Vega 56 and Vega 64 guide

[Banhammer]

Hello Desert,

Just a brief glance at the numbers, and you might have mixed up the memory OC with the core. I havent heard of vegas getting 1200 on memory. Try 1100 on core, 1000 memory with 900 core voltage, and try tuning from there.

The 1200 on memory just seems a bit high.

[Kower99]

for me my vegas sometimes start on low hashrates because they start on low states … reapplying OC solves problem … agresive OC too …

[DesertSweeper]

@Banhammer you are absolutely correct - actually not sure where I got 1200 - too much coffee. In Windows it is 930 (1860). I put it back in to my HiveOS rig and managed to get it somehow to 38.75 Mh/s with 1000-1-980-960 (a predefined template). The problem is the moment I try to change anything the rig hangs and I have to force it off as it become unresponsive. Frustrating - but I see many have issues with the Vega64 cards so I am not alone…

[ikledude]

I love the activity on this thread so I thought I would report back; Its been stable for over a week;
Do read the guide here:
“teamredminer/ETHASH_TUNING_GUIDE.txt at master · todxx/teamredminer · GitHub”

1. Flashed all Vega 64’s BIOS to Vega 56’s;
   • power saving and stability
   • “The hardcore Vega 64 Samsung setup these days is to flash it with a
     corresponding Vega 56 bios and apply advanced timings and powerplay table
     mods. This lowers the HBM2 (mem) voltage to 1.25V and therefore the power draw
     while still being able to run a high mem clock and produce a nice 50+ MH/s
     hashrate, sometimes as much as 53-54 MH/s.”
2. I’m now back on the default A mode;
   • “For Vega 56/64, the A-mode is already very capable of delivering high hashrates
     at an efficient power draw. However, the B-mode creates a slight shift in core
     vs mem clock ratio necessary to sustain a specific hashrate. It will use all
     available vram on the gpu. Early reports show that at many custom timing mods,
     the B-mode is slightly harder to keep stable over time.”
3. I start the rig with the following OC’s.
   • Core: 1100Mhz, Core Voltage: 850mV, Memory Clock: 960Mhz, Fan: 60%, Power Limit 150W
4. I then apply the straps of:
   amdmemtweak --CL 20 --RAS 26 --RCDRD 12 --RCDWR 12 --RC 38 --RP 12 --RRDS 3 --RRDL 5 --RTP 6 --FAW 12 --CWL 8 --WTRS 4 --WTRL 9 --WR 13 --REF 24000 --RFC 239

[erdigeg12]

My worker has 1 vega 56 and 2 rx 580. How can I apply that code to my vega 56 only?

[ikledude]

amdmemtweak --i 2 --CL 20…
-i ‘gpu number’ or 2,3 for multiple gpus.

[BL0xie]

Hey man i have the same card with the same memory. Been struggling to keep it stable with straps. Please keep me updated if u manage to get more MHs and stable

[Rocco1180]

Hello all,

thank you all. Im getting really good results out of my vega gpu’s.

Can anyone help me with this invalid shares? The Miner is working for 5h and three invalid shares are not that much, but what can i do to avoid this? reduce mem or core? More VDD?
is there a rule? Dead GPU = more VDD and invalid shares = to high memory clock?
I tried to reduce memory clock but after some time the inv shares are still there…

And every time i change the overclock while the miner is runing one of the gpu’s die and TRM reboot.
What can i do to avoid this?

thanks for advise

[Banhammer]

You can try a bit more voltage, but with invalids you usually need to ease off on the memory a bit. Either loosen timings slightly, or lower memory overclock. Going to 1005 and 1015 should work. Try it and see.

[Banhammer]

Sorry to only reply now, thought I would update incase it helps anyone. It turned out conflicts were a bad riser. Such a headache to troubleshoot, but got there.

Also, found a way to apply straps automatically! Will post below.

[Banhammer]

Found this forum post Automatic straps guide

Follow the instructions, and the straps will apply, even in the latest version!

Basically,

1. under worker options, enable GUI on boot
2. start hive shell
3. use the command sudo nano /etc/xdg/openbox/autostart
4. go to the bottom, and write out your straps
5. at the end of every strap, put &
6. make sure there is a blank line under the last strap
7. save and exit

There you go! I am so happy that I don’t have to keep my rig in maintenance anymore, and it will restart itself if it goes down.

Good luck!

[Rocco1180]

the way i apply straps automatically is the following.

put your straps in /home/user/xinit.user.sh by using “nano /home/user/xinit.user.sh” commant in hiveshell.

thats it. Latest version works like a charm

[JapeX]

Hi…
Is it possible to get more MH/s from my Vega 56 without modfying its BIOS?
Im getting 45MH/s

Thanks

[Rocco1180]

I’ve only had bad experiences with gigabyte Gaming oc.

• high power consumption according to the software.
• every crash of the graphics card has frozen the computer.
• more than 930mhz memory clock was impossible

you can try this straps “–CL 20 --RAS 20 --RCDRD 12 --RCDWR 12 --RC 37 --RP 11 --RRDS 3 --RRDL 6 --RTP 5 --FAW 12 --CWL 8 --WTRS 4 --WTRL 9 --WR 14 --REF 65535 --RFC 248”

should give you a little bit more mh

use “B Mode” and reduce core clock to 1020MHZ
try to lower VDD

[JapeX]

Whats B Mode… i’m know to mining world. tnx

[Rocco1180]

use command “–eth_config=B” with Team red miner.

Tuning Tab

[jssdev]

Hello.
I have the same GPU as you (Gigabyte with Samsung memory) and I have not been able to improve this. Every time I apply a strap, the miner restarts. More than the hashrate, I am concerned about consumption, I see that many people can lower it by about 40w. If you find any alternative, please comment. Thanks!

[jssdev]

PD: this is the best performance I have ever been able to achieve. As I commented, I have not been able to apply any strap.

[hwold]

Hi, whitch strap you put to get those results?

[vsimovic]

few invalids is not a problem