[ANN] YACMiner - AMD GPU miner for Scrypt-Chacha, N-Scrypt, and Scrypt coins

[Thirtybird]

Yes, I'm familiar with the readme

The config suggests 824 with 2 threads. I seem to have 1 of 4 270x's that don't mine properly with the config; its finicky (Getting like 50 hs) so I tried to let yacminer decide the buffer and it decided on 446 MB. Didn't help too much. Using -g 1 crashes the miner as well. If i restart it enough, sometimes it will start mining but it didn't make it through the night last night.

timeout /t 30
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
del *bin
yacminer --scrypt-chacha -o stratum+tcp://xxxxxxxxx -u xxxxxx -p x --nfmin 4 --nfmax 30 --starttime 1402845776 --worksize 128 --lookup-gap 8 -g 2 --buffer-size 824 -R 1600

Furthermore, I cannot get past 2 khs, and it seems to be more stable on 1.8 khs. Is this all I can squeeze out of 270xs?

If you haven't run GPUz, I highly recommend it for troubleshooting.  YACMiner is looking at how much memory it sees, and allocates 90% of it if you do not specify (512 * .9 = 460), so you're having trouble where you cannot allocate all of the cards memory.  It will also show you how much memory is allocated to dedicated memory vs dynamic memory.  If your hashing buffer winds up spilling into dynamic memory, your hashrate is going to be dismal. 

My notes for NF=14 had my 2GB R7 250's hashing at 2.7 KH/sec, so you should be able to at least match that once you get your problems sorted out.

[1714872028]

1714872028

[1714872028]

1714872028

[1714872028]

1714872028

[lmaonade80]

Yes, I'm familiar with the readme

The config suggests 824 with 2 threads. I seem to have 1 of 4 270x's that don't mine properly with the config; its finicky (Getting like 50 hs) so I tried to let yacminer decide the buffer and it decided on 446 MB. Didn't help too much. Using -g 1 crashes the miner as well. If i restart it enough, sometimes it will start mining but it didn't make it through the night last night.

timeout /t 30
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
del *bin
yacminer --scrypt-chacha -o stratum+tcp://xxxxxxxxx -u xxxxxx -p x --nfmin 4 --nfmax 30 --starttime 1402845776 --worksize 128 --lookup-gap 8 -g 2 --buffer-size 824 -R 1600

Furthermore, I cannot get past 2 khs, and it seems to be more stable on 1.8 khs. Is this all I can squeeze out of 270xs?

If you haven't run GPUz, I highly recommend it for troubleshooting.  YACMiner is looking at how much memory it sees, and allocates 90% of it if you do not specify (512 * .9 = 460), so you're having trouble where you cannot allocate all of the cards memory.  It will also show you how much memory is allocated to dedicated memory vs dynamic memory.  If your hashing buffer winds up spilling into dynamic memory, your hashrate is going to be dismal. 

My notes for NF=14 had my 2GB R7 250's hashing at 2.7 KH/sec, so you should be able to at least match that once you get your problems sorted out.

2.7?! I barely get that outta my 280x's.. What AMD driver do you recommend? 14.6 is the one I'm using.

GPUz is a cool tool, but it is showing me everything is correct as far as I can tell. Each card has 2 GB, memclock at 1400, gpu-engine at 955. I'm starting to mess with the lookup-gap, but I thought that was just set by the NF number. ..

[Thirtybird]

2.7?! I barely get that outta my 280x's.. What AMD driver do you recommend? 14.6 is the one I'm using.

GPUz is a cool tool, but it is showing me everything is correct as far as I can tell. Each card has 2 GB, memclock at 1400, gpu-engine at 955. I'm starting to mess with the lookup-gap, but I thought that was just set by the NF number. ..

13.12
My R7 240 4GB cards NF=14 at 2.8-2.9 KH/sec (depending on the card) even though they have fewer shaders.  It's all a matter of allocating the largest buffer-size you can and then testing each of the lookup-gaps with the corresponding max R for each one to see which one to settle on.  The 4GB card can run a lookup-gap of 2, whereas I had to go with LG=3 on the 2GB card.  It's all in the tuning guide I linked you to

[lmaonade80]

13.12? Why such an old driver?

I'll continue to play.

[Thirtybird]

13.12? Why such an old driver?

I'll continue to play.

It's what was out when I started this experiment and the 14.x drivers gave HW errors mining scrypt-chacha when I tried them.  Now at this point, I don't feel like tinkering with them because "they just work"

[antonio8]

@ Thirtybird

Don't laugh at me. I get bored (and sometimes I am dangerous).

I am playing with Windows 10 and trying a 750ti with your miner. Long story short, new Nvidia driver supports OpenCL so I am playing with it to see if this will work.

