2x RTX 3070 - same settings different hash rates

[fsp]

hello everyone, I’ve got a brand new 3070 for my rig. Once installed, I used the same settings for the previous one, and the oldest dropped the hashrate from 60 to 57 (see screenshot).

Any guess of what cause this? What can I do to get the same for both again?

[cryptobitch]

i would advice you to set the fan on minimal 65%

[fsp]

Why setting the fan?

[cryptobitch]

because the temperature that you see in HiveOS is the GPU temp. But it doesn’t mean the memory temp, is the same. If GPU is around 55 degrees, I assume that the memory temp is between 80/90 degrees. If memory is too hot, it will throttle. This could be why you see different hashrate. I use the autofan routine, set the gpu temp on 55 degrees and minimum fan on 65%. So the memory stays also cool.

[cryptobitch]

besides that, 63 degrees GPU is too hot

[fsp]

I’ve got some increase in stale and invalid shares today. I guess because the day was way too hot (over 30°C).

I’ve set it as you mentioned. Let’s see how it performs overnight.

[cryptobitch]

Fingers crossed…

You could also decrease core frequency, so there is more power available for memory
these are mine settings. Only 1 rejected in 6 days

[arkeyn]

@cryptobitch

63 degrees GPU is too hot

what’s your suggestion?
I just figured that mine was running 66 with fan at %75 but it was like that from the very beginning, never payed attention to it

[cryptobitch]

It better to keep it under 55 degrees

[Tim2021]

Hi there, how about this settings, some advice, i am just mining about 2 weeks ,thanks .

GeForce RTX 3070 7982 MB · Gigabyte
Samsung GDDR6 · 94.04.3A.40.DB · PL 100 W, 270 W, 270 W

63.70 Mhs 54 75% 75 -502 2950 132w

[fsp]

Somehow the overclock settings weren’t stable and I’ve got lower hashrates and lots os stale/invalid shares. I returned to the previous settings including the autofan settings, which reduced a lot in gpu/mem/room temp (higher noise tho). Thanks

[fsp]

Which pool/miner are you using with those settings?

[danielvm]

how do you use autofan? which setting are you using?

after reading this post I’m using like this:

[cryptobitch]

Yes these settings are what I also configured

[danielvm]

take a look on this

RTX 3070 settings

I was reading this post and did a chance to what the guys was talking They ( rul3s and j2h4u ) fixed the core clock, first by command line e than by hiveos console (after updating for the lastest version) From rul3s: On my research I’ve found that manually setting core clock is much better than setting a power limit to the card. They basically use the command “nvidia-smi -lgc 1150” to set the maximum and minimum core clock. I did the same and now I can get 62 Mh/s in each 3070 [Screen…

[fsp]

Where did you type that? Teleconsole?

[danielvm]

I stop the miner and type the command on hive shell

[nucknfutz]

make sure your running the latest hive OS, and try these settings.
they give me 61.25MH/s @ 120w with a temp of 38-40degrees

fan 60
core 1080 (no plus or - symbol just 1080)
mem 2450
power limit 125

[danielvm]

Where you able to solve your problem?

Here I got a solution and now I have all at the same pace

RTX 3070 settings

After lots of back and forth I nailed it. Basically the problem here was more related to the pool than the OC. I tried several OC settings and all the possibilities was giving me a lot of rejects, even if I lowered the memory clock all the way down. After some reding I tried a new pool and than it happens. Since that I changed my OC settings to the best I got and its pretty stable, solved the problem of different Mh/s in each Gpu and also the system is now cooler than it was before. the final t…

[fsp]

My problem was solved by changing the miner. Somehow, nanominer was having trouble with one GPU returning almos 100% rejected shares. The log file was relating the problem to a “memtweak” command, which was strange since there is only nvidia on rig.

I moved to Trex and there wasn’t any of the problems mentioned above.