Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 333. (Read 2347664 times)
I've never seen that alexis bug, probably it's windows only. Blame Microsoft.
There is no bug with alexis78 miner. You can start all 7 gpus at the start with a simple command. Had no issue on all of my nvidia rigs ( 6 - 7 gpus).
Here is a snapshot of the alexisminer bug:
13:45:12 start the miner (6 card celeron rig)
4 gpu thread's is starting (3 seconds delay) and produce almost nothing, and the pool is reducing the difficulty to get a response.
13:48:31 The 5th gpu thread is starting. (2.5 minutes later)
The 6th gpu thread was enabled after 5 minutes of mining.
Maybe, sp, you should invest your time into creating better code base instead of reusing old crappy ccminer? 
Why don't you opensource the Excavator?
Why would we? So you can copy paste it and resell? ^^
There is no bug with alexis78 miner. You can start all 7 gpus at the start with a simple command. Had no issue on all of my nvidia rigs ( 6 - 7 gpus).
Maybe, sp, you should invest your time into creating better code base instead of reusing old crappy ccminer?
Why don't you opensource the Excavator?
I dont get why you're so mean to my miner though... 
You want me to push any fixes and then have you around calling it "shitty gpl code" while selling my fork in the same time?
You want me to push any fixes and then have you around calling it "shitty gpl code" while selling my fork in the same time?
You fork others work and include new bugs in the code that makes the miner unstable and unusable. It's been 8 months since your last commit to github, and still you waste time in my thread and complain. My fork doesn't have this bug.
Sentences like that really makes you look dumb, sp. Hashing is not like accelerating a car. When hashing starts, it is at full speed from moment one, but some speed measuring mechanisms have a buffer and so it looks as if speed is accelerating.
Try to run the alexis's miner on a 2 core celeron. It sometimes take 2 minutes to start all the 6 gpu threads. If you combine this buggy miner with a profit switcher software that switch between coins every 5 minutes you end up earning very little. The first 30 seconds is never mining with 6 gpu's. (Nemos miner)
Well, Excavator does NOT have this issue at all and can switch from one algorithm to another instantly. And guess why? Because the most of the time spent when launching a miner is cuda context initialization. And if that is done only once, at start of mining, then there are no mining looses during switches. And from next NHM version, there are only two popular algorithms that aren't coevered by Excavator (dagger and lyra2rev2), but we are getting to that part soon.
Maybe, sp, you should invest your time into creating better code base instead of reusing old crappy ccminer?
Sentences like that really makes you look dumb, sp. Hashing is not like accelerating a car. When hashing starts, it is at full speed from moment one, but some speed measuring mechanisms have a buffer and so it looks as if speed is accelerating.
Try to run the alexis's miner on a 2 core celeron. It sometimes take 2 minutes to start all the 6 gpu threads. If you combine this buggy miner with a profit switcher software that switch between coins every 5 minutes you end up earning very little. The first 30 seconds is never mining with 6 gpu's. (Nemos miner)
Imagine what happens if you add a cheap bad internet connection too...
I dont get why you're so mean to my miner though...
You want me to push any fixes and then have you around calling it "shitty gpl code" while selling my fork in the same time?
Sentences like that really makes you look dumb, sp. Hashing is not like accelerating a car. When hashing starts, it is at full speed from moment one, but some speed measuring mechanisms have a buffer and so it looks as if speed is accelerating.
Try to run the alexis's miner on a 2 core celeron. It sometimes take 2 minutes to start all the 6 gpu threads. If you combine this buggy miner with a profit switcher software that switch between coins every 5 minutes you end up earning very little. The first 30 seconds is never mining with 6 gpu's. (Nemos miner)
Of course. And 1080 should do more, at least 35. And 1080 Ti should do 50+. If you check 1080 Ti, you can see it reaches 35, but memload is only 75% and you cannot change that no matter OC. There is something seriously NOT adjusted for GDDR5X.
It is all about latency and I'm very doubt that you can mask it in software when we talk about dagger-hashimoto.Polaris cards with stock bios are also showin < 100% mem load and poor performance until better timings resulting lower latency are applied. It is not software - but hardware feature.
