User Tag List

Page 2 of 4 FirstFirst 1234 LastLast
Results 41 to 80 of 127

Thread: UE4 - intel 7700K or AMD 1800X?

  1. #41
    0
    Quote Originally Posted by Zeblote View Post
    Still 20-30 min? Are you sure?

    My 6700k takes about that long, and the ryzen should be much faster... is it even using it fully?
    My i7 5820k @4.2 Ghz also takes 20-30 minutes, so your 6700k should take a bit longer. Compiling UE4 always uses 100% of the CPU.

  2. #42
    0
    Samaritan
    Join Date
    Sep 2015
    Posts
    123
    Can someone with Ryzen on board make benchmarks with UE4?

  3. #43
    0
    Quote Originally Posted by Zeblote View Post
    Still 20-30 min? Are you sure?

    My 6700k takes about that long, and the ryzen should be much faster... is it even using it fully?
    Same, 20-30 min for full editor rebuild on i7 6700k.

  4. #44
    0
    Quote Originally Posted by John Alcatraz View Post
    My i7 5820k @4.2 Ghz also takes 20-30 minutes, so your 6700k should take a bit longer. Compiling UE4 always uses 100% of the CPU.
    Just tried again to make sure, took 22 minutes.

  5. #45
    0
    Samaritan

    Join Date
    Mar 2014
    Posts
    115
    Quote Originally Posted by Zeblote View Post
    Still 20-30 min? Are you sure?

    My 6700k takes about that long, and the ryzen should be much faster... is it even using it fully?
    Yea. Wondering if it is using all 8 cores of the ryzen for compiling ? The multicore performance has been much faster in all benchmarks i have seen so far.
    Michael Hegemann - https://twitter.com/HegiDev

  6. #46
    0
    Quote Originally Posted by HEGI View Post
    Yea. Wondering if it is using all 8 cores of the ryzen for compiling ? The multicore performance has been much faster in all benchmarks i have seen so far.
    Does it say "16 in parallel" when it starts compiling? By default it fails to detect hyperthreading and only runs as many tasks as you have physical cores.

  7. #47
    0
    Samaritan

    Join Date
    Mar 2014
    Posts
    115
    Quote Originally Posted by Zeblote View Post
    Does it say "16 in parallel" when it starts compiling? By default it fails to detect hyperthreading and only runs as many tasks as you have physical cores.
    Not having a Ryzen yet, but i'am really interested to see how it performs in UE4 overall. Hope there will be some more benchmarks and impressions by developers in the coming weeks.
    Michael Hegemann - https://twitter.com/HegiDev

  8. #48
    0
    Ive been benchmarking Ryzen Editor performance and will probably putup the results soon.

  9. #49
    0
    Unless you guys compile the exact same project the times you individualy get don't say much.

  10. #50
    0
    Luminary
    Join Date
    Mar 2014
    Posts
    1,545
    Quote Originally Posted by masterneme View Post
    Unless you guys compile the exact same project the times you individualy get don't say much.
    Why would that matter? Couldn't you just use something free on the market place so anyone could test, replicate, and add their own results?

  11. #51
    0
    I believe that building a particular version of the engine (4.15?) from the source is the most accessible way to benchmark Ryzen for UE4.

  12. #52
    0
    Samaritan

    Join Date
    Mar 2014
    Posts
    115
    Quote Originally Posted by BlackSailSoft View Post
    Ive been benchmarking Ryzen Editor performance and will probably putup the results soon.
    Nice. How is the overall system stability so far, considering the bios versions are early ?
    Michael Hegemann - https://twitter.com/HegiDev

  13. #53
    0
    It may be that building with sufficient cores is actually I/O bound rather than CPU bound, and thus you'd be better off with a really fast SSD (M.2 PCIE or similar) ...

  14. #54
    0
    I am looking into getting new HW and this is (probably) relevant:

    https://www.techpowerup.com/231585/amd-ryzen-infinity-fabric-ticks-at-memory-speed

    In short: For memory dependent operations getting high speed ram is important. This can help to explain different results from tests that seems to be with equal HW.

  15. #55
    0
    Quote Originally Posted by ZacD View Post
    Why would that matter?
    It's pretty obvious, isn't it?

  16. #56
    0
    Samaritan
    Join Date
    May 2014
    Posts
    122
    Quote Originally Posted by ZacD View Post
    Oh, there's one thing I forgot to ask about, in 4.11 Epic added Intel Embree support to speed up lightmass about about 2x. So in theory Intel processors still might be a lot faster for light map baking. Has anyone tried doing some comparable lightmass benchmarks?
    Also interested in this

  17. #57
    0
    Samaritan
    Join Date
    Sep 2015
    Posts
    123
    Quote Originally Posted by BlackSailSoft View Post
    Ive been benchmarking Ryzen Editor performance and will probably putup the results soon.
    So where are your results? We want to see it

  18. #58
    0
    I'm also waiting for benchmarking result

  19. #59
    0
    Got mah R7 1700, but I'm waiting for a few more BIOS revisions before I try to push anything.

    Didn't have enough money left to buy a hard drive and I want to avoid putting too much stuff on my shiny new evo 960 so UE4 might be waiting another week or so.

  20. #60
    0
    It may be a good idea to select something available for everyone as a common base for benchmarks like compiling the engine or baking the lights of the Infinity Blade maps or whatever. Maybe something with the Kite Demo? I don't know...

  21. #61
    1
    Samaritan
    Join Date
    Sep 2015
    Posts
    123
    Quote Originally Posted by masterneme View Post
    It may be a good idea to select something available for everyone as a common base for benchmarks like compiling the engine or baking the lights of the Infinity Blade maps or whatever. Maybe something with the Kite Demo? I don't know...
    Yes that is good idea. @BlackSailSoft @Xerithas @Name368 @Algorithman and others with Ryzen please do some test for UE4 community with our favorite engine

  22. #62
    0
    Any news on those benches? I am really interested in 1700 vs i7, if those 8 extra threads give realistic decrease in compile times then this CPU is awesome for workstation!

  23. #63
    2
    Quote Originally Posted by vebski View Post
    Any news on those benches? I am really interested in 1700 vs i7, if those 8 extra threads give realistic decrease in compile times then this CPU is awesome for workstation!
    The 'city' I'm in didn't even have a computer store until 31st march (I actually had to drive a few hours to get my PC).
    Going in tomorrow to grab a HDD if they have any, but apparently they don't have much stock in yet so I'm not sure I'll come home with anything.

    If everything goes to plan I'll try to get results for engine compile and lightmass bake for one of the free levels.
    Would be nice to get stock + mild OC & try with different RAM speeds.

    EDIT: I'll do a release (4.15.1) compile soon, R7 1700 stock with RAM at 2400.
    EDIT2: It'll also be interesting to see the impact of storage speed.
    EDIT3: Noticing Asus' software reporting pretty different CPU usage to windows task manager. CPU chugging along at ~60c with the stock cooler, pretty happy.
    Haven't noticed task manager reporting storage at over ~7%.
    EDIT4: I notice it always does 12 tasks in parallel, not 8, not 16.

    Okay results are in:
    CPU: R7 1700 - 3Ghz (<--- All core speed)
    UE4 version: 4.15.1
    RAM: G.Skill Trident Z RGB @2400Mhz
    Using VS2015.

    Full solution build: ~38 mins
    UE4 development editor build: Total build time: 1307.32 seconds (Local executor: 1200.44 seconds), 21.78 mins | 20 mins

    I also did a real quick OC, all I did was go up to 3.5Ghz, no more voltage or anything.

    Full solution build: ~36 mins (could have been less, of course it finishes when I walk away)
    UE4 development editor build: Total build time: 1275.42 seconds (Local executor: 1172.82 seconds), 21.25 mins | 19.54 mins

    Was a little sad the OC results weren't faster, but all I did was raise the bar in Asus' Ai suite thing.
    The nice thing was that temps still never went over ~64c.

    Interested to know about the 'X tasks in parallel' thing, why 12?

    ANOTHER EDIT: Something weird going on, Asus' software reports different speeds to CPUZ.
    So for the fun of it I set it to 3.8Ghz, Asus' software still says it's at 3Ghz (even though I applied the change in that software. CPUZ is reporting 3.8Ghz.
    Wonder if CPU temp is being reported correctly, still at 61c, but I still haven't touched the voltage.

    Moar Results, at 3.8Ghz: Total build time: 1237.09 seconds (Local executor: 1138.94 seconds) 20.61 mins | 18.98 mins

    At any rate I'm a pretty happy camper, my old i5 took 6+ hours to do the full solution build.
    Pretty keen to get bios updates, maybe some more monitoring software & push things.
    Assuming the reported temps are correct I'm pretty impressed with the stock cooler.

    FINAL EDIT!:
    Okay that Asus software was not reporting things correctly (temp and possibly cpu speed), so just take those stock numbers as stock numbers and take everything else with a large grain of salt.

    Compiling may not have been maxing out the cpu, soon after the shaders started compiling for kite demo the PC shut down, and I'm sure glad it did.
    Even at stock settings AMD Ryzen Master reports temps up to 82c, while Ai Suite 3 is still reporting 55-70, remember I have the 1700x, which doesn't have the weird temp offset the X models do(did).

    Anyway it's 4:30 am and I'm going to bed, might do some more tests tomoz.
    Someone plz comment a level to run lightmass on, keep in mind I only have 16GB of RAM.
    Last edited by Xerithas; 04-08-2017 at 12:34 PM.

  24. #64
    0
    Samaritan
    Join Date
    Sep 2015
    Posts
    123
    @Xerithas thank you for delivering looks like Ryzen is good option for some indie gamedev. I think I will buy ryzen cpu too.

  25. #65
    0
    Luminary
    Join Date
    Mar 2014
    Posts
    1,545
    Quote Originally Posted by Xerithas View Post
    Someone plz comment a level to run lightmass on, keep in mind I only have 16GB of RAM.

    This guy is a pretty quick level to test. Make sure you are baking lightmass at production settings.
    https://www.unrealengine.com/marketplace/lightroom-interior-day-light

    The Infiltrator demo might be a good option if we want a longer bake time to compare again, and it bakes on 16GB of ram.

    There's also the Epic Zen Garden, Infinity Blade: Grass Lands, Infinity Blade: Fire Lands, and Infinity Blade: Ice Lands that I haven't really tested, but they might be good.

  26. #66
    0
    I've just updated my BIOS, made sure that everything was at defaults (didn't touch Ai Suite either) and rebuilt UE4 again.

    Total build time: 1491.19 seconds (Local executor: 1377.77 seconds) 24.85 mins | 22.96 mins

    Going to try the 3.5Ghz again, this time from Ryzen Master.

    EDIT for 3.5Ghz:
    Total build time: 1367.24 seconds (Local executor: 1258.25 seconds) 22.78 mins | 20.97 mins

    Wish I could remember what I changed in the BIOS when I first set up the PC.

    I'll give that Lightroom: Interior Day Light scene a go too.
    Last edited by Xerithas; 04-09-2017 at 01:31 AM.

  27. #67
    0
    Luminary
    Join Date
    Mar 2014
    Posts
    1,545
    I tried to test building on a i7 5930k but unfortunately I kept getting errors when trying to build with Visual Studio 2017.

  28. #68
    0
    Quote Originally Posted by ZacD View Post
    I tried to test building on a i7 5930k but unfortunately I kept getting errors when trying to build with Visual Studio 2017.
    Do you still have VS2015?
    I have 2017 as well but I just right slick the solution and open with Visual Studio Version Selector.

    EDIT for Lightmass:
    Production light build of Koolas Lightroom project: 1:45 min total, 122 ms importing, 28 ms setup, 4.08 sec photons, 1:41 min processing, 0 ms extra exporting [24/24 mappings]

    And OF COURSE AMD neglected to include a 'use defaults' option on Ryzen master -_- (as far as I can see)
    So everything except the swarm agent reports 3.5Ghz, Swarm says "Measured CPU frequency: 2.99 GHz".

    Also: "Processing scene GUID: CA0CB62B4EFEBFA37D4395BC01535E60 with 14 threads".
    But I'm just going to assume it leaves the first core alone, and distributes work to all other cores.

    Gehhh why does ASUS' aura software randomly decide to start using 6+% of my cpu.
    Last edited by Xerithas; 04-09-2017 at 01:54 AM.

  29. #69
    0
    Luminary
    Join Date
    Mar 2014
    Posts
    1,545
    Try opening windows task manage and see what frequency Windows says it's using at the time of the bake. Pretty much every app will use the max frequency even if it doesn't report it correctly.

  30. #70
    0
    Luminary
    Join Date
    Mar 2014
    Posts
    1,545
    EDIT: Ignore my last post, will update with correct benchmark. I forgot I had a custom lightmass.exe installed for multi bounce skylight.
    Last edited by ZacD; 04-09-2017 at 03:27 AM.

  31. #71
    0
    Mr.swarm agent would love more threads

  32. #72
    0
    Ohh it was enabling windows high performance power plan I'd forgotten I did before, not a huge boost in speed anyway (around 30 seconds).

  33. #73
    0
    Quote Originally Posted by ZacD View Post
    Try opening windows task manage and see what frequency Windows says it's using at the time of the bake. Pretty much every app will use the max frequency even if it doesn't report it correctly.
    Task manager reports 3.3-3.42Ghz, Ryzen Master reports 3.5Ghz, CPUz & Ai Suite both report 3.49Ghz.
    Swarm still reporting 2.99Ghz, but I'd guess it's grabbing the info from somewhere like in task manager where it says maximum speed 3.0Ghz?
    Last edited by Xerithas; 04-09-2017 at 03:38 AM.

  34. #74
    0
    Luminary
    Join Date
    Mar 2014
    Posts
    1,545
    Yeah if task manager is showing 3.3-3.42 it's running at the 3.5Ghz overclock. I think Windows reports the advertised speed and other apps pull from that by default.



    So I tested my laptop for fun, until I get a chance to run it on my 5930k.

    4 core 8 thread i7-4720HQ at ~2.8 GHz on the Koola scene.
    4:42 min total, 7.51 second photons, 4:34 processing.

  35. #75
    0
    Did one o them auto OC things which got me to 3.77Ghz.
    The UE4 compile time is almost exactly the same as my pre-BIOS update 3.5Ghz result (21.25 mins | 19.54 mins);
    Light build of the same scene: 1:36 min total, 126 ms importing, 30 ms setup, 3.60 sec photons, 1:32 min processing, 0 ms extra exporting.

  36. #76
    0
    Luminary
    Join Date
    Mar 2014
    Posts
    1,545
    Quote Originally Posted by Xerithas View Post
    1:45 min total, 122 ms importing, 28 ms setup, 4.08 sec photons, 1:41 min processing, 0 ms extra exporting [24/24 mappings]
    1:47 min total, 113 ms importing, 28 ms setup, 3.40 sec photons, 1:43 min processing, 0 ms extra exporting [24/24 mappings]

    So my i7 5930k (6 core) at 4.2 GHz got basically the same results as the 1700 (8 core) at 3.5 GHz.

    Probably would be a good idea to do a longer bake since they are so similar on that benchmark.

    The 5820k should run just as well as the 5930k and is $320 at Microcenter, the 1700 is $329 at Microcenter. More investigation is needed though.

  37. #77
    0
    Did a production build on my i7 6700k 4.0Ghz 4 cores 8 threads and it took 2:45 min to build the lighting for Koola Lightroom.
    2:45 min total, 236 ms importing, 39 ms setup, 3.97 sec photons, 2:41 min processing, 0 ms extra exporting [24/24 mappings]

  38. #78
    0
    With my 7700k at 5Ghz: 2:22 min total, 92 ms importing, 21 ms setup, 3.36 sec photons, 2:18 min processing

    You can see how it's much better at the single threaded parts (importing, setup, photons) and much worse at the multi threaded one (processing)

  39. #79
    0
    Luminary
    Join Date
    Mar 2014
    Posts
    1,545
    Quote Originally Posted by Zeblote View Post
    With my 7700k at 5Ghz: 2:22 min total, 92 ms importing, 21 ms setup, 3.36 sec photons, 2:18 min processing

    You can see how it's much better at the single threaded parts (importing, setup, photons) and much worse at the multi threaded one (processing)
    I wonder if the difference in performance will get larger with a longer bake, I'm going to try baking the Infiltrator demo at production settings later today, unless anyone else has a suggestion for a different scene to test.

  40. #80
    0
    Quote Originally Posted by ZacD View Post
    1:47 min total, 113 ms importing, 28 ms setup, 3.40 sec photons, 1:43 min processing, 0 ms extra exporting [24/24 mappings]

    So my i7 5930k (6 core) at 4.2 GHz got basically the same results as the 1700 (8 core) at 3.5 GHz.

    The 5820k should run just as well as the 5930k
    I'm using a 5820k @ 4.2 Ghz, and I got:

    2:09 min total, 107 ms importing, 25 ms setup, 3.94 sec photons, 2:05 min processing, 0 ms extra exporting [24/24 mappings]

    Your RAM is probably just faster than mine. I'm using two 16 GB and two 8GB modules, so 48 GB, but since its not 4 times the same RAM it probably isn't able to make use of quad channel correctly.
    Last edited by John Alcatraz; 04-09-2017 at 10:05 PM.

Page 2 of 4 FirstFirst 1234 LastLast

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •