How is it even possible to measure, unless you have two identical cards at identical speeds? running at the same time?
My experience so far with Zpool has been excellent, though I admit I'm only on day 3 of testing.
The first 24 hours I mined Lyra2v2, auto-converting to VTC, and got exactly the amount of VTC my hashrate and average difficulty predicted; yes, I understand that as an auto-convert pool there will be much more variation in payout, but it's nice to see reality match theory anyway.
The next 24 hours I mined Lyra2v2, auto-converting to XVG, and received 51 vs. 53 expected - close enough for me given the above.
Now I am running NemosMiner to test Zpool in multi-algo, auto-convert to BTC mode and will be comparing it to a similar rig running NiceHash for at least 3 days. Even if Zpool doesn't win this competition I will still use it in single algo mode because it has performed far better than similar tests done at HashRefinery, MiningPoolHub, and Zergpool.
A quick update after the first 24 hours of running - NiceHash is currently ahead with 0.33331 mBTC earned vs. 0.29274 mBTC immature + earned at Zpool.
The two rigs I have set up for doing these kinds of tests consist of a single GTX 1080 that I have tuned to deliver the same Sols/s (+/0 0.5) in Equihash. The algos I have selected in NemosMiner for Zpool are:
poly,hsr,keccak,xevan,skunk,tribus,phi,skein,equihash,groestl,sib,bitcore,x17,blakecoin,Nist5,MyriadGroestl,Lyra2RE2,neoscrypt,blake2s,lyra2z
Other settings in the NemosMiner bat file are to check profitability every 60 seconds (-interval 60) and switch algos at 5% difference in profitability (-ActiveMinerGainPct 5).
For NiceHash I have all of the single algos selected except for a few that I know have been ASIC'ed: keccak, nist5, neoscrypt, Lyra2rev2, DaggerHashimoto, Cryptonight, Equihash, Pascal, Blake2s. Oops, looks like I left Pascal enabled, even though it has been ASIC'ed... Oh well, can't change it now without invalidating the test.
One other update - I got credited for 1.5 more XVG in my earlier test so that brought the total up to 52.5 actual vs. 53 predicted; again, not the most scientific of tests, since one is mining multiple coins with a single algo, then converting everything mined to one of those coins, but you really can't complain if you get exactly the amount of coin your hashrate and average difficulty predict.*
* - I am well aware that this is not a terribly accurate means of evaluating the performance of a pool or miner, but there simply isn't any other way to do it when switching coins, much less switching algos.
