I had Russian servers enabled, my rig is in Europe, I disabled Russian servers, for now, seems like I have fewer errors than before.
I had the same issues, and was weird my Gigabyte was the one that had different performance (24MH) than my Asus (2x, 31.5MH)⌠I follow your advises and now is in the same range see below in case it can help to someone:
If any of you have updated to HiveOS version 210401 or later, it now has the option to put absolute core clock on the OC page of the card. Iâve tried it out but with mixed results.:
Note that whenever you put a number higher than 500 on core clock, this will be interpreted as the absolute core clock frequency that you want for the card. On the image above, you can see that there GPU 3 and 4 no longer need a PL value. GPU 2 is using my older OC settings.
Itâs my first time to try it out so thereâs more you can check but I hope somebody discovers better setings that can result in lower watts or even a small bump in hashrate.
I am using an Asus GTX 1660 Super with Hynx memory and was getting 20.60Mh/s, a really low hashrate. Just changed the memory to -1004 and getting now 31.34Mh/s. Awesome!
Not sure why underclocking the memory gets such a boost in hashrates, thoughâŚ
Also, I noticed that my temperature increased about 5 degrees and it is using 23w more. Not sure if it is related.
I tried this for my 1660 Super Asus that was running hot and high wattage of 88/89.
I put +930 for the Absolute Core and left memory as -1004 and my temp went from 61 to 56 and watts down to 73!
Though if the Core is at 930 I would think the opposite would happen⌠is this case it lowered temps and watts.
Whatâs recommended? All my 1660supers with Hynix Memory are at -250 and -1004 PL around 80-85 and in the mid 50s temps.
Using absolute core clock, you can try values from 900 to 1200 and see which one gets you the best efficiency. Happy hunting!
Card : NVIDIA GeForce GTX 1660 SUPER 5944 MB
Mem : Hynix GDDR6
Plarform : HiveOS 0.6-203@210503
Miner : T-Rex v.0.20.3
OC Settigs;
Core : 930Mhz
Mem : -1004Mhz
PL : 80w (Real consumption : 70-75w)
Result : 31.90 MH/s
P.S. : 96 Hours tested. STABLE.!!
You actually donât need the power limit anymore since youâve already found the sweet spot on ansolute core clock.
The fact that the power draw shows a lower wattage than the PL means you can set the PL setting to 0 and performance will stay the sane.
Yes youâre right power limit removed.
Hey mate try these. Tuning the Hynix was fairly common knowledge but still working on the samsung one
T-Rex Miner
Stable 9hrs+
Accepted 98.3%
Hey,
Iâve been running on this config +24 hours without crashes and only 99.16% accepted, no errors:
Any value above 2000 for mem oc is not increasing the MH/s
FYI: Running on T-Rex Miner
On Asus and PNY single-fan 1660S cards with Hynix memory running at a -1004 clock on GMiner, Iâm getting 31.75 MH/sâŚSeems Samsung memory can achieve 32+ MH/s with the +2000 memory clock.
any idea why my 1660S micron memory usage dint hit 100%???
=== GPU 2, 03:00.0 NVIDIA GeForce GTX 1660 SUPER 5944 MB ===
Bios 90.16.4F.00.E5, PCIE Link Gen 2, PCIE Link Width 1x
S/N -, UUID GPU-89538c96-8f07-db81-ecf3-875541439526
Power Limit: Min 70.0 W, Default 125.0 W, Max 125.0 W, Current 85.0 W
Frequency: GFXCore 930 MHz, MEMCore 6299 MHz
Memory: Total 5944 MB, Used 5476 MB, Free 468 MB, Hynix GDDR6
Utilization: GPU 100 %, MEM 100 %
PSTATE P2, PWR 79.8 W, Temp 54 °C, Fan 65 %, RPM 2145 (65 %)
=== GPU 3, 07:00.0 NVIDIA GeForce GTX 1660 SUPER 5944 MB ===
Bios 90.16.42.00.9E, PCIE Link Gen 2, PCIE Link Width 1x
S/N -, UUID GPU-e81764f0-34bb-695c-c51b-b4b876ae5f06
Power Limit: Min 70.0 W, Default 125.0 W, Max 135.0 W, Current 85.0 W
Frequency: GFXCore 1035 MHz, MEMCore 7750 MHz
Memory: Total 5944 MB, Used 5476 MB, Free 468 MB, Micron GDDR6
Utilization: GPU 100 %, MEM 86 %
PSTATE P2, PWR 79.7 W, Temp 53 °C, Fan 55 %, RPM 1930 (55 %), RPM 0 (55 %)
It works!!! many thanks man