Nvidia Ampere and AMD RDNA2, otherwise known as Big Navi, GPUs are getting the greater part of the features today, however, shouldn’t something be said about past-age equipment? The standards we’ve laid out above by and large apply to the more established GPUs also.
Note that outside of the RX 580 and 590, and the three GTX 1660 variations, every one of our tests was finished utilizing the reference models from AMD and Nvidia, which will generally be to a greater extent a gauge or the direst outcome imaginable for most GPUs. We’ll list our advanced settings underneath, yet here are the gtx 1660 ethereum overclock outcomes.
Ethereum mining execution on inheritance GPUs
For Nvidia’s Turing GPUs, execution again relates essentially straightforwardly with memory transfer speed, however with a couple of intriguing exemptions. Notice that the 2060 Super, 2070 Super, and 2080 Super all end up with almost indistinguishable execution? That is not a mistake. The odd cycle is that the 2080 Super requires significantly higher memory timekeepers to arrive.
The 2060 Super and 2070 Super both needed to run with a +1100 MHz offset in MSI Afterburner, giving a compelling rate of 15.8Gbps. The 2080 Super in the interim has a base GDDR6 speed of 15.5Gbps and we had the option to maximize the memory overclock at +1500 MHz, yielding a last speed of 18Gbps for the figure. But we wound up at a similar 44MH/s hash rate. Probably the memory timings on the GDDR6 in the 2080 Super are looser (looser), so even though transfer speed is higher, transmission capacity in addition to inertness winds up offsetting. 2080 Ti in the interim hits a similar ~60MH/s as the 3060 Ti and 3070, on account of its 352-cycle memory interface.
The GTX 16-series GPUs in the meantime offer a respectable mix of execution and power. The 1660 Ti and 1660 Super are fundamentally a similar speed, however, we would be wise to karma with memory overclocking on the Super. The vanilla 1660 with GDDR5 memory doesn’t clock very as high on the VRAM and consequently performs more terrible on Ethereum mining. Despite which GPU you’re checking out, however, all of the GTX 1660 models benefit enormously from dropping the GPU timekeepers. That lessens power use and temperatures and lifts generally speaking proficiency.
Venturing back one age further to Pascal (GTX 10-series), the methodology changes a little. Most extreme memory tickers are as yet basic, however, center timekeepers begin to issue more—the engineering isn’t tuned for the process as much as Turing and Ampere. We got our best outcomes by overclocking the GPU center and memory speed, however at that point drawing an electrical cable. But, being almost four years of age, two of our GPUs (the GTX 1070 and GTX 1080) truly weren’t content with memory overclocking. Anything over a 200MHz knock on 1080 caused a hard PC crash, and keeping in mind that 1070 oversaw +500MHz, our hashing results were still undeniably lower than anticipated.
AMD Radeon VII
Tweaking AMD’s past-age GPUs is somewhat unique, in that rather than restricting the power, the best methodology is to tune the voltage and clock speed. Of course, the more established the GPUs get, the lower the hash rates and effectiveness become. How about we start with the past age and move back from that point.