Awesome, looks like evga did cooling properly. Some cards have memory overheat while mining and lose hash rate.
Here are all the temp readings per HWiNFO after mining for 1+ hours:
Last edited:
Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
Awesome, looks like evga did cooling properly. Some cards have memory overheat while mining and lose hash rate.
That's great. A lot of cards don't have memory sensors visible to end users, so the card still throttles when memory gets hot but they only see the low core temp and don't realize. I will have to check my founders edition.
That's great. A lot of cards don't have memory sensors visible to end users, so the card still throttles when memory gets hot but they only see the low core temp and don't realize. I will have to check my founders edition.
Good info, I will have to find more on the FE thermal pad upgrade.Here's a list of card that drop hashrate due to memory getting hot: (from https://bitcointalk.org/index.php?topic=5287329.0)
Good ones that don't throttle:
EVGA FTW3
MSI Gaming X, Suprim X
Asus Strix
Palit GameRock
Some throttle, some do not:
Asus TUF https://www.reddit.com/r/EtherMining/comments/jxfoac/3080_thermal_throttling_thread/
Zotac Trinity
Palit Gaming Pro
EVGA XC3 (https://www.reddit.com/r/EtherMinin...?utm_source=reddit&utm_medium=web2x&context=3)
Those that throttle:
Gigabyte Gaming OC, Eagle OC, Vision OC (replacing thermal pads fixes issue)
Nvidia Founder's Edition (replacing thermal pads fixes issue)
MSI Ventus (need to keep under 60c - https://www.reddit.com/r/EtherMinin...?utm_source=reddit&utm_medium=web2x&context=3
you should be able to see in hwinfo64 if it is thermal throttling, applies to memory as well.Very interesting info! I’ll rig up a fan blowing at the back of the 3080 x3 card I have and see if that allows me to dip power lower without affecting hash rate.
great info. I think I am going to thermal tape some m.2 heatsinks onto the backplate of my founders 3090, as ghetto as that sounds, since I do believe it had thermal limitations on the vram (though I hadn't checked, the fan ramped way up when running phoenix)So the fan pointed at the back of my X3 does nothing to prevent power target lowering from throttling Phoenix hashrate on my card.
I grabbed a 120MM A/C high CFM fan and plugged it in pointed at the back of the X3 - plenty of airflow over the card now. Still drops hashrate at any power target below 72%. I’m not VRAM power throttling regardless of fan or no fan at these settings according to HWInfo.
Just one alternate datapoint, but here it is in pictures. At this point I can try to tick the memory overclock higher over time and see where it starts becoming error prone.
Stable all night - no incorrect shares. Power Target 72%, Mem +1402, Core doesn't matter - doesn't affect power use nor heat for this alto.
Anything above 72% power target doesn't increase hashrate - all the way up to 80% power target.
Anything below 72% starts ticking down hashrate. 72% is the magic number, the power target optimized floor, for this particular card.
Power target 65% - same everything else = 86MH/s
Power Target 60% - same everything else = 74Mh/s
So the fan pointed at the back of my X3 does nothing to prevent power target lowering from throttling Phoenix hashrate on my card.
I grabbed a 120MM A/C high CFM fan and plugged it in pointed at the back of the X3 - plenty of airflow over the card now. Still drops hashrate at any power target below 72%. I’m not VRAM power throttling regardless of fan or no fan at these settings according to HWInfo.
Just one alternate datapoint, but here it is in pictures. At this point I can try to tick the memory overclock higher over time and see where it starts becoming error prone.
Stable all night - no incorrect shares. Power Target 72%, Mem +1402, Core doesn't matter - doesn't affect power use nor heat for this algorithm.
Anything above 72% power target doesn't increase hashrate - all the way up to 80% power target.
Anything below 72% starts ticking down hashrate. 72% is the magic number, the power target optimized floor, for this particular card/sample.
Power target 65% - same everything else = 86MH/s
Power Target 60% - same everything else = 74Mh/s
Mine wont even do +1000. I was having crashes at +600. Testing right now to decide if I should get new thermal pads. Letting the fan do it's stock curve. Right now its 75% fan at 43c gpu. Ambient about 15c.Have you tried a negative core offset to see if you can "shift" some of the power from the core to the ram by lowering clocks so to speak? I am running -300 core on mine. I have not been so lucky in terms of getting my memory as high as yours! i get invalids with anything past +1000 to the point where i backed it down (in fact i completely black screened just now when trying to go +1200 and had to reboot.)
Lowering the core does nothing positive for me. Lowering my core offset from 0 to -300Mhz lowers my Hashrate by 1MH/s. Adding +50 or +100 to core offset does nothing.Have you tried a negative core offset to see if you can "shift" some of the power from the core to the ram by lowering clocks so to speak? I am running -300 core on mine. I have not been so lucky in terms of getting my memory as high as yours! i get invalids with anything past +1000 to the point where i backed it down (in fact i completely black screened just now when trying to go +1200 and had to reboot.)
that's insane how fast your ram goes. I know I lost the lottery on mine but damn, 1400+ is really ripping.Lowering the core does nothing positive for me. Lowering my core offset from 0 to -300Mhz lowers my Hashrate by 1MH/s. Adding +50 or +100 to core offset does nothing.
I’m currently at +1467 on mem and that is good for 102.6 MH/z using Phoenix. I threw it up to 1500Mhz max briefly just to see what happened and lost 2MH/s - so I’m no good at 1500, but I’m not yet topped out at 1467Mhz. It didn’t crash or submit bad shares at +1500Mhz memory, maybe it was throttling , but with the fan blowing on the back - doesn’t matter - it can’t do it.
So I’m getting close to figuring out my true max. Went back and played with power target again. Same results as previous. 72% is my number. Power use is 245 watts for my 102.6MH/s
Lowering the core does nothing positive for me. Lowering my core offset from 0 to -300Mhz lowers my Hashrate by 1MH/s. Adding +50 or +100 to core offset does nothing.
I’m currently at +1467 on mem and that is good for 102.6 MH/z using Phoenix. I threw it up to 1500Mhz max briefly just to see what happened and lost 2MH/s - so I’m no good at 1500, but I’m not yet topped out at 1467Mhz. It didn’t crash or submit bad shares at +1500Mhz memory, maybe it was throttling , but with the fan blowing on the back - doesn’t matter - it can’t do it.
So I’m getting close to figuring out my true max. Went back and played with power target again. Same results as previous. 72% is my number. Power use is 245 watts for my 102.6MH/s
I just thought of something. I’m using a Asus B250 mining motherboard.Have you tried a negative core offset to see if you can "shift" some of the power from the core to the ram by lowering clocks so to speak? I am running -300 core on mine. I have not been so lucky in terms of getting my memory as high as yours! i get invalids with anything past +1000 to the point where i backed it down (in fact i completely black screened just now when trying to go +1200 and had to reboot.)
Which card do you have. Is it one of the higher power three 8 pin cards. They might be set up to use more power in their bios. Mine is just the cheapest EVGA model 3080. No backplate, and only two eight pin.yeah every card is different. my 3080 FTW3 is pulling 240w at 63% power...
I'm running an AMD apu with onboard monitor out, so same situation here. And phoenix via nicehash.I just thought of something. I’m using a Asus B250 mining motherboard.
https://www.asus.com/us/Motherboards/B250-MINING-EXPERT/
With my dual core Pentium G4560’s onboard video off the mother board to drive the display. The x3 3080 is ONLY responsible for mining. - nothing related to the windows display. If you are using your 3080 for windows display duty too that could account for a pretty drastic difference. Or at least I found it it did with my pascal cards in 2017.
Which card do you have. Is it one of the higher power three 8 pin cards. They might be set up to use more power in their bios. Mine is just the cheapest EVGA model 3080. No backplate, and only two eight pin.
Two things to try from my past experiences over a couple years.I am having an issue where my cards stop mining overnight. I don't think it's a stability thing, but it could be. The nicehash window says I'm still mining, but phoenix miner is closed and the cards are ice cold not really doing anything. Anyone run into this?
thanks for the tips. I have only phoenix enabled, so we are good there. I looked at the profitability chart on the rig manager and saw it stopped mining at exactly 4am. Regardless, I now see you can see the rejected shares by selecting algo, and I've had some consistently. Should be 0, correct? I think it is the 1070 I have in the system causing problems so I set it back to stock and will check in 6 hours. Thanks for the tips!Two things to try from my past experiences over a couple years.
It almost certainly is a stability thing.
1) disable all algorithms under Benchmark tab except dagger hashimoto - that should leave four enabled. It’s the highest earner on average right now, but short peaks of other token contracts can and will switch you to a different algorithm temporarily. If you’ve heavily tuned your OC for Dagger chances are almost certain that your OC settings will be unstable for the other algorithms. IE with my finely tuned settings I did exactly this for my 3080. It can only mine dagger. Phoenix is the highest hashrate of the four currently for my 3080, but that doesn’t mean it’s the highest for everything. My 1080TI for instance is higher on T-Rex dagger hashimoto. If you’ve already done this and you crash overnight - then yes you have something over clocked too high. Raise power target, lower core offset, or lower memory offset until you are stable. I like testing in 50MHz increments over 8 hours until you begin to fine tune.
2) if you want to leave auto switching by profit in place you need to relax your overclock. For instance on my 1080TI. I can get up towards +750Mhz memory for dagger specifically but anything over 585Mhz (maybe 600mhz) isn’t stable for auto switching settings. You can test this by just dropping all overclocks. Put your power target to 70, 75, or 80% (whatever heat and power use you are comfortable with - during the summer I use ~66%, during winter ~75% - everything else to default and mine for 24 hours and chances are it’ll be perfectly stable. Then just start overclocking your memory by 50mhz at a time for about 8 hours at a time until you find the session it crashes during that 8 hour window, then fine tune from there. Since dagger is so much more profitable than anythingthing else right now I’d just option 1 for now. You can track what’s most profitable for your card on the nicehash profitability page. When you see the algorithms start being more of a even spread then it’s time to take the time with option 2.
https://www.nicehash.com/profitability-calculator/nvidia-rtx-3090
here you can see where I stopped mining entirely at 4am.I have some rejected shares. That’s pretty normal even without a overclock. They should be a very small percent.
There are accepted, rejected (could be stale, or network issue), and incorrect. I think only the incorrect truly indicates a big problem, unless the rejected count ends up being significant. It probably varies on the algorithm, but I remember reading a few percent rejected is perfectly normal for most any algorithm.
I think based on my testing last night, a negative offset hurts performance. I was running -500 up until this week when I had some time to tweak it, but let me explain what happens.Have you tried a negative core offset to see if you can "shift" some of the power from the core to the ram by lowering clocks so to speak? I am running -300 core on mine. I have not been so lucky in terms of getting my memory as high as yours! i get invalids with anything past +1000 to the point where i backed it down (in fact i completely black screened just now when trying to go +1200 and had to reboot.)
If you have a preffered pool use awesome miner with Phoenix miner, otherwise nicehash with phoenix miner imo.I haven't mined since back in the early btc days...What's the go to miming gide nowadays?
Figured I might as well ues my 3090
I don't have a reliable way to do that, it's in a multi card rig right now. Sorry. What do 5700s usually do MH/s / watts? This card is 58MH/s / 110 watts = 0.527Can you measure from the wall? I am curious how far apart a tweaked 5700 is compared to those
what soft u used to check and validate those values ?here are some hash rates at different settings on this 3060ti:
0/1200: pass
140w: 61.2MH/s / 0.44
130w: 61.2 / 0.47
120w: 60.4 / 0.50
110w: 55.7 / 0.51
0/1000: pass
140w: 59.6 / 0.43
130w: 59.6 / 0.46
120w: 59.5 / 0.50
110w: 56.0 / 0.51
0/800: pass
110w: 57.4 / 0.52
0/0
100w: 50.3 / 0.5
Mining with phoenix miner, monitoring with gpuz.what soft u used to check and validate those values ?
I’ve had some additional time to tune the ASUS Tuf OC card (on Performance BIOS) and as DPI suggested these Ampere cards differ.
The best settings I’ve found on the ASUS TUF OC card are 65% power target, +1456Mhz memory. This card starts ticking down has speed below 65%. Interesting it uses about 24 watts less for the same hash power than the EVGA X3 3080. 221 watts on Asus Tuf OC, and 245 on EVGA X3. (Because assumedly the power target on the EVGA can’t go below 72% target without lowering hashrate).
temps are cool, but the Asus Tuf OC puts the fans at 75% by default mining and sits at 49*
the EVGA X3 has its fans at ~60% and sits at 61
Pics of settings for both cards attached.