Vega 56 and Vega 64 guide

[S_Grim]

@ss-silver Thanks for confirmation on that BIOS.
Flashing went smooth, but I cannot seem to get higher hash rate than 45’ish MH/s.

The OC you posted just makes the rig reb00t after a few moments, even before starting the miner (TeamRedMiner). I tried lowering the Core Clock in steps, and upping the Core Voltage a bit, finally getting somewhat stable results just below 45 MH/s.

I kept the straps tho, as I have no clue as to what all those commands mean. (Please point me in the direction of learning more about straps).

Here is my current settings:

And results:

Any suggestions?
Where can I look for error logs or something similar for better troubleshooting btw?

[ss-silver]

Looks like the amdmemtweak did not work, values are all red. Try deleting the space and two dashes before CL 20 and re-typing them, sometimes copy / paste brings in weird characters. Since amdmemtweak is white that is formatted correctly, but something right after that is wrong causing the rest to show as invalid.

If re-typing that part doesnt work try setting that advanced config on the cards individually, under that card’s OCs

Not having those straps on with those settings would crash the rig, I cannot run mem that high without straps.

This link has definitions to what you are changing with those straps, biggest ones that impact hash and invalids/ crashes… in my opinion are:
Active to READ command delay (RCDRD) my cards range from 12-18
Active to Active command period (RC) I do 37 for Samsung and 36 for Hynix

Eliovp/amdmemorytweak

[S_Grim]

Oh… Thanks, still learning here

Woah, check this out!

Thanks a bunch!
This is awesomesauce!

I really hope this will work for a long time now, and not just a few minutes.
I will report back later

Thanks again!

[ss-silver]

Sick! nice ya still runs stable for me, been up 3 days minus me stopping the miner to change the strap on my new vega. That one threw a couple invalids yesterday so I dropped rcdrd to 13 then it threw 3 more in 8 hours, but none past 20 hours.
If it keeps throwing invalids I’ll try something like rcdrd 18, mem 1028 whenever this rig goes down… No way I’m turning off the rig when profits are this high

[S_Grim]

Speaking of invalids…

Should I adjust accordingly? Like you did?

[ss-silver]

Ya thats too many you can try RCDRD 18 Mem 1028 but Ive only used that setting when I was learning my 64 so not very tested.

Could also try increasing RCDRD by 2 at a time until you stop the invalids. Then see if you can lower it by one or not.

Are you using TeamRedMiner? that setting should hit 55

[oni]

I think you need to start with higher vdd, like 850, and when you’ve found your max stable hashrate (24h without rejects) try lower it in small steps of 10 and test a few hours again.
Maybe this one can’t handle the higher mem clock too, maybe lower to 1010 and test.
Vega’s are all about patience There is a lot of info in this very long topic, but it’s worth the read!

[S_Grim]

Thanks!

I updated OC for that card now, will let it run for a while and monitor it.
It even increased to 21 invalids since I posted my previous comment…

[S_Grim]

Thanks.

I have skimmed this thread for a few hours today already
Learned a lot. Especially by you two, @ss-silver and @oni. Thanks a lot guys!

Really appreciated it!

[oni]

@S_Grim

Here is my current settings:

Any suggestions?

The GPU index is not set. It is the number of the GPU for which you want to apply that strap.
Or you set the straps card by card, then the GPU index will be added automatically.
Or you set it in the overall OC for all cards, but then you have to define for which gpu you set it.
You can add more than 1 index to the same straps. Or you can add multiple different straps for different gpu’s.

Also toggle “aggressive undervolting” on.

[S_Grim]

Yes, since all cards are the same, I started with an OC template and added that to the entire rig (3 Vega 56).

But now I edited the OC for that troublesome card that gave the invalids.
I toggled the Aggressive thingie off after seeing the screenshot from @ss-silver . But perhaps I should turn it back on?

[wayne4756]

Thanks for the response.

Here are the cards at default

Cards with config form first post

Cards with config field from clipboard.

[S_Grim]

@ss-silver :
The rig crashed with Mem 1028 (and RCRD at 18)

So I edited OC a bit like @oni mentioned, going down to Mem 1010 and VDD up to 850. (and also toggle Agressiveness on)

Hasn’t been long yet, but for now it runs ok, albeit a bit slower than the other cards.

If it is still running ok when I go to bed in an hour, I will leave them on during the night to check the results tomorrow.

[Smining570]

I dont see why vega´s are not starting to mine…maybe mem too high and hive cant apply oc…try all vega´s core to 1000

[wayne4756]

Here are the stat at 1000,
Im new sure its something basic im not doing?

[Smining570]

Well, everything seens to be ok. last thing that i can think of is remove straps from all vega s and try reboot with these Clocks. the mem and core are fine for both Samsung and hynix Mem… so straps of, and if… it start mine, apply straps after you have reach the oc on all cards. when doing oc, increace values like 5to10 by time and one card at the time so u know when and what has gone wrong. if it dosent start without straps with these values, im afraid that im out of options…sorry

[wayne4756]

Increasing the OC on core did not start mining.
researching through the thread a similar issue was posted and fixed by xinit.user.sh AMDMemTweak – read/write mem timings
when I click on shellinabox Im transferred to connection is not private page where I can’t advance.
My question is do you think i’m going in the right direction?
If so Im guessing xinit.user.sh to load driver on boot? I see nvidia load in green but ADM is red on boot.

[oni]

xinit.user.sh was the old way of adding straps on boot. It used to work up to hive version 195 or so.
you have the latest hive version so the straps on boot should just work when you add them to the Tweak in the OC window.
something else must be wrong, but I also don’t see what.
maybe try TeamRedMiner instead of Ethminer ?
I’ve seen in earlier posts that you’ve already flashed a vbios on them, but what were the cards originally? brand? type? reference cards (1 fan) or aftermarket (more than 1 fan)? and did you save their original vbios prior to flashing another one?

[jokersteph083]

hello, I really need help with my Vega 56 MSI hynix memory

• either it makes invalids, or it crashes
  I tried everyone’s straps for 15 days on this discussion but nothing to do … impossible to stabilize and have at least 50Mh
  someone to help me?
  
  

[Smining570]

looks like from the screenshot that u dont have any straps applied, or? Have u tried with default hynix straps or some straps that are provided here in this thread?