Why does Ryzen 7 1800X performs so poorly in games?

Discussion in 'AMD Processors' started by _mockingbird, Mar 2, 2017.

  1. _mockingbird

    _mockingbird Gawd

    Messages:
    992
    Joined:
    Feb 20, 2017
    Why does Ryzen 7 1800X performs so poorly in games?

    I was expecting it to be able to keep up with the Intel Core i7-6900K.
     
  2. FrozenSun

    FrozenSun [H]Lite

    Messages:
    110
    Joined:
    Nov 25, 2016
    Impossible to say for those of us that aren't CPU engineers. Likely something in the architecture is just weak at doing what games apparently do. There's some potential for their to be a few kinks to be ironed out with how windows is handling their SMT, but I don't expect there to be any major improvements.

    I do find it odd that it competes so well with the 6900k at everything else, but then lags so far behind it in games, but I digress. It's still a good value depending on your usage. Feels like a shame though, if the gaming results had matched all the other results, it would have been the most exciting chip we'd seen in a long, long time.
     
  3. Simplyfun

    Simplyfun Gawd

    Messages:
    1,012
    Joined:
    Dec 17, 2016
    We'll probably get a definitive answer shortly as the hardcore architectural enthusiasts dig into this.

    I mean, it's fairly glaring and I know AMD has already focused on it as low hanging fruit for their next gen.

    It would be nice to have a clear and definitive answer though.
     
  4. shad0w4life

    shad0w4life Gawd

    Messages:
    690
    Joined:
    Jun 30, 2008
    Dont game at 720P Linus tech was 2 fps ahead of 7700 @ 4k which was only 4% but hey it's something. Total war and aots devs both commented their games weren't Ryzen optimized if I remember the remarks.

    Joker productions on YT 1700 @ 3.9Ghz vs 7700@5Ghz also showed completely different results doing 1080p and 1440.

    Lots of digging to be done but its sounding like there's a few bugs to be worked out.

    I'm very curious about 4k gaming test fps as well now after seeing the above.
     
  5. KedsDead

    KedsDead Limp Gawd

    Messages:
    128
    Joined:
    Jul 22, 2016
    It could be something simple, but its not a gaming chip.. I am a little disappointed in the game performance aswell.. Its not really that far behind in some cases.. it depends on your set up.. From the reviews I have seen so far, it performs better at 2k and higher resolutions.. well closes the gaps anyway..
     
  6. GSDragoon

    GSDragoon [H]Lite

    Messages:
    120
    Joined:
    Feb 24, 2004
    The Smart Prefetcher? I'd like to see some multi-threaded benchmarks where the workload is more random.
     
  7. damstr

    damstr [H]ard|Gawd

    Messages:
    1,362
    Joined:
    Jun 26, 2005
    I would've liked to seen different resolutions instead of GPU bottlenecked resolutions. I mean how many people are gaming @ 4K? Majority are still at 1080P and the next step being 1440P.
     
  8. cyclone3d

    cyclone3d [H]ardForum Junkie

    Messages:
    13,040
    Joined:
    Aug 16, 2004
    RAM latency.. maybe CPU scheduling.

    IF it really does have something to do with the scheduling between the units, then I want to see benches run with the affinity set to the 4 cores / 8 threads on one unit.
     
  9. gigatexal

    gigatexal [H]ardness Supreme

    Messages:
    7,237
    Joined:
    Jun 22, 2004
    Tomshardware's review had SMT turned off and it performed something like 15% better.
     
  10. 5150Joker

    5150Joker 2[H]4U

    Messages:
    3,137
    Joined:
    Aug 1, 2005
    Yep SMT definitely seems to be one of the culprits:

    [​IMG]
    [​IMG]
     
  11. Romeomium

    Romeomium Limp Gawd

    Messages:
    204
    Joined:
    Feb 9, 2017
    I saw that too! Let's see what happens in the coming weeks...
     
    Vader1975 likes this.
  12. gigatexal

    gigatexal [H]ardness Supreme

    Messages:
    7,237
    Joined:
    Jun 22, 2004
    Maybe a windows driver will fix it
     
    Vader1975 and N4CR like this.
  13. Romeomium

    Romeomium Limp Gawd

    Messages:
    204
    Joined:
    Feb 9, 2017
    Looking more and more likely. (Here's hoping!)
     
  14. _mockingbird

    _mockingbird Gawd

    Messages:
    992
    Joined:
    Feb 20, 2017
    The problem with SMT is obviously fixable, but that alone doesn't close the performance gap.

    [​IMG] [​IMG] [​IMG] [​IMG] [​IMG] [​IMG] [​IMG] [​IMG] [​IMG] [​IMG] [​IMG] [​IMG] [​IMG] [​IMG]
     
  15. Romeomium

    Romeomium Limp Gawd

    Messages:
    204
    Joined:
    Feb 9, 2017
    Thanks for those graphs - it shows SMT has a SIGNIFICANT impact on performance. It's MUCH closer to Intel than the initial reviews show.
     
    Vader1975 and Simplyfun like this.
  16. ecmaster76

    ecmaster76 [H]ard|Gawd

    Messages:
    1,152
    Joined:
    Feb 6, 2007
    I wonder if those were with OS in high performance and HEPT turned off like Tom's mentioned.

    PCGamer does not seem to mention the RAM they used but presumably the kit AMD sent. They also mention they heard the Gigabyte board was better for gaming performance for some reason (more recent UEFI release seems likely to me)
     
    Vader1975 likes this.
  17. Soularax

    Soularax [H]Lite

    Messages:
    94
    Joined:
    Feb 19, 2017
  18. Soularax

    Soularax [H]Lite

    Messages:
    94
    Joined:
    Feb 19, 2017
  19. NKD

    NKD [H]ardness Supreme

    Messages:
    7,677
    Joined:
    Aug 26, 2007
    That is a solid post. So its not across the board in every game. I mean I see 120 vs 134. I mean yea in that case its slower but does it really make a difference at the frame rates you are getting. At that point it just becomes solely a numbers game and has nothing to do with experience. That does show though SMT has issues like it was with intel hyperthreading. I used to disable it when playing games few years back, but that is all worked out now. May be it will just take a little optimizing.
     
    noko, Chimpee and Digital Viper-X- like this.
  20. Trimlock

    Trimlock [H]ardForum Junkie

    Messages:
    15,150
    Joined:
    Sep 23, 2005
    AMD is keen on blaming their competitor, probably poor wording. What it boils down to is games don't have a way to detect AMD's SMT and aren't using resources correctly much like what happened earlier with Intel and HT hurting their performance.

    AMD is going to have to get with devs quickly because I have zero intention to go through the trouble to turn off SMT just to play games only to have to turn it on again when I'm done.
     
    EchoWars likes this.
  21. Soularax

    Soularax [H]Lite

    Messages:
    94
    Joined:
    Feb 19, 2017
    I still remember the day when duo-core cpus came out and it performed worse than single-core chips. And then suddenly all new games recommended 2 cores and never gone back since.

    Zen is a completely new architecture. Give some time for devs to optimise before judging. This may very well be the future of mainstream gaming cpu.
     
  22. Lastan010

    Lastan010 Limp Gawd

    Messages:
    132
    Joined:
    Mar 2, 2017
    New hardware that needs time.
     
    Vader1975 likes this.
  23. Formula.350

    Formula.350 [H]ard|Gawd

    Messages:
    1,125
    Joined:
    Sep 30, 2011
    If you're referring to that post from AMD_LisaSu, I dunno. I definitely didn't get that vibe? I took "utilizing code optimized for our competitor" as exactly that. Though, while this is me nit-picking, I could be more onboard with that sentiment had it said "utilizing code optimizations from our competitor".

    However, I am aware of the Intel Compiler """issues""", but at the same time one can't ignore that they are a real thing that causes performance changes, and... well... wasn't it Intel who coded said compiler? (sincere question at the end, as I'm not sure if that's the case)
     
    Vader1975 and IdiotInCharge like this.
  24. SvenBent

    SvenBent 2[H]4U

    Messages:
    3,034
    Joined:
    Sep 13, 2008
    Shamlessplug:

    My program project mercury has the ability to "disable" hyperthreading on your main application while still keeping it on for background applications. it might help out on gaming performance on ryzen
     
  25. ellover009

    ellover009 [H]ard|Gawd

    Messages:
    1,846
    Joined:
    Jul 17, 2005
    I found this video pretty good.
    I noticed something, at 5.24 the ryzen cpu 1 is putting a lot of load on cpu1 anot not sharing the load as well at the other cpu.
    at 28s there's a large load on cpu16 on the ryzen compared to it's non virtualized core.
     
  26. Trimlock

    Trimlock [H]ardForum Junkie

    Messages:
    15,150
    Joined:
    Sep 23, 2005
    Hmmm maybe, but I think instead of talking about how their competitor has fixed their problem they should probably talk about their timeline to implement their fix.
     
  27. jahsoul

    jahsoul Gawd

    Messages:
    603
    Joined:
    Feb 3, 2011
    You know that mostly lies on the developers though.. *shrugs*
     
  28. Trimlock

    Trimlock [H]ardForum Junkie

    Messages:
    15,150
    Joined:
    Sep 23, 2005
    No, developers aren't magicians. You are responsible for your product being supported.

    Intel got with developers until they produced a solution in house, AMD needs to hit the ground running, not blaming.
     
  29. tybert7

    tybert7 2[H]4U

    Messages:
    2,634
    Joined:
    Aug 23, 2007
    All I know is, the gamernexus guy




    And this guy





    Are probably going to end up in a cage match based on their diametrically opposed takes. Disappointment and disgust vs pure optimism.
     
