I read in the wild some folks are utilizing the Xeon Phi for Zcash mining. If this is legitimate; what kind of hashrates are we looking at? I’ve found nothing yet.
Any ideas?
Probably less than 1.5 Sol/s per core. I’ve tested around 10 different CPUs and those with higher clocks are much much better performers (over 2.9Ghz per core in boost mode are good) .
1 Like
Thank you. I’ve been looking at hashrate distribution; and I’m boggled with seeing so many miners operating at over 4K sol…
One in particular running at over 700k
I was curious what was their secret weapon or is it just pure gpu power.
I assumed if the Xeon Phi was a heavy hitter… maybe.
Or not. I’m perplexed. It’s amazing the sheer power.
But if you compile sources with offload mic for xenon phi and avx-512 it could get better result for tromp solver, but xenoncat is in assembly so no benefits from compile on better cpu… and all that cores using the same memory… so would be low GB/s per core…
1 Like
They are probably renting hashrate from Nicehash. Im talking about Xenoncat AVX optimizations and Tromp solver mining software.
1 Like
@solardiz benchmarked the Xeon Phi 5110P at 4.2 sol/s using silentarmy: Slack It is underperforming because the software is unoptimized for it.
1 Like
If someone optimizes it, i think it can have a good sols/sec because the high memory bandwidth ! Are there more requisites for mining Zcash?
1 Like
Yes you can look up the algo. It’s pretty interesting and I’m personally impressed. I’ve seen them all since 2009.
1 Like
Long-term how well do you think it is going to hold up to the anti-ASIC goal?
Additionally:
Since it’s been more or less shown that you don’t need mining to secure a blockchain (but that it does serve other purposes including regulatory purposes) why not look into a coin or chain or similar that puts distributed computation to more immediately useful and possibly profitable ends?
1 Like
Long term if it shows resilience I believe asic architecture will be developed. Asic can be developed for any coin just depends on the demand, value, roi, etc
Long term value I’m am conservatively optimistic. Time will tell.
And you …?
I agree with your assessment.
Therefore, I think that we actually have to address a different question altogether:
When an ASIC is built, will there be a hard fork to foil it?
Also, I notice there is no ethereum ASIC. Do you know why?
1 Like
There is no ethereum asic because it’s not worth it. It’s just cost too much to develope and manifacture because it needs a lot of ram. Maybe it could be more efficient than a gpu, but overall it will not be cheaper. If a New gpu chip is coming out, millions of it will be manifactured. If you make an ethereum asic, probably won’t do millions of it, so it will be more expensive even if you can do at least as good job as AMD/Nvidia engineers.
The bitcoin asic is completly different. The SHA-256 algorithm is much simpler, it’s “easy” to make a chip for it. Btw you didn’t have competitor in the early days, because optimized sha-256 computing was not a common usecase according to ethereum algo, where gpus do a good job. This is the difference. 
But the equihash can be different. As I see CPUs can do almost as good job as GPUs, the difference is much lower than at ethereum. So I can imagine some kind of special hardware could be worth to do develop.
I highly doubt most common CPUs can do 70-120H/s and looking at power usage we still have another 20-30% to squeeze out of GPU mining software.
I’m not saying that the current cpus can do that, but I think it can be possible to create a special hardware which worth it because of the smaller difference between gpu and cpu than in the case of ethereum. Of course it’s absolutely not 100%, but I can imagine it as an engineer.
I believe that a hard fork would be necessary. We don’t want asics. This gives the small guy a slightly better chance.
As for ETH. There may be an asic in the wild… however the chance of ETH goin away from POW -work based mining, to not equates to making all of that mining equipment useless. This is probably why Asic devs have been slightly off put about going that route.
What’s your opinion?