AMD Fury X Owners' Thread

Discussion in 'Videocards - AMD Radeon' started by Fox2232, Jun 25, 2015.

  1. Fox2232

    Fox2232 Guest

    Messages:
    11,808
    Likes Received:
    3,371
    GPU:
    6900XT+AW@240Hz
    Guys, here we come. This is place for sharing experiences, not insults.
    Share your findings, don't flame.

    If you have something of value, come and let us know. If you have nothing of value (no 1st hand experience), please stay away.

    Thank You all for keeping it clean and nice.
     
    Last edited: Jun 25, 2015
  2. Texter

    Texter Guest

    Messages:
    3,275
    Likes Received:
    332
    GPU:
    Club3d GF6800GT 256MB AGP
    Yeah this should be better than some idiotic gloat thread.

    I've been looking at reviews and every time I see Fury X beating a 390X by a mere 15% margin at 1080p I get the idea that there's still some performance left that needs to be found...not in the least because it has 45% more SPs lol. Makes you wonder what performance levels will be six weeks from now.
     
  3. mR Yellow

    mR Yellow Ancient Guru

    Messages:
    1,935
    Likes Received:
    0
    GPU:
    Sapphire R9 Fury
    Has anyone done some 3dmark DX12 testing in Windows 10?
    That should be interesting and should give an indication of future performance.

    I also think there's improvement to be found in future drivers.
     
    Last edited: Jun 25, 2015
  4. thatguy91

    thatguy91 Guest

    Too true. As I mentioned in the review thread (although I think it was overshadowed by the trolls...) I believe the Fury X current driver is more of a temporary solution. There was absolutely no point in AMD fully optimising the Windows 8.1 driver for it since Windows 8.1 has 5 weeks of life left. This means the drive is effectively kind of working off the 390X optimisations (since this is a card refresh). This is why there isn't the performance different shown. It probably also can't make use of the extra memory bandwidth afforded by HBM either. Realistically, you would have to be an idiot to want Fury X, the latest hardware for the rest of the machine, and still be stubborn not to move from Windows 7 and 8.1. The ironic thing about their stubborness is if they weren't 'forced' to move, but the upgrade was still free, they would probably upgrade!

    Anyways, the focus the past few months has been on the Windows 10 driver. Not specifically just DirectX 12, but also reducing the DirectX 11 overhead and optimising the use of card resources etc. We've had a taste of this in the leaked 1040 driver build, but the real performance benefits will come to light come Windows 10 release (if not beforehand). This is in AMD's advantage, since the Nvidia drivers already got the 'DirectX 11' love in response to AMD's Mantle last year. In other words, AMD has more headroom potential!

    As above, and yes I think DirectX 12 will shine on this, and other GCN cards as well. You cant' really do any benchmarks yet, because the current leaked Windows 10 drivers don't support Fury (I believe anyway). It would also be misleading since there could be changes to the Windows 10 performance come RTM, as well as the drivers.
     

  5. Fox2232

    Fox2232 Guest

    Messages:
    11,808
    Likes Received:
    3,371
    GPU:
    6900XT+AW@240Hz
    Just for Fun of it:
    Me - lazy plagiator.
    [​IMG]
    One out, One In:
    [​IMG]
    GPU-Z
    [​IMG]
    HWiNFO64 does not recognize VRM control chip yet - version: 4.65.2540

    Side note: I ordered mine 5 minutes after it was listed. Guy in shop told me that last one was sold 10 minutes afterwards.

    Side note 2: Bundled with DP => DL-DVI-D adapter, max resolution available @120Hz for this is 1440x900. Other wise it goes over 1920x1080 @60Hz.


    Benchmarks:
    (16.15.1009-160213a = Vulkan launch driver)
    Luxmark (I have same old build from day I got HD7970):
    Code:
    Test 1		262K Trianlges		score		util.
    HD7970		AMD release		13266		100%
    HD7970	GHz	AMD 2013-10b		17641		97%
    Fury X		15.150.0.0 W8=>10	29370		93%
    Fury X		16.150.1009-160213a	33070		100%
    
    Test 2		488K Triangles
    HD7970		AMD release		1534		100%
    HD7970	GHz	AMD 2013-11b8		2523 		98%
    Fury X		15.150.0.0 W8=>10	3474		95%
    Fury X		16.150.1009-160213a	4013		100%
    
    Test 3		2016K Triangles
    HD7970		AMD release		827		100%
    HD7970	GHz	AMD 2013-10b2		1258		99%
    Fury X		15.150.0.0 W8=>10	1753		100%
    Fury X		16.150.1009-160213a	1841		100%
    Posting original driver for HD7970 score, best ever score and Fury X with driver modified from W8 to W10 by me => Therefore no W10 WDDM2.0 optimizations.
    GPU reached 45°C in continuous unlimited run after 10 minutes. Was stable @1050MHz. Running 1.2V.
    API overhead Test vs HD7970:
    API-Overhead
    1st is 15.200.1023.5 (no optimization driver on W10) HD7970 @925/1375
    2nd is 15.200.1040.0 (optimized driver on W10) HD7970 @925/1375
    3rd is 15.150.0.0 (no optimization driver for W7/8.1 hacked into W10) Fury X @1050/500
    4th is 15.200.1040 (optimized driver on W10) But not Fury X code - using Tonga code instead. Fury X @1050/500 - But not reaching even HD7970 scores.
    5th is 15.200.1040 (optimized driver on W10) But not Fury X code - using Tahiti code instead. Fury X @1050/500 - Got even closer to HD7970 scores.
    It is quite clear that release driver is not able to feed Fury X properly even with Tonga code path. I hope to see Fury X code in W10 test driver soon.
    API Overhead Test Fury X with Different Drivers:
    API Overhead
    1st launch driver for 8.1, no DX12.
    2nd 15.200.1040.0 Tonga code path.
    3rd 15.200.1040.0 Tahiti code path.
    4th 15.200.1023.10 Fiji code path. (quite best in everything. On DX11 just a little bit)
    5th 15.200.1046.0 (15.7 WHQL)
    6th 15.200.1055.0 (original does not have code path for Fiji, but inf config added by asder00. So not fully Fiji aware.); While it does best in draw calls, its gaming performance is under 1023.10/1046.0 and crashes firefox in case you have game open and then open new tab.
    7th 15.200.1046.2 Seems like it has improvements from 15.200.1055 but does not cause crashes.
    8th 15.200.1062.1002 (15.7.1 W10 release driver)
    9th 16.150.1009-160213a (Vulkan release driver)
    Heaven benchmark
    [​IMG]
    Witcher:
    Everything Maxed.
    1080p: 73~120fps (limit set in RP)
    1440p: 68~120fps (limit set in RP)
    2160p: 59~115fps
    Not really worth even 1440p since 4xAA on 1080p looks exactly same as there are no detail textures/geometry. GPU mostly idles on 1080p/1440p even when fps is close to lower limit.
    Witcher 2:
    Everything Maxed, includes AA+Ubersampling.
    1080p: 40~90fps
    1440p: 67~113fps (I disabled Ubersampling here, but kept AA and all other things.)
    2160p: 35~60fps (I disabled Ubersampling here, but kept AA. Had to disable DoF/Cinematic DoF, otherwise conversations crumble to 10~15fps)
    Not really worth 4k here, rather enjoy all eye candy and fluid movement on 1440p which looks better than 1080p with Ubersampling.
    Witcher 3:
    At very good custom settings (visually) I used with HD7970, it was running outside in villages/forest 40~48fps at 100% utilization.
    Fury X does at very same settings 94~105fps. No complains here.
    Found nicely heavy location (Forest area, standing/running right next to river; small village across the river.):
    1080p: 93~107fps
    1440p: 68~73fps
    2160p: 34~40fps
    -------
    W10 - 15.200.1023.10:
    1080p: 93~109fps
    1440p: 65~77fps
    2160p: 33~42fps
    Practically no improvement since GPU was 100% utilized on all resolutions.
    Battlefield 4:
    Some review stated that mantle does in BF 4 now worse than DX11. That I can confirm for release driver meant for W7/8.1. No W10 native driver yet.
    But how it runs on Mantle: 1080p maximum details (including max AA, HBAO, everything possible) + maximum FoV: 125fps +
    4k (1080p with downsampling from 4k) again maximum details, but AA turned down to 2x: 55fps+
    On DX11 it runs better like 19~30%. I would say there is something wrong, but well, that is what I measured.

    On test range in menu, which adds additional blur, I tested FPS which is greatly reduced due to this blur. I got (all maximum, including fov):
    Code:
    	DX11	Mantle	difference
    1080p	126	106	19%
    1440p	64	52	23%
    2160p	41	31	32%
    And I must say I loved how stable frame-time graph looks like. Which corresponds with very smooth gameplay feeling.

