[Guide] Solo mine testnet bitcoins with bfgminer, Bitcoin Core, and a CPU/GPU

[n0nce]

I just catched an instance where the last block was 20 minutes ago and bfgminer never adjusted the difficulty setting from its '125M' value.

It sounds like this could be the reason it doesn't mine any blocks: if it never knows the difficulty is low, it won't find a block at low difficulty.
From your screenshot:

125M (894.7T)

I'm starting to think difficulty=1 doesn't mean it takes picoseconds to mine a block:

Minimum difficulty of 1.0 on testnet is equal to difficulty of 0.5 on mainnet.

At difficulty 1, that is around 7 Mhashes per second.

That would explain why ASICs still beat any CPU miner as it would take me about a second, while it takes the ASICs a fraction of a millisecond.

You may be up to something. I don't know why, but I assumed that difficulty=1 meant the hash just had to be smaller than 2²⁵⁶-1.
This is where I was wrong:

The highest possible target (difficulty 1) is defined as 0x1d00ffff

The average time to find a block can be approximated by calculating:

Code:

time = difficulty * 2**32 / hashrate

The expected number of hashes we need to calculate to find a block with difficulty D is therefore

D * 2**256 / (0xffff * 2**208)

Actually, if we use this formula, it appears that with difficulty=1, you need on average 4295032833 or ~4.3GH to find a valid one. So on a 7MH/s machine, that's 613 seconds or 10 minutes. Impossible..

Code:

>>> 1 * 2**32 / 7000000
613.5667565714285

But on the other hand, it can't be right since people here did succeed with CPUs. What if the article is outdated / wrong?

If we work with the figure of 7MH for 1 valid hash, it would mean the minimum network hashrate must have always been over 11.6kH/s; otherwise it wouldn't be possible to mine a block in 10 minutes (11,600*600 ≈ 7,000,000). That seems realistic for 2009. According to the blockchain ^[1], it was in the single-digit Megahash range all the way back in 2009.

Do we know @ETFbitcoin 's hash rate?

^[1] https://www.coinwarz.com/mining/bitcoin/hashrate-chart/2009


Quickly using the same formula for a single Compac F, it becomes clear it has much better chances:

Code:

>>> 1 * 2**32 / 300e9
0.014316557653333332

Code:

>>> 4.3e9 / 300e9 # hashes needed on average / hashrate of compac f
0.014333333333333333

With just 300GH/s and difficulty=1, it only takes it 0.0143s to go through those 7 million hashes.
This would also mean that the formula expecting @LoyceV's CPU to take on average 10 minutes, is correct.

[1714717537]

1714717537

[1714717537]

1714717537

[1714717537]

1714717537

[1714717537]

1714717537

[LoyceV]

Do we know @ETFbitcoin 's hash rate?

It looks like he used his GPU:

IIRC my GPU has speed about 200 Mh/s, so you should already mine at least a block within a month.

It also looks like my CPU mining is utterly useless I'll give it another week, then I'll wipe it.

For ASICs owners like yourself: would it be possible to only turn it on when the difficulty is 1? Or use it for something else, and then switch to testnet mining when the last block was for instance 19 minutes and 50 seconds ago?

[n0nce]

Do we know @ETFbitcoin 's hash rate?

It looks like he used his GPU:

IIRC my GPU has speed about 200 Mh/s, so you should already mine at least a block within a month.

It also looks like my CPU mining is utterly useless I'll give it another week, then I'll wipe it.

Indeed, that's a pity. @nullama may want to edit the original post / title and include the calculations I posted.
It seems that CPUs are too weak even for difficulty=1, and even GPUs may struggle.

An old 1GH/s USB mining stick would take 4 whole seconds to mine a block and a 200MH/s graphics card takes 20 seconds. That's really stretching it, if you ask me, as ASIC miners will have a 20 second head start.

For ASICs owners like yourself: would it be possible to only turn it on when the difficulty is 1? Or use it for something else, and then switch to testnet mining when the last block was for instance 19 minutes and 50 seconds ago?

Yeah, that should definitely be possible. Although I don't think it's advised to heat cycle hardware a lot, so maybe it's worth modifying cgminer to monitor the testnet blockchain and switch over whenever the last block approaches 20 min old. The ASIC doesn't care; it just hashes whatever it's given. But that way it stays hot and mines actually valuable BTC whenever the difficulty on testnet is higher than 1.

On the other hand, old / weaker ASICs make you such few sats per day that it's almost not worth the hassle to set this up. I'd just leave Compac F sticks on testnet for a day if I need tBTC and rest of the time on solo pool.

[LoyceV]

It also looks like my CPU mining is utterly useless I'll give it another week, then I'll wipe it.