When running the miner I have just a basic setting, besides the nfMin = 4 nfMax = 30 nTimeStamp = 1388949883, I just run the -I and change it higher until I get a HW error.

It does find the card and mines but it does not connect to the pool. The miner shows it and running but the pool does not show the miner active and I do not get hash rate at the pool.

Is there some setting I might try that might be specific.

Again, I am just experimenting to kill time and see if it works.

EDIT: Under Windows 7 with the exact same bat and Nvidia driver it does not find the card. Says no OpenCL device found or something like that.

[antonio8]

@ Thirtybird

Don't laugh at me. I get bored (and sometimes I am dangerous).

I am playing with Windows 10 and trying a 750ti with your miner. Long story short, new Nvidia driver supports OpenCL so I am playing with it to see if this will work.

When running the miner I have just a basic setting, besides the nfMin = 4 nfMax = 30 nTimeStamp = 1388949883, I just run the -I and change it higher until I get a HW error.

It does find the card and mines but it does not connect to the pool. The miner shows it and running but the pool does not show the miner active and I do not get hash rate at the pool.

Is there some setting I might try that might be specific.

Again, I am just experimenting to kill time and see if it works.

EDIT: Under Windows 7 with the exact same bat and Nvidia driver it does not find the card. Says no OpenCL device found or something like that.

Show me the output of yacminer -n, would you? Also, for Win 7, try diff OCL platforms.

Wolf0

Here is what I think you mean. Again I only have a setting of -I 8

[Thirtybird]

@ Thirtybird

Don't laugh at me. I get bored (and sometimes I am dangerous).

I am playing with Windows 10 and trying a 750ti with your miner. Long story short, new Nvidia driver supports OpenCL so I am playing with it to see if this will work.

When running the miner I have just a basic setting, besides the nfMin = 4 nfMax = 30 nTimeStamp = 1388949883, I just run the -I and change it higher until I get a HW error.

It does find the card and mines but it does not connect to the pool. The miner shows it and running but the pool does not show the miner active and I do not get hash rate at the pool.

Is there some setting I might try that might be specific.

Again, I am just experimenting to kill time and see if it works.

EDIT: Under Windows 7 with the exact same bat and Nvidia driver it does not find the card. Says no OpenCL device found or something like that.

Is it possible that it just was minign so slow that it never found any shares for submitting?  I seem to remember something about the nVidia cards supporting OpenCL (Even from the beginning), but they were horribly slow at it.  Post a screenshot of the thing mining?  That will show connection status and mining speeds and the like

[antonio8]

@ Thirtybird

Don't laugh at me. I get bored (and sometimes I am dangerous).

I am playing with Windows 10 and trying a 750ti with your miner. Long story short, new Nvidia driver supports OpenCL so I am playing with it to see if this will work.

When running the miner I have just a basic setting, besides the nfMin = 4 nfMax = 30 nTimeStamp = 1388949883, I just run the -I and change it higher until I get a HW error.

It does find the card and mines but it does not connect to the pool. The miner shows it and running but the pool does not show the miner active and I do not get hash rate at the pool.

Is there some setting I might try that might be specific.

Again, I am just experimenting to kill time and see if it works.

EDIT: Under Windows 7 with the exact same bat and Nvidia driver it does not find the card. Says no OpenCL device found or something like that.

Is it possible that it just was minign so slow that it never found any shares for submitting?  I seem to remember something about the nVidia cards supporting OpenCL (Even from the beginning), but they were horribly slow at it.  Post a screenshot of the thing mining?  That will show connection status and mining speeds and the like

 Here is a screen shot but something I did not notice until I put the log in the bat (thanks Wolf0). The miner shows that it is at a N-Factor 17 but here is my exact bat that I run for N-Factor 16:

yacminer --scrypt-chacha --nfmin 4 --nfmax 30 --starttime 1388949883 -d 1 -o stratum+tcp://pool:port -O xxxxxxx:x -D -l 5

[Thirtybird]

@ Thirtybird

Don't laugh at me. I get bored (and sometimes I am dangerous).

I am playing with Windows 10 and trying a 750ti with your miner. Long story short, new Nvidia driver supports OpenCL so I am playing with it to see if this will work.

When running the miner I have just a basic setting, besides the nfMin = 4 nfMax = 30 nTimeStamp = 1388949883, I just run the -I and change it higher until I get a HW error.

It does find the card and mines but it does not connect to the pool. The miner shows it and running but the pool does not show the miner active and I do not get hash rate at the pool.

Is there some setting I might try that might be specific.

Again, I am just experimenting to kill time and see if it works.

EDIT: Under Windows 7 with the exact same bat and Nvidia driver it does not find the card. Says no OpenCL device found or something like that.

