There are a lot of different and competing digital currencies. The big names like Bitcoin, Litecoin, and Dogecoin are the most commonly used coins for benchmarking purposes. But in the view of SemiAccurate a quality benchmark has to meet three major criteria: it needs to be a real-world workload, the results have to be consistent, and the results have to be reproducible. To this end we’ve done quite a bit of research and using Vertcoin as a measure of OpenCL-based GPU compute performance in our benchmarking suite is the result of all that work.
As a viable crypto currency with a network hashrate of about five Gigahashes per second it clear that mining Vertcoin is a meaningful real-world workload. Consistency and reproducibility are concepts that go hand in hand when it comes to crypto currency. Vertcoin is a Scrypt-N coin; unlike SHA-256 coins or other Scrypt-based coins, Vertcoin offers comparatively low hashrates. Using one of AMD’s HD 7970’s will net you about 650 Megahashs per second mining a SHA-256 coin. That same GPU will mine a Scrypt-based coin at about 700 Kilohashes per second and it will mine Vertcoin at about 350 kilohashes per second. We believe that extreme difficulty of the algorithm behind Vertcoin will allow it remain relevant through successive generations of new GPUs.
Vertcoin is unlike a lot of other coins in that the user’s options for tuning their miners are far more limited than with other coins. There are really only two relevant settings when it comes to tuning a Vertcoin miner: intensity and thread concurrency. As a good starting point to find the right thread concurrency number for your GPU you should take the number of cores and then multiply that number by four. Because Vertcoin’s algorithm requires double the amount of memory as other Scrypt coins we then are going to double that thread concurrency number. In the case of your HD 7970 that means we’re using a thread concurrency of 16384.
Now that we have that nailed down, we have to tune the miner’s intensity. This is done by trial and error, but most miners find themselves using a setting between 10 and 20. For our purposes we’re going to find the setting that provides the maximum hashrate on a given piece of hardware and then use that average hashrate as our benchmark.
So by changing merely two settings we’ve effectively tuned our miner to its maximum hashrate. This is the biggest advantage of using Vertcoin as a benchmark; it’s easy to tune and performance is very consistent. As an effect of that trait it also becomes very easy for our readers to replicate our results. Thus we can achieve our last two goals: consistency and reproducibility.
Now with all this said, Vertcoin is still one of the smaller coins available if you judge it by the total volume of transactions per day. But our Vertcoin results are applicable to the wider mining market in two ways. First off the performance differences between GPUs on a relative basis are very similar when comparing the performance of two GPUs mining Litecoin to two GPUs mining Vertcoin. Thanks to this trend getting a good estimate of a given GPU’s performance mining other scrypt-based coins can be easily achieved by merely doubling its Vertcoin hashrate. Thus we’re killing two birds with one stone by benchmarking with Vertcoin as we have a benchmark that will retain its relevancy thanks to its difficulty, and we’re getting a general gauge for how a card will perform in other currencies, without incurring the consistency and reproducibility issues associated with more tunable currencies.
Hopefully you now have a better understanding of why we’ve chosen to benchmark with Vertcoin over Dogecoin or Litecoin. Moving forward we will have Vertcoin benchmarks in all of our APU/GPU review articles and it will be our go to crypto currency benchmark for gauging the performance in that category of workloads. In general benchmarking crypto currencies is a messy endeavor, but we’re confident that we’ve found a meaningful way to quantify the relative performance of GPUs with our Vertcoin benchmark.S|A
Latest posts by Thomas Ryan (see all)
- Intel Admits to Broadwell Delays in its Q2 Call - Jul 18, 2014
- Intel’s Core i7-4790K: Devil’s Canyon - Jun 10, 2014
- AMD Launches its 2nd Generation of G-Series Embedded APUs - Jun 5, 2014
- Nvidia’s Tegra K1 is the Heart of Google’s project Tango - Jun 5, 2014
- AMD Goes All Out to Win with FirePro - Jun 5, 2014