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

AMD 6600xt in hiveOS

You want the lowest possible core clock, so you can lower core voltage as much as possible. Higher clock speeds will require more voltage, both core and memory, any chip really. 900 MHz is the lowest core clock a 6600XT will accept. But at 900, if the rig was rebooted, the miner wasn’t respecting the preset core voltage - it was using the default voltage, which was way higher and the power draw was thus higher. Only once you changed any setting and applied the OC, the core voltage would go down to your preset level. A setting of 901 is therefore the lowest core clock I found, that was still supporting full hashrate and did keep the preset voltage, if rebooted. The miner shows 900 MHz in both cases, just behaves differently. Perhaps this has been fixed now in HiveOS and a setting of “900” does keep voltage at the preset setting, I haven’t tried 900 since, to be honest.
EDIT: Another thing I did, that I would also recommend, is to lock the ETH config in the miner to a specific value. Otherwise the rig might get unstable, if settings are pushed to their limits. Notice I locked all configs to, in this case, A512:


A512 just so happens to be a near-perfect config for my 8 cards, but it could have been a combination of various config values separately for each card, just one that never changes. This is because during auto-tuning, when the miner is starting, it can assign different configs at each reboot and different configs will have slightly different power requirements and suddenly a core voltage you set, that SEEMED stable for days/weeks, gets unstable soon after rig reboot. I found out the hard way :wink:

1 Like

Thank you. Very useful info!

How did you lock in?

Auto turn seems to override my --eth_config=A512

Best I can do while keeping my power under 90 watts


This works on TRM 0.9.4.2. Perhaps you’re using an older version?

hahah

I’m dual mining TON and I put it under the wrong coin

Rocky mistake :stuck_out_tongue:

Turns out if you manually set this the TON side won’t mine

All my 6600XT’s are drawing 53-57 watts. Running TRM, Core 1000, VDD 700, Mem 1130. All cards are mining 32.0x MH/s. Got one 6600 in the mix.

Hey Area51

That’s only running a single coin…

I’m Dual mining TON

image

How do I do this for just one gpu? Using your example, what do I type to get config A512 for just GPU 1?

I tried:
–eth_config=A512 1

It didn’t work.

You chose the order of the cards for your config
Like:
–eth_config=B608,A628,A480,A1152

So card 0 is with config B608
Card 1 is A628
And so on

ok, but how do I just tune 1 out of 4 cards? Let’s say I want to tune gpu 0, but let auto tunning tune the rest.
–eth_config=B608,0,0,0 doesn’t work

For cards that should auto tune don’t set any value.
Like:

–eth_config=B608, , ,A,B506
Without space between, ,
So gpu 1,2 would Auto config with A and B mode, gpu 3 would autoconfig in A mode and
0 and 4 have specific configs.

It’s crazy how even after many months of owning a card and thinking you know it inside out, you see something that makes you rethink your strategy completely. An earlier post by @niexasadas got me thinking, since his SoC voltage was much lower than mine and he said it was stable, so I retested SoC voltage on all my cards. To my delight, all cards could run on much lower settings than what I had and wattmeter showed a noticable drop. After that I also retested my VDDCI setting and it turned out that one also had some room for improvement. I ended up with this efficient and stable (48+hrs) config:



Some cards, as it turned out, easily dropped to as low as 745mv SoC. This lowered my power draw by around 10W, to 571W at 260 MH (455 kH/Watt).
Note how the micron-ram-based GPUs run on much lower VDDCI than their samsung-based counterpart. That 610 ended up being arbitrary, they would take any value, without any visible drop in power draw below 650mv. I just stopped decreasing it at 610, seeing no effect. The samsung-based one, was a different story. Below 675mv VDDCI, hashrate would start dropping by about 0.04 MH with each 5m drop. No idea how I never caught that.

I doubt that going under 765 for SoC VDDmax makes a difference, or might be negative impact. Could you please check that? I have tried for some cards and it seems nothing gets improved. Also it is a pitty, that we need to restart the rig after those changes - if not soc values goes to default.

Also guys.

I wound out that using TRM R-mode makes a bit difference in thermal throttling. my hashrate would go down by 0.2mhs if mem temp goes from 48 to 52, with using R-mode it stays almost the same using same values.

I am using teamredminer-ver. 0.9.4.7
argument --kernel_vm_mode=RR

and also used command in console:

echo -e “amdgpu.vm_block_size=11\namdgpu.vm_size=2048” > /hive/etc/grub.custom && selfupgrade --force -g && sreboot

I’ve just tested the difference between my current SoC settings and having them all 765+ mv SoC - and yes, it was basically negligible, around 1W. It almost gets lost in power fluctuations on the wattmeter. But even 1W, if stable, is a gain :slight_smile:
I was very curious about this 0.2 MH drop you experienced beyond 50 C mem temp. I cannot replicate it at the moment, as all my GPUs report slightly above 50 C. But I might open all windows on a cold night and see if that’s really the case, just out of curiousity. Haven’t tested the new R mode yet, but sounds promising.

Yeah I really need to look into that. Is there a HOW-TO somewhere?
ETA: figured it out :slight_smile:

Constant crashing, I give up. I tried LOLMINER with TON, 1000 CORE, 700 VDD and 1130 MEM overclock, my usual for ETH with TRM. That’s 6 x 6600XT and 1 x 6600 on this rig.

Hi , i need help with the GPU 0, is a 6600xt Asrock, i have 2 but only one works on 32 mhz, gpu 0 cant do more :C… any suggestion?

Seems to have settled down, working great.