Nice to be part of this community. I would like to thank everyone who shares their experience and screenshots! (Love them - pictures/screen shots are very useful IMO)
Please see my OC screen shots for 1x 1660 Super and a RX480 8GB. Initially tweaked and thought what can I get for the lowest power setting:
Thanks to all the member who share their OC setting as it helps us newbies with some baseline numbers and a starting point to find our own sweet spot. I could probably squeeze another 1MH/s but do I want an extra increase of 10-20W usage… not really… My goal was to try and keep each card under 70W with an average of 30MH/s and now looking for more 1660 card and 480 8GB cards LOL.
Thanks to all the HIVEOS team for making this happen…
(My previous experience is mostly buying low and selling high on exchanges)
Thanks - I will try that and see what I get and run a full day tomorrow with 800 first and slowly bump it up a day at a time as I’ve learnt not to just go with reported and I think I saw more stale shares at 850 @ 29.53 MH/s when I ran it for 2 hours last weekend but that could have been due to using wireless connection which I’ve changed to wired now.
I would like to see some Reported vs Mean Vs Realtime graphs from you and the rest of the great people of HIVEOS if you all don’t mind sharing. As it helps to get a good idea what to expect.
On a side note I’ve just managed to find a Asus DUAL-GTX1660S-O6G-EVO for £300 which seems to have very similar specs as the MSI 1660 Super Gaming X.
definitely a learning curve and a shock to find out that not it’s not just the Type of memory that’s different across the GTX 1660 super but between brands and sub models there is subtle difference in Bandwidth, Boost clocks, Theoretical Performance (FP16,32,64) and found these card to be very similar in spec For my own reference and for any one else:
MSI GTX 1660 SUPER GAMING X (Samsung Memory)
ASUS DUAL GTX 1660 SUPER EVO OC (DUAL-GTX1660S-O6G-EVO) (Hynix Memory)
Colorful iGame GTX 1660 SUPER Ultra V (I’d be interested to know more if anyone owns one of these? - looks like a beast with 3 fans which is a good indicator could run cooler)
As there is so many variations so I don’t expect to use the same numbers from the pre-set OC profiles. I tried most from page 1-5 and the best I got was around 28-29MH/s, so I used the based numbers from the card specifications and I can’t seem to find any posts that is using 1750 for memory clock.
Happy mining folks - I’m looking forward to my new card.
he is what happens with my rig for 1 day. well i have only one GPU. anyways.
i get around 2% stale shares , i am using wireless connection. So it varies but around 2 or less %. Stales are due to connectivity.
You can see however that the average hashrate ovr 24H is lower than reporded. Its due to two things. i circled in red the first one, its when i didnt have electricity for some reason for some time and second thing is the luck factor. Check the realtime hashrate, its lower than the reporded that means i dont have luck now and if this day is like this all day, the average will be lower that the reporded. i had days it was above the reported hashrate so in time they should even out.
About the 1660S, its a very good GPU. I personally like a bit more the 1660TI but they do more or less the same. It doesnt matter much the maker of the GPU( well atleast get one thats not from some unknown maker ). To my knowledge Micron memory is best but all others will OC good too.
RX 480 - 800 voltage gets quite low yeah. i thought with 1100 it might run with 800. now that you use 850 , have you tried upping a bit the core clock, maybe a test to 1150 with 850 voltage? also you can test mem to 2050. Its a matter of finding the sweet spot. mine is 1131, 825 , 800 , 2050, with amdmemtweak --REF 20. if i up a bit the core and voltage i get more MHs but too much watts up also. And my card is a fuking power demanding beast , is Asrock 570 8GB samsung like your memory but it pulls 95 watts with these settings and it makes 30.80 only.
Give my settings a go and see what you get across 6 hours or so… I tried 850 for a few hours this morning and I noticed average dropped and the peaks over reported were smaller. Here’s a comparision.
It looks to me that it by having a more optimal setup the gpu processes and returns the data sooner hence better real times rather than later and my average dropped using 850. I’m guessing it might be that at times when the pool is harder real time drops and then continues to adjust for a more suitable difficulty hence lower real-time. With the setting I had yesterday I was probably being allocated a lower difficulty hence better real time. I’ve also read some where that actual share is not important it’s the difficulty of that share and the share from harder pools are of higher value… but I might be wrong need to do more research.
Motherboard might also play a part in data handling as I’m using a BTC S37 and why one setting may work for you but not for me.
I enjoy learning about all this gobbledigook
Using my usual setting now for the rest of the day then when the Asus turns up I’ll be cranking that in and up.
dudes the important is that the reported is as much as possible, meaning your GPUs proccessing speed. average with gain on it and be equal or a little higher or lower , depends on the luck factor. so dont look at real time hashrate , look at your reported and average. about the 850 core voltage, it doesnt influence the hashrate. the core voltage keeps the gpu chip running. the idea is to get as lower core voltage as you can while keeping as higher core clock as you can. low core voltage = lower temp and watts consumed. higher core voltage means you can run the gpu with higher clocks but that increases also temp and watts used. so thats why while experimenting you get the sweet spot where you get the best temp and consumption for your MHs. you can easily increase core voltage and clock but at some point it just provides more heat and power usage than MHs.
My reported with 850 and 900 was @ 61.63 +/- 0.01 but saw only maximum real-time peak of 70 with 850 and now seeing peak of 85+ with 900, Using -1004 for memory drops my reported to about 28/29.
I’m looking at the numbers from a Metrologist (Professional background) point of view because the reality is there is no such thing as 100% certainty. Main thing I’m looking at are the +/- of real-time because this is from the work that is done.
What I’m trying to say is on my current settings I’m fine with the extra 7w of usage because I’m seeing more higher real time peaks which in turn improves the average. I’ve seen a few threads where the real-time was much lower than the reported and after reading HIVEOS FAQ I’m understanding it as the closer the average (Based on Real Time) is to your reported (Estimated) the better.
I built my rig last Saturday and tweaking and stopping and starting the miner I’m not giving it a chance to run a full cycle. I’d like to think If my average ends up higher than than reported it’s a good setup to I should just leave the ball keep rolling.
hey thank for your input. quick question can you help. my 480 keeps crashing it nbminer and none of my o.c setting seen to be working. any advise would be helpful
You have Samsung memory, and zajcew has Hynix memory.
OC settings are totaly different according to the memory you have.
For Hynix : Mem -1004 and Core 930 or 1000
For Samsung : Mem between 1700 and 2500, you need to try what works the best for you, and Core 1000 or 1050.
Here is my setting with Samsung memory
More 1660 Supers added to my rig I did get an opportunity to get my hands on a RTX 3090 but I couldn’t justify ROI ( >£2K for 120MHs) 4 x 1660 Supers Cost me £1500 for >120MHs. TBH very impressed with the BTC 37 S37 Motherboard which has 8 GPU slots.
Circa 220Mhs @ > 500W
RX480 was running hot because the Thermal paste and Pads needed a service as it was pre-owned - (can’t complain for £250) Now running sweet between 60°C>70°C.
3 x Inno3d 1660 Super added and they preferred memory clock @ -1005 anything more or less I’d drop hash rates.
1 x MSI 1660 Super Gaming X added - This one was weird - Comes with Hynix Memory my other one that looks exactly the same aesthetically has Samsung Memory) , The other weird thing was if I used similar setting to MSI Super Gaming X (Samsung) I get the same Hash Rate but an Increase of 10W consumption and I could over clock the Memory to 2500 and get Hashrates of 35Mhs @<70°C/77W<85W. But when I use similar setting to the Nvidia (Inno3D) Supers Wattage drops to 68W<73W
I sacrificed maximum hash rate for the lowest achievable consumption <72W for 30MHs<32MHs per card.
Max possible across rig was 235MH/s @ 525W but wasn’t keen on running @ 70°C<80° (With Invalid/Stale Shares seen + Reboots - possibly due to the heat)
Stable and Happy at 220MHs @480W<490W / 49°C<65°C (Stale shares <1%/ No Invalids so far)
Need to let it run for at least a week now to get a better picture and before any more OC adjustment As I can squeeze a little more from the MSI 1660S (Samsung) with a slight increase in stale shares under still <2% and under 70W definitely my favourite card at the moment.
for the RX 480 , have you tried different settings ? like these:
1150 core clock, no DPM , VDD 850, memory controler voltage 800 , no memory state , 2050 mem clock or higher. aggressive undervolting on.
800 put the card to sleep - But 850 hit the sweetest spot yet.
). To my knowledge Micron memory is best but all others will OC good too.
