The Division 2

Discussion in 'Games, Gaming & Game-demos' started by UnrealGaming, Jun 10, 2018.

  1. JonasBeckman

    JonasBeckman Ancient Guru

    Messages:
    17,004
    Likes Received:
    2,535
    GPU:
    AMD S. 5700XT Pulse
    The rogue agents in Warlords were a little bit too effective in stopping the player...
    (At least they don't ability spam like a certain boss did.)

    Could be any number of things, unless it could be narrowed down as more likely to happen on certain areas of the map or around certain encounters it's pretty hard to specifically get a lead on what is causing things to crash plus open world and random encounters and events it could even be some slight glitch or other problem.
    (What was that one crash bug in Half-Life 2 some QA tester found again, run in a circle 40 something times in one area of this one map and then one more and it would immediately crash. Somehow someone found that.)


    EDIT: For the game specifically there used to be a few areas with missing geometry and see through little holes and such take that with culling or lack thereof or hitting some math or other error and some small pixel sized hole could be messing everything up.

    Not entirely likely to be the case but part of the nature of how difficult it is to go over a large map and open world game and check on everything.

    EDIT: Post TU8 there was this one skyscraper in the Washington map that would fail to render properly from one side so you'd see right through it and part of the rest of the map, didn't crash but caused a hefty framerate hit for how much the GPU had to be doing suddenly that shouldn't be visible normally though usually these glitches are smaller or you just have minor clipping or floating objects that look off but don't do anything problematic.
    (Well other than wonky collision boxes and animations and yeah that didn't count as being in cover apparently..)

    Probably fixed now, I stopped playing shortly after that was released and wrapping up the Warlords expansion and content.
     
    Last edited: Jun 13, 2020
  2. AsiJu

    AsiJu Ancient Guru

    Messages:
    6,644
    Likes Received:
    1,713
    GPU:
    MSI RTX 2070 Armor
    Wacky DX12 aside how's the game these days?

    I suppose there's not much point in returning to game without the WoNY DLC?
     
  3. Corrupt^

    Corrupt^ Ancient Guru

    Messages:
    6,962
    Likes Received:
    296
    GPU:
    ASUS 1080GTX STRIX
    That kind of describes it, sometimes I can run for an entire week with no crash, only to suddenly crash 3 times on an evening. Anyhow, I'm experimenting now with a tool called nvoclock to lock the clock speed in place (because even if you underclock a factory overclocked GPU, it still often goes over NVIDIA reference specs due to GPU boost 3.0) and will see from there. But I can only be sure if it's stable after 2 / 3 weeks because the crashes are so difficult to reproduce.
     
    Last edited: Jun 13, 2020
  4. AsiJu

    AsiJu Ancient Guru

    Messages:
    6,644
    Likes Received:
    1,713
    GPU:
    MSI RTX 2070 Armor
    ^ yeah same for me, sometimes days without crashing, at others crash after 15 min of play and everything in between.

    Mind I did get lots of CTDs with Division 1 and increasing SOC voltage did resolve them (with 1st gen Ryzen) but that didn't help here (3rd gen Ryzen).
    Neither did any downclocking etc.

    While I suppose it's theoretically possible all the crashes are hardware caused, I find it very unlikely, given how many people have these while having no issues anywhere else.
    Even if some games are very good at triggering instabilities and Division games seem to belong to that group.

    As an example I still - sometimes - use consecutive runs of Metro 2033 benchmark to stability test my system, it seems to crash way easier than anything else when doing 10 ... 30 runs in a row.
    Probably as it stresses CPU and GPU simultaneously while also using memory and audio hardware.
    With a heavy yet realistic load.
     

  5. Astyanax

    Astyanax Ancient Guru

    Messages:
    8,027
    Likes Received:
    2,672
    GPU:
    GTX 1080ti
    And you're not correct.

    Welcome to the stability issues of modern silicon electronics, until you've plugged in a serial debugger and watched the data moving between components you've got no idea what you're on about and your dismissals of the issue being rooted in hardware / cec are just useless hot air.

    You cannot exist as an educated technician or hardware engineer and proclaim that an issue that does not exist for a majority of users with a given set of hardware but does for some portion of that same set of hardware is on its own a driver issue.

    I suspect as more and more usage of DX12 happens more and more cases will begin to appear, where dx12 compute is used.

    Also you're over-estimating the effected users based on a statistically insignificant amount of internet posters.

    For just about all driver caused bugs you get a consistent reproduction of the issue once you've got the STR nailed down.

    That Divison 2 behaviors being so eratical implies either randomised process corruption as a result of inconsistent data generation in the game itself, or the hardware.

    And to make it absolutely clear, no two runs of a process and no two boots of an OS are exactly the same, the display driver could be mapped to a different region of kernel memory between boots, a specific shader could execute on a region of the chip and crash or corrupt but on a subsequent run it executes on a different region that has no problem - and the task operating in the affected region completes just fine because its not critical data.

    Take Gears 4 for example, the pascal crashes on that were a result of drivers and nobody was left unaffected.
    Pascal firefox crashes, everyone was affected but the content of websites determined the crashes, more likely to see it on twitch or google image search when decoding images than if you were on Wikipedia or google site search.