  30. Trimlock

    Trimlock [H]ardForum Junkie

    Messages:
    15,150
    Joined:
    Sep 23, 2005
    Was watching a video of one guy talking about the CPU, he both loved it and hated and also ... didn't have it hahah. I wish I was at my computer, this release has the most emotion tied to it..
     
    Vader1975 likes this.
  31. Luka

    Luka Limp Gawd

    Messages:
    130
    Joined:
    Aug 25, 2016
    Maybe because it is not that good....
     
  32. Ocellaris

    Ocellaris Ginger @le, an alcoholic's best friend.

    Messages:
    18,821
    Joined:
    Jan 1, 2008
    Holy shit get this guy a robe and a crystal ball, clearly he thinks he can see the future. His whole premise is "Well I see these numbers, and they are bad, but also I know what's going to happen next year. I swear". Also he never shuts up about "It has more cores" and he even says "AMD needs sales more than Intel does, support the underdog".
     
  33. Luka

    Luka Limp Gawd

    Messages:
    130
    Joined:
    Aug 25, 2016
    Also i said AMD went cheap with memory and PCIe controller and overall chipset which is just pile of garbage. AMD was cutting the cost in order to sell top end CPU for $499. By doing that AMD really fucked it up with gaming performance. There is a reason why 8 core Intel CPU costs $900...quad memory channel, good PCIe controller and so on.
     
  34. N4CR

    N4CR 2[H]4U

    Messages:
    3,833
    Joined:
    Oct 17, 2011
    Seen things like this before. It's either AMD design isn't quite as good (we are splitting hairs over a few percent in a some games here), or there is a patch/compiler/driver/bios/whatever fix needed.
    I remember when the compiler wars were going on towards the end of AMDs' CPU reign (yes a five year CPU reign over Intel which people like to forget about), you'd see some pretty big disparities between each camp.
     
  35. Trimlock

    Trimlock [H]ardForum Junkie

    Messages:
    15,150
    Joined:
    Sep 23, 2005
    Well yea, there's a reason it's prices like that but even then AMD couldn't have competed with the x99 platform at all so they targeted the z270 and I think they did a good job!

    Once board makers get more time I hope to see better options, the Asus Pro looks like the best for the price right now.
     
  36. Simplyfun

    Simplyfun Gawd

    Messages:
    1,012
    Joined:
    Dec 17, 2016
    How's the shilling pay today? I mean seriously, add to the conversation or something. ffs. Running around sniping your bitchfest into every thread is about as useful as an Intel dGPU.

    People are thinking based on other factors it should make a slightly better show in games... apparently just turning off SMT achieves that sometimes and it's led other people in the know to start doing some investigation.
     
    JustReason, rgMekanic, Reus and 3 others like this.
  37. N4CR

    N4CR 2[H]4U

    Messages:
    3,833
    Joined:
    Oct 17, 2011
    Yup and not the first time even some people running Intel for gaming turn SMT off...

    Either way I'm happy with it. Not any slower than my old 2600k at worst case and faster at anything threaded/WS which I also do quite heavily in some workflows.
     
  38. tybert7

    tybert7 2[H]4U

    Messages:
    2,634
    Joined:
    Aug 23, 2007

    The SMT issue begs some questions.

    I've heard that when intel first implemented hyperthreading they also had some teething issues where hyperthreading actually offered less performance than that feature disabled in some games. With time the issues were sorted out and it no longer caused performance decreases. But what about performance increases? Is hyperthreading a clear performance boost in most games? How many?


    The reason I am asking is because it seems clear that SMT is either causing minor performance issues, or larger ones where the framerates for a game get a decent jump with it disabled. But IF hyperthreading can actually offer a boost if the software and hardware and os (or whatever) is tuned correctly, then might AMD be leaving performance on the table by not having effective SMT for games? Such that the performance hierarchy would look something like:

    ryzen in games = working smt > disabled smt > broken smt

    ????
     
    whateverer likes this.
  39. Soularax

    Soularax [H]Lite

    Messages:
    94
    Joined:
    Feb 19, 2017
    And so does not the cpu manufacturer. They need to work together in order to achieve optimisation. Especially when you are not monopolising the market like Intel does. Just wait and see and give them a chance. Premature judgement helps no one since it is mostly bullshit.
     
  40. funkydmunky

    funkydmunky 2[H]4U

    Messages:
    2,359
    Joined:
    Aug 28, 2008
    6900K at what frequency? Stock? Surely the 1800X looks quite nice.
    In fact at stock the Ryzen line pumps it quite well. It OC's minimally so that is a factor for [H]OCP, but for everyone else, pure win if you need cores.
    2017, I need some CORES!
     
    Spirit_Retro and N4CR like this.