Thanks for the very useful info.

AMD Radeon Vega Frontier Edition (Air)
OS = Windows 7 Professional SP1
Driver = Radeon Vega Frontier Edition 17.6
Average Solutions per Second = 310 (throttles if you let it run for 10+ minutes)

AMD Radeon Vega Frontier Edition(Liquid) / RYZEN 7 1800X or Core i7-7700K
Will try later tonight with Windows 10 Professional or Ubuntu 16.04.3 LTS with beta drivers and report back

Could I please have your launch configuration + OS + driver version?

After 11.5 minutes:
ZEC: 08/08/17-15:04:13 - SHARE FOUND - (GPU 0)
ZEC: Share accepted (140 ms)!
ZEC: 08/08/17-15:04:23 - New job from eu1-zcash.flypool.org:3333
ZEC - Total Speed: 260.951 H/s, Total Shares: 25, Rejected: 1, Time: 00:11
ZEC: GPU0 260.951 H/s
ZEC: 08/08/17-15:04:29 - New job from eu1-zcash.flypool.org:3333
ZEC - Total Speed: 261.559 H/s, Total Shares: 25, Rejected: 1, Time: 00:11
ZEC: GPU0 261.559 H/s
ZEC: 08/08/17-15:04:48 - SHARE FOUND - (GPU 0)
ZEC: Share accepted (156 ms)!

AMD Radeon Vega Frontier Edition @ stock clocks:

AMD Cards available: 1
GPU #0: gfx900, 16192 MB available, 64 compute units
GPU #0 recognized as Vega
POOL version
GPU #0 algorithm ASM, intensity 6
Total cards: 1
Watchdog enabled
Remote management (READ-ONLY MODE) is enabled on port 3333

ZEC: Stratum - connecting to 'eu1-zcash.flypool.org' <94.23.60.113> port 3333 (unsecure)
ZEC: Stratum - Connected (eu1-zcash.flypool.org:3333) (unsecure)
ZEC: Authorized
Pool sets new share target: 0x00083126 (diff: 8000H)
ZEC: 08/08/17-13:54:52 - New job from eu1-zcash.flypool.org:3333
ZEC - Total Speed: 311.773 H/s, Total Shares: 0, Rejected: 0, Time: 00:00
ZEC: GPU0 311.773 H/s

I am interested in this GPU Prime Miner bounty, it is one way to discourage botnets.

They both run at the same speed, how did you measure performance?

Same here, this was a bandwidth exhaustion attack targeting Riecoin nodes.

@gatra
Here’s a suggestion to make Riecoin botnet resistant while keeping the classic proof-of-work.
Add double SHA-256 (SHA256d) as an additional constraint to the proof-of-work.
SHA256d’s difficulty should be 150s seconds of computation time for an i7 4790k, approximately  4.2 billion hashes.

1). SHA256d(target || nonce) -> offload to GPU
2). Add digest to target
3). Solve proof-of-work puzzle.
4). Submit(256-bit offset,  64-bit nonce).

The idea is simply to add 150 seconds of preprocessing time on a CPU, but only 3-5 seconds on a single GPU.

Yes, it is as simple as flipping on a single bit. But you will also need to modify the controller software.

GPUs and FPGAs
Exactly the same mining algorithm used in Bitcoin and you can use existing Bitcoin mining software with trivial modifications. To be more specific, a one bit change to the second SHA256 chunk's padding (h0280 -> h0680). You will also need to modify the host/controller software to pass the difficulty and optionally increment the time stamp.

Bitcoin SHA256d data:
80 byes fixed.

Pascal SHA256d data:
208 byes fixed - but variable.

FPGA:
Ztex 1.15y hash rate 890 MH/s @ 45 Watts from the outlet.

How are you calculating these numbers, they seem way off.

Wait a minute, the calculations are off from this image. The nonce is in sequential order until a new job arrives, then it resets to zero.

Now keep in mind that the solution rate per run for Equihash parameter (n=200, k=9) is approximately 1.87905. Therefore, a million runs of the solver with a random nonce should produce 1,879,050 solutions on average. This is actually consistent with the number of runs reported in the image above. The only problem is that it is 12 Sol/s; I am not sure how the 134 Sol/s is being calculated. Also, as far as ZCash’s implementation of Equihash is concerned, H/s = Sol/s. A hash is actually a hash of your solution set which is then compared against the target difficulty.

I think that we should all be on the same page when measuring solution rate, it would create less confusion.

That would be nice but the Steemit platform has some other major security issues:

Depending on how a transaction was constructed, you can change the Security Keys of another user's account without their permission. See error log below:

    {"witness_owner":"supercomputing5"}
    th_a  database.cpp:700 operator()
3471006ms th_a       witness.cpp:383               maybe_produce_block  ] Clearing pending transactions and attempting again
3471006ms th_a       witness.cpp:382               maybe_produce_block  ] 10 assert_exception: Assert Exception
witness_obj.signing_key == block_signing_private_key.get_public_key():
    {}
    th_a  database.cpp:721 _generate_block

    {"witness_owner":"supercomputing5"}
    th_a  database.cpp:700 operator()
3471006ms th_a       witness.cpp:383               maybe_produce_block  ] Clearing pending transactions and attempting again
3471006ms th_a       witness.cpp:286               block_production_loo ] Failure when producing block with no transactions
3471999ms th_a       witness.cpp:274               block_production_loo ] Not producing block for supercomputing5 because I don't have the private key for STM5b3wkzd5cPuW8tYbHpsM6qo26R5eympAQsBaoEfeMDxxUCLvsY
3472999ms th_a       witness.cpp:274               block_production_loo ] Not producing block for supercomputing5 because I don't have the private key for STM5b3wkzd5cPuW8tYbHpsM6qo26R5eympAQsBaoEfeMDxxUCLvsY

# name of witness controlled by this node (e.g. initwitness )
witness = "supercomputing"
witness = "supercomputing1"
witness = "supercomputing2"
witness = "supercomputing3"
witness = "supercomputing4"
witness = "supercomputing5"
witness = "supercomputing6"
witness = "supercomputing7"
witness = "supercomputing8"
witness = "supercomputing9"
witness = "supercomputing10"
witness = "supercomputing11"
witness = "supercomputing12"
witness = "supercomputing13"
witness = "supercomputing14"
witness = "supercomputing15"
witness = "supercomputing16"
witness = "supercomputing17"
witness = "supercomputing18"
witness = "supercomputing19"
witness = "supercomputing20"
witness = "supercomputing21"
witness = "supercomputing22"
witness = "supercomputing23"
witness = "supercomputing24"
.
.
.


# name of miner and its private key (e.g. ["account","WIF PRIVATE KEY"] )
miner = ["supercomputing","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing1","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing2","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing3","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing4","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing5","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing6","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing7","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing8","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing9","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing10","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing11","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing12","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing13","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing14","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing15","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing16","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing17","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing18","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing19","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing20","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing21","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing22","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing23","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
miner = ["supercomputing24","5Jxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"]
.
.
.

Not so fast, the AMD (R9) implementation is just now catching up to NVidia’s (Maxwell) implementation for Niro mining in terms of Performance per Watt.

Wolf0 has done an amazing job optimizing/vectorizing the OpenCL kernel for AMD. It’s well written, clean, and easy to follow: 62 MH/s on my R9 280X. If we use his code on NVidia with some minor tweaks, we should see some decent performance gains and retain the Performance per Watt crown. So in a way, he is also indirectly optimizing the code for NVidia’s Maxwell.

Thanks to Wolf0 and Viz.

@Wolf0

Miner version: 1.2 (with Wolf0 OpenCL kernel)
OS: Ubuntu 14.04 LTS
Driver: 14.9
GPU: R9 280X (stock)

[MASTER] Coinshield Network: New Block 511975
60.8 MH/s | 0 Blks ACC=0 REJ=0 | Height = 511975 | Diff = 36 0-bits | 00:06:29
60.8 MH/s | 0 Blks ACC=0 REJ=0 | Height = 511975 | Diff = 36 0-bits | 00:06:40
60.9 MH/s | 0 Blks ACC=0 REJ=0 | Height = 511975 | Diff = 36 0-bits | 00:06:51
60.9 MH/s | 0 Blks ACC=0 REJ=0 | Height = 511975 | Diff = 36 0-bits | 00:07:02
61.0 MH/s | 0 Blks ACC=0 REJ=0 | Height = 511975 | Diff = 36 0-bits | 00:07:13
61.0 MH/s | 0 Blks ACC=0 REJ=0 | Height = 511975 | Diff = 36 0-bits | 00:07:24
61.0 MH/s | 0 Blks ACC=0 REJ=0 | Height = 511975 | Diff = 36 0-bits | 00:07:35
61.1 MH/s | 0 Blks ACC=0 REJ=0 | Height = 511975 | Diff = 36 0-bits | 00:07:46
61.1 MH/s | 0 Blks ACC=0 REJ=0 | Height = 511975 | Diff = 36 0-bits | 00:07:57
[MASTER] Coinshield Network: New Block 511976
61.1 MH/s | 0 Blks ACC=0 REJ=0 | Height = 511976 | Diff = 36 0-bits | 00:08:08
61.2 MH/s | 0 Blks ACC=0 REJ=0 | Height = 511976 | Diff = 36 0-bits | 00:08:19
[MASTER] Coinshield Network: New Block 511977

Awesome, I will take a look at it sometime tonight or tomorrow and report back.

