sapphire rx 5700xt nitro+ underperform

Discussion in 'Videocards - AMD Radeon' started by Drthrax2k16, Jul 22, 2020.

  1. Drthrax2k16

    Drthrax2k16 Member

    Messages:
    14
    Likes Received:
    0
    GPU:
    MSI GTX1080 GAMINGX
    i saw many results on superposition results with same cpu and rx 5700xt with 5800+ extreme score, mine not reach even 5300 what is the problem? maybe psu or windows?
     
  2. The_Amazing_X

    The_Amazing_X Master Guru

    Messages:
    554
    Likes Received:
    188
    GPU:
    Red Devil V64
    when you do a post like this:

    1 - system specs
    2 - issue
    3 - test in more then 1 benqmark
    4 - make sure you have max power always on the gpu
     
  3. Drthrax2k16

    Drthrax2k16 Member

    Messages:
    14
    Likes Received:
    0
    GPU:
    MSI GTX1080 GAMINGX
    ryze 3600x , msi mpg x570 gaming plus , 16gb 3000mhz, seasonic focus gx gold 80+ 550w , i tested superposition and time spy same result , under with similar spec
     
  4. Undying

    Undying Ancient Guru

    Messages:
    13,903
    Likes Received:
    3,137
    GPU:
    Aorus RX580 XTR 8GB
    Your cpu or gpu are probabaly clocked lower. 500 points is not a world of difference.
     

  5. Drthrax2k16

    Drthrax2k16 Member

    Messages:
    14
    Likes Received:
    0
    GPU:
    MSI GTX1080 GAMINGX
    why its cloked lower ? if i modify any setting in radeon , crash or reset wattman , only setting its working is power limit
     
  6. Drthrax2k16

    Drthrax2k16 Member

    Messages:
    14
    Likes Received:
    0
    GPU:
    MSI GTX1080 GAMINGX
    anyone solution ?
     
  7. SpajdrEX

    SpajdrEX AMD Vanguard

    Messages:
    2,474
    Likes Received:
    772
    GPU:
    Gainward RTX 3080
    Set power limit to 50%, set fan to 100%, try to slowly increase GPU frequency / VRAM frequency until you see any gfx anomalies.
    DRAM frequency set to 3600Mhz would help also, if you can make it stable.
     
    mas pras likes this.
  8. JonasBeckman

    JonasBeckman Ancient Guru

    Messages:
    17,004
    Likes Received:
    2,535
    GPU:
    AMD S. 5700XT Pulse
    The GPU and CPU will differ in binning and how much this then affects overall boost and AMD's fairly aggressive system here.
    CPU's a 3600 so I believe that's a lower TDP meaning something like precision boost overdrive could have effect by pushing the TDP value higher and potentially allowing a higher maximum boost whereas for the 3800 and 3900 with a higher default power target and power draw it's less of a factor.

    GPU's trying to hit something like 2200 but the actual value can be anything from 1700 to 1900 or around 2 - 2.1 Ghz factoring in lets see here.

    Heat, especially the junction temp although the Nitro has a solid triple slot cooler that should keep temps below the critical 110 degree limit though mounting and ambient temps could still see it hover around 90 - 100 degrees Celsius.
    Drivers and Wattman behavior especially the later 2020 drivers, increasing power draw around 10 - 20% can help but from memory the Nitro is already pushing into 200+ watt and above that you have diminishing returns plus AMD's default 1.2v GPU power too.

    I can do around 1800 - 1900 Mhz on my 5700 XT Pulse, dual fan instead of triple and a somewhat lower power target but nothing that can't be fine tuned via Wattman but above that I need to really increase the power draw and voltage for a very minimal performance gain although benchmark wise it could make a few points of difference as these are more sensitive.

    Memory is a thing too, 1800 - 1900 can be a important factor but binning and error correction or outright driver crashes might mean your GPU could have a difficult time reaching above 1800 or even much above the stock 1750 already.
    My own card can do around 1850 from testing but it barely changes performance at all so chances are it's running into error correction from GDDR6 and the result is that performance is unaffected or actually gets worse.



