Vega 64 16+14
1630 core 1190 hbm
2470 H/s  

Somehow, my watercooled vegas have better hashrate than normal vega 64 even when ODN settings are the same. I have only 2 watercooled vegas so the sample size is not so big, yet if I use for example 1450/1100mhz, normal vega 64 will be around 2150 and liquid vega at around 2250. I will try in following days to flash liquid vega bios to normal vega and see what happens

New AMD gpu drivers are out and with so many new things I just gotta use them. So far mining speed on teamred seems to be a bit better and now one can undervolt well enough without registery powertables.. Not sure if the combination of new drivers + teamred or just teamred but whole system seems to be laggy while mining. It wasn't before Probably need new CN-values.

Release notes https://www.amd.com/en/support/kb/release-notes/rn-rad-win-18-12-2
Download as usual.

Sapphire Pulse Vega 56 & Cryptonight V8 algo.

Well I tried automatic functions like undervolt, oc gpu and oc memory but the increased hashrate is from the same manual settings I used before(just used wattman this time instead of overdriventool). Now I just don't need powertable regfile for it. You need to pick one from the automatic options and I just rather do undervolt + oc memory at once so manual works for me better. Also automatic oc never did over 800mhz on memory(it's max value on default settings but with manual one can adjust like before) and I rather have 900mhz at least. There are two memory timing levels but too busy to test them out now(wouldn't know if it could crash and no time for that).

EDIT: Both memory timing levels give a bit better hashrate for me than default value "Automatic". First tier a bit better, very very little. Tried every other gpu miner I know there is and no OS lag(whole windows and videos) with any of them. I hope this is something that can be fixed within teamred so I can go back to it.

i can verify the lagginess with teamread and 18.12.2. drivers with my vega56/64.  hashrate is higher without having to mess with powertables, but there's a substantial lag on the system now.

i7-5930k, one vega56 pciex16 slot with monitor connected, two vega56 on risers, one vega64 on riser... all set to 14+14, win10x64

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:


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.

For who is interested:

Running on a RX570 4GB ITX, CC 1075 MC 775 MC 1910 MV 787, getting 870 h/s reported, maybe 860 h/s at the pool, using 86.5 watt at the wall (minus mobo draw without card installed) (CNv8). (max hash is about 920 h/s or so for this card, but it uses about 112 watt (just without mem errors etc)

about 3 mem errors hour (2.5 hour test run), non hw in the miner. Voltages are ridiculously low, especially for TRM. Determined these odd clocks and voltages using SRB to give the most hash/watt, (and no mem errors), and then tried them as well for TRM. SRB 1.7.0 gets 9.4 h/w, TRM 0.3.8 almost 10.0 h/w. Was surprised TRM did not crash on these low voltages, as at higher core clocks it needs the power.

Cost wise it would save me 10 dollar/year, as my power is about 0.20 dollar kWh.

So TRM can be a decent competitor, when some more features will be introduced  , and the dev fee goes down just a tickle more 

 

You are totally right! although it saves on heating my home. Guess atm you only earn a bit when you got 'free' power from somewhere.

Hi!

There are a number of skilled Vega tuners in the forum, some also hang out in the supportxmr chatango chat room (pbfarmer, derp4vega, saltednutz, etc). I know derp4vega has been kind enough to share his ppt mods earlier, but of course it's entirely up to him. Personally, I've done a fair amount of mining on my own Vega rigs, but I'm nowhere near these guys in experience with ppt mods.