miner battle: z-enemy 1.17 vs t-rex 0.6.1 vs spmodgit-9 [x16r] FINISHED

[kerzNOberz]

miner battle: z-enemy 1.16 vs t-rex 0.5.7 vs spmodgit-7 [x16r / ravencoin]


I've been wanting to compare performance between miners on the X16R algorithm for some time. This will be a test done in the vein of JackIt's tests in the past, where I ask which miner puts more coins in my wallet?



Setup: For this first test, I'm using a 3 GPU rig with Zotac 1080 Ti Minis running Windows 10 x64. I started an instance of each miner at the same time (using 1 GPU each) running on the same pool (Ravenminer) and the same machine. The miners will run for ~6 hours, I'll stop them simultaneously, and we'll see how many coins are in the 2 wallets. I will likely run this test 2-3 times to ensure that the results are consistent.

Overclock: 100% TDP, +110 core clock, stock memory


Drivers: 398.82


Miners: Z-enemy 1.16 x64,  CUDA 9.2, STOCK intensity (thread here)  vs.  T-rex 0.5.7 , CUDA 9.2 , STOCK intensity (thread here)  vs.  spmodgit7 (suprminer), STOCK intensity (github here) . z-enemy and t-rex have 1% dev fee, spmod has no dev fee. Running auto (server set) difficulty on both miners. I'll switch which of the 3 GPUs each miner uses for each round to account for any variations in hardware.



Results: Experiment #1: default intensities

ROUND #1: 660 minutes

z-enemy 1.16 : GPU2 : pool wallet : 25.63 RVN [+17.1%]
t-rex 0.5.7 : GPU1 : pool wallet : 25.64 RVN  [+17.2%]
spmodgit7 : GPU0 : pool wallet : 21.88 RVN

ROUND #2: 610 minutes

z-enemy 1.16 : GPU0 : pool wallet : 32.48 RVN [+12.7%]
t-rex 0.5.7 : GPU2 : pool wallet : **HUNG CLIENT**
spmodgit7 : GPU1 : pool wallet : 28.83 RVN

ROUND #3: 650 minutes

z-enemy 1.16 : GPU1 : pool wallet : 25.51 RVN [+26.7%]
t-rex 0.5.7 : GPU0 : pool wallet : 23.51 RVN [+16.7%]
spmodgit7 : GPU2 : pool wallet : 20.14 RVN

ROUND #4: 500 minutes (ended early b/c i clicked on z-enemy, trex/spmod kept running briefly)

z-enemy 1.16 : GPU0 : pool wallet : 25.79 RVN [+32.0%]
t-rex 0.5.7 : GPU2 : pool wallet : 23.44 RVN [+20.0%]
spmodgit7 : GPU1 : pool wallet : 19.54 RVN

Conclusions: Experiment #1

The big takeaway from this experiment is that both z-enemy and trex are dramatically faster than spmod (suprminer) at the default intensities, even with the 1% developer fees. It sounds like spmod may be faster with OC'd memory and a static difficulty, so I may test that eventually. For now, I would reccomend both z-enemy and t-rex over spmod, unless you stand firm on only using open source software.

If we averaged the coin gains I would rank below, my estimation being that z-enemy is 3-5% faster than t-rex with the default settings:

#1: z-enemy 1.16 +22%
#2: t-rex 0.5.7 +18%
#3: spmodgit 7


Results: Experiment #2: comparing t-rex intensities

Here, I plan on testing various t-rex intensities to find the optimal intensity for t-rex (and I will do the same for z-enemy) and then to eventually compare those optimal intensities head to head.

ROUND #1: 840 minutes

t-rex 0.5.7 : default intensity [20] : pool wallet : 24.11 RVN
t-rex 0.5.7 : intensity 21 : pool wallet : 25.44 RVN [+5.5%] ** WINNER **
t-rex 0.5.7 : intensity 23 : pool wallet : 24.90 RVN [+3.3%]

ROUND #2: 600 minutes

t-rex 0.5.7 : intensity 21 : pool wallet : 16.19 RVN
t-rex 0.5.7 : intensity 22 : pool wallet : 16.39 RVN [+1.2%]
t-rex 0.5.7 : intensity 24 : pool wallet : 17.58 RVN [+7.3%] ** WINNER **

Conclusions: Experiment #2

I stopped the test a little early because a new version of t-rex was released (0.6.1). It seems higher than default intensities are fastest for t-rex. I will likely revisit this with the new version, but currently it seems 24 intensity is the fastest on my machine.



Results: Experiment #3: comparing z-enemy intensities

ROUND #1: 720 minutes

z-enemy 1.16 : intensity 20 : pool wallet : 23.21 RVN
z-enemy 1.16 : intensity 21 : pool wallet : 23.27 RVN [+0.3%]
z-enemy 1.16 : intensity 23 : pool wallet : 25.12 RVN [+8.2%] ** WINNER **


Results: Experiment #4: t-rex 0.6.1 vs z-enemy 1.16 [optimal intensities]

ROUND #1:  1440 minutes

Going to let this one run 24 hours per round. From my previous tests intensity 24 is the fastest on my machine for t-rex and 23 intensity for z-enemy. the z-enemy internal tests show intensity 20 is the fastest, so i'll compare that as well

z-enemy 1.16 : intensity 20 : pool wallet : 53.98 RVN [+6.8%]
z-enemy 1.16 : intensity 23 : pool wallet 54.08 RVN [+6.9%] ** WINNER **
t-rex 0.6.1 : intensity 24 : pool wallet : 50.56 RVN

z-enemy is clearly the winner here, the higher intensity doesn't seem to change much. going to run one final 24 hour run and then i'll finalize z-enemy as the winner.

ROUND #2: 2000 minutes

z-enemy 1.16 : intensity 20 : pool wallet : 71.83 RVN  [+1.1%]
t-rex 0.6.1 : intensity 22 : pool wallet : 75.09 RVN [+5.7%] ** WINNER **
t-rex 0.6.1 : intensity 24 : pool wallet : 71.03 RVN

ROUND #3: 1200 minutes

z-enemy 1.16 : intensity 20 : pool wallet : 44.73 RVN
z-enemy 1.17 : intensity 20 : pool wallet : 49.63 RVN [+11.0%] ** WINNER **
t-rex 0.6.1 : intensity 22  pool wallet  : 45.10 RVN [+0.8%]

ROUND #4: 500 minutes

spmodgit 9 : intensity 24 : pool wallet : 26.40 RVN
z-enemy 1.17 : intensity 20 : pool wallet : 36.86 RVN [+39.6%] ** WINNER **
t-rex 0.6.1 : intensity 22  pool wallet : 35.45 RVN [+34.3%]


Final Conclusions

1. z-enemy 1.17 is the fastest x16r miner in my tests. While a recent entrant into the x16r mining field, t-rex, was an impressive foe, z-enemy consistently won out. The latest version of z-enemy, 1.17, is faster than the previous version, 1.16, by a small amount. It's worth upgrading if you're running 1.16 or earlier.

2. intensity matter and is unique to each miner. While z-enemy appears to run fastest on the default intensity, 20, t-rex performance improves with raised intensity. In my tests, the optimal intensity for t-rex was ~22.

3. open source x16r miners have a ways to go. Even with the 1% dev fees, both z-enemy and t-rex were signifncantly faster (15%-30% faster) than spmodgit. Apparently spmodgit can improve in performance with increased memory speed, but I didnt get a chance to test performance under those conditions.



Winner: Faster X16R Miner: z-enemy 1.17 ( thread here)

[1714047600]

1714047600

[1714047600]

1714047600

[1714047600]

1714047600

[1714047600]

1714047600

[kerzNOberz]

suprminer starting (appears it sets stock intensity @ 24):



trex starting (appears it sets stock intensity @ 20):



z-enemy starting (appears it sets stock intensity @ 20):

[GeePeeU]

Glad to see someone doing this. There's been a lot of new X16r miners recently.