Big latency can be properly hidden with correct programming approach. The current one used by Genoil's CUDA kernel is obviously not the correct one. At the end, big mem bandwidth is more important when proper coding is applied.
constant switching and miners that need 2 minutes to run at full speed
Sentences like that really makes you look dumb, sp. Hashing is not like accelerating a car. When hashing starts, it is at full speed from moment one, but some speed measuring mechanisms have a buffer and so it looks as if speed is accelerating.
with Zen you are making NOTHING... since wallet at trex is offline.
Speaking of daily profits when you cant trade the coin is nonsense.
Speaking of daily profits when you cant trade the coin is nonsense.
Zen up... Holy shit finally making remotely decent profits with mainstream cards. Still, comparatively speaking, a 1070 should be earning a $18 a day.
amd was lucky that ETH is still increasing so much, still i would choose nvidia because of better option, when ETH will go pos or diff boom there will be no more profit for amd
Zen up... Holy shit finally making remotely decent profits with mainstream cards. Still, comparatively speaking, a 1070 should be earning a $18 a day.
now with zencash about 9$
True words djeZo..have you improved it like zawawa fpr excavator? I was missing any word about the CUDA version ..but instead only read about 10% on the opencl one.
I achieve with 1.1.7 around 38MH/s on y 1080Ti ..so not that bad anyway.
I achieve with 1.1.7 around 38MH/s on y 1080Ti ..so not that bad anyway.
38MHASH is slow when the gtx 1070 can do 36 MHASH (genoils miner samsung mem oc/ windows 7). If you want to mine ethereum with the 1080ti you should dual mine siacoin with the claymore miner and use a high -dcri in the bat file like -dcri 128 or -dcri 256
Boah you talking shit because you compare GDDR5 to GDDR5x that many said are so bad. I just wanted to show that GDDR5X isn't that bad and since Ethash hasn't been optimized for long time its better than nothing.
Profitable because zpool hasn't added it yet. The beginners (nicehashminer/nemos miners) seems to be happy with making less than $4 a day. Since the profit switching isn't working properly (constant switching and miners that need 2 minutes to run at full speed), and because the pools steal 20-30% of the profit they end up with $2-3 a card.
Zen up... Holy shit finally making remotely decent profits with mainstream cards. Still, comparatively speaking, a 1070 should be earning a $18 a day.
i ahd that result when i ran 3x gainward 1070 1 year ago... samsung memory. + 1000 mhz on mem
Even with +1000 you aren't going to get it and unless you're extreeeeeemely lucky it wont be stable either.
i ahd that result when i ran 3x gainward 1070 1 year ago... samsung memory. + 1000 mhz on mem
Never seen more then 33.3 mhs eth even with my record-breaking palit sjs samsung running 2200/2400 gpu/mem. Any proof about 36mhs with genoil, sp?
True words djeZo..have you improved it like zawawa fpr excavator? I was missing any word about the CUDA version ..but instead only read about 10% on the opencl one.
I achieve with 1.1.7 around 38MH/s on y 1080Ti ..so not that bad anyway.
I achieve with 1.1.7 around 38MH/s on y 1080Ti ..so not that bad anyway.
38MHASH is slow when the gtx 1070 can do 36 MHASH (genoils miner samsung mem oc/ windows 7). If you want to mine ethereum with the 1080ti you should dual mine siacoin with the claymore miner and use a high -dcri in the bat file like -dcri 128 or -dcri 256
Yup, the only way Nvidia can compete in the Eth arena is with dual mining, but since ASM upgrades, AMD does so much better then Nvidia even at that. Claymore will probably port them in six months when AMD has had it's fill or there is something better to mine.
Also not sure where users are getting these insanely high Eth mining numbers for 1070. 1070 will get about 29Mhs with tweaked memory (+400), which is why you dual mine. And no, of course 1080 and 1080ti don't have their memory fully utilized because that's not what's holding back the performance (it's not the raw bandwidth, it's the latency).
Jump to: