Zcash Alpha Testnet Update

Last night the Zcash team released an update to the Test Network.

The Zcash developers have natively implemented a Hard Fork increasing the block size to 2MB. I believe you will have to download z6 and re-compile to be on the new version. Testing now…

My favorite part of this release is the Equihash optimization, now those of us like @Voluntary and @Austin-Williams will be able to run much more accurate benchmarks :sunglasses:

Hm… My z5 install is still downloading blocks from where it left off last time… I might take the precaution of saving my wallet before this upgrade.

I had heard from some miners that they had lost connection to the network, and the Github was updated with the new z6 download link so I recommended just doing a re-compile to be safe.
So your z5 may still be working because it’s a native fork, but I don’t know how else you would get the updated Equihash unless you download z6.

I think this version is pretty close to 1.0 launch. Will there be any significant changes on testnet beside pumping Equihash n & k parameters to 144 and 5?

“Equihash n & k parameters to 144 and 5”

Will this increase the recommended memory requirement
above 1 GB per core?

I haven’t compiled yet but if you use a program like System Monitor, it’ll give you a rough idea of the resource usage of running tasks.

–Edit–
I’ve compiled z6 and done the tests. I’m doing a small benchmark and the zcashd process is using between 400 to 457 MB…

After downloading and re-compiling my average Equihash solve time is down about 5 seconds overall.

Solving:
I did notice that solveequihash still only runs on a single thread and zcash-cli peaks at approx: 600MB + zcashd 400MB for 1 to 1.3GB total while running a solve 1. You can shave about another 5 seconds off your solve time if gen is set to 0 (not mining at the same time). Since its only solving on one thread adding to genproclimit does nothing to reduce times.

Memory:
Idle zcashd (not mining) uses about 40MB, While Mining zcashd uses about 400 - 435MB per thread assigned to genproclimit, so 1=400MB, 2=800MB, 3=1.2GB etc.

At this point I don’t quite understand why when Mining (gen=1) zcashd uses so little memory. Isn’t Equihash supposed to be the algo for solving blocks? During Mining and finding several blocks the memory never peaked to 1.3GB like in the solve Equihash test so that tells me zcashd is not using Equilash. Or am I misunderstanding how Equihash is supposed to work?

I tend to set gen=1 in zcash.conf and I’m seeing a fairly consistent 455 MB in use by zcashd - and I’ve managed to get a few blocks so far during this session.

The benchmark graph is unfortunately inaccurate in this regard. The current parameters use about 440MB of RAM, which you can see in the raw memory benchmarks on the CI server. But the benchmark graph only shows peak memory, and on the CI server Equihash is reliably spiking to double or triple the expected memory in the final round. I haven’t yet had time to track that down.