Here you go:
https://www.dropbox.com/s/jotw7xgp2re8qwi/Nov-24-2015_17-24-21.zip?dl=0

@Wolf0

Downloaded from: https://github.com/BitSlapper/SKMiner

Apart from fixing a few bugs and adding some debugging code, I’ve only made a one line change to the OpenCL (sk1024_djm.cl) kernel code.

It would take a few hours to optimize the OpenCL kernel (using CodeXL) to reduce the VGPRs usage thereby increasing the occupancy. But since you’ve already done so and your miner is significantly faster, would you mind sharing your code?

Miner version: 1.2 (cleaned up to compile and run on Linux)
OS: Ubuntu 14.04 LTS
Driver: 14.9
GPU: R9 280X (stock)

[MASTER] Coinshield Network: New Block 511517
39.5 MH/s | 0 Blks ACC=0 REJ=0 | Height = 511517 | Diff = 37 0-bits | 00:29:00
39.5 MH/s | 0 Blks ACC=0 REJ=0 | Height = 511517 | Diff = 37 0-bits | 00:29:11
39.6 MH/s | 0 Blks ACC=0 REJ=0 | Height = 511517 | Diff = 37 0-bits | 00:29:22

[Block Nonce = 989773809]
00000000035f7878e09d8e2c442be919182bf0f6caed9b14e2ef1252e40b6925f30bfed7aa6c70\
e444fba536bd560b643886ca5a83fe35d453367bfc1c0c687cee6d815eeab873cfea8acdd2b887\
9560e8b8fa81c1339c1bff614ac7113abdac2085c684582845cfca6728eb848709e7d77411e141\
a6126122709105b5da1cda


******* BLOCK with 38 Leading Zero Bits FOUND *******

             ...Preparing for Block Submission...

Submitting Block...
[MASTER] Block Accepted By Coinshield Network.
[MASTER] Coinshield Network: New Block 511518
39.3 MH/s | 1 Blks ACC=1 REJ=0 | Height = 511518 | Diff = 37 0-bits | 00:29:34
39.3 MH/s | 1 Blks ACC=1 REJ=0 | Height = 511518 | Diff = 37 0-bits | 00:29:45

Ah, this reminds me of a girl at the university who came from a very wealthy family and thought that some rules did not apply to her. There was a reserved parking area which she did not hold a permit for, but she parked there anyway. She would pay her fines and repeat the offense. This went on for quite some time before the university was fed up and made it a point to tow the vehicles of repeate offenders.

The purpose of the reserved lot wasn’t there to bring in additional revenue for the university from those who can afford to violate the rules and pay the fines. It was there to serve the needs of those who were designated to use it.

The same can be applied to Bitcoin; abusing the system because you can afford to pay the fines does not make it OK and should be discouraged.

Benchmark: ASUS Radeon R9 280X R9280X-DC2T-3GD5
Using platform: AMD Accelerated Parallel Processing
Using device: Tahiti(OpenCL 1.2 AMD-APP (1573.4))

GROUP_SIZE = 64

DAG_SIZE = 8388593

Trial 1... 28043946
Trial 2... 27688960
Trial 3... 27688960
Trial 4... 28043946
Trial 5... 27688960

min/mean/max: 27688960/27830954/28043946 H/s
inner mean: 27807288 H/s


Actual mining: ASUS Radeon R9 280X R9280X-DC2T-3GD5
GPU 0: core=1070MHz mem=1600MHz util=99% powertune=-1 fanspeed=70 temp=64C

ethminer  Mining on PoWhash #d217ded5⦠: 27557610 H/s = 911605760 hashes / 33.08 s
ethminer  Mining on PoWhash #d217ded5⦠: 27558745 H/s = 925450240 hashes / 33.581 s
ethminer  Mining on PoWhash #d217ded5⦠: 27559038 H/s = 939294720 hashes / 34.083 s
ethminer  Mining on PoWhash #d217ded5⦠: 27560120 H/s = 953139200 hashes / 34.584 s


Network Difficulty:
125 GH/s

Average Block Time:
15 s or (Avg. 31,000 ETH per day)

R9 280X Hash Rate:
27.5 MH/s


Average ETH per day for an R9 280X at 125 GH/s network difficulty:
(27.5 / 1000 / 125) x 31000 = 6.82 ETH or 6.37 ETH on stock frequencies