Search Unity

  1. Unity support for visionOS is now available. Learn more in our blog post.
    Dismiss Notice

Question Anyone knows if 5800X3D vs 5800X makes a difference in workflow in Unity?

Discussion in 'Editor & General Support' started by ClosingTime, Nov 18, 2023.

  1. ClosingTime

    ClosingTime

    Joined:
    Nov 23, 2016
    Posts:
    230
    I am looking to upgrade my CPU from 2700X to either 5800X or 5800X3D.
    Does 5800X3D has an advantage in working in the editor?(not running apps/gaming) over 5800X?
     
  2. CodeSmile

    CodeSmile

    Joined:
    Apr 10, 2014
    Posts:
    5,108
    Probably negligible. The question really is, given general performance measurements, is a 5-15% increase in computational performance worth the extra money?

    Note that the CPU alone is just one factor. Is it still +$100 extra? If so, that's totally not worth it!

    Instead I would double the installed memory if it's only 16 GB, or quadruple it if it's only 8 GB. 32 gb does help over 16 gb, especially if you want to run other 3d apps like Blender at the same time.

    Next I would opt for an SSD that's +$100 more expensive, but not necessarily bigger, just a LOT faster, specifically write speeds. It matters a lot more if you choose between a cheap run-of-the-mill SSD for $100 or get the same size Samsung Pro EVO for $200 (prices VERY roughly). This will speed up EVERYTHING noticably, including boot times.

    Lastly, a +$100 more expensive GPU will be more beneficial than spending an extra $100 on the CPU.

    CPU almost always is the least important in these four hardware categories that determine how responsive a system is: CPU, GPU, RAM, SSD (HDD). Exceptions notwithstanding, eg heavy parallel processing intensive tasks such as C++ compilation or video encoding (if it cannot be done by the GPU) benefit most from a faster CPU.
     
  3. CodeSmile

    CodeSmile

    Joined:
    Apr 10, 2014
    Posts:
    5,108
    I did a quick test with Unity 2023.3.13f1. Created a project with built-in render pipeline and dropped in my script recompile measuring scripts.

    I got 3.1s for script recompile time after making a change as the fastest time.

    I then disabled all built-in modules, which did not affect script compile times (they do affect build size though).
    After that I removed all packages, including the IDE packages, so there were 0 packages in the project.

    The script compile time was 2.5s now. Adding the Jetbrains Rider package increased the script compile time by about 0.05s.

    So disabling packages does help with script recompile time but removing all of them (unrealistic) only makes it 20% faster. Adding more packages, especially when you leave them in there unused, will certainly waste time. So it's generally a good idea to remove any packages you don't need or stopped using.
     
  4. ClosingTime

    ClosingTime

    Joined:
    Nov 23, 2016
    Posts:
    230
    I asked about 5800X vs 5800X3D because 5800X3D has more cache. On the other hand 5800X has higher clocks.
    So 5800X is actually faster in almost all tasks, but gaming.
    So I wondered if the bigger cash might also help in certain productivity tasks too.
    But I think it doesn't.
    The higher clocks of the 5800X is more significant, it's also cheaper.
    But an upgrade from 2700X is definitely helpful.

    I don't agree that CPU is the least of your issues.
    Of course if you use HDD, you are limited to HDD performance.
    But with SSD or tasks that don't require a drive, you are most likely limited by CPU/RAM combination.
    I have 64GB of RAM, so there is no issue with capacity.

    Compilation times are very important.
    You can compile scripts hundreds or even thousands times in a day of work. So every second you shave of a compilation will save you time,
     
  5. CodeSmile

    CodeSmile

    Joined:
    Apr 10, 2014
    Posts:
    5,108
    I agree but C# compilation is a little different from C++ compilation, which as I remember is heavily threaded and CPU bound.

    These two CPUs are relatively close to each other, about 20% difference at most in theoretical tests, and much less in real world apps. Will the faster CPU drop the script recompile time from 10s to 8s? I doubt it. A lot of this is file access, and a lot is single-threaded.

    You'd have to actually get both CPUs, test intensively on the same system (they use the same socket I believe?) and measure times using scripts, then return the slower CPU.

    I bet the difference will be negligible. Something like 9.5s over 10s.
     
