Your better off just over clocking the 5820k and not waste a bunch of money on a minor upgrade, the 8700k is roughly ~25% faster at a ~25% higher clock speed.
Lightmass scales pretty well, but cannot split one object across multiple cores, so it can get stuck finishing the last bit of a bake using one thread. So generally the more cores the better.
Also with lightmass, there’s generally things you can do to get a better bake without cranking the settings up to 11, you can get great results with a quick bake (depending on the size of the level of course).
You can get great results with a quick bake. but only if you know exactly which settings to use… for which you have to make 100 test bakes, hahaha (at least, this is what I’m doing at the moment)
Wait… That’s a very big difference with the Pugetsystems benchmark (188s on high, 444s on production settings for Ryzen 1700x, UE4.16.2) so I guess the driver updates / newer UE version is having a big impact there…
Is there any that you can perform the same test with 16GB RAM only? I would be very grateful :)
Lightmass on DESKTOP-A78O7FL: 1:40 min total, 887 ms importing, 145 ms setup, 3.96 sec photons, 1:34 min processing, 0 ms extra exporting [734/734 mappings]. Threads: 23:41 min total, 20:48 min processing.
15:18:20: Lighting complete [Startup = 887 ms, Lighting = 1:39 min]
So my system has arrived, and I can report the following out-of-the-box results for Zen Garden:
Processor: Threadripper 1950X
Memory: 32GB at 3333MHz
Version: UE4.18.2 (on Windows 10)
For HIGH settings: 58 sec:
Lightmass on DESKTOP-PQNGBVS: 58.1 sec total, 893 ms importing, 156 ms setup, 3.96 sec photons, 53.0 sec processing, 0 ms extra exporting [734/734 mappings]. Threads: 27:20 min total, 23:07 min processing
.
For PRODUCTION: 135 sec:
Lightmass on DESKTOP-PQNGBVS: 2:15 min total, 1.60 sec importing, 157 ms setup, 6.89 sec photons, 2:06 min processing, 0 ms extra exporting [734/734 mappings]. Threads: 1:05:14 hours total, 56:40 min processing.
Not too bad. It seems the memory speed is having an impact. With a complicated map (not Zen Garden) I’m using some 14GB memory (so I guess I could have gotten by with less memory). The system is running around 3.5GHz.
Nice, it’s faster than mine. Maybe the other guy just had his memory set up wrong?
For comparison, I used the same engine version on i7 7820X @ 4.6 Ghz:
High:
1:29 min total, 696 ms importing, 106 ms setup, 4.00 sec photons, 1:24 min processing, 0 ms extra exporting [734/734 mappings]. Threads: 21:08 min total, 17:57 min processing.
Production:
3:32 min total, 705 ms importing, 113 ms setup, 7.04 sec photons, 3:24 min processing, 0 ms extra exporting [734/734 mappings]. Threads: 51:12 min total, 44:58 min processing.
I have a Ryzen 1700, it is blazing fast, and I have no problems with UE4. The main reason I choose Ryzen is because you have twice more speed for the same price and that’s because Intel sells its s*itty GPU included with all its CPU that you won’t need if you have a desktop computer.
Ryzen is great and I have a Gaming Gigabyte Motherboard that is doing great too AX370-GAMING k3
I am doing 60 FPS no matter what with the editor (that’s the max speed from my monitor).
Why would you go with the 7820x now that the 8700k is available for $150 less yet is faster overall? 7820x uses the new mesh bus which is optimized for server level tasks, but for anything graphics related, like games, it’s considerably slower than the ring bus used in the 8700k. Those 2 extra cores won’t make any difference to the speed of your PC for anything other than shader/lightmass compilation and slows you down in nearly everything else.
Here’s some proof of that:
If you compare the 8700k with the 1800x, it’s a bit cheaper but the performance results speak for themselves:
Core count is not the only thing to consider when buying a CPU, individual core speed is still the most important aspect for gaming & game development. You might save a small amount of time in compilation but for the other 98% of development work you aren’t getting the best performance possible for the money you spent.
If you can wait a little bit before purchasing Intel is coming out with an 8 core Coffee Lake chip (same generation as the 8700k) to add to their lineup, when they do it should have the best performance in nearly every game development aspect at a fairly reasonable price (unlike the Intel X series chips).
I have the 8700k and it is a beast, especially when overclocked to 5.0ghz (mine will do 5.2ghz stable but I don’t want to shorten it’s life too much). Shader compilation times are very fast and gaming performance is even better. I highly recommend it.
Compare with the new 2700X, not with the old 1800X
That new XFR (higher boost clocks) and some more recent production improvements on 12nm give about 10-15% more performance in average, compared to the 1800X.
Also higher multicore results in pretty much all benches, but for these you shouldnt rely on “UserBenchmark”, instead use better known sites like Guru3d, Computerbase or Toms Hardware. They test real world cases, not only synthetic corner cases.
Since I’m interested in the upcoming Threadripper 2, i would love to see some TR1 results in case anyone here is using that beast!
