Zeropond Cloud Mining

Mining
61

0.03 h/s i can get from cpu market for just half of 0.0125, so i undestand that cloud want to earn but not 100%!
And dont forget that any crypto cloud markets will never give you profit!

62

Each solve on the benchmark is 2 hashes per reported solve time. A new i7 with DDR4 (assuming it’s at least 3 GHz) should give about 30 second solve times which is 1 hash per 15 seconds. Times 4 cores gives 4 / 15 = 0.26 H/s i.e. 8x faster than their 0.03 H/s. I see what you’re saying: if you just invert the 30 s/H you get what they are wanting us to pay for, about 0.03 H/s. But they did not say 0.03 x 4 cores x 2 hashes per solve is what you’ll be getting. I am assuming their 0.03 H/s is exactly the correct and literal metric, per core, with the 2x factor correctly applied.

[edited to be more clear]

63

They can if they have the GPU mining software and you don’t. That’s the whole business plan here.

64

Even without gpu they will get x2 profit at start but not investors with such prices, when hashrate rises to heaven 0.03h/s will be nothing!
Gpu can give max at x8 speed because of 8gb memory limit.

65

I see, it would be clearer if they had stated “Equivalent to one core in a Modern PC”

66

No, 0.067 H/s per core is a new PC getting 30 seconds per “solve” which is 15 seconds per hash. 0.03 H/s per core is my 2010 PC. Remember, there are 2 hashes per solve time reported by the benchmark, unless they fixed it in beta. If this cloud service made the same mistake, maybe they are getting 10x a CPU instead of 20x.

67

If I have a 12GB GPU can I run ~12cores?

68

12 / 0.8 = 15 threads. You might need 5 cores per thread to be equivalent to 1 CPU core. Maybe 1 GPU core per thread is enough, if each GPU core is working on a 384 bit channel instead of 64 bit at a typical 1.3 GHs core clock speeds.

69

Please link to where we can see the 2 solutions per solve info, I was not aware of this and would like to put this info in the wiki.

EDIT- With Zawys new information I have re-calculated the Wiki and added Hash/sec to the table. benchmark.minezcash.com

70

It’s been mentioned about 5 or 10 times, but not nearly as much as people just reporting their seconds to solve. Here’s the most recent from someone who should know: Hashrate for dummies

71

Tim, did you make a mistake in calculating the 20x factor? As I pointed out above, those solve times on the Wiki are for 2 hashes, not 1. So maybe you’re 10x faster? Even so, you might still get 20x more blocks because the CPU miner has a minimum solve time of 35 seconds “to get started” for the upper 1/3 of CPUs and even longer for slower machines. If you make up 33% of the network using the 10x factor, your average solve time will be 450 seconds. But 8.5% of your blocks would be under 35 seconds where you would get nearly 100% of blocks. This works out to getting 58% of all blocks.

72

0.03 H/s is based on our tests, and it’s one core, SORRY, not a whole high-end CPU. We will fix the language on the pricing page.

The 20x number is based on speedup over one core, so if you are getting 0.06 after you saturate your memory bus with a bunch of cores, then we are only 10x. Salt got almost 0.1 on his dev laptop, and then we are only 6x better.

We based the 0.03 number on this test of the highest-end desktop we could find:

The blue line is the high-end box and the orange line is a consumer laptop.

Tower:
{{1, 0.031404}, {2, 0.0624906}, {4, 0.115088}, {6, 0.172909}, {8, 0.206423}, {10, 0.226736}, {12, 0.259101}, {14, 0.268521}, {16, 0.316972}, {18, 0.335401}, {20, 0.340362}, {22, 0.333263}, {24, 0.324177}}

Laptop:
{{1, 0.04909691301846127}, {4, 0.09760332693323176}}

Here are the specs for the big box:

processor    : 47
vendor_id    : GenuineIntel
cpu family    : 6
model        : 63
model name    : Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz
description: DIMM Synchronous 2133 MHz (0.5 ns)
            product: 9ASF51272PZ-2G1A2
            vendor: Micron (date:14/4p)
            physical id: 0
            serial: 0D6ABEE6
            slot: P1-DIMMA1
            size: 4GiB
            width: 64 bits
            clock: 2133MHz (0.5ns)
description: Motherboard
      product: X10DRT-PIBF
      vendor: Supermicro
      physical id: 0

CPU specs: Intel Xeon Processor E52680 v3 30M Cache 2.50 GHz Product Specifications

Motherboard: X10DRT-PIBF | Motherboards | Products | Super Micro Computer, Inc.

Note that this mobo has two CPU’s, so there are 24 cores (48 cores with hyperthreading) and 8 memory channels total.

To run this test, we used str4d’s multithreaded benchmark branch found in this pull request:
Measure multithreaded solveequihash time per-thread by str4d ¡ Pull Request #1376 ¡ zcash/zcash ¡ GitHub We ran solveequihash 30 then calculate hashrate as 2 * threads / avg_time_per_thread

The performance scales linearly only up to the number of memory channels, then drops off as the memory bus gets saturated.

You also see this in the orange line. Running 4 threads on a 4-core laptop does NOT get you 4x speedup because the membus is saturated. We got 1.99x speedup which is almost exactly the effect we’d expect from a dual memory controller system.

