|
0.19.7 works far better for eth mining with my 3060ti's than phoenixMiner.
Just switched, happy with that !
Thanks
|
|
|
|
rumours 3070ti to have 320 bit gddr6x 8gb (like 3080) and 3080 super to have 384 bit gddr6x (like 3090). Any thoughts?   this 3070ti will be better than 3080 if that the case, less power consumed less hot..and we know there is an issue about the temps of those gddr6x. 3080 super will be like 3080 super "hot" LOL ...hmm this will do above 100MH if that's the case. and since it will be cheaper than 3090 and will be x3 of 1080ti at least in mining algos..that's irresistible for an upgrade. If 3070ti has ddr6x, the problem of mem temps too hot remains here too no ? |
|
|
|
|
Reducing power limit also reduce voltage, so is it useful to manually undervolt with curve points ?
I do it on my 3060ti cause I OC the core too for gaming.
But for eth mining, wich only memory freq matters, is there a need for manually undervolt ?
My MSI 3060ti TrioX : PL 55%, fan 40%, core -502, mem +1400 : 63 mhs@130W
Temp : 45
But I don't play with curve for the mining profile in AB, only for the gaming profile.
|
|
|
|
|
No we can say that when GDDR6X (3080 or 3090), you'll have throttling even with a fine core temp due to memory temps who are a concern ?
No problem so far with GDDR6 of 3700 / 3060TI
Ps : Maybe the Suprim X version of the 3080 would not have memory temp problems, cause of superior cooling ?
|
|
|
|
The 3080ti will have a 320-bit memory bus like the 3080.. So expect similar hash rates to 3080.
yeah, it's funny because is TI, 3060ti implies same bit memory as 3070, 3080ti should have the same bit memory as 3090 and its not, so what we see is that 3080ti is a lie. If the 3070ti has the same bit memory than the 3080 and is $100 cheaper then 3070ti could be a winner on eth. Except the fact GDDR6X on 3070TI and 3080TI runs too hot on memory intensive algos like eth, especially if OC |
|
|
|
for this gpu to be worthwhile, it had to be 170 mhs or more, nvidia could have done it, decided not to. The 3090 line are the top hashers but not having the best cost to hash ratio of the 3000rtx series cards. That belongs to the 3060ti hashing at 66mh/s at 125watts which is even better than the 5700xt cards were doing. Those were the king of hashrate for a while until these ti cards came along. Maybe the 3080ti will do 170mh/s mark you are saying to be worth it to mine with. As they were announced to be between the 3090 and the 3080 in performance but not costing as much for the 3090 at a price tag of $1499 MSRP. https://www.techradar.com/reviews/nvidia-geforce-rtx-309066mhs@125W ? oO I can do nothing more than 63mhs@130W (PL 55%) with my 3060ti TrioX |
|
|
|
|
What about memory temps on 3090 ?
I heard GDDR6X warms a lot.
Maybe mining an other algo less mem intensive would be better for 'memory life' ?
|
|
|
|
Yes, if overclocked too much and not in a well ventilated case.
Even some of the cards that have water blocks fitted are still having this overheat issue on the memory chips.
Wowww.. It's damn serious ! Thanks for the info I assume the problem will be worse for next 3080Ti / 3070TI |
|
|
|
The RTX 3060 Ti and the RTX 3070 use the older GDDR6 memory, so they don't get as hot and is not an issue 99.98% of the time.
The higher spec'd cards use the GDDR6X memory (Double Rate) and need more power to drive them, this is why these RTX 3080 / 90 cards get hot.
Yeah, I got that. Even a highly cooled custom card like MSI Supprim 3080 would get too hot on mem temps ? |
|
|
|
The cards also have their own curves which can not be controlled with AB. NVIDIAs still have temp sensor on memory - just not showing public - so when mem temps get too high, fans are maxed out.
Very instructive, thanks ! So I assume if my fans never get maxxed out, my memory temps are 'considered' fine ? |
|
|
|
You know mem temp is too high when fans start spinning like hell even when the temp of the core is low.. then hash drops too
3060 ti-s are great
Yeah it is for sure. Fan spinning is regulated from core temp via afterburner, so how can they spin very high if core is low / normal ? |
|
|
|
|
I have a fantastic 3060ti MSI TrioX wich gives 63.5mhs@PL 55%, 130W, core -502, mem + 1400 : fan 45%, core temp 48
I heard that GDDR6X from 3080 runs too hot when pushing memory OC high like I did for my little 3060ti
Is it the case ?
How can we know the mem temp ?
|
|
|
|
Unfortunately still incorrect shares with my radeon RX6800. To bad, switch back to lolminer, they runs quite good ....
Nothing related, but.. what hashrate can you get with your (wonderful) 6800 ? And congrats for having grabbed it  |
|
|
|
PhoenixMiner 5.4c (release candidate) is ready. You can download PhoenixMiner 5.4c from here (both Windows and Linux versions): - Validated support for Nvidia RTX 3090, 3080, 3070, and 3060Ti GPUs. Those still use the current kernels and there are no speed increases from the previous PhoenixMiner release
Thanks for this ! Anyway, wich % hashrate boost can we expect with future versions with next kernels for rtx 3xxx ?  |
|
|
|
Same technique here. I start miner for DAG with +1000 And then +1500 with PL55% on my 2 3060ti's Result here : 130W, 62.7 mhs Can't get any higher. I'm satisfied enough though with this You have to see if that translates to that hashrate at the pool. Yes, hashrate is the same on pool (2miners) |
|
|
|
It's time to build with a newer version of CUDA. NBMiner and T-rex use this. They have a higher speed with less power consumption. Especially on RTX3000 cards
isn't CUDA 11 the latest? I was getting higher HAsh with Trex but I would get more failed shares. with Phoneix I can raise the mem all the way to +1500 on rtx 3060 ti but only after the Dag is done. I have to start the phonex on mem of max +1000 or I end up with failed shares following that I can run at 63.4MH on 130W Same technique here. I start miner for DAG with +1000 And then +1500 with PL55% on my 2 3060ti's Result here : 130W, 62.7 mhs Can't get any higher. I'm satisfied enough though with this |
|
|
|
It's time to build with a newer version of CUDA. NBMiner and T-rex use this. They have a higher speed with less power consumption. Especially on RTX3000 cards
Are you sure of this ? I get exactly the same hashrate with t-rex latest or phoenix with my 3060ti |
|
|
|
Dear Mr.Phoenix , I have problems mining with 3060ti cards. Whenever I start the miner with an applied OC profile - it's getting only incorrect shares. I have to start it with default settings for the cards and apply OC later. It's especially frustrating when the DAG epoch changes, because that leads to new DAG generation with OC already applied and incorrect shares. Lost hours of work last night. Please address the issue. Thank you!
I don't have this kind of problem with my MSI trioX 3060TI I start mining with a +1200mhz mem applied, and all goes well, no incorrect shares I receive today my second 3060ti (Aorus ), will try with and let you informed. Hard to grab those cards anyway |
|
|
|
Using the 5.4b on my RTX 3060 ti
It is very stable with full over clocking at 1500memory and -500 core , power of 66% 130W hashing at 63Mhs
using older driver
Problem is if I start the Phonexminer with full over clock settings it will keep getting me Bad Shares
I have to set the MSI afterburner to memory of around +1000 all other settings the same, after few minutes I can raise it to 1500 memory and see my Hash increase.
Is there a fix for this?
Hoping for a stable restart and everything working fine instead of me intervening with each restart , windows 10 pro x64
I do 59.5 mh/s but at 50% PL, 119W + 1000Mhz, -502 core I think it's more efficient this way, give it a try and tell me your results  Strange thing, if I go PL 65% like you, I have more than 130W consumed.. |
|
|
|
This post contains the instructions on how to fix the latest 460.79 Nvidia Windows driver. The reason for the problem is that one of the DLLs (nvml.dll) that the driver installs is corrupted. Here is what you have to do in order to fix the problem: File: nvml_460.79_win10.zip
---------------------------
SHA-1: 749ae64e4a5044994bec61d6d94459aa4c4cc719
SHA-256: 541ccdbd77c693d1de2919adab9eaa7e2b502c9ec2f722544002abbcf643539e
SHA-512: 80562b59e0a4f3c47557abc8742d9af6b119102f6fecfb3aa38112084ba0edd06451a631f633b667d57ec0417eb0db551d4cc5df90e0b835bce212b3ee6b75ce
- Extract the file nvml.dll from the .zip archive you have downloaded in the previous step, and then copy it to the folder C:\Windows\System32. There will be a file with the same name here, so you have to overwrite it (you will be prompted by Windows are you sure, blah, blah, blah)
Note that this .dll is the original .dll from Nvidia (it is still digitally signed by them). If you are really paranoid, you can download 460.79 drivers from Nvidia, then use 7zip to open the .exe and extract the dll (you will find it at 460.79-desktop-win10-64bit-international-dch-whql.exe\Display.Driver\nvml.dll), and then copy it to the C:\Windows\System32 folder. I can't extract the dll from archive to System32, win10 says file is already used by another process. I killed nvidia controlPanel and MSI afterburner, but same result Any ideas ? Rename the existing nvml.dll in System32 to something else (e.g. nvml_old.dll) then copy the one from the archive in System32. After that restart Windows, and you should be able to delete the nvml_old.dll from System32. Works like a charm ! Now I'm able to mine with latest driver with my 3060TI's rig Thanks chief |
|
|
|
|
Show Posts
