Pages:
Author

Topic: WildRig Multi 0.35.1 beta 2 multi-algo miner for AMD & NVIDIA - page 31. (Read 92440 times)

legendary
Activity: 1797
Merit: 1028



Try 1150/1750, so memory will be stock. No sense to reduce core(core can provide wrong results as I remember only on Vega).

DEFAULT/STOCK IS 1350/2000--

I tried 1150/1750 and the hash rate dropped to 1.6MH/s per GPU, ~10MH/s for the rig.  After changing the configuration to 1150/2000 and rebooting, the rig seemed to behave well.  While I was outside it was mining ok at 1100/2000 clocks, with 345 accepts and 6 rejects at 11.5MH/s for the rig.

Currently, the rig shows 100 accepts with 1 reject at 1150/2000, and the rig is hashing at 11.75MH/s.  It has been mining for about 40 minutes.  Maybe it can be pushed harder somehow, but it seems ok.  The wattage per card is less than 90W in the console report.       --scryptr
Can you press p(this will pause mining) and then r(resume mining) when you get rejects again, and check will it reduce new rejects for next hour? If so, probably I found a problem.

I WILL TRY THAT--

Just getting my coffee today, and taking a look at the rig.  Generally, I am seeing more rejects than with WildRig 10.5.  I will download the latest release and work with it rather than my currently installed v16.0 WildRig-Multi.

Bottom line: I get higher hashrates but produce less BoolBerry since using WildRig-Multi.       --scryptr

PAUSE AND RESUME DO NOT HELP--

I am connected by ssh and my rig is headless.  The use of Pause (p) and Resume (r) does not seem to work.  The miner keeps mining, and rejects are not lessened.    I am gettig at least 10% rejects.

If  I start with only "auto" sttings, the hash rate is about 11MH/s, and less than the stable 11.4MH/s when running WildRig 10.5.  With clocks at 1100/2000 and intensity at 21x0, the hash rate is 11.5MH/s with v16.2, but the rejects are a problem (10%+).  I have not run the miner at just "auto" and default intensity for a long period, I will do so overnight just to get a baseline.

--scryptr
jr. member
Activity: 225
Merit: 1
Got it stable on beenode. 1380 core 930mv 41mhs.
Have no idea if its good or not but doesn't look too bad profit wise
My rx550s get 4.5mhs so I will keep them on XMR instead
jr. member
Activity: 225
Merit: 1
Any tips on where to start with vegas on honeycomb. I have tried different clocks on memory and core but crash all the time. Might be the auto intensity maybe.
This algo needs stable core, so check your voltage. As mentioned in OP I use 1000mV for my Vega 64. Also you can try to decrease the intensity.
Thanks that helps. I lowered intensity so I could get it to run overnight. I will play with voltage now
legendary
Activity: 1797
Merit: 1028



Try 1150/1750, so memory will be stock. No sense to reduce core(core can provide wrong results as I remember only on Vega).

DEFAULT/STOCK IS 1350/2000--

I tried 1150/1750 and the hash rate dropped to 1.6MH/s per GPU, ~10MH/s for the rig.  After changing the configuration to 1150/2000 and rebooting, the rig seemed to behave well.  While I was outside it was mining ok at 1100/2000 clocks, with 345 accepts and 6 rejects at 11.5MH/s for the rig.

Currently, the rig shows 100 accepts with 1 reject at 1150/2000, and the rig is hashing at 11.75MH/s.  It has been mining for about 40 minutes.  Maybe it can be pushed harder somehow, but it seems ok.  The wattage per card is less than 90W in the console report.       --scryptr
Can you press p(this will pause mining) and then r(resume mining) when you get rejects again, and check will it reduce new rejects for next hour? If so, probably I found a problem.

I WILL TRY THAT--

Just getting my coffee today, and taking a look at the rig.  Generally, I am seeing more rejects than with WildRig 10.5.  I will download the latest release and work with it rather than my currently installed v16.0 WildRig-Multi.

Bottom line: I get higher hashrates but produce less BoolBerry since using WildRig-Multi.       --scryptr
member
Activity: 725
Merit: 49



Try 1150/1750, so memory will be stock. No sense to reduce core(core can provide wrong results as I remember only on Vega).

DEFAULT/STOCK IS 1350/2000--

I tried 1150/1750 and the hash rate dropped to 1.6MH/s per GPU, ~10MH/s for the rig.  After changing the configuration to 1150/2000 and rebooting, the rig seemed to behave well.  While I was outside it was mining ok at 1100/2000 clocks, with 345 accepts and 6 rejects at 11.5MH/s for the rig.

Currently, the rig shows 100 accepts with 1 reject at 1150/2000, and the rig is hashing at 11.75MH/s.  It has been mining for about 40 minutes.  Maybe it can be pushed harder somehow, but it seems ok.  The wattage per card is less than 90W in the console report.       --scryptr
Can you press p(this will pause mining) and then r(resume mining) when you get rejects again, and check will it reduce new rejects for next hour? If so, probably I found a problem.
member
Activity: 725
Merit: 49
Any tips on where to start with vegas on honeycomb. I have tried different clocks on memory and core but crash all the time. Might be the auto intensity maybe.
This algo needs stable core, so check your voltage. As mentioned in OP I use 1000mV for my Vega 64. Also you can try to decrease the intensity.
jr. member
Activity: 225
Merit: 1
Any tips on where to start with vegas on honeycomb. I have tried different clocks on memory and core but crash all the time. Might be the auto intensity maybe.
legendary
Activity: 1797
Merit: 1028



Try 1150/1750, so memory will be stock. No sense to reduce core(core can provide wrong results as I remember only on Vega).

DEFAULT/STOCK IS 1350/2000--

I tried 1150/1750 and the hash rate dropped to 1.6MH/s per GPU, ~10MH/s for the rig.  After changing the configuration to 1150/2000 and rebooting, the rig seemed to behave well.  While I was outside it was mining ok at 1100/2000 clocks, with 345 accepts and 6 rejects at 11.5MH/s for the rig.

Currently, the rig shows 100 accepts with 1 reject at 1150/2000, and the rig is hashing at 11.75MH/s.  It has been mining for about 40 minutes.  Maybe it can be pushed harder somehow, but it seems ok.  The wattage per card is less than 90W in the console report.       --scryptr
member
Activity: 725
Merit: 49
HIGH REJECT RATE IN BOOLBERRY--

I am running the new  (v16.0) WildRig-Multi on Linux, and mining BoolBerry.  Up until yesterday, I was using WildRig v10.5.  However, the new version runs about 10% rejects!

Any suggestions?  My cards are RX580, 8GB.       --scryptr
What pool? Did you try re-download scratchpad? How many cards?

LUCKYPOOL, NO, 6 CARDS--

The scratchpad has been updating successfully, I did not change it from the pad left by WildRig v10.5.  I could "prime the pump", and restart after a wget download, I suppose.  The cards are running faster on the v16.0 release, near 2.0MH/s per card.

I am running on ethOS, and the rig reports 12MH/s (rounded).    Previously, it was running at 11.4MH/s.  I have the intensity at 22x256, and have tried lowering the intensity with neutral results.  The reject rate is not lessened, but the hashrate is lower.

Any suggestions are welcome, and thanks for adding BoolBerry!       --scryptr

EDIT:  After downloading a fresh scratchpad, and no other changes, the rejects are still present.       --scryptr


hmm, what if try --opencl-launch auto? Also how that rejects appear. Between accepted shares, or just starting to be rejected until new job? I was running my small rig(4xRX460) for more than 3 days and no problem %)

LOWERING CLOCKS--

I lowered my clocks to 1100/2000 (core/mem) from 1150/2050.  Default is 1350/2000.  Initially, I now see fewer rejects, but the hash rate is now 11.5MH/s for the rig. and 1.91MH/s for each card.

I need to mow the lawn!  Will check back in a while for a more reliable readout.  Thanks!       --scryptr
Try 1150/1750, so memory will be stock. No sense to reduce core(core can provide wrong results as I remember only on Vega).
legendary
Activity: 1797
Merit: 1028
HIGH REJECT RATE IN BOOLBERRY--

I am running the new  (v16.0) WildRig-Multi on Linux, and mining BoolBerry.  Up until yesterday, I was using WildRig v10.5.  However, the new version runs about 10% rejects!

Any suggestions?  My cards are RX580, 8GB.       --scryptr
What pool? Did you try re-download scratchpad? How many cards?

LUCKYPOOL, NO, 6 CARDS--

The scratchpad has been updating successfully, I did not change it from the pad left by WildRig v10.5.  I could "prime the pump", and restart after a wget download, I suppose.  The cards are running faster on the v16.0 release, near 2.0MH/s per card.

I am running on ethOS, and the rig reports 12MH/s (rounded).    Previously, it was running at 11.4MH/s.  I have the intensity at 22x256, and have tried lowering the intensity with neutral results.  The reject rate is not lessened, but the hashrate is lower.

Any suggestions are welcome, and thanks for adding BoolBerry!       --scryptr

EDIT:  After downloading a fresh scratchpad, and no other changes, the rejects are still present.       --scryptr


hmm, what if try --opencl-launch auto? Also how that rejects appear. Between accepted shares, or just starting to be rejected until new job? I was running my small rig(4xRX460) for more than 3 days and no problem %)

LOWERING CLOCKS--

I lowered my clocks to 1100/2000 (core/mem) from 1150/2050.  Default is 1350/2000.  Initially, I now see fewer rejects, but the hash rate is now 11.5MH/s for the rig. and 1.91MH/s for each card.

