core voltage is not being applied at boot, any change in the oc will apply core voltage. Seems to be a hive isue
1 Like
Hi!
Could you tell us, what hashrate/watt ratio do you get with these settings?
Yes
2 Likes
Does anyone know if it’s possible to lock DCEF state?
Sometimes I have underperforming 6800s that hash at ~56MH vs ~62MH. From comparing it to a properly performing card I’ve consistently noticed that the DCEF state ends up like this:
Typically this happens when I apply new overclocks (as simple as the adjusting the fan speed).
I also got the same behavior…
Same problem over here… i simply can’t change the DCEF state from o to 1.
I’ve seen somewhere that maybe UPP can change that. (exactly as I’ve changed the F state to 1551 as you can see here.
To write in another topic a few weeks ago.
1 Like
I thought with newer versions of hive you weren’t required to manually modify f-states for 6800 cards?
Why are people still needing to run these upp commands?
Some folks are attempting to set DCEF values for improvements.
F State for 6800’s are set at 1551 in the latest stable kernel.
its not really in the kernel. Hive is not doing the upp by itself, you can see it in the screen when hive boots.
Reason I ask is that I have a 6800 that is still pretty finicky.
Frequently when I reboot the rig, power consumption on all my cards is about 5-10% higher than it should be. If I simply make a minor adjustment to any voltage value on any of the cards that triggers them all to run in a slightly lower power state. For all cards except the 6800, their hash rates continue to be normal.
For the 6800 though the hash rate crashes to 50MH. It can be a struggle to get it to snap out of this and sometimes I just end up having to reboot and try again until eventually when I make the minor voltage value adjustment to drop the power states it doesn’t crash back to 50MH and maintains 60MH.
I have no idea why it acts in this way.
Fair enough, but why suggest to folks to run it again when that was included in the kernel release?
Take the power reported to the OS as a guideline, measure at the wall and expect values far higher than reported.
In my sample of 6800 GPUs, every finicky unit wanted more power to maintain stability. Yours may be in the same boat as this one:
Sure, I’m aware that the value reported in the software is around 10-15% under what you actually see at the wall for AMD cards. It’s the same in Windows.
Once I make that slight voltage mod to any card in the rig, if things work right, this is how it ends up here. All cards in a slightly lower power state and operating at normal hash rate. The 6800 is also hashing normally, and at only 111w in the software. Note, I have checked, and this is an actual reduction in power at the wall not just some software trickery.
Attaching a photo.
Just as what happened to me, ALWAYS !
i dont konw why either.
Hi,
Why hiveos limit Mem Clock to 1075?
I already read that on linux we need to configure half value than windows mem clock configuration, but I notice that when I configure 2122 Mhz on windows for mem clock, on miner status chart shows 2116 on GMClk column. And when I configure 1075 Mhz on Hiveos OC for 6800 AMD, miner status chart shows 1075 on GMClk column using the same miner.
With the above example it does not make sense to limit Memclock on hiveos as the card can use more memory.
On Windows
On Hiveos
This is because hiveos is working very bad with 6800 cards. Core voltage is not set at boot, you can verify this with amd-info. Changing anything in the oc makes it be applied
1 Like
You can always try with other Linux based OS and distributions if you like, but I’d suggest staying <=1075 for 6800 or 6700XT’s for best results.