All About BTC, LTC, ETH mining as well as other alternative crypto currencies
Time for another windows binary release based on the latest code of the ccMiner 1.5-git SP-MOD fork of the Nvidia miner optimized for Maxwell GPUs by SP (source). Since the last windows binary we have published a few days ago there have been a lot of fixes and improvements made to the code and we were pleasantly surprised today to see a significant performance improvement. The results we are getting in terms of hashrate on our stock test GTX 980 GPU for X11 have gone up from about 7900 KHS with the previous release to over 8300 KHS and for X13 up from a bit over 6200 KHS to about 6700 KHS. There are also hashrate optimizations for other algorithms – NIST5 is up to about 26500 KHS, so performance should increase not only for X-based coins, so definitely a recommended update for owners of Maxwell-based Nvidia GPUs such as GTX 750 series and GTX 970/980. There seems to be a bug with the launch config of X15 for Computer 5.2 cards as on GTX 750s it works fine, also the improvements are also present for GTX 750 and GTX 750 Ti users, but the hashrate increase is not so big. Do note that this fork of ccMiner is supports only Maxwell cards as it is optimized and compiled for Compute 5.0 and 5.2 architecture.
Back in September with the release of the new Maxwell-based NVidia GPUs, the GTX 970 and GTX 980 we did some performance testing of the GeForce GTX 980 hashrate using the available at that time ccMiner and CudaMiner miners. Today, a bit over two months later we’ve decided to check how hings have progressed in terms of performance improvements of the various crypto mining algorithms with the latest optimizations available in ccMiner. We’ve used the just posted Maxwell-optimized ccMiner 1.5-git ccMiner fork from SP to get the performance results from the same GTX 980 GPU running at stock frequencies (overclocking can further increase performance).
There is no surprise that there are some significant performance improvements in various crypto algorithms that have been achieved using various optimizations, but to our surprise there ware a few algorithms that did perform worse. The results we got from Groestl, HEFTY1 and JHA were a bit lower with the latest ccMiner than what we got in terms of performance back in September. Some other algorithms also don’t seem to have very big performance improvements such as Keccak, Luffa512 or NIST5, but the more popular and still profitable to mine algorithms do come with nice improvements. You can thank for the increase in performance to all of the developers that continue to work actively and optimize the performance out of which SP and tpruvot seem to be the most active lately, djm34, tsiv as well as others who have contributed code and also Christian Buchner and Christian H. – the creators of the ccMiner in the first place. Their code contributions and efforts in releasing their work publicly deserve some appreciation from the miners who are using Nvidia GPUs in order to keep them motivated. Below you can find the official donation addresses where you can send some coins for a beer or more to the each of the developers:
We have compiled a windows binary from the latest code of the ccMiner 1.5-git SP-MOD fork of the Nvidia miner optimized for Maxwell GPUs by SP (source). This miner includes the latest features available from tpruvot’s fork of ccMiner such as extra nonce support, however the main difference is that it is a version designed to provide the maximum performance on Maxwell-based Nvidia GPUs using Compute 5.0 (GTX 750, GTX 750 Ti) and Compute 5.2 (GTX 970 and GTX 980). If you are using an older Nvidia-based video card for mining crypto currencies you might want to stick with the tpruvot’s ccMiner fork that works with cards supporting Compute 3.0 or newer. The optimizations by SP do provide some extra hashrate and our test has shown a bit higher performance as compared to the latest code from tpruvot, testing with GTX 980 at stock frequency on X11 algorithm we are getting about 7900 KHS or slightly faster with the SP fork. The windows binary was compiled from the latest source with VS2013 and uses CUDA 6.5,s o make sure you have the latest video drivers installed.