To conclude this: I give up After wasting another week of CPU time without finding a block, I wiped the testnet installation. Using a testnet faucet is much faster if you need some.

[LoyceV]

it looks like mining tBTC (at difficulty 1) is more difficult than i expected.

In a way it makes sense: mining on one computer produces one block per 10 minutes. That sounds exactly like the beginning of Bitcoin.

[n0nce]

it looks like mining tBTC (at difficulty 1) is more difficult than i expected.

In a way it makes sense: mining on one computer produces one block per 10 minutes. That sounds exactly like the beginning of Bitcoin.

The 10 minute figure is correct:

The average time to find a block can be approximated by calculating:

Code:

time = difficulty * 2**32 / hashrate

Actually, if we use this formula, it appears that with difficulty=1, you need on average 4295032833 or ~4.3GH to find a valid one. So on a 7MH/s machine, that's 613 seconds or 10 minutes. Impossible..

Code:

>>> 1 * 2**32 / 7000000
613.5667565714285

Indeed, yes. The whole network hash rate in the beginning of 2009 was relatively stable around 4-5MH/s ^[1]. A new CPU from Q1 2009 like the Intel Pentium E5400 ^[2] was already able to output about 2.2MH/s ^[3], so there were literally just 2 CPUs on the whole network for almost a year back then. It somewhat surprises me that more people did not spin up a rig just for fun, but on the other hand the coin was worthless and energy & hardware did cost something.

[1] https://www.coinwarz.com/mining/bitcoin/hashrate-chart/2009
[2] https://ark.intel.com/content/www/us/en/ark/products/40478/intel-pentium-processor-e5400-2m-cache-2-70-ghz-800-mhz-fsb.html
[3] https://bitcointalk.org/index.php?topic=1628.0


@LoyceV, after browsing a bit online, it seems like there are modern CPUs now that can compete with something like ETFBitcoin's 100MH/s GPU.
It's just that your CPU is old.. old old..

I just mined 27 blocks over the last ~24h using a single Compac F; those were always 1-difficulty-blocks, so it should have been possible with CPU, just as well.

Your post got me motivated, so I've been mining Testnet on CPU (Intel(R) Xeon(R) CPU E3-1270 V2 @ 3.50GHz) for the past 19 hours. So far, not a single block mined.

From 2011 to 2023, well, things changed. An i9-9900K from Q4 2018 even manages over 100MH/s, it seems like. ^[4]
I will verify with some relatively modern Desktop hardware, later myself directly with bfgminer and report back.

Sadly, @nullama's assumption (I guess he assumed that, like myself?) that difficulty=1 means an immediate hit on any hardware, was wrong. However, with a modern computer you should definitely get something, if you have a few hundred MH/s.

[4] https://forum.level1techs.com/t/whats-the-fastest-processor-for-single-threaded-single-process-of-running-sha256sum/157464?page=3

[LoyceV]

It somewhat surprises me that more people did not spin up a rig just for fun, but on the other hand the coin was worthless and energy & hardware did cost something.

Don't remind me I should have done that when I first read about Bitcoin, which was years before I joined Bitcointalk.

@LoyceV, after browsing a bit online, it seems like there are modern CPUs now that can compete with something like ETFBitcoin's 100MH/s GPU.
It's just that your CPU is old.. old old..

It's old indeed, but works very well for it's task: pushing bytes to the outside world.
This Xeon CPU is already faster in finding Bitcoin vanity addresses than my laptop GPU.

From 2011 to 2023, well, things changed. An i9-9900K from Q4 2018 even manages over 100MH/s, it seems like. ^[4]

Buying a CPU just for hashing is a waste of money (and electricity). That's 95W that could be used for better things.

[n0nce]

It somewhat surprises me that more people did not spin up a rig just for fun, but on the other hand the coin was worthless and energy & hardware did cost something.

Don't remind me I should have done that when I first read about Bitcoin, which was years before I joined Bitcointalk.

I think many folks here share this sentiment.

@LoyceV, after browsing a bit online, it seems like there are modern CPUs now that can compete with something like ETFBitcoin's 100MH/s GPU.
It's just that your CPU is old.. old old..

It's old indeed, but works very well for it's task: pushing bytes to the outside world.
This Xeon CPU is already faster in finding Bitcoin vanity addresses than my laptop GPU.

From 2011 to 2023, well, things changed. An i9-9900K from Q4 2018 even manages over 100MH/s, it seems like. ^[4]

Buying a CPU just for hashing is a waste of money (and electricity). That's 95W that could be used for better things.

Sure, but if you have it already and need a bunch of testnet BTC, it should work nicely. I will try on the weekend with some modern desktop general-purpose hardware that I can access. I was surprised to see a >100MH/s hashrate on CPUs, since yours only managed single digits. Therefore, I was about to call this guide half-failed (the CPU part), but those benchmarks give me hope.

