For me it's just a single Vega 56 alone right on pcie 16x. Dual "monitor" setup(144hz freesync + 60hz tv).
Asus Prime B350 Plus / Ryzen 1600 / Win10 Pro 1809(Os Build 17763.194, should be latest) / 2x4GB 3000mhz and 16000MB virtual memory on Samsung EVO850 SSD / CN 16+14(tried few others but no change other than hash speed decreased)
Just ran teamred for the night as I wasn't using the PC; otherwise it works wonderfully except the system lag.
So, I upgraded my gpu dev workstation to 18.12.2 today, hoping it'd reproduce the system lag issue. Unfortunately, everything is running just fine:
[2018-12-15 21:02:53] Stats Uptime: 0 days, 07:14:02
[2018-12-15 21:02:53] GPU 0 [60C, fan 60%] cnv8: 1.011kh/s, avg 1.010kh/s, pool 1.007kh/s a:212 r:0 hw:0
[2018-12-15 21:02:53] GPU 1 [64C, fan 17%] cnv8: 1.039kh/s, avg 1.040kh/s, pool 1.086kh/s a:229 r:0 hw:0
[2018-12-15 21:02:53] GPU 2 [54C, fan 29%] cnv8: 2.081kh/s, avg 2.081kh/s, pool 2.109kh/s a:448 r:0 hw:0
[2018-12-15 21:02:53] Total cnv8: 4.131kh/s, avg 4.132kh/s, pool 4.203kh/s a:889 r:0 hw:0
The box is a pretty standard Dell Win 10 Pro Ryzen 7 1700X, 16GB ram, 48GB overkill swap. It runs a Rx580 at PCIe 3.0 x8, Vega 64 LC at PCIe 3.0 x16 and a Rx570 on a riser at PCIe 2.0 x1.
So, when running the miner using standard configs (8+8,8+8,16+14) and getting the results above after 7h, I experience zero lag or sluggish feeling. No difference logged in remote vs physically. However, there's no chance in hell I can run my Vega 64 at 16+14 when it's driving monitor(s), hashrate drops significantly.
I'd love to be able to reproduce and nail this issue, I have a few ideas about things that could make TRM stick out compared to other miners. Right now we don't have a workstation or rig exhibiting the problems described, but we'll continue the search.
Just curious if any of you guys having this problem could try a compute algo as well and check if you still have the issue? For example, you can just run the provided start_phi2.bat in the 0.3.8 release.