There is an upp command to lock memory too, I believe. What miner are you using, btw?
Indeed, found about it on the github page.
upp -p /sys/class/drm/card2/device/pp_table set smc_pptable/FreqTableUclk/3=1075 --write
But once done my F state is locked to 0
and not to 1 like on the other GPU
No now… how do i force the F state to be 1?
Btw teamblackminer
I made the script work for 6900 XT but hashrate didn’t change any. Using TeamRedMiner.
Hello again. 
I’ve been pretty busy lately. But I’m glad the topic is going 
. Now the strange thing is try to limit the SOC frequency to a minimum value (800MHz) and check the hashrate again.
Hey I have a 6800 and I had super low hash rate on Hive OS. Windows 62MHS/ Hiveos 54 mhs. So I did the fix of using 6800XT drivers on to my vanilla 6800 and voila 62 MHS. Now with the patch today which talks about fixing SOC settings on rx6*00 I am back down to 51 MHS on this card. Am I correct in assuming I can use my original 6800 bios and adjust the SOC now to squeeze out the hash rate?
The state doesn’t matter, as long as you have that little * next to 1941 it’ll run at 1941.
The F state script is pointless for cards that already run at their max.
Only the 6800 (non xt) suffered from that issue.
Not all cards will run that high mem. Try lowering it to 1050 and start working your way up slowly, 5 by 5 per 24h.
The update today also messed up my cards hashrates. One 6800 dropped to 51, another to 60 and a third to 61. My 6800XT’s dropped to 58…
Downgraded again but still not getting my hashrates from before the patch where all were >62.
Probably have to do all the fine-tuning all over again 
1 Like
What miner are you using?
THANK YOU! My cards started off at 63MH months ago then started falling to 58-60MH. Now they’re back to 62-63MH
Upgraded the kernel with hive-replace to #140, then upgraded to the latest HiveOS version 0.6-211@211029.
At my baseline, with no scripts, and no aggressive overclocking, power consumption is up when they were all under 100w, but they are shockingly similar in performance(GPU2 always runs hot in this system):
Edit: Upgraded (4) rigs. Baseline configs, no scripting, and not one of the 4 brands of 6800s is running less than 61MH/s. Where was this combination of Kernel&HiveOS months ago?! 
Note: some of these were the formerly pain in the rear 57/58MHs units.
Edit2: none of the GPUs are running F state less than 1551 with (again, this is quite a few GPUs and across four brands):
0.6-211@211029
5.4.0-hiveos #140
A20.40 (5.11.0701)
2 Likes
I haven’t updated yet but my cards have been running solid for a couple of weeks now at over 62 MHs, with several over 63 MHs. Should I still upgrade the kernel and update Hive?
If yours are stable for weeks on end, I would not touch them until folks run the new version a few weeks.
fwiw: Those temperatures bug me, so I’d increase the fans. Fans are cheap, GPUs, not so much.
1 Like
At least I know its not just me. My worst fear is that I need to be forever downgraded because I have a anomalous issue.
Yes finally HiveOS team integrate our method for F state. Today I check it and everything looks fine. Now next point is to see how we can make “fast timings” works in Linux
2 Likes
TRM got it(c mode) done for VII’s, it will happen ;>)
Appreciate your efforts to bridge the gaps. Thank you.
Well, something is wrong with the new upgrade 0.6-211@211029 kernel 140.
It seems that now both my 6800’s as my 6800XT’s are running FCLK 1551.
For the 6800XT that should have remained 1941. Now they have lower hashrate due to lower FCLK of 1551…
What miner? How are they so efficient in electricity?
Hi, so what do you suggest, to remove “our” script out of system a keep newly integrated Fstate mod from HiveOS?
And btw, what is this?
- Reduced power usage on AMD RDNA2 RX 6x00 series (Recommended SoC ranges to reduce power usage: clock: 800-960MHz, voltage: 800-900mV. WARNING Only change these settings if you know what you’re doing!)
It really means an option to set up core at ~900Mhz @800mV? It seems as a very strange setup to me, do I understand it wrong or it is really working? :]