[nullama]

~snip~
Sure, but if you have it already and need a bunch of testnet BTC, it should work nicely. I will try on the weekend with some modern desktop general-purpose hardware that I can access. I was surprised to see a >100MH/s hashrate on CPUs, since yours only managed single digits. Therefore, I was about to call this guide half-failed (the CPU part), but those benchmarks give me hope.

CPU mining absolutely works for testnet with relatively new devices.

Just did a quick benchmark with bfgminer on a mac mini M1(2020 device) with CPU only(you can also use OpenCL to use the GPU), and in less than 5 minutes I got 2 shares over Diff 1:

Code:

 [XXX-XX-XX XX:12:17] New best share: 1
 [XXX-XX-XX XX:12:17] Accepted fe912210 CPU 7  Diff 1/1
 [XXX-XX-XX XX:12:21] 20s: 8.46 avg: 8.56 u:30.08 Mh/s | A:1 R:0+0(none) HW:0/none
 [XXX-XX-XX XX:12:42] 20s: 8.61 avg: 8.60 u:26.34 Mh/s | A:1 R:0+0(none) HW:0/none
 [XXX-XX-XX XX:13:02] 20s: 8.57 avg: 8.59 u:23.44 Mh/s | A:1 R:0+0(none) HW:0/none
 [XXX-XX-XX XX:13:22] 20s: 8.59 avg: 8.59 u:21.09 Mh/s | A:1 R:0+0(none) HW:0/none
 [XXX-XX-XX XX:13:43] 20s: 8.69 avg: 8.61 u:19.17 Mh/s | A:1 R:0+0(none) HW:0/none
 [XXX-XX-XX XX:14:03] 20s: 8.73 avg: 8.63 u:17.58 Mh/s | A:1 R:0+0(none) HW:0/none
 [XXX-XX-XX XX:14:23] 20s: 8.64 avg: 8.62 u:16.23 Mh/s | A:1 R:0+0(none) HW:0/none
 [XXX-XX-XX XX:14:44] 20s: 8.70 avg: 8.63 u:15.07 Mh/s | A:1 R:0+0(none) HW:0/none
 [XXX-XX-XX XX:15:04] 20s: 8.74 avg: 8.64 u:14.07 Mh/s | A:1 R:0+0(none) HW:0/none
 [XXX-XX-XX XX:15:24] 20s: 8.71 avg: 8.64 u:13.19 Mh/s | A:1 R:0+0(none) HW:0/none
 [XXX-XX-XX XX:15:44] 20s: 8.56 avg: 8.62 u:12.42 Mh/s | A:1 R:0+0(none) HW:0/none
 [XXX-XX-XX XX:15:53] New best share: 291
 [XXX-XX-XX XX:15:53] Accepted 00e0a123 CPU 5  Diff 291/1

You can see that it is absolutely possible to find a block when Diff is 1 with a CPU. It's just that it's way less probable than with an ASIC.

If you want to test a device without having to setup a local Bitcoin Core, you can do this:

Code:

./bfgminer -S cpu:auto --benchmark-intense

Also, notice the M1 hashrate hovering around 20Mh/s or so. It should be higher than that because I did the benchmark while using the device at 100% for other intensive tasks that I didn't want to stop.

[BitcoinSoloMiner]

For comparison purposes a single RX 6600 achieves 1500Mh/s in cgminer/bfgminer

[alexeyneu]

CPU

did you use sse at least? say sse3.1 have some stuff for that. avx too

[nullama]

CPU

did you use sse at least? say sse3.1 have some stuff for that. avx too

SSE is not supported in an ARM device like the M1.

Code:

benchmarking all sha256 algorithms ...
"c"        : benchmarking algorithm ...
"c"        : algorithm runs at 1.52071 MH/s
"cryptopp" : benchmarking algorithm ...
"cryptopp" : algorithm runs at 0.47593 MH/s
"c"        : is fastest algorithm at 1.52071 MH/s

That's per core, the M1 has 8 cores.

In an Intel CPU:

Code:

benchmarking all sha256 algorithms ...
"c"        : benchmarking algorithm ...
"c"        : algorithm runs at 1.82270 MH/s
"cryptopp" : benchmarking algorithm ...
"cryptopp" : algorithm runs at 1.11739 MH/s
"sse2_64"  : benchmarking algorithm ...
"sse2_64"  : algorithm runs at 1.42778 MH/s
"sse4_64"  : benchmarking algorithm ...
"sse4_64"  : algorithm runs at 1.31795 MH/s
"c"        : is fastest algorithm at 1.82270 MH/s

That's per core, the CPU had 20 cores. Also, I've seen the sse4_64 algo being the fastest at other times in the same CPU.

Note that this is bfgminer, with CPU. This is not supposed to have the latest CPU optimizations enabled as it's not used for that in years, maybe a decade or more.

[n0nce]

For comparison purposes a single RX 6600 achieves 1500Mh/s in cgminer/bfgminer

That's good to know. How about compiling a list of modern desktop hardware just for fun? Then it could be added to ^[1] the Bitcoin Wiki, which as of now only has very old hardware listed.
I think it would be good to post at least a console output from bfgminer as proof.

[1] https://en.bitcoin.it/wiki/Mining_Hardware_Comparison


Very odd. I quickly logged into a VM of mine with 7 cores (underlying CPU is pretty new Intel model boosting to almost 5GHz); the benchmark only results in ~20MH/s (probably due to virtualization), yet still I got an accepted hash after 7 minutes, another 3 minutes later and another one minute later.

Code:

[XXXX-XX-XX XX:51:31] Accepted 11128a4f CPU 4  Diff 14/1
[XXXX-XX-XX XX:54:51] Accepted bcd92259 CPU 0  Diff 1/1
[XXXX-XX-XX XX:55:27] Accepted f5f3501e CPU 1  Diff 1/1

I will test on bare-metal hardware with Bitcoin Core on the weekend, as mentioned before. Maybe there is hope for Loyce's 7MH/s second machine (although it makes no sense on paper - even getting accepted hashes that fast with 20MH/s...).

[alexeyneu]

Note that this is bfgminer, with CPU. This is not supposed to have the latest CPU optimizations enabled as it's not used for that in years, maybe a decade or more.

if there's demand now what's really needed is instructions for cpu with AES support. for sure regular sha256() ,openssl, with hardware aes turned on there , is faster than yours now.

[LoyceV]

I got an accepted hash after 7 minutes, another 3 minutes later and another one minute later.

Did you create your own Fork that stays on difficulty 1?

[nullama]

I got an accepted hash after 7 minutes, another 3 minutes later and another one minute later.

Did you create your own Fork that stays on difficulty 1?

When you use the benchmark option the difficulty by default is 1. It doesn't connect to any network.

[BitcoinSoloMiner]

For comparison purposes a single RX 6600 achieves 1500Mh/s in cgminer/bfgminer

That's good to know. How about compiling a list of modern desktop hardware just for fun? Then it could be added to ^[1] the Bitcoin Wiki, which as of now only has very old hardware listed.
I think it would be good to post at least a console output from bfgminer as proof.

[1] https://en.bitcoin.it/wiki/Mining_Hardware_Comparison


Very odd. I quickly logged into a VM of mine with 7 cores (underlying CPU is pretty new Intel model boosting to almost 5GHz); the benchmark only results in ~20MH/s (probably due to virtualization), yet still I got an accepted hash after 7 minutes, another 3 minutes later and another one minute later.

Code:

[XXXX-XX-XX XX:51:31] Accepted 11128a4f CPU 4  Diff 14/1
[XXXX-XX-XX XX:54:51] Accepted bcd92259 CPU 0  Diff 1/1
[XXXX-XX-XX XX:55:27] Accepted f5f3501e CPU 1  Diff 1/1

I will test on bare-metal hardware with Bitcoin Core on the weekend, as mentioned before. Maybe there is hope for Loyce's 7MH/s second machine (although it makes no sense on paper - even getting accepted hashes that fast with 20MH/s...).

16/02/2023

RX6600 (bfgminer):

Code:

 OCL 0:       |  1.18/ 1.50/ 1.46Gh/s | A: 5 R:0+0(none) HW:0/none

RX6700XT (cgminer):

Code:

GPU 0:                | 2.198G/2.663Gh/s | A:6468 R:0 HW:0 WU:    39.5/m I:10

I would enjoy seeing what a 4090 can achieve

[n0nce]

Note that this is bfgminer, with CPU. This is not supposed to have the latest CPU optimizations enabled as it's not used for that in years, maybe a decade or more.

if there's demand now what's really needed is instructions for cpu with AES support. for sure regular sha256() ,openssl, with hardware aes turned on there , is faster than yours now.

Since openssl is a dependency, I suppose it already uses that. Furthermore, this isn't nullama's software. I'm pretty sure it was first released (and maintained - last commit in '21) by Luke Dashjr ^[1-2].

[1] https://github.com/luke-jr/bfgminer/
[2] https://en.bitcoin.it/wiki/BFGMiner

I got an accepted hash after 7 minutes, another 3 minutes later and another one minute later.

Did you create your own Fork that stays on difficulty 1?

I ran it like suggested by nullama just for benchmarking:

Code:

./bfgminer -S cpu:auto --benchmark-intense