I have 2x vega 56 with samsung memory. Right now I have them running around ~190-194mh @ 175w with 1400 core and 930 mem and voltage at 950mv.
Using the straps from the OP
If I try to lower the voltage much they tend to get unstable. Are the straps from the OP still the best option? Should I try something else?
More worried about max hash vs efficiency right now but would still be nice to either lower power or increase the hash.
If youâre not running mem at 847 then there is no use in soc 860; so delete the soc value for starters.
Try with a vdd of 900 just to test if it keeps running. If it does, then you can start lowering vdd step by step; by 10 for example and if stable for 24h then down with 10 again until you get dead gpu again; then increase by 5 by 24h until you keep it alive.
The Sapphire bios is just the same as any other reference card; they even mention it in the description:
Same Bios: AMD RX Vega 56, Gigabyte RX Vega 56, MSI RX Vega 56, Powercolor RX Vega 56, XFX RX Vega 56, Dataland RX Vega 56, HIS RX Vega 56, Yeston RX Vega 56, ATI Vega 56, ASRock RX Vega 56, Asus RX Vega 56
So that should be a good one for your card.
bro how do i reach 150 Mh/s at vega 56, i only get 90Mh/s is it required to PBE one time patching like rx570/580 Series? please help guys i want to reach 150at ergo or 50at eth
bro how do i reach 150 Mh/s at vega 56, i only get 90Mh/s is it required to PBE one time patching like rx570/580 Series? please help guys i want to reach 150at ergo or 50at eth
This topic will close a year after the last reply.
Reply
You will see a count of new replies because you posted a reply to this topic.
Suggested Topics
HiveOS not installing Amd-drivers?
0
AMD Cards
May 26
RX 5600 XT - VRAM Bios Flash - Size 1.024 kB? Bios Source?
9
AMD Cards
2d
System unstable with 6 GPUs , all normal and stable with 5 GPUs 5700 XT
2
AMD Cards
Dec '20
Phonixminer reboot
4
AMD Cards
amd
miner
reboot
Apr 11
6800 - Phoenixminer 5.6d -clKernel 1 Stopped working
3
AMD Cards
Aug 2
Want to read more? Browse other topics in
AMD Cards
or view latest topics.
I had this same issue (same type of card). Everytime i moved RCDRD down below 18 it would crash. So i tried something different. Moved CL down (this aparently increases voltage to memory - i found this link a good read Complete BIOS Tuning Guide - "Memory Timing Setting" - ASUS Striker II Extreme: Mucho Bang, Mucho Bucks) and increased RCDRD to 21 ⌠this seamed to stabalise the card at a hashrate of 186 (i was on 175 before the changes).
amdmemtweak --CL 18 --RC 35 --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 21 --RCDWR 11
Not sure if this will help your situation, but worth a try. Yeah, these cards are frustrating ⌠suppose this is what makes it interesting.
I think lowering CL doesnât increase memory voltage; lowering CL means memory needs more voltage.
And if Iâm not mistaken, the default CL for Hynix is already 18, so adding --CL 18 has no real use.
I also have one that canât get lower than RCDRD 20, but as described somewhere above, it is built technically differently than the other reference cards, so I guess due to that difference it isnât possible to get this particular one a better hashrate.
Thanks this is so helpful! At first these straps performed worse than the Ergo straps but I noticed they enable higher mem clocks and therefore better performance. How did you tune cards with this? I havent been able to reach 1027 on any of my cards (top 1010, some 990), and 1027 seems oddly specific reminiscent of the 847 mem for the 860 soc MHz. Appreciate the help!
1027 is highest value you can run on MEM and stay in SOC state 1028, if you set mem to 1028 you move to 1107 SOC state so just needs more vdd.
here is a slightly slower strap but still runs 55Mh with Core 1040, VDD 825, Mem 1027.
amdmemtweak --CL 20 --RC 38 --RP 11 --WR 14 --CWL 8 --FAW 12 --RAS 20 --REF 65535 --RFC 248 --RTP 5 --RRDL 4 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 12 --RCDWR 12
On ERG If you set mem to 846 it sets you to SOC state 847, but since HiveOS allows you to override the standard SOC states I get a little more hash with the 847 mem 860 SOC.
Thanks! I was able to get above 55 MHs with the other strap, but by upping VDD and core clock. Havent fully optimized yet but liking the results.
Here is my tested running straps Contribution to the community after a lot of you helped me too.
Note: All cards are Vega 64 Cards Flashed to Vega 56 bios but Vega FE cards are not.
