50 days left to claim 3x$500 in bounties for Cuckoo Cycle proof-of-work

[tromp]

(erroneously posted to Marketplace yesterday)

Since mid-2014,
  https://github.com/tromp/cuckoo
has offered bounties for disproving that Cuckoo Cycle,
myProof-of-Work that spends most time on memory latency,
- has an optimal reference miner
- has no feasible memory-time trade-off
- offers no advantage for GPUs

These bounties are set to expire at the end of this year,
so get coding for a modest Xmas bonus!

[MaxDZ8]

I've found the bounties irrelevant some time ago and I find them irrelevant now, just FYI.

What truly pisses me off about them is:

• you unilaterally set conditions about testing performance target. This is not how it works. I speculate you're not truly interested in being informative or you would offer bounties to investigate scaling.
• Just getting the hardware you specified mandates I waste most of the prize. FYI, an hi-end i7 is ~450USD here implying you only accept candidates who already got the hardware... let me elaborate: in my experience, most ppl buying anything over i5 does not even know what a GPU is, let alone how to use it. CPUs end at 100 bucks for me. OFC, to compare to a 450 bucks CPU one would need at least a 300 bucks GPU for proper comparison...
• you haven't got anyone to adopt your algo, thereby people with the hardware will most likely not find it very interesting (I sure don't).

As a side note, I was asked to provide an Ethereum miner. I didn't find that interesting either as I think requiring such amounts of memory is unacceptable for consumers - sure it is acceptable for the elitist 'core miner' mindset around here. No idea how much memory you need... I guess this a good indicator of how much your offers are interesting me.

My experience in graphics seems to collide with your statements. GPUs have extended 'hyperthreading-like' capabilities and natively out-of-order memory... oh I see you CPU guys got that as well at least... Leaving aside that some of the properties you mention can be proven theoretically as you already have an implementation you claim to be 'about as fast' I honestly wouldn't be so confident.

But just to be clear, I still take some time to inform you. After all, your job is original and you're trying to move things forward.

[tromp]

you unilaterally set conditions about testing performance target.

That's how bounties work. I offer some money for meeting some goal of my choice.

Just getting the hardware you specified mandates I waste most of the prize.

You don't need to buy any hardware. You only need to code an improvement,
and submit it as a candidate. I'll test it on an i7 and let you know if it meets the criteria.
Even if you don't meet the criteria but still make a substantial improvement,
I may offer a partial reward.

you haven't got anyone to adopt your algo

The whole point of these bounties is to make miner improvements happen before rather than after adoption, so that mining will be fair from the start.

[tromp]

These bounties are set to expire at the end of this year,
so get coding for a modest Xmas bonus!

Someone informed me they have achieved 6x speedup with new GPU code,
and is preparing to submit the code for verification. Stay tuned...

[tromp]

Someone informed me they have achieved 6x speedup with new GPU code,
and is preparing to submit the code for verification. Stay tuned...

I verified the (GPU over CPU) speedup as 2.05x, so the $500 bounty is claimed.

At least the money stays in Dutch hands:-)

The improvement will be on github in a few days,
while the bounty winner tries to squeeze out some more performance.

[Genoil]

Someone informed me they have achieved 6x speedup with new GPU code,
and is preparing to submit the code for verification. Stay tuned...

I verified the (GPU over CPU) speedup as 2.05x, so the $500 bounty is claimed.

At least the money stays in Dutch hands:-)

The improvement will be on github in a few days,
while the bounty winner tries to squeeze out some more performance.

To clarify a bit...the ~6x speedup was in the GPU kernel itself. If I'd known it was only 2.05, I'd made sure to add some more speed before submitting . "With the heels over the ditch" we call that here.

I have just achieved another 9% improvement over my submitted work.

[tromp]

Someone informed me they have achieved 6x speedup with new GPU code,
and is preparing to submit the code for verification. Stay tuned...

I verified the (GPU over CPU) speedup as 2.05x, so the $500 bounty is claimed.

At least the money stays in Dutch hands:-)

The improvement will be on github in a few days,
while the bounty winner tries to squeeze out some more performance.

The new and improved cuda_miner.cu is now up on https://github.com/tromp/cuckoo
I also link to Genoil's https://github.com/Genoil/cuckoo as the Cuckoo Cycle repo of choice for Windows developers.

I'd like to collect Cuckoo Cycle performance data on various graphics cards.
I propose to use the wall-clock running time of a single proof attempt:

$ make cuda30
nvcc -o cuda30 -DSIZESHIFT=30 -arch sm_35 cuda_miner.cu -lssl -lcrypto
$ time ./cuda30 -h 0 -n 7 -t 16384
Looking for 42-cycle on cuckoo30("0") with 50% edges, 7 trims, 16384 threads
Using 64MB edge and 128MB node memory.
final load 48%
  58-cycle found at 0:98%
  60-cycle found at 0:99%
 426-cycle found at 0:99%

real    0m8.100s
user    0m7.515s
sys     0m0.576s

That 8.1 seconds was on a

$ ./cuda_query
Device Number: 0
  Device name: GeForce GTX 980
  Memory Clock Rate (KHz): 3505000
  Memory Bus Width (bits): 256
  Peak Memory Bandwidth (GB/s): 224.320000

Please post times for your favorite GPU,
and I can add a little table to the project page.

Information on power use and temperatures are also welcome.
For the latter, you may prefer a longer running version like

$ make cuda32
nvcc -o cuda32 -DSIZESHIFT=32 -arch sm_35 cuda_miner.cu -lssl -lcrypto
$ time ./cuda32 -h 0 -n 7 -t 16384

or a loop like

$ time for i in {0..99}; do ./cuda30 -h $i -n 7 -t 16384; done