  6. ClosingTime

    ClosingTime

    Joined:
    Nov 23, 2016
    Posts:
    230
    I doubt that an AMD processor 2 generation ahead, is only 5% faster.

    https://www.cpu-monkey.com/en/compare_cpu-amd_ryzen_7_5800x-vs-amd_ryzen_7_2700x
     
  7. CodeSmile

    CodeSmile

    Joined:
    Apr 10, 2014
    Posts:
    5,108
    I refer to 5800X vs 5800X3D. Those are maybe 5-15% apart. ;)
    This is a performance difference you will only be able to measure but you won't notice it.
     
  8. spiney199

    spiney199

    Joined:
    Feb 11, 2021
    Posts:
    7,358
    As far as domain reloads go, which is the more time consuming part of Unity, single-threaded speed and disk speed are the biggest contributing factors - in my experience.

    The newer intel CPU's with their hybrid dual and single-threaded cores are very good for this. Namely the i9s make domain reloads absurdly fast. I use an i9 12900kf and domain reloads are about 1-2 seconds.

    Not sure if AMD has anything like this, but I would go with pure single-threaded speeds when choosing a CPU.

    Honestly CPUs are so fast these days the biggest limiting factor is drive speeds and the speed at which information travels through your mobo.

    There are massive differences in domain reload when comparing a HDD, SSD, and high-speed NVME drive. Right now I'm using Unity on a Samsung Evo 970, though I have a 990 on order (Black Friday specials, hoo-ah) so I'm curious if that improves things any.

    As for other day-to-day Editor usage, a fast CPU is a fast CPU. I doubt you'd see much other notable difference.
     
  9. ClosingTime

    ClosingTime

    Joined:
    Nov 23, 2016
    Posts:
    230
    Is there a benchmark for domain reload?
     
  10. spiney199

    spiney199

    Joined:
    Feb 11, 2021
    Posts:
    7,358
    Domain reloads are logged to your editor.log file for each session, showing how long it takes and how long each part of it took.

    Looking at my log, the last one I did before closing Unity for today took 2.3 seconds (nice).
     
  11. ClosingTime

    ClosingTime

    Joined:
    Nov 23, 2016
    Posts:
    230
    Sure, but a Benchmark is like the same scene a lot of people can try out to see and compare their hardware performance?
     
  12. spiney199

    spiney199

    Joined:
    Feb 11, 2021
    Posts:
    7,358
    There's no standard benchmark. Most folk, including Unity staff, have just used a new URP or HDRP project.
     
  13. Ryiah

    Ryiah

    Joined:
    Oct 11, 2012
    Posts:
    20,758
    A larger cache will have a positive impact on performance but it's barely noticeable in most cases. The 5900X has 66% of the cache of a 5800X3D while having 50% more cores and a 300MHz higher clock speed. Currently in the US they're within a few dollars of each other.

    Unity provides a tool for profiling the iteration times of the editor if you want to see before and after for some tasks like compiling scripts. On my 5950X (with DDR4-3600) my compile time for an empty project in Unity 2021 is about 3 seconds.

    https://forum.unity.com/threads/introducing-the-editor-iteration-profiler.908390/
     
    Last edited: Nov 26, 2023
  14. ClosingTime

    ClosingTime

    Joined:
    Nov 23, 2016
    Posts:
    230
    The 5800X3D has more cache than the 5900X or the 5950X, that's the whole point of the "3D".
    It's the 3D stacking of cache.
     
  15. Ryiah

    Ryiah

    Joined:
    Oct 11, 2012
    Posts:
    20,758
    Right, I'm aware, I'm saying that the 5900X is a better deal for workstation usage than the 5800X3D since it's the same cost but it trades a minor improvement in single-threaded performance for a massive improvement in multi-threaded performance.

    AMD's 5800X3D is mostly a proof of concept release moreso than a practical one. At least for anything that isn't gaming where the benefit is just okay. If you want an increased cache you should look at the 7000 series where you can get it on models more practical for work like the 7900X3D and 7950X3D.

    That way you're not sacrificing multi-threaded performance for a slight boost to single-threaded.
     
    Last edited: Nov 26, 2023
  16. ClosingTime

    ClosingTime

    Joined:
    Nov 23, 2016
    Posts:
    230
    5800X and 5900X are better than 5800X3D in some single threaded scenarios too.
    Because they have higher clock rates.