Ploting with two NVIDIA Quadros M6000 using Direct mode.
-
wow nonces Speed :O
-
@nixxda said in I'm having trouble with GPU Plotting using Direct mode.:
Plotting in direct mode first creates the file and then writes sorted data into it
Ok so it finish creating the file and after that it going really really slow creating the data into it.
Im getting 116.35 nonces/minutes. Can someone please tell me what is going on?
-
@Wolf post screenshots of your devices.txt and the command used with gpuplot
-
@Wolf said in I'm having trouble with GPU Plotting using Direct mode.:
@nixxda said in I'm having trouble with GPU Plotting using Direct mode.:
Plotting in direct mode first creates the file and then writes sorted data into it
Ok so it finish creating the file and after that it going really really slow creating the data into it.
Im getting 116.35 nonces/minutes. Can someone please tell me what is going on?It will increas as it goes along, mine does the same thing, writes the file at 40-50K nonces per minute and once done it optimizes it which starts off at ~100-300 nonces /minute and goes up to about 4000
-
@Focus Will it be faster to GPU Plot first and then optimize after?
-
@iKnow0 Here is my devices.txt configuration
0 0 2048 128 2048
0 1 2048 128 2048
-
@Wolf said in I'm having trouble with GPU Plotting using Direct mode.:
@Focus Will it be faster to GPU Plot first and then optimize after?
Just do it the way you are and be patient, I tried optimizing after with so-so results
-
@Wolf Uhm... 23GB GPU memory, oh my god :)... Maybe experiment mad things like increasing 3rd number:
0 0 65536 32768 600 (16GB of memory here)
If it works, increase the last value. I don't know if it works, but maybe this is the way...
-
@Dario's-wallet I have made a post about the GPU Plotter not using all of my gpu memory and the differance between increasing the third number here..
https://forums.burst-team.us/topic/2656/gpu-plotter-is-not-using-my-gpu-memory/2
but I see that you increase the 4th number too I haven't did that one. Also 65536 suppose to be 64GB right?
-
@Wolf 65536 it's the memory space measured in number of nonces: 65536*256KB(size of one nonce)=16GB
In my post I mean set the 2nd number at 65536 WHILE you set the 3rd at 65536/2=32768 ( a lot of parallel threads), then increase the last number starting from 600 (typically 600 is a low value)...
-
@Dario's-wallet said in I'm having trouble with GPU Plotting using Direct mode.:
@Wolf 65536 it's the memory space measured in number of nonces: 65536*256KB(size of one nonce)=16GB
In my post I mean set the 2nd number at 65536 WHILE you set the 3rd at 65536/2=32768 ( a lot of parallel threads), then increase the last number starting from 600 (typically 600 is a low value)...Dario, what do you suggest on a system that has a RX480 and 256GB of ram ?
-
@Focus Hi mate, I'm so sorry for late reply :(, when I wrote my last post was 3 am and I fell asleep...
However for me is better always start with default values, then tweak. Every system can be a bit different. A good value for global work size could be 32768 (8GB RX480 version, without screen output). For the last two values is more difficult to say... Maybe a solution could be set the hash number at a middle-high value (like ~5000), find the higher local work size that works, and then increase the last number...
However a card like the RX480 I think can top every hdds write speed: if you have to plot more than one hard drive (@AdamGuerbuez system?:)), I suggest you to parallelize plotting (two drives at same time, maybe more, you have to try...)
-
@Dario's-wallet with that said.. Do you know a good stagger size for a 10TB Harddrive? Because every time that I increase the global work size(the third number in the device.txt file) for some reason I have to decrease the stagger size.
gpuPlotGenerator generate direct D://3225037558013223906_0_38075392_105472
pause"Updated"
I'm unable to go with this configuration0 0 65536 32768 600
0 1 65536 32768 600Here is the error for this configuration. Even if I divide it by 2 all the way till the third number is 4096...
[ERROR][-54][CL_INVALID_WORK_GROUP_SIZE] Error in step1 kernel launch
-
@Wolf No probs for the stagger, 105472 is ok, in direct mode the output file will be 3225037558013223906_0_38075392_38075392
About the devices configurations...mmm...maybe try lowering the local work size number increasing the hashes per call (6200 is just an example, maybe also try with 8192):
0 0 65536 32768 6200
0 0 65536 16384 6200
0 0 65536 8192 6200
And so on...
-
@Dario's-wallet said in I'm having trouble with GPU Plotting using Direct mode.:
0 0 65536 8192 6200
I just set it to...
0 0 65536 2048 8192
0 0 65536 2048 8192I get this error
[ERROR][-54][CL_INVALID_WORK_GROUP_SIZE] Error in step1 kernel launch
-
@Wolf 1024, 512, 256,...?
-
@Dario's-wallet If i set the last number to 8192 it throws me an error
Even if i set the configuration to thisexample..
0 0 1024 128 8192
0 1 1024 128 8192This is the error for this configuration.
[ERROR][-5][CL_OUT_OF_RESOURCES] Error in synchronous readThe highest that I can configure the last number with no problems is 7168. But I have to leave the third number to 1024
example..
0 0 1024 128 7168
0 1 1024 128 7168But if I leave the last number to 7168 and go up on the third number to 2048 it throws me a another error.
example..
0 0 2048 128 7168
0 1 2048 128 7168This is the error for this configuration.
[ERROR][-36][CL_INVALID_COMMAND_QUEUE] Error in step2 kernel finish
-
@Dario's-wallet
btw.. The highest I can configure the third number with no problems is 4096. But I have to leave the last number to 1024
example..
0 0 4096 128 1024
0 1 4096 128 1024But with this configuration I lose 20000 nonces/minutes,.
-
@Wolf Uhm... Interesting... Seems you cannot use your entire GPU memory... Seems there's an inviolable line for this card memory in burstcoin plotting. However to figure out what's your fast configuration don't trust direct mode nonce/minute values, they're funky. Use buffer mode for that.
-
@Dario's-wallet said in I'm having trouble with GPU Plotting using Direct mode.:
Seems there's an inviolable line for this card memory in burstcoin plotting.
Do you have a rx480? If so what's the nonces/minutes? Also do you think its faster to plot first and optimize after?





