Pages:
Author

Topic: [BBR] Boolberry GPU Miner Discussion - AMD & Nvidia - OpenCL & CUDA - page 2. (Read 82987 times)

newbie
Activity: 69
Merit: 0
There has to be something that one can do with the scratchpad related stuff.
member
Activity: 720
Merit: 49
GTX 1080 and only 1200kH/s? Better to mine something profitable for this card and buy boolberry ) What about hashrate - it jumps a lot, so on pool you won't see stable hashrate.
newbie
Activity: 5
Merit: 0
Been mining all night at 1200Kh/s
https://i.imgur.com/eZvxNgW.png

But for some reason on mining.blue it shows that I'm mining slower than that.. Has anyone else experienced this?
https://i.imgur.com/IG2caYF.png
full member
Activity: 202
Merit: 104
is there any way to mine boolberry solo on gpu?
You can get the https://github.com/mbkuperman/boolberry-opencl compiled and run using your existing blockchain if it's synced. I don't know if it will sync from scratch, though.
sr. member
Activity: 463
Merit: 256
is there any way to mine boolberry solo on gpu?
sr. member
Activity: 520
Merit: 253
555
Fun times: I'm getting 1400 kh/s on a Radeon 7970 with opensource Linux drivers. Default clocks of 950/1450 MHz and I estimate the power draw at 150 W max.
Are you able to set clocks with that?
Yea, but not any higher so far. The "high" profile gives the same frequencies as "default". What's worse, all non-default profiles go to "low" speeds when the monitor is off.

For details see https://wiki.gentoo.org/wiki/Radeon#Power_management and in this case, dynamic power management must be disabled with the kernel option radeon.dpm=0. The dynamic setting doesn't have the monitor (DPMS) issue.

For comparison, using fglrx gives 875 kh/s at 1100/350 MHz, about the same power judging from similar temperatures.
full member
Activity: 202
Merit: 104
Fun times: I'm getting 1400 kh/s on a Radeon 7970 with opensource Linux drivers. Default clocks of 950/1450 MHz and I estimate the power draw at 150 W max. (For comparison, my GTX 1070 does about 1600 kh/s at 100 W). I've seen my share of miners reporting wrong numbers, but this one is verified by a pool over a full day, so it's looking legit.

This is with Linux 4.15.12, mesa 18.0.0_rc5, llvm 6.0.0 and xf86-video-ati 18.0.1. It should be noted that the newer opensource AMDGPU driver isn't available for the 7970, but the real action for OpenGL/OpenCL is happening with Mesa and LLVM, and there have been some great developments.

I've been testing the opensource driver stack for a while, as I've sold most of my Radeons and I've stopped patching the old binary driver (fglrx) for newer kernels. So far, it's been slower, around 650 kh/s vs. 875 with fglrx. This incredible leap came after the latest Mesa and LLVM updates. It's been known for a while that the opensource stack outperforms the old binaries in many applications, but this is a first for me with miners.

While great for BBR, there aren't many other miners that work at all with the OS drivers. Newer miners use OpenCL 2.0 which isn't implemented yet, and there are other issues. So I guess I'll keep this machine for BBR only for now Wink

Also noting this is a rather slow machine with 1x PCIe bus, but it's probably irrelevant for BBR mining.
Are you able to set clocks with that?
sr. member
Activity: 520
Merit: 253
555
Fun times: I'm getting 1400 kh/s on a Radeon 7970 with opensource Linux drivers. Default clocks of 950/1450 MHz and I estimate the power draw at 150 W max. (For comparison, my GTX 1070 does about 1600 kh/s at 100 W). I've seen my share of miners reporting wrong numbers, but this one is verified by a pool over a full day, so it's looking legit.

This is with Linux 4.15.12, mesa 18.0.0_rc5, llvm 6.0.0 and xf86-video-ati 18.0.1. It should be noted that the newer opensource AMDGPU driver isn't available for the 7970, but the real action for OpenGL/OpenCL is happening with Mesa and LLVM, and there have been some great developments.

I've been testing the opensource driver stack for a while, as I've sold most of my Radeons and I've stopped patching the old binary driver (fglrx) for newer kernels. So far, it's been slower, around 650 kh/s vs. 875 with fglrx. This incredible leap came after the latest Mesa and LLVM updates. It's been known for a while that the opensource stack outperforms the old binaries in many applications, but this is a first for me with miners.

While great for BBR, there aren't many other miners that work at all with the OS drivers. Newer miners use OpenCL 2.0 which isn't implemented yet, and there are other issues. So I guess I'll keep this machine for BBR only for now Wink

Also noting this is a rather slow machine with 1x PCIe bus, but it's probably irrelevant for BBR mining.
member
Activity: 720
Merit: 49
Does anyone have info about hashrate for GTX 750(not Ti)? Searched through whole topic, but didn't find anything.
newbie
Activity: 3
Merit: 0
So, could anyone Rebuild the miner to support more GPUs at once?

I could not, even thoroughly following the instructions on GitHub...
newbie
Activity: 3
Merit: 0
Quote
Using "-d 8" only selects a single GPU, starting with 0 for the first GPU. It looks like there's a hardcoded limit of 8 GPUs since it ignores the 9th one.

That is absolutely true.
I just shown that to illustrate that devices with numbers higher than 7 are ignored.

Quote
so you could change this and rebuild the miner.

Exactly. The MAX_GPU is checked in several places in source code.
I tried to binary-patch the check, but did not succeed.

Rebuilding also failed for me. I feel I could not set up the environment correctly.

I suppose in need a lot higher skill than just copy-paste from instruction.

