Tromp's solvers

Mining
330

This if I’m not mistaken is the root of the confusion. Some are referring to equihash runs - and there are ~2 solutions per equihash run but this isn’t how it is now defined.

331

I am about to go crazy with all this sol/s and h/s…

Can someone please explain.

https://www.internetsociety.org/sites/default/files/blogs-media/equihash-asymmetric-proof-of-work-based-generalized-birthday-problem.pdf

332

I made some tiny optimizations to CUDA.
Now doing 19.3 Sol/s on GTX980…

If anyone gets 0 solutions on first run after compiling with

make eqcuda
./eqcuda

then please report it here. And also whether it happens if you
make and run eqcudah instead.

333

Tested 980ti with eXTremals miner using the tromp solver,
25 Sol/s (in windows, will test in linux later)

You can Watch the results here:
GTX 980ti Tromps Solver in eXTremals Miner 25 Sol/s

1 Like
334

Can you share the Win binary? Or give how to use extremals miner with tromp solver?

335

it uses tromps Cuda solver for Nvidia… when i said “using the tromp solver” i mean it uses the tromp solver, thats all
and for AMD “This is a OpenCL fork of CUDA John Tromp’s miner.” (first reply on bitcoin talk") - just like how Zogminer is an opencl fork of johns solver

1 Like
336

Is there a miner that integrates directly the original CUDA version of tromp’s solver and can work with Nvdia cards?

337

so is 2 slots of ddr4-2133 with an I7 or I5 CPU going to mine faster than the same with ddr3-1333?

338

I’ve installed and compiled zcash (with tromp’s solver).

How can I get the Sol/s of my machine?

339

Zcash developer here, and that is incorrect. 1 Sol/s == 1 H/s.

340

Adding a separate post so the clear message above isn’t lost.

The confusion stems from several areas:

  • Bitcoin uses SHA256d(block_header) -> difficulty check for its PoW
  • People who are used to mining Bitcoin think of H/s as being equivalent to SHA256d runs/s (which it is).
  • Zcash uses Equihash(most_of_block_header) -> SHA256d(block_header) -> difficulty check for its PoW.
  • Equihash produces on average approximately 2 solutions per run.
  • Thus some people approximate 1 Equihash run/s = 0.5 H/s (== 0.5 Sol/s), but that is an approximation.
  • Some developers (in particular, NiceHash) assume that 1 Equihash run is equivalent to 1 SHA256d run, and therefore that 1 Equihash run/s == 1 H/s. This is incorrect.
  • The correct way to calculate the solution rate is to count the actual number of solutions found, which is exactly equal to the number of times the block hash is checked against the target. Thus 1 Sol/s = 1 H/s.
  • The reason Zcash uses Sol/s is because it is unambiguous what solutions per second means, whereas H/s clearly causes confusion with developers (and therefore with users).
1 Like
341

Common (200,9) Equihash implementations, which have negligible discarding, produce on average 1.88 solutions per run.

1 Like
342

Thanks, edited my post to clarify it was only approximately 2.

343

Solution rate per run:
1.87905 (n=200, k=9)
1.99997 (n=144, k=5)

Any plans to switch to (n=144, k=5) or Cuckoo anytime soon? Botnets drive down profitability; the low memory consumption for n=200, k=9 is not enough of a deterrent.

344

Cuckoo Cycle would have even lower memory footprint, since a size 2^30 instance requires only 128MB, and takes around 5 seconds to solve on a fast GPU, and over 20 seconds on any CPU.

As I said before, we shouldn’t expect any PoW changes in the next several months.

345

Is there a guide anywhere to implementing the xenoncat miner in linux to mine on the 28th? I am currently using the official zcash miner but there is all this talk about how xenoncat’s is so much faster yet I cant seem to find how to set it up in linux! help!

346

@sarath-hotspot is working on the integration and it should be ready zoon. However, Tromp’s CPU solver is almost as fast.

3 Likes
347

There’s some confusion due to me having two versions, only the first of which is widely available in a miner.

tromp equi1: 4.5 Sol/s
tromp dev1 : > 6 Sol/s (requires AVX2)

xenoncat: > 6.4 Sol/s (with AVX2)

1 Like
348

Hi Tromp, can you please clear this up for me.

Speed: 18.1613 H/s
Speed: 33.998 Sols/s

Above is an example of a benchmark test. How can everyone be saying that H/s and Sols/s are the same when they are showing two different results?

I just want to know what performance I am actually getting.

Thanks.

350

You produce two solutions on 1 equihash run and each solution is used against target therefore you run 2 hashes. basically 1 solution =1 hash.