    People have forgotten how the Asic quality of fermi gpu's was a factor when 300 branch drivers were getting less stable. People in the 70 and 80%'s+ had no issues, meanwhile people less then 50 were constantly seeing browser TDR's when they didn't lock the performance state.


    Also that's another factor: Some components have decayed area's of silicon that don't respond well to dynamic voltage adjustment's. Clock generated processing is sensitive to noise patterns, and this is for both the data being generated as well as the data being stored.

    Today there are CPU's getting RMA'd for various reasons,
    a PCIE controller being fine on pascal chips, but multiple different 2080's would crash and TDR on it.
    Ryzen 1 segmentation fault (claimed not to affect windows but.... eh, errata usually find their ways into everything)
    Ryzen 2 and 3 pcie faults / high ISR latency and whea errors
    Skylake HTT corruption

    "CPU Errata Are the Rule, Not the Exception"

    GPU's also have Errata, they also have cases where math (especially FP) generates as expected with a chip at X.0 Ghz, but not at X.5Ghz, one of the reasons why the CUDA performance state decreases vram speed btw.

    And this is a throw back to the critical code statement, a chip can produce defective results that do not affect the overal impact of the application, since some deviations are acceptable (see the results for cinebench's opengl and luxmark for instance)

    Point of fact: The only reason why we see Artifacts on defective chips is because the math is NOT in critical code in these cases the chip is still generating the wrong results.

    Games that use Async Compute are doing more critical work on the gpu.
     
    Last edited: Jun 13, 2020
    JonasBeckman likes this.
  6. JonasBeckman

    JonasBeckman Ancient Guru

    Messages:
    17,004
    Likes Received:
    2,535
    GPU:
    AMD S. 5700XT Pulse
    Will be interesting to see if more stability and benchmark tests can incorporate more modern API usage (Meaning based on DirectX 12 or Vulkan primarily.) and scenarios doing more heavy work and stressing the cards to possibly help work out instability and even faulty hardware including factory overclocks or even GPU's that seemingly do work OK in many games but then there's this one or two cases were it crashes.

    As a AMD GPU user on the Navi series of the 5700 XT specifically it looks like a handful of D3D11 titles are really good at wringing out memory issues like Witcher 3 being sensitive and for D3D12 games like Battlefield apparently hit problems very early making for a particularly good stability tester when doing overclocking especially under volting which is popular here on the AMD side of things.

    More random and alternate modes for how to utilize the VRAM rather than fixed testing or specific stress testing or even system RAM seem to also hit more problems compared to running a set of software doing some specific patterns although a few of these do have a good reputation too. (Not stuff that AMD and NVIDIA has profiles for where it holds the GPU back or has other specialized stuff applied too though that isn't going to be very effective other than possibly testing power consumption or raising the GPU's heat output.)


    EDIT: Not the best as usual at formulating it but it'd be interesting to see a more clear result against software or hardware problems though with silicon variance and different card models and bios adjustments or various components and other differences it's going to be tough to say for certain even if stuff like this might help improve stability testing.

    Microcode or other tweaks from the drivers affecting or altering the hardware could also be a thing, AMD's tuned components on their hardware and how the software (drivers.) work with it for a long time now and not always talking much about these sort of changes other than users noting alternate behavior whether good or bad. (Or a bit of both for what gets changed and how.)
     
  7. JonasBeckman

    JonasBeckman Ancient Guru

    Messages:
    17,004
    Likes Received:
    2,535
    GPU:
    AMD S. 5700XT Pulse
    Suppose neither NVIDIA, AMD nor Massive has much to say about Division 2 and D3D12 or the current state of the API on the software implementation in the engine. Pretty sure patch logs haven't said much about DirectX 12 for some time and there's little more from NVIDIA and AMD in the driver release notes.

    AMD's work led to utilizing async compute more extensively otherwise their tech video says much the same as low-level API usage overall did initially pushing the CPU and GPU performance mostly.
    I should remember but I don't think the games implementation of where it uses async compute or the usage of async compute can be disabled either.
    (Might be disabled on some AMD and NVIDIA hardware for compatibility but that would be older cards only or so I'd expect.)


    Low level API and developers hopefully knowing what they are doing especially with a in house engine although they already got some work with utilizing it from the patch for the first game so this isn't some early adaption or implementation scenario here.
    Would assume NVIDIA and Massive also did some more extensive in-house testing through the earlier stability reports and resulting driver fix which could be reliably replicated.

    Not sure what the state of the game is either, was expecting support to dwindle down after Warlords if the player count and sales didn't pick up as Ubisoft prioritized keeping the game alive but mostly maintenance although there's been at least two actual "Title Updates" and some other fixes but I have no idea if the actual Year 2 content and prior planned stuff like the second raid materialized or if there's more planned after this so it could be a smaller crew doing minor work mostly server side by this point complicating more extensive patching including testing and various vendor or API specific and also platform specific issues.
     