Anyone can help?
(I wrote an email to author but had no response).
sr. member
Activity: 520
Merit: 253
555
The only problem I have right now is inefficient mining. I got 11 (eleven) GPUs in my rig, all of them AMD.
But the miner accepts only 8 of them (and that includes integrated graphics, with 27  Hashes per second). 4 more card just stay idle.

If I try to set minerd option "-d 8" or higher, it responds me with just little help info.

Using "-d 8" only selects a single GPU, starting with 0 for the first GPU. It looks like there's a hardcoded limit of 8 GPUs since it ignores the 9th one. Indeed, there's a line in gpu.h:
Code:
#define MAX_GPU 8
so you could change this and rebuild the miner.

newbie
Activity: 3
Merit: 0
Hello, guys!

Just discovered the coin for myself. Looks interesting.

The only problem I have right now is inefficient mining. I got 11 (eleven) GPUs in my rig, all of them AMD.
But the miner accepts only 8 of them (and that includes integrated graphics, with 27  Hashes per second). 4 more card just stay idle.

If I try to set minerd option "-d 8" or higher, it responds me with just little help info.
How can I increase the number of GPUs involved in mining?
Or is there a way to use, say, sgminer for minining (seems like it knows nothing about scrathpad files).
Hope to here from you guys soon Smiley
copper member
Activity: 51
Merit: 0
Open source Boolberry OpenCL miner
https://github.com/mbkuperman/boolberry-opencl.git

Read README.md for details and build instructions.

Windows x64 executables:
https://github.com/mbkuperman/boolberry-opencl/releases/download/v1.0/BoolberryOCL-20140719.zip

2% developer donation (you can change or disable it in the configuration file)

Solo version: boolbd.exe (best performance/reliability)
Pool (JSON RPC) version: simpleminer.exe (please, use solo version if you can!)

Personally, I think it doesn't make any sense to use pools while the block value is about $7. Just give enough time to the solo miner.

Tested on AMD GPUs only!

Expected performance:
6870, R270x - 500-600 khs
6970, R280X - 900-1000 khs

OpenCL code is available under the terms of the GNU Public License version 3.
Author: Mikhail Kuperman ([email protected])
BBR donation address: 1EmWGnwhydr3S2vRWQbbefh1hgDKgMjdMGe43ZgdPhdARhNBRkUMuD4YzLA2nyYG8tg2HKCCBg4aDam JKypRQWW1Ca2kSV8
BTC donation address: 1Lns6UjL3sw77DJ5z1EKJZy6SnqriqvVGK


MBK:  I've seen new versions of XMR miners that were built on Open CL 2.0 that are getting some pretty good efficiency gains, especially on the new AMD VEGA architecture with HBM2 memory.  I know you have probably moved on to greener pastures, but I wanted to reach out to you and see if you would be willing to look into the current minerd and see if there are some efficiency gains that are achievable with some modifications similar to what CAST-XMR and XMR-STAK have accomplished.  As interest is growing in the BBR community, I think either an open source version with a bounty, or a closed source with a fee would be amenable, as I think there is quite a bit of efficiency that could be achieved.  Thank you in advance.
newbie
Activity: 1
Merit: 0
So with minerd you can't set 2 different intensities for 2 different cards? Am I correct?
ie  -d 0,1  -i 14,19  should give intensity 14 to device 0 and intensity 19 to device 1... but it doesn't?
wondering the same, what is the vega hashrate?
can you also split threads like with XMR then apply different intensities to each?
mine is 2330 kh/s with minerd.exe  Smiley
vega 56 flashed to 64 GPU@1300MHz; MEM@1025MHz...
i think that even power consumption is lower than with cast or stak miners...i can't say for sure how much, because i don't have any instrument to be precise, but in my opinion blower is somewhere at ~2500rpm instead 2800+rpm with other miners and currencies...
today is my first day when it comes to mining boolberry and that's why i'm not such a good informer  Grin

edit: of course i forgot to use gpu-z and he reported lower cooler speed than 2400rpm
legendary
Activity: 3164
Merit: 1003
http://mining.blue/#
Congratulations ..the first pool that knows how to set the stats for Nvidia and ccminer.  Smiley
Working good now.

Can you tell me what options you have set in your miner, please?
Sure. Smiley This is for 2 cards on this computer.. same on my rig...but rig has 6 cards. My other computer has 1 card.

ccminer-x64-2.2.2-cuda7.5.exe -a wildkeccak  -t 2 -l 60x256  -o stratum+tcp://usbbr.mining.blue:8888 -u myaddress -p x --max-log-rate 30  -k http://mining.blue/scratchpad.bin

EDIT: And my high end card rig has a blown mobo. Cheesy Be back online in a week or 2. Smiley
full member
Activity: 202
Merit: 104
http://mining.blue/#
Congratulations ..the first pool that knows how to set the stats for Nvidia and ccminer.  Smiley
Working good now.

Can you tell me what options you have set in your miner, please?
jr. member
Activity: 88
Merit: 1
So with minerd you can't set 2 different intensities for 2 different cards? Am I correct?
ie  -d 0,1  -i 14,19  should give intensity 14 to device 0 and intensity 19 to device 1... but it doesn't?
wondering the same, what is the vega hashrate?
can you also split threads like with XMR then apply different intensities to each?
newbie
Activity: 1
Merit: 0
i mining error retry after 10 second. Help me, help me!
legendary
Activity: 3164
Merit: 1003
http://mining.blue/#
Congratulations ..the first pool that knows how to set the stats for Nvidia and ccminer.  Smiley
Working good now.
Pages:
Jump to: