With the rapid expansion of the engine footprint, my Windows PC is taking about 73 minutes to build the 4.24.1 engine from clean. This is too slow so I am investigating upgrading my hardware to build faster. I’ve done some timing studies in various configurations that might interest others. Note, I recommend excluding your development disk area from Windows Defender since this sometimes has a big impact, although it only makes about a 2% difference for me building 4.24.1.
Generally, the build scales really well with real cores. Hyperthreading only makes an incremental difference of 8% or so. Fast IO seems to help the per-core scaling. I would love to see the performance of the new AMD 32-core Threadripper 3970 system compared to this, especially with 4 M2 SSD’s in RAID 0 to see if bandwidth or NTFS is the limiting factor.
Unreal Engine 4.24.1 complete build, (Clean Solution, then Rebuild Solution in Visual Studio 19 Version 16.4.0)
Intel(R) Xeon(R) W-2133 CPU @ 3.60GHz (6 real cores)
Samsung SSD 860 QVO 4TB
12 processes: 73 minutes
6 processes: 77 minutes
5 processes: 83 minutes
4 processes: 85 minutes
3 processes: 97 minutes
2 processes: 127 minutes
1 process: 234 minutes
New iMac Pro doing Mac build:
18/36 2.3 Ghz Intel CPU
2TB custom Apple PCIE SSD
iMac Pro 18/36, 36 processes - 20 minutes
iMac Pro 18/36, 24 processes - 20.8 minutes
iMac Pro 18/36, 18 processes - 21.7 minutes
iMac Pro 18/36, 17 processes - 22.3 minutes
iMac Pro 18/36, 16 processes - 23.2 minutes
iMac Pro 18/36, 15 processes - 24.3 minutes
iMac Pro 18/36, 12 processes - 28.7 minutes
iMac Pro 18/36, 9 processes - 36 minutes
iMac Pro 18/36, 8 processes - 39 minutes
iMac Pro 18/36, 6 processes - 49.8 minutes
iMac Pro 18/36, 4 processes - 71.8 minutes
iMac Pro 18/36, 3 processes - 93.6 minutes
iMac Pro 18/36, 2 processes - 136 minutes
iMac Pro 18/36, 1 process - 263 minutes
The same iMac Pro doing the Windows build using Parallels:
18 processes 45 minutes 30 seconds
16 processes 48 minutes
14 processes 52 minutes
6 processes 97 minutes