    Multiplayer - Lancang Dam (64-man; Everything maxed except Deffered AA):
    1080p: 88~119 fps
    1440p: 72~97 fps
    2160p: 55~75 fps
    --------
    Multiplayer - Lancang Dam (64-man; Everything maxed except Deffered AA): W10 - 15.200.1023.10
    1080p: 101~146 fps (74% GPU utilization)
    1440p: 87~128 fps (96% GPU utilization)
    2160p: 58~75 fps (100% GPU utilization)
    Another big boost, not due to graphical optimizations but GPU utlization. Not much space left on 1440p, but 1080p has still space for improvement.
    WH 40k: DoW 2 - Retribution:
    2160p: maximum details + Adaptive AA + texture LOD settings "-1" enforced via RadeonPRO: 74.33fps average in build it benchmark.
    2160p: Clean windows 10 (10162) install, maximum game details, no external settings enforced: 86.57fps
    2160p: W10 driver 15.200.1023.10: 92.67fps
    2160p: W10 driver 15.200.1062.1002: 91.92fps
    WH 40k: Space Marine:
    Maximum details + Adaptive AA + texture LOD settings "-1" enforced via RadeonPRO.
    1080p: 120fps (limit set in RP)
    1440p: 108~120fps (limit set in RP)
    2160p: 96~120fps (limit set in RP)
    --------
    W10 Driver 15.200.1023.10 Fiji
    1080p: 128~205fps
    1440p: 130~202fps
    2160p: 95~192fps
    Improvement on lower resolution is notable. At 2160p card is quite utilized, so no improvement there:
    [​IMG]
    It is 3times same section of game where hordes of smaller orks are sieging camp on ground, and then at end there is 1st ground boss fight.
    Killer Is Dead - Nightmare Edition:
    Maximum details + Adaptive AA + texture LOD settings "-1" enforced via RadeonPRO.
    1080p: 120fps (limit set in RP)
    1440p: 115~120fps (limit set in RP)
    2160p: 105~120fps (limit set in RP)
    This thing runs on Unreal Engine 3, horrible controls.
    Payday 2:
    Everything Maxed, including DoF, Blur and stuff I hate (just for testing).
    1080p: 100~120fps (limit set in-game)
    1440p: 92~120fps (limit set in-game)
    2160p:69~95fps
    --------
    W10 Driver 15.200.1023.10:
    1080p: 109~135fps (limit set in-game)
    1440p: 107~135fps (limit set in-game)
    2160p: 87~123fps
    Quite a change too.
    CS:GO
    Everything Maximum, 8x SSAA+FXAA (just about everything possible).
    1080p: 155~280fps
    1440p: 152~255fps
    2160p: 123~180fps
    And this is in spectate mode flying around. Because in 1st person view fps is like 25% higher, and I wanted to put some load to see some worst case scenario since it did so well.
    DOTA 2:
    Maximum everything, tested in spectate mode with maximum zoom out to see most geometry.
    1080p: 78~120fps
    1440p: 73~120fps
    2160p: 75~120fps
    All lowest were in teamfights and GPU was not fully utilized, CPU bottleneck/bad coding.
    Side note: DOTA 2 does not like downsampling as mouse cursor is not at proper place. At 2160p game thinks it is only at top left quarter of screen. (easily fixed if you start game after setting desktop to 2160p too. Fix delivered by CalculuS.)
    --------
    W10 - driver 15.200.1023.10: (again spectating with max zoom out + unlocked fps with: fps_max 300)
    1080p: 99~144fps ; around 25% GPU utilization
    1440p: 87~135fps ; around 45% GPU utilization
    2160p: 74~118fps ; around 95% GPU utilization
    Now there is again improvement in min framerate, but seems like there is other than driver limit too. Because CS:GO (same engine) just runs unbelievable fps even on older driver.
    Warframe:
    Everything maxed out.
    1080p: 270fps+
    1440p: 250fps+
    2160p: 138fps+
    Dead Island:
    Maxed out. Texture LOD settings "-1" enforced via RadeonPRO.
    1080p: 120fps (limit set in RP)
    1440p: 107~115 (locked by game, even main menu before loading any geometry runs on that fps)
    2160p: 107~115 (locked by game, even main menu before loading any geometry runs on that fps)
    In all resolutions fps was limited by software and GPU not fully utilized. DirectX 9 game.
    --------
    W10 Driver 15.200.1023.10 Fiji code path:
    1080p: 207fps+
    1440p: 198fps+
    2160p: 124fps+
    What a change. And btw, that famous Stutter this game had is completely gone.

    BTW: Following of games above are in TOP20 currently played on steam:
    DOTA2 - 1st; CS:GO - 2nd; Payday 2 - 11th; Warframe - 12th; Witcher 3 - 14th.
    Untested from TOP20 (which I have): Team Fortress 2; Civilization V; Elder Scrolls V: Skyrim; CS 1.6; Left 4 Dead 2
    Highest VRAM Utilization (OS+Game = total): The Witcher 2 - 3311MB; 2160p all settings maxed.
     