I need to mow the lawn!  Will check back in a while for a more reliable readout.  Thanks!       --scryptr
member
Activity: 725
Merit: 49
HIGH REJECT RATE IN BOOLBERRY--

I am running the new  (v16.0) WildRig-Multi on Linux, and mining BoolBerry.  Up until yesterday, I was using WildRig v10.5.  However, the new version runs about 10% rejects!

Any suggestions?  My cards are RX580, 8GB.       --scryptr
What pool? Did you try re-download scratchpad? How many cards?

LUCKYPOOL, NO, 6 CARDS--

The scratchpad has been updating successfully, I did not change it from the pad left by WildRig v10.5.  I could "prime the pump", and restart after a wget download, I suppose.  The cards are running faster on the v16.0 release, near 2.0MH/s per card.

I am running on ethOS, and the rig reports 12MH/s (rounded).    Previously, it was running at 11.4MH/s.  I have the intensity at 22x256, and have tried lowering the intensity with neutral results.  The reject rate is not lessened, but the hashrate is lower.

Any suggestions are welcome, and thanks for adding BoolBerry!       --scryptr

EDIT:  After downloading a fresh scratchpad, and no other changes, the rejects are still present.       --scryptr


hmm, what if try --opencl-launch auto? Also how that rejects appear. Between accepted shares, or just starting to be rejected until new job? I was running my small rig(4xRX460) for more than 3 days and no problem %)
newbie
Activity: 4
Merit: 0
Uploaded hotfix.

algo xevan rx560 on zergpool work properly

vega56/64 only invalid shares
 
Ok, fixed now, thanks! Will prepare new build soon, with another optimization to x16 family.

thanks!
legendary
Activity: 1797
Merit: 1028
HIGH REJECT RATE IN BOOLBERRY--

I am running the new  (v16.0) WildRig-Multi on Linux, and mining BoolBerry.  Up until yesterday, I was using WildRig v10.5.  However, the new version runs about 10% rejects!

Any suggestions?  My cards are RX580, 8GB.       --scryptr
What pool? Did you try re-download scratchpad? How many cards?

LUCKYPOOL, NO, 6 CARDS--

The scratchpad has been updating successfully, I did not change it from the pad left by WildRig v10.5.  I could "prime the pump", and restart after a wget download, I suppose.  The cards are running faster on the v16.0 release, near 2.0MH/s per card.

I am running on ethOS, and the rig reports 12MH/s (rounded).    Previously, it was running at 11.4MH/s.  I have the intensity at 22x256, and have tried lowering the intensity with neutral results.  The reject rate is not lessened, but the hashrate is lower.

Any suggestions are welcome, and thanks for adding BoolBerry!       --scryptr

EDIT:  After downloading a fresh scratchpad, and no other changes, the rejects are still present.       --scryptr

member
Activity: 725
Merit: 49
HIGH REJECT RATE IN BOOLBERRY--

I am running the new  (v16.0) WildRig-Multi on Linux, and mining BoolBerry.  Up until yesterday, I was using WildRig v10.5.  However, the new version runs about 10% rejects!

Any suggestions?  My cards are RX580, 8GB.       --scryptr
What pool? Did you try re-download scratchpad? How many cards?
member
Activity: 725
Merit: 49
0.16.2 beta
- fixed xevan on Vega
- another improve to x16-like algorithms for specific hashorder(SIMD at start)
legendary
Activity: 1797
Merit: 1028
HIGH REJECT RATE IN BOOLBERRY--

I am running the new  (v16.0) WildRig-Multi on Linux, and mining BoolBerry.  Up until yesterday, I was using WildRig v10.5.  However, the new version runs about 10% rejects!

Any suggestions?  My cards are RX580, 8GB.       --scryptr
member
Activity: 725
Merit: 49
Uploaded hotfix.

algo xevan rx560 on zergpool work properly

vega56/64 only invalid shares
 
Ok, fixed now, thanks! Will prepare new build soon, with another optimization to x16 family.
member
Activity: 725
Merit: 49
This is a huge surprise. SHA512 for 128 bytes doesn't work on Vega, while same code compiled for Polaris - works, and same code(less calls) for 64 bytes works too O_O I "like" AMD OpenCL so much )
member
Activity: 725
Merit: 49
Uploaded hotfix.

algo xevan rx560 on zergpool work properly

vega56/64 only invalid shares
 
hmm, pretty strange bug O_o Will check what hashfunc produces wrong result.

ok, may be in future you will describe this in first page of this thread or in help section?
yeah, will do an update
newbie
Activity: 4
Merit: 0
Uploaded hotfix.

algo xevan rx560 on zergpool work properly

vega56/64 only invalid shares
 
Pages:
Jump to: