Mining ERGO with Vega 56 & 64

hotness · October 25, 2021, 11:43am

In my testing Navi GPUs (5000 and 6000 series) under-report total board power by 21%. While Vegas under-report by 31%. Unless you manage to find a Gigabyte Gaming OC Vega then you’ll get the full power draw of like 170W+, while all other Vegas are in the 120s. Then on top of that you have to add the efficiency of the PSU, which is generally 85-90%.

Micron 5700 can go up to 950Mhz for 58+ MHs but use up to 150W (real) to get there. Samsungs rarely do over 910Mhz so cannot break 56 MHs, but the big benefit is they do it with much lower power of ~110W. So for example if you had a 1000W PSU you could have 6 Microns @ 348 MHs, or 7 Samsungs at 385 MHs.

isthoger1 · October 28, 2021, 11:25am

Anyone can help me to touch stable 180 mh+ ? i got 30 mh when i increase above 900 memclock or GPU dead when i decrease rcdrd under 16.

can someone give me better tweak ?
i already tried to increase CC etc and still capped at 175mh max

846875 · October 28, 2021, 12:36pm

Raise the voltage, I need 930mv or more to reach 190mh+ BUT in hive it would limit my max power to ~165w and that would limit the hashrate to around 180mh.

isthoger1 · October 30, 2021, 3:22am

it keep crash bro when i increasing memory. should i experiment in tweaking ?

ss-silver · October 30, 2021, 3:25am

Hey try dropping mem to 846, remove the SOC, then increase core to 1300

isthoger1 · October 30, 2021, 4:13am

but my red devil cant RCDRD under 18. it will detected instadead

dukeofdream · October 31, 2021, 8:35pm

yes | rocm-smi -d 0 --setpoweroverdrive 245

This will take care of your 165w problem i’ve gotten 220mh/s from my vega but i need my waterblock to get here before i can push it for 24/7 or even try to get more out of it

846875 · October 31, 2021, 9:01pm

You are the first person to even know that this is an issue lol.

Where do I enter that command though?

The sad part is that it is looking like I will have to setup proxmox and run hive as a VM in order to actually use it since for some silly reason it can’t run 2 instances of the same miner.

I can’t even get TRM setup as a custom miner so I can run it separately. I need TRM x3 for ETH, ERGO and RVN.

dukeofdream · November 1, 2021, 8:22am

That’s an interesting setup, never had to do it even though i’m pretty sure i will have to once i put most of my cards on one rig… My 5700XT sucks for ergo but my vega should be better of mining ergo at 220mh/s than eth at 58… I’m interested on what you find about the matter… Feel free to ask on TRM discord channel As to where you should put the command, you will either have to press the command button that looks like this >_ or start hive shell and type it there… Don’t forget to replace the 0 with the number of your card though… I can tell you right now that it’s not difficult to reach that 245 limit though I wish there was a way to fully unlock it for ergo mining… Pretty sure that once you go full waterblock you should be able to max the card… Whereas on Eth you are limited to 58mh/s since changing your memory from 1060 and up won’t affect the hashrate for some reason

846875 · November 1, 2021, 1:18pm

I asked about making a custom miner “copy” of TRM in the discord but the only person that replied didn’t even know custom miners were a thing. That would still be the easiest option by far. It is possible, I just have no idea how to actually do it.

I gave it my best shot but could not get the custom miner to show up in hive and when I did it said the miner was already running, even if there was no miner running at all.

Someone that knows what they are doing should be able to rename TRM and create a copy of it so it can run officially I would think.

Proxmox adds a lot of unnecessary complication but is the only real option I can think of. I have an old server board that should be able to handle a lot of GPU’s (56 PCIe lanes?) and due to space constraints it would be a lot easier to run them all off a single system.

Is there no startup script for running commands at boot automatically?

dukeofdream · November 1, 2021, 7:06pm

There is a startup script but i don’t know how to use it… I’ve googled it before but it seemed like too much hassle… Ergo mining is pretty stable and if you use the command button on hiveos interface you actually have a history of previous used commands so it’s not that big of a deal… By the way i’m about to open a ticket about hiveos having problems while running multiple miners… I have 2 rigs that both show weird behavior at times… Like if i check my hashrate on the pool, sometimes one of the 2 miners has stopped reporting hashrate even though on hiveos it’s mining normally… This is a huge deal cause hashrate is effectively going to waste… I have a rig that runs a 390X on phoenixminer and a vega on TRM… Sometimes the vega hashrate is missing and sometimes the 390X is missing… Same with another rig that has a 5700xt on TRM and 2x3080tis on Nbminer… Though most of the time its the TRM that’s not reporting the hashrate on the pool… Really frustrated about why is that and what i should do about it… I think i will rename the worker name on the miners so i get 2 different hashrates on the pool instead of a combined one…

846875 · November 1, 2021, 7:23pm

yeah, seems really odd when there is only one truly viable miner for AMD cards that they would not be able to handle multiple miners better. I tried all the others when trying to make a mixed rig work but they all had noticeably less hashrate if they worked at all.

dukeofdream · November 1, 2021, 7:28pm

I know right? Not only that but some cards are different than others… In my TRM tuning settings i have different modes for the vega and 5700xt…If i was to put them all in the same rig i’d have trouble setting the b-mode right for both… Not only that but the 8.6 version of TRM affects stability on vegas for some reason… It really made me think my vega has degraded or something… Havent tested the 8.6.2 update but if the 8.5 works fine why bother with the 8.6.2?

846875 · November 1, 2021, 7:43pm

Which is why I am thinking about working around the issue with proxmox. The issue then is that it will think I have 3-4 rigs when I actually only have 1 and that will kick me out of the free tier I think.

Not to mention possible issues with the extra layer of complexity. Such as a card dying and that VM restarts but it doesn’t restart the card and just loops.

Proxmox is not the answer, just the only thing I can figure out unless someone can create a custom TRM (or 3) miner that we can use with the normal TRM. It is possible, I just have no idea what I am doing.

dukeofdream · November 1, 2021, 7:44pm

I think it would be better to create a thread dedicated to it…

846875 · November 1, 2021, 7:58pm

I did, it didn’t get any replies: Need to create a copy of TRM so I can run it 2x but get errors?

846875 · November 2, 2021, 2:58pm

Made a new thread with a better title to see if anyone would reply this time. I am even considering offering a bounty for someone that can get a working TRM custom miner.

SmokeOne · November 4, 2021, 10:41pm

Thank you to all the contributors of this post, and @ss-silver for starting the thread and helping many along the way. I’ve learned quite a bit getting my rig stable and optimized. I’m sure I could push my cards a bit further but I’ve tried to balance power consumption with performance. Figured I’d share my settings to give back to the community. Cheers!

Straps below:
Sapphire Vega 56 (Samsung) -

amdmemtweak --CL 20 --RC 37 --RP 11 --WR 14 --CWL 8 --FAW 12 --RAS 20 --REF 65535 --RFC 248 --RTP 5 --RRDL 6 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 12 --RCDWR 12

ASUS Vega 56 (Hynix) -

amdmemtweak --CL 20 --RC 37 --RP 11 --WR 14 --CWL 8 --FAW 12 --RAS 20 --REF 65535 --RFC 248 --RTP 5 --RRDL 6 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 12 --RCDWR 12

Red Dragon Vega 56 (Samsung) -

amdmemtweak --CL 20 --RC 36 --RP 11 --WR 14 --CWL 8 --FAW 12 --RAS 20 --REF 65535 --RFC 248 --RTP 5 --RRDL 6 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 12 --RCDWR 12

AMD/ATI LC Vega 64 (Samsung) -

amdmemtweak --CL 20 --RC 44 --RP 14 --WR 14 --CWL 8 --FAW 12 --RAS 30 --REF 17000 --RFC 249 --RTP 5 --RRDL 6 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 14 --RCDWR 12

Sapphire Nitro+ Vega 64 (Samsung) -

amdmemtweak --CL 20 --RC 44 --RP 14 --WR 14 --CWL 8 --FAW 12 --RAS 30 --REF 17000 --RFC 249 --RTP 5 --RRDL 6 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 14 --RCDWR 12

gkhosv64 · November 6, 2021, 9:45pm

This is absolutely the best OC for my Sapphire Vega 64 Nitro+ flashed to Vega 56, its working for days without problems, with stable 55.65MH for ETH. Just want to say BIG THANK YOU for sharing with us.

If you have OC for ERGO in range from 180-190MH and for 190-200MH please share with us!!!

Thank you!

getloppy · November 7, 2021, 7:29am

Happy it helped! Contributions that SS-Silver and the likes made and helped me get started makes me want to help others and optimize here. Sapphire are far and away my favorite vega’s Ive worked with. I would add Ive been able to get all Vega stable at 57 MHs so you may be able to eek some more performance out, not at the cost of a lot of power. Newest strap i use is here: amdmemtweak --RC 38 --RP 11 --WR 14 --CWL 8 --FAW 12 --RAS 20 --REF 65535 --RFC 248 --RTP 5 --RRDL 6 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 12 --RCDWR 12. RC 38 is something I have been playing around with and am happy with. It allows you to go to 1060 mem clock which unlocks 57 MHs. I usually have had to set memory voltage to 1300 mV to get this. Memory voltage is a huge lever for both power consumption, stability, as well as unlocking higher clocks. Ive also noticed that hashrate will jump as soon as certain memory speeds are reached. Its unlike other cards where adding +5 clock adds X% more hash. Nothing will happen until you reach about +15 and then it will jump. Here is my current sapphire rig: (still trouble shooting the one with invalids - likely voltage too low)