More
referral
Increase your income with Hive. Invite your friends and earn real cryptocurrency!

Vega 56 and Vega 64 guide


R-Mode, Sapphire Radeon RX Vega 64’s with 56 BIOS (all Samsung memory chips)
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
I’ve lowered about 20w per card on the software, at the wall almost 200w saving if my TPLINK Kasa is correct.
Hope this helps others as a good start.

1 Like

My Vega reference rig was running too hot; at 55MH/s they ran at almost 80°C HBM. It is probably still a safe temperature, but I don’t like it and try to keep my HBM’s closer to 60°C. Off course during summer this is very difficult.
So below test is not about the highest hashrate, but to lower temperatures.
Although they’re only at 50MH/s now, they run about 10-15°C lower and 200W less at the wall (1220W now, 1410W before), so that’s a big win for me, and they’ll survive summer like this :sweat_smile:

SoC Frequency = 910Mhz

amdmemtweak --CL 20 --RC 36 --RP 10 --WR 14 --CWL 8 --FAW 12 --RAS 20 --REF 65535 --RFC 216 --RTP 5 --RRDL 6 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 11 --RCDWR 12

Going lower than 755mV VDD didn’t work; as in the rig didn’t crash, but in the background (saw on mining screen) set the VDD to 1000mV (and thus higher Watts again). Don’t know why. Maybe a built in safety?

During coming days I’ll try to do my non-reference rig with Asus Strix’es and Gigabyte Gaming OC’s.

My non-reference rig :


0 = Asus Arez Strix 56
1 & 2 = Asus Rog Strix 64 > 56
3 to 7 = Gigabyte Gaming OC 56

They mostly work with the same settings as my reference rig; being core 925, mem 900 and soc 910.
I do have some exceptions in the other settings:
All but 4 works with:
amdmemtweak --CL 20 --RC 36 --RP 10 --WR 14 --CWL 8 --FAW 12 --RAS 20 --REF 65535 --RFC 216 --RTP 5 --RRDL 6 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 11 --RCDWR 12
4 works with:
amdmemtweak --CL 20 --RC 36 --RP 11 --WR 14 --CWL 8 --FAW 12 --RAS 21 --REF 65000 --RFC 216 --RTP 5 --RRDL 6 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 13 --RCDWR 12
And 4, 6 & 7 need a but higher mvdd to lessen the errors. Still finetuning there.

Again, these are not to have the highest possible hashrate, but to lower the HBM temperatures and survive summer.
Before R mode this rig was running at 426.5MH/s @ 1300W.
Now with R mode it is running at 399.5MH/s @ 1165W.
And most importantly; HMB temperature dropped around 10 to 15°C.

EDIT: I’ve changed some values on the Gaming OC’s as below are more for Hynix memory and above are more for Samsung:
3, 5, 6 & 7:
amdmemtweak --CL 20 --RC 36 --RP 12 --WR 14 --CWL 8 --FAW 12 --RAS 22 --REF 65000 --RFC 216 --RTP 5 --RRDL 6 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 11 --RCDWR 12
4:
amdmemtweak --CL 20 --RC 36 --RP 11 --WR 14 --CWL 8 --FAW 12 --RAS 21 --REF 65000 --RFC 216 --RTP 5 --RRDL 6 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 13 --RCDWR 12

And then the last Vega’s on my finicky test rig:

0 & 1 = MSI Airboost 56
3 = Gigabyte Gaming OC 64 > 56

Same as above rigs, core 925, mem 900, soc 910.
0 has mvdd 1320
2 has mvdd 1200
(I don’t know why these aren’t showing in the overview as in my non-reference rig)

0 is running:
amdmemtweak --CL 20 --RC 36 --RP 11 --WR 14 --CWL 8 --FAW 12 --RAS 21 --REF 65000 --RFC 238 --RTP 5 --RRDL 6 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 14 --RCDWR 12
1 is running:
amdmemtweak --CL 20 --RC 36 --RP 12 --WR 14 --CWL 8 --FAW 12 --RAS 22 --REF 65000 --RFC 238 --RTP 5 --RRDL 6 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 18 --RCDWR 12
2 is running:
amdmemtweak --CL 20 --RC 36 --RP 10 --WR 14 --CWL 8 --FAW 12 --RAS 20 --REF 65535 --RFC 216 --RTP 5 --RRDL 6 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 13 --RCDWR 12

Before R mode this rig was running at 148.5MH/s @ 496W.
Now with R mode it is running at 145.5MH/s @ 470W.

Still not satisfied about these ones; but hey, that’s why they’re on the test rig :sweat_smile:

Really struggling with GPU3 below. It was running fine then I removed a 580 which died and GPU3 went into full meltdown. Settings, straps etc all the same but not I keep getting 'teamredminer: GPU 3: detected dead". It ran stable for months and now refuses to run for 10minutes. Tried many things.

Current strap: amdmemtweak --RC 37 --RP 11 --WR 13 --CWL 8 --FAW 12 --RAS 22 --REF 65535 --RFC 239 --RTP 6 --RRDL 5 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 12 --RCDWR 12

That’s a Hynix, but those straps are more for Samsung.

If it’s a reference card, try:
amdmemtweak --CL 20 --RC 36 --RP 13 --WR 14 --CWL 8 --FAW 12 --RAS 23 --REF 65000 --RFC 238 --RTP 5 --RRDL 6 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 15 --RCDWR 12

If it’s a Gaming OC, try:
amdmemtweak --CL 20 --RC 36 --RP 11 --WR 14 --CWL 8 --FAW 12 --RAS 20 --REF 65000 --RFC 238 --RTP 5 --RRDL 6 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 13 --RCDWR 12

Both with higher or lower RCDRD depending on what yours can handle.

Or check your riser and power cables :wink:

Thanks for the suggestion Oni.

I tried those straps but card comes up as ‘dead’ crashes after a couple minutes.

Tried your suggestion on upping/lowering RCDRD and gave up at 20/10. In total around 50 (i’m not joking) combinations of straps and core/vdd/mem without success as the card keeps crashing in minutes. Nothing on, the card runs without crashes.

This POS is about to get binned lol.

@cuthy2k I have a couple suggestions. First I would suggest updating to the latest version of hiveos and using teamredminer version 0.10.0 if you haven’t already. You should also enable r-mode. You’ll use less power and generate less heat. To do that, add the following to the miner section of your flight sheet in the extra config arguments:
–kernel_vm_mode=RR
Also delete these options from the config if they’re in there:
–eth_big_mode_adjust=64
–eth_config=A
–eth_config=B

You’ll need to reboot the rig once the config is set in the flight sheet, if it doesn’t reboot on its own. And if it doesn’t start hashing after the reboot, you may need to manually select TRM version 0.10.0 in the flight sheet instead of “the latest”.

Read the tuning guide here:

Reduce the core clock to 900-950. Reduce the VDD to 750-800. Maybe start on the high end and reduce core/VDD by 5-10 until you lose stability. The guide says 925-950 should be enough. I’m currently running at 965 core and 800 VDD on my Vega 56s / 1020 core and 820 VDD on my Vega 64s, but I’m dual mining Eth+Ton which requires a higher core/VDD.

For the straps, I always start with the following:

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

If you get reboots or invalids, increase the RC by 1 and restart. If you’re still getting reboots or invalids, increase the RCDRD by 1. All of my Vegas (with the exception of two) are running with RC 37 and RCDRD 13. I’ve gone up to RC 38 and RCDRD 14, but I think you get a better hashrate if you reduce the memory clock instead of going above 37/13. But it’s also fine to use 38/14 if you’re having significant problems with a GPU.

Edit:
After messing around with my config a bit more, I’m getting the best hashrate keeping the RCDRD at 12 and the RC at 36-37. Start with RC at 36, if you get invalids or reboots, increase to 37. If you still get invalids or reboots, reduce the memory clock (as opposed to increasing the RCDRD). With r-mode: for Vega 64s I’m using 1020 core and 820 VDD, starting at 1060 memory clock. For Vega 56s, I’m using 965 core and 800 VDD, starting at 950 memory clock. I’ve found that some GPUs need a higher core clock and VDD, so if these settings don’t work try increasing each by 10 until it starts hashing. With these settings I’m averaging 51 MH/s with my Vega 56s, and 55 MH/s with my Vega 64s (which are all flashed with a Vega 56 bios). Again, this is dual mining Eth+Ton (with the max_dual_impact setting at 0.5), so you may be able to go lower on core/VDD, and you may also get a better hashrate.

2 Likes

After much mucking about I’ve got RR mode to load. Now I just need to play with the settings… I’m too tired lol as Vega’s really take it out of you! Will try another night and see how I get on!

Thanks for the help so far guys, will keep you posted.

1 Like

so complicated.


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
hynix
amdmemtweak --CL 20 --RC 36 --RP 12 --WR 14 --CWL 8 --FAW 12 --RAS 22 --REF 65535 --RFC 249 --RTP 5 --RRDL 6 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 13 --RCDWR 12

Obviously you start with higher VDD and higher CORE.

The only values I need to change is RP and RCDRD.

I would start with RP 12-13 and RCDRD 15. If card works with RCDRD 12, then you can reduce RP to 11.

Also use --high_sample_mode=8 argument - this will show invalids much faster.

I also noticed that you can achieve same hashrate by many many different cores and straps, but you will not be able to stay stable over the card limit.

2 Likes

hi,

I have some vegas, and I wanted to try R-mode on them and bring the power even more lower but if I bring the core voltage under 760mV, the power goes up rly high. I guess it is some kinda limit. Any idea how could I bring it down?

original setup core:1000/760mV/960mem
new setup core:925/760mV/960mem
the power draw on wall is the same 141W

I’ve never messed with RP before. I’ll have to give this a shot.

Gone with RR mode and GPU3 continued to be a b!tch. GPU 3 is Gigashyte Gaming OC.

Took a few goes but running stable for 2hr 30min so far and counting.

amdmemtweak --CL 20 --RC 38 --RP 11 --WR 14 --CWL 8 --FAW 12 --RAS 21 --REF 65000 --RFC 216 --RTP 5 --RRDL 6 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 14 --RCDWR 12

Now to tackle the 5600XT! :joy:

Glad you were able to get it working. Also looks like you brought your power usage down quite a bit with r-mode.

managed over 5hrs stable before GPU1 went dead. Some would dive right into fixing that but after 2 days of tackling one Vega 56 appearing dead every 2-3minutes I’m happy with 5hrs and 8min of uptime! Thanks for helpl!

if one goes dead; upp that one’s vdd by 5 and see if it helps; if crash again; upp again by 5 until stable :slight_smile:

Hello everyone,
share my vega’s for R mode
For the straps, someone RC 38
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

I will try more low power

1 Like

That’s great power usage. Are these 64s or 56s?

Sorry, I am late
Some are vega 64, some are vega 56

I was able to tweak a bit more, 36+ hours running up with those settings:

0 - gigabyte ref
1 - sapphire nitro+
2 - gigabyte gaming oc
3 - sapphire pulse

1 Like