  8. Party Poison

    Party Poison Maha Guru

    Messages:
    1,445
    Likes Received:
    299
    GPU:
    Rtx 2080Ti
    I was literally on the QA team at Leamington, But sod it I really can't be arsed to argue.

    It is what it is.
     
    AsiJu likes this.
  9. DocStr4ngelove

    DocStr4ngelove Master Guru

    Messages:
    588
    Likes Received:
    400
    GPU:
    MSI RTX2080 Super G
  10. Corrupt^

    Corrupt^ Ancient Guru

    Messages:
    6,962
    Likes Received:
    296
    GPU:
    ASUS 1080GTX STRIX
    yay I can use AR's efficiently again...

    On a sidenote: haven't crashed yet in DX12 with the GPU clock speed & power state locked using that tool (locked at 1733, within NVIDIA spec).
     

  11. DocStr4ngelove

    DocStr4ngelove Master Guru

    Messages:
    588
    Likes Received:
    400
    GPU:
    MSI RTX2080 Super G
    Sooo ... within 15mins of playtime i found an exotic while liberating a level 4 commandpost and had 2 rogue agents attacking. Loot wasn't too bad so i guess that's good.

    Another thing i noticed: they opened up some of the latest skin outfits for payment with textiles (points you earn through playing). Esp interesting are those Midnight outfits like the Reaper guy.
     
  12. Enmity

    Enmity Maha Guru

    Messages:
    1,309
    Likes Received:
    0
    GPU:
    EVGA 1080ti FE SLI
    Anyone having big issues with FPS after the title 10 update? 1080ti user here - Usually sit around 80fps, now im getting 30ish indoors and 18fps outdoors. Restarted, fiddled with settings which are a mixture of high and a couple on mediums, tried switching between dx11 and dx12 - the behaviour is the same. Anyone experiencing this? I am trying to work out if its the title 10 update, or the windows update that also was installed within the last couple of days.
     
  13. DocStr4ngelove

    DocStr4ngelove Master Guru

    Messages:
    588
    Likes Received:
    400
    GPU:
    MSI RTX2080 Super G
    I still have my normal fps and also hardly any stuttering as before. Noticed no visible difference. Game runs great.
     
  14. PrEzi

    PrEzi Master Guru

    Messages:
    402
    Likes Received:
    268
    GPU:
    Sapphire Vega 64 LC
    To add a third to the equation / possibility ---- It might be also an issue with the drivers /// not an expert in the green camp, but I remember that the 9xx and 1xxx series (Kepler and Pascal) are lacking HW Scheduler and the drivers need to be tweaked per game. Maybe TU10 tweaked some engine internals that those optimisations no longer apply.
    For me it runs good, even a tick better than TU 9.1.
     
  15. AsiJu

    AsiJu Ancient Guru

    Messages:
    6,644
    Likes Received:
    1,713
    GPU:
    MSI RTX 2070 Armor
    Maxwell and Pascal ;)

    Both actually have a HW scheduler but Maxwell has a static one, meaning the game or driver would need to handle job queueing.

    That's why Maxwell never got async compute support, it wouldn't benefit anything practically when the GPU cannot efficiently allocate load between render and compute (and copy).

    Pascal has a dynamic scheduler and Turing I think made some further improvements to it.

    Turing can also execute INT and FP operations concurrently.
     
    Last edited: Jun 18, 2020

  16. PrEzi

    PrEzi Master Guru

    Messages:
    402
    Likes Received:
    268
    GPU:
    Sapphire Vega 64 LC
    Ahhh ok, I stand corrected then.... never the less... you cannot exclude the drivers -- might need some re-tweaking aswell...
    What wonders me only /// Enmity writes he faces it with both DX11 and DX12.... what kind of windows update has been installed? The 2004 (2020 1H) one ? If it is the case then some Windows settings have been defaulted (e.g. I needed to disable the Xbox overlay and other trash again etc. might be worth looking at too).
     
  17. Astyanax

    Astyanax Ancient Guru

    Messages:
    8,027
    Likes Received:
    2,672
    GPU:
    GTX 1080ti
    We can exclude drivers.

    Good hardware doesn't crash.
    Borderline hardware crashes unless you underclock or lock the p-state.

    It's that simple.

    only nvidia's diagnostics lab can ultimately rule out decayed silicon.
     
  18. LocoDiceGR

    LocoDiceGR Ancient Guru

    Messages:
    2,239
    Likes Received:
    749
    GPU:
    Gigabyte R9 380 4G
    Is the texture bug (textures not loading for an area) fixed yet?

    It was driving me crazy.
     
  19. AsiJu

    AsiJu Ancient Guru

    Messages:
    6,644
    Likes Received:
    1,713
    GPU:
    MSI RTX 2070 Armor
    Try reinstalling graphics driver.

    Sounds to me like you're either stuck at a low power state or your GPU is throttling. Suspect the former.

    You could still check with Afterburner OSD on your GPU temperature, usage and clocks with Division 2 running.
     
  20. DocStr4ngelove

    DocStr4ngelove Master Guru

    Messages:
    588
    Likes Received:
    400
    GPU:
    MSI RTX2080 Super G

Share This Page