Looking for some suggestions.
Would anyone have any idea why my windows asus 1070ti rig is rock solid stable on V0.2 cuda9 ,,,,,,,,,,,,but "crashes within 10 minutes on V0.3 or V0.4 cuda 8 or 9 using same settings"
currently V0.2
68 power limit
205 core
650 memory
approximately 225 sol from 6 cards
Thanks
Topic: EWBF's Cuda Equihash Miner 0.6 - page 21. (Read 85776 times)
So far single cards work best. Going to try running separate batch files for each card over night (6 hours approx) and see how it fairs. Seems to be a strong miner for nivida though when it's hashing.
Can you confirm if this works? Is the CPU usage higher with multiple batch files?
Short answer is yes and yes.
CPU usage % fluctuates dramatically 11- 100% given any other applications being used, windows open, broswers, etc. since I'm running video off of the cpu as well. Plus side is I can trouble shoot the cards giving/causing errors, keeping uptime on the other cards hashing perfectly. Since this is a dedicated mining rig and uses only a 960i, I'm not CPU mining with it either. Works pretty well for the application.
Place in the batch file " --cuda_devices X" in reference to each GPU as EWBF's miner relates to it, "0" being GPU 1, etc.
I'm having flawless run time so far, I"ll let run though the day today and see if some slight down clocks on a couple cards will keep from crashing and see if I can up clocks on those that aren't. Once I feel stable I'll try and run a single batch file again for all cards.
So far single cards work best. Going to try running separate batch files for each card over night (6 hours approx) and see how it fairs. Seems to be a strong miner for nivida though when it's hashing.
Can you confirm if this works? Is the CPU usage higher with multiple batch files?
Well... it works for my "palit jetstream"s and "evga ftw"s but makes unstable the rest ones (KFA2 HoF, ASUS Dual, ASUS Strix, Gigabyte F1, Zotac AMP! — they all start to crash miner if there's just PL set to 60% and voltage isn't locked at individually chosen value).
And their "60% PL" is aaaaaaaaaaall different... For example KFA2's 60% PL is 170+Watt (and more than 300Watt at 120%
...it obviously can't really "eat" so much but manufacturer setting are that strange!), at the same time ASUS Dual at 60% PL can't reach 1600MHz (!) even with down-volting. That's why I chose my way.
Thanks for sharing your experience anyway
And their "60% PL" is aaaaaaaaaaall different... For example KFA2's 60% PL is 170+Watt (and more than 300Watt at 120%
...it obviously can't really "eat" so much but manufacturer setting are that strange!), at the same time ASUS Dual at 60% PL can't reach 1600MHz (!) even with down-volting. That's why I chose my way.Thanks for sharing your experience anyway
Why everybody overclocking MEM on 144 one? It works with deafult equihash but it seems that "144" mode is MEM-independent...
My 1070th work at 2025MHz/950mV (some of them need 975mV, and 2 of 15 need above 1000mV) with -200 on memory and give about 36~38 each.
1070ti (bought one recently) with same settings gives 40~41.
1080s (not "ti") work at 2050MHz/1025mV (and -200 on MEM) give about 43~44 each.
Why overclock memory then?..
My 1070th work at 2025MHz/950mV (some of them need 975mV, and 2 of 15 need above 1000mV) with -200 on memory and give about 36~38 each.
1070ti (bought one recently) with same settings gives 40~41.
1080s (not "ti") work at 2050MHz/1025mV (and -200 on MEM) give about 43~44 each.
Why overclock memory then?..
My clocks at 100% power will yield close to 200 sol with 4 1070 ti's but are more efficient with the power drop.
I find overclocking easiest setting core first, mem second, and then slowly dropping power.
I'm getting the same hash with mem OC's at 60% power.
Wow! 0.03$/kWh! How did you manage that?
Live in Siberia... I lock voltage so there's no point for me to use PL, 'cause it does almost the same thing (limits top voltage of your f/v curve but leaves it unstable on under limit levels). So I prefer voltage locking instead of PL.
At "144" mode average S/W of my GPUs are about 0.27~0.32. Not the best maybe, but electricity costs only 0.03$/kWh for me... almost doesn't matter.
Only one of all my GPUs get higher than 65 Celsius (lucky me! this summer leaves RIG room's temperature slightly below 35 Celsius — when all GPUs turned off).
My hottest ones (right now, at night, when every card is 3~4C cooler): 1070 HoF KFA2 (61C), 1070 ASUS Dual (60C) and 1070 ZOTAC Amp! edition(63C). But they all need just a thermal grease change I think.
P.S.
At first my mem was like +200 ... +400 (+600 on palit jetstream) and temps of all GPUs were much higher.
At "144" mode average S/W of my GPUs are about 0.27~0.32. Not the best maybe, but electricity costs only 0.03$/kWh for me... almost doesn't matter.
Only one of all my GPUs get higher than 65 Celsius (lucky me! this summer leaves RIG room's temperature slightly below 35 Celsius — when all GPUs turned off).
My hottest ones (right now, at night, when every card is 3~4C cooler): 1070 HoF KFA2 (61C), 1070 ASUS Dual (60C) and 1070 ZOTAC Amp! edition(63C). But they all need just a thermal grease change I think.
P.S.
At first my mem was like +200 ... +400 (+600 on palit jetstream) and temps of all GPUs were much higher.
Wow! 0.03$/kWh! How did you manage that?
I lock voltage so there's no point for me to use PL, 'cause it does almost the same thing (limits top voltage of your f/v curve but leaves it unstable on under limit levels). So I prefer voltage locking instead of PL.
At "144" mode average S/W of my GPUs are about 0.27~0.32. Not the best maybe, but electricity costs only 0.03$/kWh for me... almost doesn't matter.
Only one of all my GPUs get higher than 65 Celsius (lucky me! this summer leaves RIG room's temperature slightly below 35 Celsius — when all GPUs turned off).
My hottest ones (right now, at night, when every card is 3~4C cooler): 1070 HoF KFA2 (61C), 1070 ASUS Dual (60C) and 1070 ZOTAC Amp! edition(63C). But they all need just a thermal grease change I think.
P.S.
At first my mem was like +200 ... +400 (+600 on palit jetstream) and temps of all GPUs were much higher.
upd: I cleaned my RIG
all GPUs are lower than 60C now
except that asus (I bet thermal grease change will change it))
At "144" mode average S/W of my GPUs are about 0.27~0.32. Not the best maybe, but electricity costs only 0.03$/kWh for me... almost doesn't matter.
Only one of all my GPUs get higher than 65 Celsius (lucky me! this summer leaves RIG room's temperature slightly below 35 Celsius — when all GPUs turned off).
My hottest ones (right now, at night, when every card is 3~4C cooler): 1070 HoF KFA2 (61C), 1070 ASUS Dual (60C) and 1070 ZOTAC Amp! edition(63C). But they all need just a thermal grease change I think.
P.S.
At first my mem was like +200 ... +400 (+600 on palit jetstream) and temps of all GPUs were much higher.
upd: I cleaned my RIG
all GPUs are lower than 60C now
except that asus (I bet thermal grease change will change it))
Why everybody overclocking MEM on 144 one? It works with deafult equihash but it seems that "144" mode is MEM-independent...
My 1070th work at 2025MHz/950mV (some of them need 975mV, and 2 of 15 need above 1000mV) with -200 on memory and give about 36~38 each.
1070ti (bought one recently) with same settings gives 40~41.
1080s (not "ti") work at 2050MHz/1025mV (and -200 on MEM) give about 43~44 each.
Why overclock memory then?..
My 1070th work at 2025MHz/950mV (some of them need 975mV, and 2 of 15 need above 1000mV) with -200 on memory and give about 36~38 each.
1070ti (bought one recently) with same settings gives 40~41.
1080s (not "ti") work at 2050MHz/1025mV (and -200 on MEM) give about 43~44 each.
Why overclock memory then?..
What is your power setting (watts/TDP) for those numbers? Do you monitor Sols/W?
I have found at higher power settings, memory O/C has little affect on Sol/s.
However when you lower power settings, memory O/C then becomes somewhat more effective.
It is hot summer here so I am running @104 watts (58% TDP) to keep the GPU temps below 63c without running AC in the room.
I usually adjust power first, core second and then memory last.
Give me Readme.txt please!
Why everybody overclocking MEM on 144 one? It works with deafult equihash but it seems that "144" mode is MEM-independent...
My 1070th work at 2025MHz/950mV (some of them need 975mV, and 2 of 15 need above 1000mV) with -200 on memory and give about 36~38 each.
1070ti (bought one recently) with same settings gives 40~41.
1080s (not "ti") work at 2050MHz/1025mV (and -200 on MEM) give about 43~44 each.
Why overclock memory then?..
My 1070th work at 2025MHz/950mV (some of them need 975mV, and 2 of 15 need above 1000mV) with -200 on memory and give about 36~38 each.
1070ti (bought one recently) with same settings gives 40~41.
1080s (not "ti") work at 2050MHz/1025mV (and -200 on MEM) give about 43~44 each.
Why overclock memory then?..
Finally got a batch file to run, and run stable with four 1070's. Hashing all night about 5 hours.
So far my clock are the same as posted before. 1080 ti is maxed I feel with 60 pwr / 150 core / 780 mem, haven't checked the draw but it's hopefully around 150w estimated.
I feel I can get more out the 1070's once stability is reached on my end.
(Note: one is a hobby rig and the other gaming desktop. Not much time for gaming lately .. .)
So far my clock are the same as posted before. 1080 ti is maxed I feel with 60 pwr / 150 core / 780 mem, haven't checked the draw but it's hopefully around 150w estimated.
I feel I can get more out the 1070's once stability is reached on my end.
(Note: one is a hobby rig and the other gaming desktop. Not much time for gaming lately .. .)
0.4 works fine and stable for me on 3 PCs. Algo is Equihash 144.5
6x 1060 - 3 & 6 GB mixed (Power 68, Core +120, Mem +500) - Win 10 - Nvidia 398.36
1x 1080 Ti - OC-Version +100 Core default! (Power 65, Core +70, Mem +600) - Win 10 - Nvidia 398.36
2x 1080 (Power 65, Core +120, Mem +500) - Win 10 - Nvidia 398.36
6x 1060 - 3 & 6 GB mixed (Power 68, Core +120, Mem +500) - Win 10 - Nvidia 398.36
1x 1080 Ti - OC-Version +100 Core default! (Power 65, Core +70, Mem +600) - Win 10 - Nvidia 398.36
2x 1080 (Power 65, Core +120, Mem +500) - Win 10 - Nvidia 398.36
So far single cards work best. Going to try running separate batch files for each card over night (6 hours approx) and see how it fairs. Seems to be a strong miner for nivida though when it's hashing.
Yes, winx here and latest drivers. Been looking at a roll back but been super busy with LN and BTCpay . And remote connection is helping uptime currently.
Yes, this is a common problem, on the miner's side I will do everything possible to fix it.
Yea, this the error I'm getting as well. 2 crashes of this type in a day and a half (second one over night).
FWIW, I am not seeing this nor am I receiving any reports from my Linux users in ZhashOS using the cuda 8 binary.
So this might be a cuda 9.1 and/or windows issue.
Yes, this is a common problem, on the miner's side I will do everything possible to fix it.
Yea, this the error I'm getting as well. 2 crashes of this type in a day and a half (second one over night).
@EWBF do you monitor: https://github.com/nanopool/ewbf-miner/issues
Or is it best to post bugs/issues/suggestions within your forum topics?
Thanks!
Or is it best to post bugs/issues/suggestions within your forum topics?
Thanks!
please work your maigc for a fully functional amd miner....
I thought about it, but right now I do not have time for it, maybe later.
Jump to: