- Joined
- Dec 14, 2010
Well, I've had a few RTX 3060 Ti's now. I felt like putting some of my thoughts down, hopefully this post is helpful for someone.
All these stats for Ethash (ETH).
FHR RTX 3060 Ti (-500 Core, +900 Mem Clock, 60% Power Target, 55% Fan Speed) Samsung Memory ~58.5MH/s (crashes at any Mem Clock above +1100 or so, with no real performance to be gained from doing so) This card tends to run hotter, and basically always draws more power than the other two FHR Samsung cards, while also having a lower hash rate (even if in a system by itself). 52-57°C typically with 124-127W draw displayed
FHR RTX 3060 Ti (-500 Core, +900 Mem Clock, 60% Power Target, 55% Fan Speed) Samsung Memory ~59MH/s (can go to +1400 Mem Clock, but doesn't gain much performance and the efficiency drops) 48-55°C typically with 119-124W draw displayed
FHR Asus RTX 3060 Ti (-500 Core, +900 Mem Clock, 60% Power Target, 55% Fan Speed) Samsung Memory ~59MH/s (can go to +1400 Mem Clock, but doesn't gain much performance and the efficiency drops) 48-55°C typically with 119-124W draw displayed
FHR Asus RTX 3060 Ti (-500 Core, +900 Mem Clock, 60% Power Target, 55% Fan Speed) Samsung Memory ~59MH/s (can go to +1400 Mem Clock, but doesn't gain much performance and the efficiency drops) 48-55°C typically with 119-124W draw displayed
The two cards with Samsung Memory were actually like clones of each other, although they were actually from different brands. At the same settings, they performed basically identically all the time.
LHR RTX 3060 Ti (-500 Core, +900 Mem Clock, 70% Power Target, 55% Fan Speed) Hynix Memory ~58-58.25MH/s (or 54.8-55MH/s at 65% Power Target) 512.77 DCH Driver used
Interestingly, for FHR cards with Hynix memory it was required to lower the clock rate to increase performance (thereby forcing the memory to run at tighter timings, which increased performance). With this LHR card with Hynix memory though, it behaves exactly the opposite and lowering the memory clock continuously lowers the hash rate on ETH and it was only once I started raising the memory clock that the performance started going up. So, evidently with cards from the LHR series/lines Hynix made some change which modifies how their memory is effected by adding or taking away from Memory Clockspeed.
I had actually ordered what was claimed to be an FHR card, but what I received was an LHR card. Figured I may as well see what it can do since I have it and I'd never tested an LHR card for any sort of mining. I'd actually been intentionally avoiding them, as a sort of rebellion to NVIDIA's locking them down to artificially limit what a customer can do with the product. As in my mind once you've bought the hardware you should be able to do whatever you want with it, as long as it doesn't break some law, but that's a bit of a tangent.
I'd heard that these LHR cards required more power to unlock the same performance, and that definitely seems to ring true, I'm at ~140W on this one, whereas the FHR cards I've tested were all around 119-125W per card.
All these stats for Ethash (ETH).
FHR RTX 3060 Ti (-500 Core, +900 Mem Clock, 60% Power Target, 55% Fan Speed) Samsung Memory ~58.5MH/s (crashes at any Mem Clock above +1100 or so, with no real performance to be gained from doing so) This card tends to run hotter, and basically always draws more power than the other two FHR Samsung cards, while also having a lower hash rate (even if in a system by itself). 52-57°C typically with 124-127W draw displayed
FHR RTX 3060 Ti (-500 Core, +900 Mem Clock, 60% Power Target, 55% Fan Speed) Samsung Memory ~59MH/s (can go to +1400 Mem Clock, but doesn't gain much performance and the efficiency drops) 48-55°C typically with 119-124W draw displayed
FHR Asus RTX 3060 Ti (-500 Core, +900 Mem Clock, 60% Power Target, 55% Fan Speed) Samsung Memory ~59MH/s (can go to +1400 Mem Clock, but doesn't gain much performance and the efficiency drops) 48-55°C typically with 119-124W draw displayed
FHR Asus RTX 3060 Ti (-500 Core, +900 Mem Clock, 60% Power Target, 55% Fan Speed) Samsung Memory ~59MH/s (can go to +1400 Mem Clock, but doesn't gain much performance and the efficiency drops) 48-55°C typically with 119-124W draw displayed
The two cards with Samsung Memory were actually like clones of each other, although they were actually from different brands. At the same settings, they performed basically identically all the time.
LHR RTX 3060 Ti (-500 Core, +900 Mem Clock, 70% Power Target, 55% Fan Speed) Hynix Memory ~58-58.25MH/s (or 54.8-55MH/s at 65% Power Target) 512.77 DCH Driver used
Interestingly, for FHR cards with Hynix memory it was required to lower the clock rate to increase performance (thereby forcing the memory to run at tighter timings, which increased performance). With this LHR card with Hynix memory though, it behaves exactly the opposite and lowering the memory clock continuously lowers the hash rate on ETH and it was only once I started raising the memory clock that the performance started going up. So, evidently with cards from the LHR series/lines Hynix made some change which modifies how their memory is effected by adding or taking away from Memory Clockspeed.
I had actually ordered what was claimed to be an FHR card, but what I received was an LHR card. Figured I may as well see what it can do since I have it and I'd never tested an LHR card for any sort of mining. I'd actually been intentionally avoiding them, as a sort of rebellion to NVIDIA's locking them down to artificially limit what a customer can do with the product. As in my mind once you've bought the hardware you should be able to do whatever you want with it, as long as it doesn't break some law, but that's a bit of a tangent.
I'd heard that these LHR cards required more power to unlock the same performance, and that definitely seems to ring true, I'm at ~140W on this one, whereas the FHR cards I've tested were all around 119-125W per card.
Last edited: