 4 results. Teh Fury X gets lower FPS at 1080p than 1440p. Makes no sense of course.

Now about pricing. This is were the main problem is. Also to show how expensive GPUs are outside of US. Currency is worth more than USD so actually $ prices would be even higher.

Fury x: 731

Cheapest Reference 980 Ti: 690
Cheapest Custom 980 Ti: 700 (EVGA GTX 980 Ti Superclocked+ )
Fast Custom 980 Ti: 762 (ZOTAC GTX 980 Ti AMP! Extreme Edition)
Hybrid 980 Ti: 879

980s go for around 550, 390x for 450, 390 and 970 for 380.

So if you want cool and quiet (water) Fury X is the obvious choice. I mean this thread is just as it was predicted. If Fury X is cool and quiet and uses less power, the goalpost will be shifted so that this doesn't matter anymore and only performance does. And look what happened? Exactly that. For cool & quiet Fury X is the clear winner. Also for small cases where no 980 Ti fits.

But I with this price and Benchmarks all over the place (very bad for some games) it's hard to recommend. I guess price will come down unless supply really that limited so they can sell all to niche users with small cases.

Time to wait for Fury and Nano. But still, for me going a cheap used 290(x) seems the best route. Just mist a huge deal couple days back. 3 290s with waterblocks for 400.

 

[Udgnim]

hopefully, the performance issues are largely driver based because AMD will have additional chances to place its Fury brand in a better light with the Fury release in July, Fury Nano some time after, Fury X2 release, and possibly some sites doing W10 testing to verify whether there are any tangible performance increases for gaming

if the Fury line ends up being limited by its ROP count, then it's poor engineering & planning by AMD

 

[krumme]

hopefully, the performance issues are largely driver based because AMD will have additional chances to place its Fury brand in a better light with the Fury release in July, Fury Nano some time after, Fury X2 release, and possibly some sites doing W10 testing to verify whether there are any tangible performance increases for gaming

if the Fury line ends up being limited by its ROP count, then it's poor engineering & planning by AMD

The drivers need far more time than 3 weeks. As happened with 7970 and 290x. 3 month plus at the minimum imo to review it properly.

 

[wege12]

The drivers need far more time than 3 weeks. As happened with 7970 and 290x. 3 month plus at the minimum imo to review it properly.

Let's say drivers need to be optimized further to reveal fury x's true performance. Realistically, how much performance gain could we expect from the fury x through driver optimization and dx12?

 

[AtenRa]

Euro prices,

Fury-X from 692.00 Euro (Sapphire)

GTX980Ti from 677.00 Euro (Gigabyte)

GTX980Ti custom from 710.00 Euro (Gigabyte G1)

GTX980Ti custom 768.00 Euro (MSI Gaming)

GTX980Ti custom from 807.00 Euro (EVGA Hybrid WC)

I sure hope the Air-Cooled model will be at 549-599 Euro and Nano at 400-449 Euro.

 

[Azix]

Same specs on all the retail cards. If they aren't any different from the review samples all we are left to wait for are drivers. Bios update maybe later I guess. I would have expected something since it has a dual bios switch

http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&IsNodeId=1&N=100007709 600566292

Not really a hard launch. lets see if people get their hands on them.

 

[Erenhardt]

Water cooling comes at the premium. Superior cooling solution is worth a lot, we were told when 290X came out.

Second grade chips are always best perf/$ parts. FuryX is not a second grade chip. There will be a Fury pro, which should offer performance close to Fury X for a much better price. With comparable air-cooling solutions, any comparisons with other similar cards will be valid.

Also, it seems it has a Tonga effect - its quite fast in some titles - like where 285 beats 280X, and tanks badly in other. They need to work on their drivers. Seems like they have a long way to go.

 

[AtenRa]

https://www.facebook.com/EKWaterBlocks

 

[3DVagabond]

power. Improvement is pretty huge

Power, heat, and noise are no longer considerations. Because there is no compatible O/C'ing software, that is the key consideration.

As someone else mentioned, there are compromises in every design. What the nVidia marketing machine does, and it starts with reviews and forums, is make their strengths the talking points vs AMD's weaknesses. Next thing you know, that's all anyone is talking about.