Using the Following Straps will consume around 150 W to 180 W
Stable High Perfomance Straps (Samsung Memory) :
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
OR
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
Using the Following Straps will consume around 110 W to 130 W
Stable Medium Perfomance Straps (Samsung Memory) with Low power :
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 10 --RCDWR 8
Using the Following Straps will consume around 150 W to 180 W
Stable High Perfomance Straps (Samsung Memory) :
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
OR
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
Using the Following Straps will consume around 110 W to 130 W
Stable Medium Perfomance Straps (Samsung Memory) with Low power :
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 10 --RCDWR 8
General NOTE:
If you are getting a lot of invalids with --RC 36 then change it to --RC 37
Avoid any freaking Gigabyte Vega Card 56/64 or Gigabyte Nvidia 30xx or 20xx cards they run really HOT and consume gazillion of Power .
NOTE: For Vega FE Cards (Samsung 16 GB HBM2 Memory ) Below are the straps
For Godâs sake i could not get this cards to go above 166 MH/s otherwise they would crash.
amdmemtweak --CL 19 --RC 44 --RP 14 --WR 14 --CWL 8 --FAW 18 --RAS 28 --REF 65535 --RFC 300 --RTP 5 --RRDL 5 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 12 --RCDWR 10
You should get around 55 MH/s @ 100W to 125W
Overclock Template:
CORE @ 1040
VDD @ 810
MEM @ 1025
Overclock Template:
CORE @ 1040
VDD @ 825
MEM @ 1027
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
QUESTIONS:
For ERGO Mining : Can we get STABLE straps that consume around 110 W to 130W on Vega 64 (Flashed to 56 Bios ) and get around 200 MH/s ???
For ERGO Mining : Also anyone with a better Vega FE straps ??
I also want to share my ERGO oc & amdmemtweak with this wonderful community, special thanks for @ss-silver , @Smining570 , @oni
i have 2 sapphire vega (64 & 56) 64 are flashed to sapphire 56 bios
canât remember which GPU is Vega 64, but i think itâs GPU 0 because itâs need 880 VDD and RC 37 in order not to crash
GPU 0 Tweak
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
GPU 1 Tweak
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
can still push it to achieve 200 MHS but i rather prefer efficiency
Guys! I need some help. Iâm struggling with my MSI Vega 56 Airboost and I canât get more than 150mh.
Any advice?
(Bottom one is a Gigabyte vega 64 with 56 bios and it runs ok, maybe a little too high power consumption, but stillâŚ).
every card itâs different, for example iâve 2 power color 56 ref , same bios and same model but have different oc setting to achive same hashrate.
try to up msi core to 1400-1450 and target the mem core about 930 - 960, if you have achieve satisfying hashrate, try to lower the vdd core until it canât be lower anymore (crash)
Yes, with Ethereum this wasnât much of an issue, but as Ergo is more power hungry, some cards get up to their (bios) power limit.
So please check your bios version; some of the MSI Airboost biosses are limited to 150W like this one for example:
If thatâs the case, you should flash one that has a higher limit.
Or if you canât check it (or donât know how), then just flash below bios.
Most reference cards (yes, MSI Air Boost is also a reference card, even if it has some extra red lines on itâs outer shell) have a bios with limit of 165W. I prefer below one for all my reference cards, being 56 Hynix, 56 Samsung and 64 Samsung:
As you can see, on first sight, they look the same, but the biosses differ enough inside for you to end up at oneâs power limits.
Thanks! You solved it 
Good to know!
Please also share your results after the bios change so we know the difference.
Hi, what timings do you use for vega 64?
Youâre again at the max wattage of 165W.
Youâll need more efficient straps if you want to go higher in hashrate.
Plenty of straps to test in this topic 
I have good results with these:
amdmemtweak --RC 36 --RP 11 --WR 14 --CWL 8 --FAW 12 --RAS 20 --REF 65535 --RFC 248 --RTP 4 --RRDL 6 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 10
But not all my Vegaâs accept it. If it works for you, good, otherwise try different straps to find the best one for your card.
PS: some other stuff to think about for those that are running bigger rigs.
I donât mine at max hashrate but around 180MH/s as my 13 Vega rig pulls ~2400W from the wall on a 10A fuse at 240V, so I shouldnât go higher or the fuse might blow (10A x 240V = 2400W).
If I want to go higher, I need another fuse box. My electric wires are already 2.5mm² so should be able to run higher.
My 2nd rig and my test rig run on a separate 10A fuse.
Donât go over your electric limits or your house might burn 
Youâre better off flashing it to a 56 bios and apply the 56 straps.
