AMD Radeon R9 Fury X Video Card Review

Jump To:

Final Thoughts and Conclusions

980ti furyx

The AMD Radeon R9 Fury X was one of the most hyped up desktop graphics cards as it is the first consumer graphics card to ship with High-Bandwidth Memory on it. AMD was the first company to ship products using GDDR5 memory back in 2007 with the introduction of the Radeon HD 4870 graphics card, so this is the first major memory technology change to happen in nearly a decade. By going with HBM Gen 1 technology on the Radeon R9 Fury X video card it limited AMD to just 4GB of memory, but it gave the gobs of memory bandwidth and the card is certainly not memory bandwidth limited! We’ve only had the Radeon R9 Fury X for a few days, but we didn’t see any major issues with 4GB of memory while gaming on a 4K monitor.

HBM is costly right now since it’s so new, so don’t expect to see HBM come out on sub $300 video cards anytime soon. NVIDIA will be using HBM memory as well when they release the Pascal GPU sometime in 2016. NVIDIA might be late to the HBM game, but they made the decision to wait until HBM Gen 2 memory to come out as there are a number of improvements and it allows for up to 32GB of HBM2 to be used. AMD and NVIDIA will both be moving to HBM2 memory when it becomes available. The only bad news about that is Fiji was only designed to work with HBM1 memory, so we’ll have to wait until AMD’s next GPU offering in 2016 to see that come to market for the red team.

AMD Radeon R9 Fury X performance

When it comes to performance of the AMD Radeon R9 Fury X we were expecting for a clean sweep since AMD showed us that they were able to defeat the NVIDIA GeForce GTX 980 Ti in 12 different game titles at 4K Ultra HD settings. We didn’t test nearly as many titles since we don’t have automated testing like NVIDIA and AMD, but our manual testing testing showed that the AMD Radeon R9 Fury X was slightly slower than the GeForce GTX 980 Ti in most of the benchmarks despite performing significantly faster than the Radeon R9 390X 8GB video card in all the benchmarks. The AMD Radeon R9 390X wasn’t able to keep up with the GeForce GTX 980 Ti, but luckily the Radeon R9 Fury X can keep it competitive. AMD needed a really fast flagship offering and while they weren’t able to knock the GeForce GTX 980 Ti out, they certainly got NVIDIA’s attention with this product launch.

When it comes to overclocking, the AMD Radeon R9 Fury X was able to overclock up to 1130MHz before we started having issues. This is an overclock of 7.6%, but we only saw about a 4-5.5% increase in the game titles. We didn’t expect to see 1:1 performance scaling from the overclock, but this was a little lower than expected. It makes us wonder if keeping the ROP count at 64 on the Fury X GPU is actually a performance bottleneck now. AMD has kept the ROP count at 64 since releasing the Hawaii GPU in 2013. AMD will likely be able to get more performance out of this new GPU with driver enhancements though, so they might be able to get some more performance there.


The water cooler used on the AMD Radeon R9 Fury X did a great job keeping the card cool as it never got over 54C on our open air test bench. As long as you don’t die and need to restart the game, the GPU will be clocked at 1050MHz while playing games sine the GPU is kept cool and throttling to prevent overheating isn’t needed on this card!

AMD Radeon Lineup 2015


The AMD Radeon R9 Fury X runs $649, the same price as the GeForce GTX 980 Ti.

  • AMD Radeon R9 Fury X2 (What we are calling the Dual-GPU Fiji card)
  • AMD Radeon R9 Fury X – $649
  • AMD Radeon R9 Fury – $549
  • AMD Radeon R9 Nano
  • AMD Radeon R9 390X – $429
  • AMD Radeon R9 390 – $329
  • AMD Radeon R9 380 – $219
  • AMD Radeon R7 370 – $149
  • AMD Radeon R7 360 – $129

If you have been waiting on the AMD Radeon R9 Fury X video card for months the wait is finally over and we hope the card lived up to the hype. The AMD Fiji GPU used on this card will be used on the AMD Radeon R9 Fury and Radeon R9 Nano air cooled cards that are coming out later this year. There will also be a pair of them on a Dual-GPU Fiji card that is also slated to be released this year on another water cooled card.

