ETH Mining with Nvidia 1070 GTX MSI Gaming X 8G and Overclocking Performance Results

It's time to put aside the old 980 GTX and upgrade to the 10 series of Nvidia's newest Pascal series GPUs, let's see what kind of hash rates we could get with these cards while mining Ether.

MSI Gaming X 1070 GTX

First the facts, the old 980 GTX overclocked and hitting 20-21 MH/s while mining Ether would use about 160 Watts. The new Nvidia 1070 GTX does about 25-27 MH/s stock and only uses 140 Watts. Efficiency and speed are significantly improved and so is compatibility.

Here's a scatter graph I put together using 3 sets of values from each of the 980 GTX and 1070 GTX, so that we can see visually how they spread in terms of performance metrics and results.
Ethmining 980 GTX Vs 1070 GTX Power and Effeciency
Ethmining 980 GTX Vs 1070 GTX Power and Effeciency
While I had to use old rolled back 347.52 Drivers on the 980 to make it mine at a reasonable speed, the new 1070 worked with the latest Nvidia drivers right away and worked well too, giving solid 25-27 MH/s without any tweaking at all.


The 1070 was running very cool by itself as well, never crossing 65 C while running Genoil Ethminer. Well worth the upgrade from the 980, which is surprising because I always felt the 980 was still an excellent card, it's less than a few years old.


The 1070's core clock can do 1797 out of the box with a memory speed of 8108 MHz,  compared to the 980's 1216 MHz core speed and 7010 MHz memory speed.

Overclocking the 1070 GTX was also quite easy, without much trouble we were hitting 31 MH/s while running the core at 2088 MHz and the memory 4498 MHz, which is almost 9000 MHz in "box label" speeds. To achieve this I set NVinspector (Version # 1.9.7.8 on Windows 10 v. 1703) to P0, P-state, then 115% Power and Temperature target, + 700 MHz Memory Clock Offset and +140MHz Base Clock Offset. The temperature never climbed above 67 C.


Comments

Anonymous said…
how is your power draw from the wall?
Art Gen said…
Really not bad, about 550-600 Watts on average.