Crypto Mining Blog

Recently we have tested the new Nvidia Pascal-based GeForce GTX 1080 Founders Edition GPU for crypto mining and it is now the time to see what the little brother in the form of GTX 1070 is also capable of. The Nvidia GeForce GTX 1070 Founders Edition is a little stripped down version of the current top model in the form of GTX 1080, but generally the two are not that different. The GPU on the GTX 1070 is a bit less powerful and the memory is regular GDDR5 instead of the GDDR5X used in the 1080. The cooling solution, the backplate, even the 8-pin power connector are the same on both and that should signal a good overclocking potential for the GTX 1070 and the significantly better price makes it the more interesting of the two for crypto mining. Do note that the TDP rating of the GTX 1070 is just 151W and the power limiter allows you for just 12% increase over the stock limit, so you get 169 Watts max without having to resort to something like BIOS modification for example in order to be able to get serious in terms of overclocking. We are going to cover the overclock for the GeForce GTX 1070 and the mining hashrate with OC in a follow up article, for now we are comparing only the performance at stock settings.

We have already discussed the not so good situation with Ethereum mining on the GTX 1080 and unfortunately things are the same with the GTX 1070. Mining Ethereum under Windows is pointless at this point as the hashrate is very low (issue with the video driver) and although this might be fixed at a later time with a driver update, for the moment the only option you have is Linux. The results under Linux are not great though the GTX 1070 apparently does perform better because of the GDDR5 memory even at lower frequencies as compared to the GDDR5X, especially when you overclock the GTX 1070. We are going to be doing some more testing for this under Linux at a later time in order to see what the new Nvidia Pascal-based GPUs are really capable of for mining crypto currencies based on the Dagger-Hashimoto algorithm like Ethereum, but for now we are focusing on the other algorithms.

The results we’ve got for other popular algorithms besides Ethereum’s Ethash are pretty interesting as you can see from our summary in the table. It is interesting to note that the GTX 1070 is a bit faster in all algorithms besides Decred than the GTX 980 Ti and it does it with a significantly lower power usage. Price wise the GTX 1070 is still a bit more expensive than GTX 980 Ti, but the power savings on the long term should more than justify the difference. With that said we should note that the GTX 980 Ti has never been one of the popular Nvidia video cards among miners due to the pretty high price that it has been keeping. Compare to GTX 970 the new GTX 1070 is significantly faster, and the results of the 1070 were not that close to the ones from GTX 1080 as we’ve hoped they could be. With some overclocking however we expect that the gap in performance between a stock GTX 1080 and the OCed GTX 1070 cloud become pretty insignificant, but that would need some more testing to confirm.

Do note that we have not included results from Lyra2RE and NeoScrypt this time, because with both algorithms the GTX 1080 and GTX 1070 did not perform well and would need some optimizations. In fact there is already an optimized version of ccMiner that fixes the performance issues with Lyra2RE that also brings quite big performance boost for older GPUs as well – ccMiner 1.7.6-r6 fork With Faster Lyra2RE, but for Neoscrypt the performance of the Pascal GPUs needs some work.

We continue our series of tests of the new Nvidia GeForce GTX 1080 Founders Edition (reference design) for crypto currency mining after yesterday we have checked the situation for mining Ethereum with the new Pascal GPUs and have seen the not so great results. Ethereum and the Dagger-Hashimoto algorithm it uses is still doing better on AMD GPUs, but Ehereum mining is set to end at some point with the altcoin switching to PoS only and the forks that use the same algorithm are not yet providing a real alternative. So with the growing difficulty and the switch to PoS at some point in probably less than a year you should also be interested in how the GTX 1080 performs in other new GPU mineable algorithms as well as in the more popular ones that. This is the reason we are now going to compare the GeForce GTX 1080 Founders Edition with GTX 980 Ti and GTX 970 to see the difference in terms of performance in other algorithms as well. As you will probably see the GTX 1080 does provide a nice performance boost compared to the two other older generation alternatives, but with its currently high price even the power saving does not make it great option for building multi-GPU mining rigs for the moment. The more interestingly priced GTX 1070 will most likely be the better choice for building 6 GPU mining rigs succeeding the GTX 970 as the best Nvidia-based GPU for crypto mining, though we are yet to test the 1070 in order to confirm that.

The ccminer forks we have used for testing:
– Blakecoin (Blake256-8rounds), Decred (Blake256-14rounds), Vcash (Blake256-8rounds) ccminer
– Lyra2REv2 ccminer
– X11evo ccminer
– All other algorithms

Tests were done under Windows 7 with already existing ccMiner releases that are not compiled with CUDA 8.0 and with support for Compute 6.1 that the GTX 1080 uses, though without optimizations specific for GTX 1080 or the Pascal architecture in general we are probably not going to see much of a difference. Anyway, we do have planned to do additional testing comparing the difference in terms of performance with existing ccMiner forks compiled with CUDA 8.0 and Compute 6.1 against the results here achieved with CUDA 6.5/7.5 and Compute 5.2 releases that were tested here.

We had almost problem-free experience testing the above algorithms on the GTX 1080, though there are some things that we need to note. For example the Nist5 default intensity (21?) was crashing the miner, so tested with 20 where it was working fine. The very low Neoscrypt performance on the GTX 1080 is because the algorithm probably needs special optimizations to take advantage of the GPU. The Lyra2RE performance was not that much faster on the GTX 1080 as compared to GTX 980 Ti, so here some optimizations will most likely result in increased performance. Note that we have also tested X11, though now that is has moved to the ASIC phase you probably won’t want to GPU mine it anyway.

What is clearly seen from the comparison above is that the new GTX 1080 performs about twice as fast (on average) than a single GTX 970 and we are using Gigabyte Windoforce 970 OC GPU comparing to stock 980 Ti and GTX 1080. The GTX 1080 uses about the same power as a single GTX 970, but offers twice the performance, but the real issue here is the price at the moment. If the GTX 1080 was about twice the price of GTX 970 it might’ve been an interesting option for miners that are currently using Nvidia GTX 970 mining rigs, but it is unfortunately more like 2.5 and not two times. Comparing the GTX 1080 to the GTX 980 Ti shows an average performance advantage of almost 50% in favor of the GTX 1080, with not that great price difference between the two cards (the GTX 1080 is about 30% more expensive). So definitely no point in going for GTX 980 Ti instead of GTX 1080 for mining, though still both cards are not the best choice for multi-GPU mining rigs, then again if you are buying just a single GPU for gaming and want to also mine with it when not gaming the GTX 1080 might be much more interesting that for use on multi-GPU mining rigs.