Is it possible that it just was minign so slow that it never found any shares for submitting?  I seem to remember something about the nVidia cards supporting OpenCL (Even from the beginning), but they were horribly slow at it.  Post a screenshot of the thing mining?  That will show connection status and mining speeds and the like

 Here is a screen shot but something I did not notice until I put the log in the bat (thanks Wolf0). The miner shows that it is at a N-Factor 17 but here is my exact bat that I run for N-Factor 16:

yacminer --scrypt-chacha --nfmin 4 --nfmax 30 --starttime 1388949883 -d 1 -o stratum+tcp://pool:port -O xxxxxxx:x -D -l 5

That's kind of interesting.  If you look at the pool line it is showing "diff 2 NF 16" but your debug output shows NFactor 17.  My miner on my system (that I know submits shares) actually shows "Nfactor: 17" in the debug output too, so that's likely inaccurate, or based on 2^NF instead of the scrypt-chacha funky math of 2^(NF+1) or something like that - it's been a while, but it's always been confusing (thanks pocopoco!).  I think you've got it working, but at that hash rate, you're likely not submitting shares, or are hitting HW errors instead of accepted shares.  

Since you did not specify buffer_size or lookup gap on the command line, the default values are going to pick ~90% of the GPU VRam it can see and a lookup-gap of 4, which is not ideal for NF=16, and is probably giving you HW errors.  What buffer size does it tell you it's allocating for the card when it starts up?  (likely need to turn debug off first)

[antonio8]

@ Thirtybird

Don't laugh at me. I get bored (and sometimes I am dangerous).

I am playing with Windows 10 and trying a 750ti with your miner. Long story short, new Nvidia driver supports OpenCL so I am playing with it to see if this will work.

When running the miner I have just a basic setting, besides the nfMin = 4 nfMax = 30 nTimeStamp = 1388949883, I just run the -I and change it higher until I get a HW error.

It does find the card and mines but it does not connect to the pool. The miner shows it and running but the pool does not show the miner active and I do not get hash rate at the pool.

Is there some setting I might try that might be specific.

Again, I am just experimenting to kill time and see if it works.

EDIT: Under Windows 7 with the exact same bat and Nvidia driver it does not find the card. Says no OpenCL device found or something like that.

Is it possible that it just was minign so slow that it never found any shares for submitting?  I seem to remember something about the nVidia cards supporting OpenCL (Even from the beginning), but they were horribly slow at it.  Post a screenshot of the thing mining?  That will show connection status and mining speeds and the like

 Here is a screen shot but something I did not notice until I put the log in the bat (thanks Wolf0). The miner shows that it is at a N-Factor 17 but here is my exact bat that I run for N-Factor 16:

yacminer --scrypt-chacha --nfmin 4 --nfmax 30 --starttime 1388949883 -d 1 -o stratum+tcp://pool:port -O xxxxxxx:x -D -l 5

That's kind of interesting.  If you look at the pool line it is showing "diff 2 NF 16" but your debug output shows NFactor 17.  My miner on my system (that I know submits shares) actually shows "Nfactor: 17" in the debug output too, so that's likely inaccurate, or based on 2^NF instead of the scrypt-chacha funky math of 2^(NF+1) or something like that - it's been a while, but it's always been confusing (thanks pocopoco!).  I think you've got it working, but at that hash rate, you're likely not submitting shares, or are hitting HW errors instead of accepted shares.  

Since you did not specify buffer_size or lookup gap on the command line, the default values are going to pick ~90% of the GPU VRam it can see and a lookup-gap of 4, which is not ideal for NF=16, and is probably giving you HW errors.  What buffer size does it tell you it's allocating for the card when it starts up?  (likely need to turn debug off first)

You are correct on the look-up gap of 4 and the buffer is being set at 448. After a while of mining I do get a HW error.

If I change the starttime I get a high hash rate but all rejects.

[lmaonade80]

13.12? Why such an old driver?

I'll continue to play.

It's what was out when I started this experiment and the 14.x drivers gave HW errors mining scrypt-chacha when I tried them.  Now at this point, I don't feel like tinkering with them because "they just work"

My 270x max out at 2.2 khs on Nfactor 14...

setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
del *bin
yacminer --scrypt-chacha  --worksize 128 -g 1 --lookup-gap 8 -R 3200 --buffer-size 1696 --gpu-memclock 1400 --gpu-engine 1000

I've tried 13.12, 14.6, and 15.3 drivers, no noticeable changes. The buffersize is the largest I can make it. Is there anything else I can play with to get more juice outta these cards? Thanks in advance.

[Thirtybird]

You are correct on the look-up gap of 4 and the buffer is being set at 448. After a while of mining I do get a HW error.

If I change the starttime I get a high hash rate but all rejects.