LR Recommended Award

Legit Bottom Line: The AMD Radeon R9 Fury X video card is the fastest single-GPU card from AMD and is able to battle the NVIDIA GeForce GTX 980 Ti at 4K Ultra HD gaming resolutions!

Jump To:
« Prev 1 2 3 4 5 6 7 8 9 10 11 12 13
  • Freedom Miner

    Maybe its the particular setup used to test, or maybe something else. These benchmarks conflict with others. The bencchmarks at Tom’s Hardware, for example, shows the Fury X at parity with the Titan X. Consider, too, that AMD’s budget is far lower than nVidia’s. Give AMD time to work on drivers, and allow some time for some non-reference Fury x’s to be released, and the potential to beat the Tiatan x and 980ti with substantial numbers will reveal itself. Allow for DX 12 to enter the picture, (not 12.1, which nVidia seems to be hoping will be chosen as the main DX version used in the future, probably by paying to make it so regardless the cost) and all DX 12 AMD compatible products will likely show a far better picture that we’ve seen with their cards using DX11.

    AMD needs is to gain enough market share to give them a chance to further innovate via drivers and future design, and at long last we’ll have a healthier gpu market with competition which will benefit the consumer. If nVidia becomes a monopoly, we’ll have no one but ourselves to blame. IF we find ourselves with a one brand gpu market in the near future, it’ll be because of those among us who were uncompromising and bought yet another green team card based mostly on marginal gains due to the larger company’s hype and underhanded strategies and tactics which its known for. I rather doubt any enthusiast will be happy if that time ever comes.

  • Ed Starkey

    Why are there no OpenCL or compute test performed in this review? Every single site that has reviewed this card has omitted OpenCL and compute test. Is it because the compute performance is as bland as the gaming performance, or perhaps it’s even worse!

    • Nathan Kirsch

      Ed, what OpenCL or compute test would you like to see? If it’s free or something we already have, we can most certainly run it for you and post up the results here in the comment section or add it to the review.

    • Nathan Kirsch

      Ed, what OpenCL or compute test would you like to see? If it’s free or something we already have, we can most certainly run it for you and post up the results here in the comment section or add it to the review.

      • Nwgat

        perhaps LuxMark?

  • Rene Stanneveld

    Damn funny, there is no way a 120 rad can dissipate 500 wats.
    Thats what a 3×120 from a good brand does.

    • Nwgat

      fury is 275W, 500W is the rated cooling power of the water unit

  • Hooligan1976

    This clearly shows the 980Ti as a massive winner. We have to take into account that most 980Ti cards come factory overclocked with warranty. Some cards easily at 1250Mhz out of the box. We see that an 980Ti OC spanks the Fury X OC’ed card by a vast margin, on air cooling.

    HBM seems to make zero difference. Furthermore the 980Ti is even better in 1440p.
    Reality check neither 980Ti/Titan X or Fury X are true 4K cards, in many games they barely push 30fps. These cards are 1440p cards, that is a fact. In that category 980Ti run circles around the Fury X which oddly perform even WORSE.

  • linuxnutcase

    first off thanks for the legit review, nicely done, in my view you guys do the best in reviews, be great if you could get into youtube.
    I’m hoping that 120 rad will give us in the future tons of overclocking room, to bring the fps up.
    Either way amd did do a very good job on looks this time around.