    Last edited: Feb 20, 2016
  6. Undying

    Undying Ancient Guru

    Messages:
    25,206
    Likes Received:
    12,611
    GPU:
    XFX RX6800XT 16GB
    Nice one. I'm waiting for a Nano or Fury (non-X). Dont really have that much to spend. Next month maybe? :p

    I'm hoping Fury (non-X) will have similiar performance for less cash, just like 290 did vs 290X.
     
  7. Ryu5uzaku

    Ryu5uzaku Ancient Guru

    Messages:
    7,546
    Likes Received:
    608
    GPU:
    6800 XT
    And it pretty much beats the 390x on 4k with 45% margin so yeah.
     
  8. Ryu5uzaku

    Ryu5uzaku Ancient Guru

    Messages:
    7,546
    Likes Received:
    608
    GPU:
    6800 XT
    Gz on the card. I am getting Nano for htpc when that thing hits the streets.
     
  9. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    8,125
    Likes Received:
    969
    GPU:
    Inno3D RTX 3090
    Congrats! Report back with GPU-Z stats asap! :p

    As for the driver, you just need to wait for AMD to find a way to feed this thing fast enough. Get ready to jump to Windows 10 really fast :D
     
  10. Fox2232

    Fox2232 Guest

    Messages:
    11,808
    Likes Received:
    3,371
    GPU:
    6900XT+AW@240Hz
    I am on Windows 10 already for quite some time. On my 1080p it will be underutilized, but at least I can use downsampling in more games and it will be put in proper use with Valve's Vive.
    Will post some stuff from monitoring tools (GPU-z / HWiNFO64) if they will actually read it properly. Because for now HWiNFO could read only GPU temperature, not voltages, nor VRMs temperatures or power draw... (I am on bit older release, so new beta may already be able to read this chip.)

    Will get to it as I get home... have one more errand to run... My old Accelero Xtreme cooled HD7970 has to travel 40km to cousin's wife's gaming PC, and I am one to install it.
     

  11. Texter

    Texter Guest

    Messages:
    3,275
    Likes Received:
    332
    GPU:
    Club3d GF6800GT 256MB AGP
    No it doesn't. It's 25% (at TPU and Computerbase) AND that's the highest margin it gets over 390X. (in case you were serious)

    I think I did see the ~45% show its face in Unigine Heaven at Normal settings somewhere...
     
    Last edited: Jun 25, 2015
  12. mR Yellow

    mR Yellow Ancient Guru

    Messages:
    1,935
    Likes Received:
    0
    GPU:
    Sapphire R9 Fury
    My supplier is only getting stock on the FuryX on the 30 June but guess what... the dual GPU is expected on the 15th July

    SC-S9FX2 Sapphire r9-Fury X2 dual-gpu 14999.00
     
  13. lexer98

    lexer98 Guest

    Messages:
    660
    Likes Received:
    2
    GPU:
    GTX 1070 - WC
    If someone have time, can upload some Compute benchmarks ?. Im really curious how this card will perform in programs like Adobe Premiere , Vray, 3Dmax ect.
     
  14. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    8,125
    Likes Received:
    969
    GPU:
    Inno3D RTX 3090
    There is no way that on the same clocks a 4096 shader Tonga is only 25% faster than 2816 shader Hawai. It is obviously a driver issue.
     
  15. Ryu5uzaku

    Ryu5uzaku Ancient Guru

    Messages:
    7,546
    Likes Received:
    608
    GPU:
    6800 XT
    That was the best case scenario in like 1 game! :D but it shows it could if everything gets to it's place.
     

  16. kybeer

    kybeer Guest

    Messages:
    20
    Likes Received:
    0
    GPU:
    2x Sapphire 7950
  17. undeadpolice

    undeadpolice Master Guru

    Messages:
    203
    Likes Received:
    0
    GPU:
    EVGA GTX980 K|NGP|N (SLI)
    Lower the price, improve the driver support, improved aftermarket water cooling.
    I might purchase Fury X over 980ti.
     
  18. treetops

    treetops Guest

    Messages:
    7
    Likes Received:
    0
    GPU:
    GTX 670
    Hi guys, I got mine delivered this morning after grabbing the last one at the retailer. I'm not an enthusiast, just got it for VR gaming (Elite dangerous) which I can't play at the mo until my DVI-DisplayPort adaptor arrives tomorrow, but if anyone needs info I will see what I can do. A guy just asked me to post VRM temps in another thread. Where do I find those? Thanks :)
     
  19. lexer98

    lexer98 Guest

    Messages:
    660
    Likes Received:
    2
    GPU:
    GTX 1070 - WC
    GPU-Z but ... i think is not supported right now, because is a new GPU. But give a try
     
  20. mR Yellow

    mR Yellow Ancient Guru

    Messages:
    1,935
    Likes Received:
    0
    GPU:
    Sapphire R9 Fury
    Please respect OP request and keep thread clean.
     

Share This Page