Do not play with starttime - that is specific to the coin!  You need to work on setting the highest buffer-size you can and get a higher lookup gap going.  Stop using Intensity and use rawintensity instead I=8 is equivilent to R=256, at that lookup gap and nFactor, your -rI should be 112, so no wonder you're getting HW errors.

You really need to read the tuning guide if you want any chance at being successful
https://github.com/Thirtybird/YACMiner/blob/master/SCRYPT-README

[Thirtybird]

13.12? Why such an old driver?

I'll continue to play.

It's what was out when I started this experiment and the 14.x drivers gave HW errors mining scrypt-chacha when I tried them.  Now at this point, I don't feel like tinkering with them because "they just work"

My 270x max out at 2.2 khs on Nfactor 14...

setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
del *bin
yacminer --scrypt-chacha  --worksize 128 -g 1 --lookup-gap 8 -R 3200 --buffer-size 1696 --gpu-memclock 1400 --gpu-engine 1000

I've tried 13.12, 14.6, and 15.3 drivers, no noticeable changes. The buffersize is the largest I can make it. Is there anything else I can play with to get more juice outta these cards? Thanks in advance.

Uncertain - I would try every lookup gap and the corresponding Raw Intensity and see which nets you the biggest hashrate.  There is a way to calculate your maximum R given buffer size and lookup gap, but I'll leave that for you to find   (it's published in at least 2 separate places in 2 different programming languages)

Your target to beat is 2.7 KH/sec, which my $100 R7 250 2GB card did at NF=14

[lmaonade80]

13.12? Why such an old driver?

I'll continue to play.

It's what was out when I started this experiment and the 14.x drivers gave HW errors mining scrypt-chacha when I tried them.  Now at this point, I don't feel like tinkering with them because "they just work"

My 270x max out at 2.2 khs on Nfactor 14...

setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
del *bin
yacminer --scrypt-chacha  --worksize 128 -g 1 --lookup-gap 8 -R 3200 --buffer-size 1696 --gpu-memclock 1400 --gpu-engine 1000

I've tried 13.12, 14.6, and 15.3 drivers, no noticeable changes. The buffersize is the largest I can make it. Is there anything else I can play with to get more juice outta these cards? Thanks in advance.

Uncertain - I would try every lookup gap and the corresponding Raw Intensity and see which nets you the biggest hashrate.  There is a way to calculate your maximum R given buffer size and lookup gap, but I'll leave that for you to find   (it's published in at least 2 separate places in 2 different programming languages)

Your target to beat is 2.7 KH/sec, which my $100 R7 250 2GB card did at NF=14

I understand the -R flag is multiplicatives of the worksize. I have it at 3328. I can get the miner at 3 khs, but it crashes the driver (13.12) after a few minutes. I'll probably need to dial down the engine.

[hoper]

I understand the -R flag is multiplicatives of the worksize. I have it at 3328. I can get the miner at 3 khs, but it crashes the driver (13.12) after a few minutes. I'll probably need to dial down the engine.

Nfactor 14 = 4Mb per hash if lookup-gap=1   and 0,5Mb if lookup-gap=8

if buffersize=1696 and lookup-gap=8 rawintensity must be = 1696/(4/8)-1 = 3391

try maximum increase buffer size first
then calculate rawintensity for selected lookup-gap

try another lookup-gap

and try this
setx GPU_MAX_HEAP_SIZE 100
setx GPU_FORCE_64BIT_PTR 1

example --lookup-gap 4 -R 1695 -B 1696

[reb0rn21]

Tested yacminer with latest win 10 preview + amd driver from windows update 15.5 beta and it works just alocating memory is low 2800MB for 4GB card ... in win 8.1 no go with new drivers

[homeplus]

I would like to mine scrypt jane N Factor and I want to know what downloads I need?  Also, what I need to use?  Like a Graphics card or a Miner like Titan?  Let me know...I Mine for a Hobbie and Haven't seen anyone mine scrypt jane N Factor as Good as you...Thanks

[bankmyrik10]

I have ASUS R 9 290 4Gb. This is my bat file
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
yacminer --scrypt-chacha --worksize 256 --lookup-gap 2 -g 1 --buffer-size 2656 -R 6411  -o stratum+tcp://wk10.botpool.net:4339 -u bankmyrik.bankmyrik -p x --nfmin 16 --nfmax 30 --starttime 1375817223
Bat i have HW errors. What my problem?

[omnik]

I have ASUS R 9 290 4Gb. This is my bat file
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
yacminer --scrypt-chacha --worksize 256 --lookup-gap 2 -g 1 --buffer-size 2656 -R 6411  -o stratum+tcp://wk10.botpool.net:4339 -u bankmyrik.bankmyrik -p x --nfmin 16 --nfmax 30 --starttime 1375817223
Bat i have HW errors. What my problem?

Go with lookup-gap 6 or 7