So no, you cannot simply say a 4-core machine will perform at 4 times the single-threaded equihash benchmark.

We will make the pricing page more clear about what 0.03 H/s means. Or maybe we should just scale our prices to 1 H/s so there’s no confusion? What do you think?

73

Thank you for clarification @tim_olson it looks like you guys put a lot of work into the numbers

74
  1. It looks like your laptop was at 0.05 H/s per core which is more like my 2013 desktops. When making comparison, I think you should start it with a 0.06 H/s per core as a modern CPU comparison.
  2. The last I saw, the benchmark was not accurate for multicores. You might have to throw out that data.
  3. At least older PCs are not saturating the memory bus at 4 cores. I have not seen anything to indicate newer ones are either. I can reduce my memory cards from 4x 8GB to 1x 8Gb and seem to get 4 times as many blocks on the testnet verses using 1 core. My 2013 PCs on the testnet appear to be 0.18 H/s when running 4 cores.

I see a run-up in the difficulty this afternoon. Maybe you’ve begun testing? It was a nice little pick up, of about 20 equivalents to my desktops. My electricity expenses per block might be nearly 1/2, and my hardware costs less.

75

Note that we did use a multithreaded benchmark for this test, which hasn’t yet been merged into master.
See this pull request:

76

I need to know when you start mining beta testnet and what you’re throwing into it. I have 5 identical 2013 PCs running, and I just started some single core tests. I tested the single cores back on z7 and z8, but not hardly on z9 and none on beta.

77

Hey everyone, really interesting reading over all the posts here. @zawy you’re actually correct on the 2x speedup on solver vs hash. Not sure what happened but I think I copied the wrong column from our benchmark sheet while making the page. I guess that’s what happens when you barely sleep the night before. :slight_smile:

To be clear, yes, 0.03 is derived from time from solveequihash, and that rate is actually doubled (0.06) for real mining. I am updating the page now. Thank you for pointing that out to us, we really appreciate it.

It’s clear to me that the pricing is a bit complicated, and I’m going to work on simplifying it (how it’s displayed). I think at this point it’s hard for the average user to understand exactly what hashrate they are getting for solveequihash vs with threads vs real mainnet mining.

As for the threaded speedup, it’s still clear to us that it’s not linear. All our tests show memory bus saturation (and therefore a big limit on multiplier), but I’d love to see more data from anyone - 1 vs 4 cores on 30+ runs would be ideal.

Along with our goals to make pricing simple for someone visiting each unit is sold at speed in ideal conditions (memory totally available.)

We’re really busy but also still fiddling with optimizations.

Also, no, we have not run a mainnet mining test yet. Any hashrate you’re seeing right now is not us. We will post before we run a test.

The 50% off you mentioned is not for the whole Genesis Contract (20%) but rather for the first month. We are swallowing the reduced profits from Slow Start. Hope this makes sense.

78

[edit: with further testing, I believe my estimates here are too low]
It looks like you’re 2.5x a moderate DDR4 CPU. On an equipment cost basis, you’re about 2x less costly than used desktops. I think it is fair to say you are 3x CPUs, like Ampy said. [edit: Ampy also said he hope to get higher, last we heard] On an electrical expense basis, you’re probably 2x more costly.

Your big box Xeon will not drop off like that on testnet. In a previous thread z7 feeds we got the info on a 64 core Xeon setup. Despite the DDR4, the 2.3 GHz CPU made him nearly as slow as my 2010 desktops on a per core basis. When he went on testnet, he got exactly the number of blocks expected if the 64x cores scaled linear. I’m not sure how many memory buses it was.

The following rule has not yet been broken:

blocks gained = CPU speed x RAM bandwidth x cores
[edit: with further testing, I believe “x cores” here is too high, in partial agreement with Tim Olson.]

The GPUs will face the same problem. If you go only for RAM, it’s going to cost you in core speeds. It’s not going to be as efficient electricity wise.

Make sure nothing else is running on the laptop. I would not trust the multithread benchmark.

Actually, 0.06 H/s seems a little too fast for that 2.5 GHz Xeon. I do not think it is faster than 0.05 H/s and 0.03 H/s might be correct.

79

Hi All,

In case others are interested, I did the following to validate the claim on zeropond.com of generating 300,000 XMR (Monero) coins.

  1. Downloaded and started running the monero daemon: bitmonderod.exe (Downloads | Monero - secure, private, untraceable)

  2. Let it sync for a few hours.

  3. create a view wallet with the keys provided

    simplewallet --generate-from-view-key proofWallet.dat

     Enter in address addresskey when prompted
 Entered in the viewkey when prompted

  4. Let it sit at the [wallet xxx] prompt until the wallet starts to sync, this sync process takes a while.

  5. After the sync is done, issue the ‘balance’ command.

.
[wallet 43GoDc]: balance
Balance: 323021.862646320626, unlocked balance: 323021.862646320626
[wallet 43GoDc]:

Is Zeropond cloud mining a scam or trustworthy?
80

@farallon Thank you! Great to have third party validation. Some of the devs on #monero on irc have also done verification and would probably be happy to confirm.