Anyone know if absolute core clock will work on the 3090? For all of my other cards it massively improves thermals and power consumption.
I was trying to think of a way to validate- Haven’t come up with anything yet… I saw the update to T-Rex this morning.
Do you happen to know what -200 and -300 translate to (roughly) in absolute clock speed terms? I might be the guinea pig for everyone here to see if it works.
Sorry - I don’t.
using suggested 990 absolute core clock lowers my Hashrate significantly for my 3090…No thx…LOL
It’s a range though, you tweak from there. If you have it like you want it, definitely don’t change - but for those just starting, it offers massive power savings. This is what I am seeing just to start with (after letting my rig cool down, as it had been throttling due to poor airflow - I moved a fan and it should be better).
First set the memory for safe and stable 2000Mhz. After this, adjust the core around 950-1150, looking for hashrate drop/core temp/stable. Next you can find max memory setting, depends of sillicon lottery, temp of memory on tour card.
If you have drop hashrate after a few minutes you need to replace thermal pads on memory, add some additional air circulation on card (some big fan on rig) , set lower frequency on memory. Ideal if you doing all of it thinks.
^^Exactly this - I hadn’t seen this last post before testing, but Tony is exactly right. Just a note for the crowd, HP 3090’s (custom made, they come in the Omen 30L or at least used to) run hot. I’m pretty certain they have more headroom, but until I swap out the thermal pads + paste, this is the best I can do:
Stable ranges for my HP card were: 990 - 1060 absolute core clock; 1700 - 2100 memory.
I had to keep the fan high (around 85%) or it would quickly pass the 55 degree Celsius mark and start throttling drastically (e.g. from 118 MH down to 107 or 109 MH while the wattage was reduced automatically to cool it down). While I could circumvent this with a bios update, I wouldn’t recommend it - there is something wrong with the thermal constructions on the card (paste + pads most likely) that is causing run away thermals. I’ll update here if/when I swap those out.
I never set a power limit, as I wanted to really test the card to see where it peaked - 302 - 304 was about where it is at during peak hashrate.
Hope that helps anyone trying the same - like the guy above said, it’s silicon lottery (and also in this case the thermals of the board partner).
I have had the following RTX 3090 cards and used them in identical environments. Here are the resultats at 300w, -300 core, 1200 memory
MSI Suprim X: 123 MH/s
Gainward Phantom OC: 115 MH/s
KFA Hall of Fame: 115 MH/s
Gigabyte Gaming OC: 88 MH/s
This is after running the cards for more than 24 hours
mine same starts nice then drops to 24MH/s (verry bad for mining) since 24gb and 10k cudacores
can’t find right overclock settings. anyone have link for tarining video how optimaly overclock?
MSI Suprim X: 123 MH/s No thermalthrottling
Gainward Phantom OC: 115 MH/s thermalthrottling
KFA Hall of Fame: 115 MH/s thermalthrottling
Gigabyte Gaming OC: 88 MH/s more thermalthrottling
Your memory are too hot.
hashing about 100+ try these settings in overclock
Core clock -100
Mem Clock 1000
Fan 0
Power limit 260
it depends a lot on you ROOM TEMPERATURE to prrevent overheating you might want go as low as ¬85Mh/s
Did you have problems with the thermal throtling? change pads or something? I got 2 RTX3090. One gives me 120mhs, the other 105mhs. Can’t figure out why.
Thanks!
when hashing speed slow down after a few minutes from start is therm.thro. on memory.
Would it be safe for the GPUs? I am just wondering if I would have any problem overclocking it like this! Thanks!
how are you setting absolute core and mem freq on the 3090?