  • Beelzy

    Your going to hafta wait for AMDs driver updates to get good performance out of any game that isnt already optimized for the 4GB of HBM or see your performance bottlenecked badly ohhh and 8gbs on the 390x/390 an only 4 gb here is just plain goofy especially with plenty of benchmarks showing games hittin almost 8gb usage its hardly gunna be futureproof. i would hafta find room for that quite large watercooler??i have a big enuff case now LOL no ty itd be apain too install and would force significant tradeoffs in my cases layout for airflow/fans setup.Been a huge AMD fan for years and it is nice to them get very competitive again…but titan/980ti killer this sure as hell isnt it……imagine nvidias next card in 6mnths with a die shrink and hbm2…….lolz

  • Etienne Boutet boucher

    Really deceiving, i was expecting alot more with all this hype around fury…

  • agentbb007

    Well definitely not the 980ti killer rumored but at least AMD is in the ball park now.

    • nem

      i disagree, still Fury can perform with better drivers meanwhile the TI is full performing with already good drivers , and too here let you this..¬¬

      • nobodyspecial

        1. Devs support the largest userbase period. They will support DX features of NV before AMD as the share is 75% to 25%.

        2. NV can wring more out of their card too as Extremetech says Metro LL up to 25% faster with 353.30 vs. 352.90:
        They will keep doing it as Maxwell 2 is a new gen too. TitanX and 980ti just came out…LOL. Keep living the fantasy.

        3. DX12 features are coughed up by Microsoft NOT Nvidia or AMD and these are MISSING from AMD:
        Feature Level 12_1 has Raster Order Views, Conservative Raster and Volume Tiled Raster enabled on the API. They won’t just NOT matter because 25% market share wishes it to be so.

        4. Driver overhead means almost nothing in reality and you left out the fact that there are ZERO Dx12 games and won’t be for a while so you’ll be dealing with DX11 driver overhead which shows the EXACT opposite:
        LOL – in singlethread and multithreaded. Even if Win10 massively takes over (doubtful with many happy win7/8 users and even more missing dx12 support on their current cards anyway), games won’t massively go there until the hardware numbers are there. You don’t write games for people who don’t massively exist yet.
        WOW, driver overhead 3-4x faster on METAL on apple devices vs. OpenGL ES. So games will magically be 3-4x faster now…But in reality…It generates 10% in Manhattan/TrexHD…ROFL. How much does 33% end up being when apples is 300-400% faster on the same devices and only ends up 10% at the end of the day in a game situation? It’s kind of like saying WOW, HBM1 is bandwidth is massive and AMD now kicks butt in all things. But then bandwidth really isn’t the issue today so NV still wins everywhere…ROFL.

        No review site says buy FuryX over 980ti especially for 95% of us which run 1440p or less and no matter what when considering OCing (even 4k stupid then for FuryX). Futuremark even says don’t use their tool as a benchmark for this. Driver overhead is the WORST case scenario that COULD happen if a dev decided to throw as many draw calls as possible at the screen in a game, and ZERO games do it, and even star swarm is NOT a game and never will be. Draw calls/driver overhead is a small part of a game. As the metal tests show, even a 400% improvement nets 10% today. AMD’s cards won’t magically speed up massively in gaming when win10 hits. Heck they still have to put out a WHQL driver (not one since Dec8 2014!) while NV already has one for win10.

        See pcper, techreport, maximumpc, hardocp, hothardware, hexus, hardwarecanucks, techpowerup, etc etc. The best you get is not a bad buy if you’re an AMD 290 owner 😉 Not that it’s a bad card, just that it isn’t as good as 980ti. Checking those also gets you a larger bunch of games to judge and see how often 4K even goes NV (more often then not and many times by >20%). You need to be beating the competition not your old gen.

    • Nwgat

      its more like a ninja “small, cool, quiet and sneaks up on you”
      meybe fury had a mad mood? tomorrow it will be 1% faster than 980ti?
      3% is within of the margin of error

      Fury is around 3% slower yet its water cooled, almost half the size of 980ti…

  • tacoslave

    I was actually expecting it to be slower than 3% average compared to the 980ti not bad,plus its already watercooled, shame about the over clocking though.