So far looks like T-rex is winning, and SP/Zenemy are about even.

I use SPmod because its open source.

[kerzNOberz]

round #1 is done, z-enemy and t-rex almost dead even (and both way in front of spmodgit7).

round #2 begins now.

[abudfv2008]

Of cause its better than nothing. But 1gpu and 10hours - at least 10-15% is luck difference.

[Xazax310]

I was running SPmod with my 13 GPU rigs and ran into many crashes possible because of high CPU usage. Using Z-Enemy 1.16 now 1 day+ no crashes.

[kerzNOberz]

Of cause its better than nothing. But 1gpu and 10hours - at least 10-15% is luck difference.

Luck is not a factor in these tests, all 3 miners are on the same pool mining identical blocks — their luck is the exact same .

[JW_B]

awesome try!

I think dev fee switching is the factor.

should do test longer period.

[nsummy]

awesome try!

I think dev fee switching is the factor.

should do test longer period.

How is that a factor?  The 2 miners with dev fee are outperforming the free miner.

[nsummy]

bitcointalk needs more posts like this.  Scientific and to the point.  If possible would you be able to include power usage?  I know some miners outperform others but sometimes its at a significant wattage increase.

[kerzNOberz]

t-rex hung in ROUND 2, unclear why, but I won't post a result for t-rex for this round (and I'll add a 4th round to get adequate data). z-enemy was in the lead (by ~1%) before the crash.

this is what the client looked like on the machine:

[kerzNOberz]

bitcointalk needs more posts like this.  Scientific and to the point.  If possible would you be able to include power usage?  I know some miners outperform others but sometimes its at a significant wattage increase.

excellent question, i'll look into this

[abudfv2008]

Of cause its better than nothing. But 1gpu and 10hours - at least 10-15% is luck difference.

Luck is not a factor in these tests, all 3 miners are on the same pool mining identical blocks — their luck is the exact same .

Are you serious? You can run 3 instances of the same miner on these gpus and easily get 10% difference in shares.

[Bananana]

x16r algo is too random, I have notice my hashrate jump from 15 to 25 depending on luck.

[kerzNOberz]

Of cause its better than nothing. But 1gpu and 10hours - at least 10-15% is luck difference.

Luck is not a factor in these tests, all 3 miners are on the same pool mining identical blocks — their luck is the exact same .

Are you serious? You can run 3 instances of the same miner on these gpus and easily get 10% difference in shares.

This is randomness of the Algo, not luck of hitting blocks. This randomness *should* even out over time. This is wha I am doing many rounds and will only make conclusions based on a large dataset. If one miner is always 10%+ ahead of another, will you really argue this is “luck” and not a real advantage of hat miner?

If it were just just luck/stochasticity, wouldn’t you expect not to see the same winner each time?

Also, share number is completely meaningless. It can be manipulated by changing share difficulty in command line.

[kerzNOberz]

Anyway, there is no point in arguing with you all, I’ll post the results and people can (or not) take the results for what they are.

[dragonmike]

The reason trex hung is because you fiddled with the mouse and randomly selected something in the mining window (look at your cursor selection in the upper right corner). Believe it or not, this actually freezes the miner.

[sp_]

You need to add +500 on the memclock or more. spmod-git is using memory in the simd implementation. Simd is one of the slowest algos in the chain. You also need to run with a fixed difficulty. After a fast sequence, the difficulty can rise alot, and if yout get a slow sequence, the miner might not find any solutions, then timeout, and the pool-speed reset to 0.

[kerzNOberz]

round 3 complete, onto round 4

[Jounouchi]

t-rex hung in ROUND 2, unclear why, but I won't post a result for t-rex for this round (and I'll add a 4th round to get adequate data). z-enemy was in the lead (by ~1%) before the crash.

this is what the client looked like on the machine:

https://i.imgur.com/T75ZW8L.png

It is because you have click on the blank screen, see your windows title bar showing select.