 

[RampantAndroid]

Power, heat, and noise are no longer considerations. Because there is no compatible O/C'ing software, that is the key consideration.

As someone else mentioned, there are compromises in every design. What the nVidia marketing machine does, and it starts with reviews and forums, is make their strengths the talking points vs AMD's weaknesses. Next thing you know, that's all anyone is talking about.

Seriously? Am I really reading this? AMD has *no* part in this huh? They didn't hype it us as if it'd crush the Titan? Weren't people not long ago on here saying it'd get 295x2 perf? If NV hadn't released the 980ti what would they have priced the Fury? Closer to 1k? I bet certain people would have been calling it a great value at that price...

 

[Silverforce11]

Actually AMD didn't hype anything. They didn't even said anything until Computex & E3, where they claimed world's fastest GPU.. but obviously they forgot about all the GameWorks titles that reviewers would bench. Including those, oops, no longer fastest!

If there's any hype, it would be from gamers like myself who were looking forward to a big performance leap to justify an upgrade. Unfortunately it fell short, but certainly would be very competitive, at $599. They expected too much premium on cool & quiet / benefits of water cooling. AMD marketing drank the NV koolaid actually to put such a premium on water cooling.

 

[SolMiester]

and............AT goes live with a motherboard review instead of being on time with Fury X review

EDIT: Apparently Ryan Smith is sick and should have it up in the next few days

He was sick when the GTX960 released too!...

 

[SolMiester]

Actually AMD didn't hype anything. They didn't even said anything until Computex & E3, where they claimed world's fastest GPU.. but obviously they forgot about all the GameWorks titles that reviewers would bench. Including those, oops, no longer fastest!

If there's any hype, it would be from gamers like myself who were looking forward to a big performance leap to justify an upgrade. Unfortunately it fell short, but certainly would be very competitive, at $599. They expected too much premium on cool & quiet / benefits of water cooling. AMD marketing drank the NV koolaid actually to put such a premium on water cooling.

Huh?

 

[wand3r3r]

It's common knowledge, GPU-Z (the normal gpu-z window) reports the base boost clock instead of the actual clock for any modern NV card without a modded bios.

What?

Show me proof that GPU-Z reports clocks incorrectly. I want images where GPU-z says one thing, and (something else?) which demonstrates the actual/correct speed.

It's a simple question, which should be simple to prove since you are the one claiming it's inaccurate (and common knowledge at that). So far you haven't demonstrated a shred of proof to dispel the accuracy of hwbot. Basically you haven't proven that 980 ti's overclock any higher than 1294. It's a simple thing to disprove, if as you claim, GPU-Z lies for NV cards.

Don't bother responding if you don't have proof, "common" knowledge claims aren't proof. Obviously I will take any proof, but so far nobody has been able to prove the "mythical" overclock statistics that are thrown around as the norm, contrary to conventional forum wisdom they appear grossly exaggerated. If I am wrong I would like to know.

tldr; Read the bold.

 

[3DVagabond]

Power, heat, and noise are no longer considerations. Because there is no compatible O/C'ing software, that is the key consideration.

As someone else mentioned, there are compromises in every design. What the nVidia marketing machine does, and it starts with reviews and forums, is make their strengths the talking points vs AMD's weaknesses. Next thing you know, that's all anyone is talking about.

Seriously? Am I really reading this? AMD has *no* part in this huh? They didn't hype it us as if it'd crush the Titan? Weren't people not long ago on here saying it'd get 295x2 perf? If NV hadn't released the 980ti what would they have priced the Fury? Closer to 1k? I bet certain people would have been calling it a great value at that price...

I'm failing to see how your post addresses what I said?

Anyone too concerned with GM200 running 20°C hotter and 10db louder than Fury? It used to be important when nVidia dominated those parameters. Not so much any more.

 

[AtenRa]

If i was in AMD, i would release the Fury Nano first to win the title over GTX980 and then with Windows 10 release, i would launch the Fury-X and Fury aircooled for the high-end segments.

 

[ShintaiDK]

If i was in AMD, i would release the Fury Nano first to win the title over GTX980 and then with Windows 10 release, i would launch the Fury-X and Fury aircooled for the high-end segments.

After the Fury X, I think we all want to see Nano benchmarks before making any conclusion about how it performs.