    CPU boost wise I'm not entirely sure how it works between the various Ryzen CPU models. My own 3900X hits up to 4.5 on a single core but during more regular workloads across a couple of cores it'll be 4.2 and I assume if something could load up all the cores it'd be under 4.0 with some fine tuning possible through the latest bios and chipset driver combinations but it's a complex affair to set each of the CCX units and overdrive on precision boost or PBO isn't that useful on the higher-end CPU's for a automated system.

    RAM at 3600 should give a good infinity fabric speed and then it'd be down to timing tweaks to tune latency and how this translates to performance, small gains but benchmarks will be a bit of a different situation as these are more sensitive to such changes and the resulting total score but again every memory module is a bit different and even the higher-end Samsung or Micron based chips have their variances plus the bios profile and memory compatibility and training plus voltages and also if it's 4x or 2x modules and resulting strain on the memory controller by having additional memory modules. :)



    EDIT: Looks like SpajdrEX pretty much covered most of it in the above post.

    Small tweaks and adjustments and how benchmarks can be more sensitive whereas games or applications will show mostly minor differences though there are room for overall gains too but binning and factors both hardware and software wise will mean different results for every system even with the same base components installed.
    (Same drivers and OS and all that.)


    EDIT: Even the display driver can give or take a few hundred points on it's own come to think of it, matters little overall but it's a factor if you are trying to go for a close to maximum score in specific benchmarks.
     
  9. mas pras

    mas pras New Member

    Messages:
    1
    Likes Received:
    0
    GPU:
    Sapphire RX 5700 XT
    Also i think your psu is not suffice. My spec almost same like you but i’m using 3600mhz ram and 700watt antec neo gold. Sapphire rx 5700 xt is really power hungry and using 8+8 pcie cable. my system (which more or less the same like you) take 5200watt at peak (synthetic bench, measured using ac wattmeter). use dedicated pcie cable instead daisy chain cable also help.

    in fact i bricked my rx 5700xt nitro+ when do superposition ultra when i’m using my corsair Cx 600watt gold. had to RMA And upgrade my psu to 700watt
     
  10. Chastity

    Chastity Ancient Guru

    Messages:
    2,557
    Likes Received:
    825
    GPU:
    Nitro 5700 XT
    I'd like to concur here. @JonasBeckman has already revealed that the 5700XT can hit transient peaks of 400w of power consumption. Probably most of the issues people have with their units black screening and such is either the Windows 10 issue with Alt-Tab'ing out of a game, or their potato PSU isn't up to snuff because they tried to get the smallest PSU size possible for what they thought would work. By going cheap, you do not leave any headroom for the PSU to work with. Some makers are conservative and make the PSU capable of more than what it's rated for, others do not.

    Also, by pushing the PSU to near max capacity, you are increasing the stress on the PSU, generating more heat, and being less efficient with power consumption, all leading to earlier PSU failure. I have a 1200w PSU that is 12 years old, and still going strong because I never really ever stressed it, having never gone over 50% capacity. ($189 at MicroCenter)

    I see people online recommending 600w PSU for new systems, and I just shake my head and cringe. A friend is making a new system soon, and I convinced him to get at least a 750w PSU. After pricing, he found a good 850w for only $10 AUS more. This should last him awhile, and be good for future systems.
     

  11. Exodite

    Exodite Ancient Guru

    Messages:
    1,991
    Likes Received:
    198
    GPU:
    Sapphire Vega 56
    Re: The PSU.

    It's a Gold-rated Seasonic unit, it's fine.

    I'm running a Vega 56 on their older 460W fanless, also Gold-rated, unit and I have 100W to spare under a max CPU+GPU stress test.

    That said, Vega has issues with power spikes and I can't speak to Navi in that regard. The way to test that would be set strict power limits and try UV/OC as far as it goes. Wouldn't really address the OPs problem, however.

    TLDR; It could be worth testing a more powerful PSU if you have one but don't buy a new one, as it's quite unlikely to be the issue unless it's faulty.
     

Share This Page