RTX 3080 TI LHR Settings

[wishZor]

Hi,

Finally managed to make an account.
I ended up here for following reason, until recently I was running my modest rig with 3 cards quite stably. In this particular case talking about the 3080TI. Up to more or less a week ago it was stably running at 80MHs with -200 cclock, 2600 mclock, 280 pl and auto fan.

Over the weekend I noticed the MHs dropped to 50MHs. Found out that I missed a few nbminer updates so ran those. Now I can’t consistently go higher than 70Mhs. (lower no problem) It does often start out with 80MHs, but drops rapidly.

I’ve tried going back to nbminer 39.5-7. I’ve tried trex miner.
I’ve tried adding various arguments to no avail.

I’ve tried various cclocks between 0 - 1350, various mlocks between 0 and 3000, various power limits etc etc.

Right now even running all at 0 which seems to be the most stable staying around 70.

I’ve tried flashing the bios now running GeForce RTX 3080 Ti 12053 MB · EVGA
Micron GDDR6X · 94.02.71.80.BF · PL 100 W, 350 W, 366 W (It’s actually an MSI card)

Right now when I have everything on zero it’s giving following nvidia-info:

=== GPU 2, 06:00.0 GeForce RTX 3080 Ti 12053 MB ===
Bios 94.02.71.80.BF, PCIE Link Gen 3, PCIE Link Width 1x
S/N -, UUID GPU-8f33ad94-8a89-432f-15ec-93d1a07ec154
Power Limit: Min 100.0 W, Default 350.0 W, Max 366.0 W, Current 350.0 W
Frequency: GFXCore 1500 MHz, MEMCore 9251 MHz
Memory: Total 12053 MB, Used 5029 MB, Free 7023 MB, Micron GDDR6X
Utilization: GPU 100 %, MEM 100 %, Throttle: SWThermal
PSTATE P2, PWR 229.0 W, Temp 56 °C, Fan 54 %, RPM 1135, RPM 1135

When I change it to cclock 1000-1350 , mclock 2000-2600 it won’t say Throttle: SWThermal

cclock 1150, mlock 2200, pl 280, fan 80%:

=== GPU 2, 06:00.0 GeForce RTX 3080 Ti 12053 MB ===
Bios 94.02.71.80.BF, PCIE Link Gen 3, PCIE Link Width 1x
S/N -, UUID GPU-8f33ad94-8a89-432f-15ec-93d1a07ec154
Power Limit: Min 100.0 W, Default 350.0 W, Max 366.0 W, Current 280.0 W
Frequency: GFXCore 840 MHz, MEMCore 10351 MHz
Memory: Total 12053 MB, Used 5029 MB, Free 7023 MB, Micron GDDR6X
Utilization: GPU 100 %, MEM 100 %, Throttle: PowerLimit
PSTATE P2, PWR 218.7 W, Temp 50 °C, Fan 80 %, RPM 2358, RPM 2353

Anyone any other bright idea’s? One thing I’ll be trying this week is move the rig to another (colder) location. I would say the current room is slightly above room temperature.

Before I used the settings I’m also using on gaming desktop on that rig I am getting around 88-90MHs.

Thanks in advance.

[mathelm]

Little hard to follow but a pic is worth a thousand words, so here’s my setup running T-Rex.
If I read it right, I think your memory is too high (2600) so it may be throttling. Either way, take fan off auto and jack it up to 100% for testing, then prob 75 or 80 for running in case backplate memory is getting too hot.

Ignore the red underlined bios, that was for another post.

[wishZor]

You’re right of course!

These are the settings it ran over last night:

Should admit, I’m not sure how reliable the screenshot is at the moment. I tried to switch to stats and I got an internal server error from hiveos itself.

[wishZor]

After having moved to a slightly colder room and taking some of the feedback in mind I’ve gotten to following results:

If there’s any further hints / tips I’d gladly take them onboard.

[keaton_hiveon]

locked core clocks, like @mathelm is using, will increase stability and typically lower power draw as well.

[wishZor]

I’ll retry those again during my work day tomorrow so I can monitor, so far I’ve experienced when I go above 0 the MHs vary between 50 to 75. Where 75 is at 0 and 50 when at 1200.

I’ve ran the above setting for the last 4-5 hrs consistently at 85MHs. Which is still far from the 90+ some people seem to have, but at least 5 more than what it was running stable at before.

[BigPoppa]

I used to have a 3080TI VISION from GIGABYTE. It was unstable AF. Some other guys manage to get lucky and get a decent card, mine was so bad tho. I was locking at 79-82mh with a lot of thermal throttling from the memory temps.

I was trying to get rid of the card. One day some one messaged me to give me some money and a 3070 non lhr for my 3080ti. So I agreed upon that and now I am very happy :P. So, if it’s not working as excepted just throw it away…

[BigPoppa]

Also, try using GMINER 2.7 version. That miner gave me some kind of hashing stability…

[keaton_hiveon]

have you tried trex? i dont think ive seen anyone getting 90+ on anything besides trex 24.7 or 24.8

[wishZor]

lol

I was unhappy when it went unstable out of the blue.
Happy with todays results. Still 85MHs stable. 3070 non LHR runs at 60-62MHs.
Not ideal settings but it works.
Would trade my current 3070 for another 3080TI too

[wishZor]

I have tried trex. Wasn’t so successful.
Redpanda did a video on nbminer reaching 91mhs
On my gaming pc I reach 88-89.

[keaton_hiveon]

heres what i run, trex 24.8, bounces between 89-94mh

[mathelm]

Funny I had to lower my FTW3 mem from 2350 to 2300 because it had dropped down to 84-85mh after a few days at 90+. It’s now back doing 88-89.

old pic before 2300 mem

[cryptoworx]

stable, trex 0.24.8

[mchtakblt43]

Hello,

I have Rog strix 3080 ti, using t-rex 0.24.8, i think temp too high for rog strix. How can i lower the temp ?

[everon1218]

@mchtakblt43, Try cooler room

[cryptomoriss]

hi what is the item you underlined? bios version? batch series?

[mathelm]

Just the bios number, and that pic was for a different post. But it’s a EVGA FTW3 3080 TI, Part Number:12G-P5-3967-KR

[WaspyHalo]

I"m getting about 359 Eff/Watt on my 3060Ti and right at the 310 Eff/Watt on my 3080Ti. Wish I could figure out some settings in Hive to get the 3080Ti up there a little higher ron the efficiency but no luck thus far.

[Pjm77]

Hi! Your EVGA 3080 ti is PIN:12G-F5-3967-KR? Thanks.