Author

Topic: [ANN][BURST] Burst | Efficient HDD Mining | New 1.2.3 Fork block 92000 - page 849. (Read 2170900 times)

newbie
Activity: 56
Merit: 0
How much memory do you have in your GPU?
The most I can gen @ is 3000 on a higher end card.

The lower end ones with 1GB constantly fail with the error posted above.


It's a 3GB R9 280x...  Undecided

Hi guys,

I am aware of this issue. The fact is that I have to create two full size buffers on the GPU side to reduce thread-local memory consumption. Thus the memory amount needed on the CPU side has to be doubled to get an estimate of what is needed on the GPU side.
As an example, for a stagger size of 4000 you will need 1GB RAM on CPU side and more than 2GB (exactly (PLOT_SIZE + 16) x stagger) on GPU side (doesn't include here the local buffers and the kernel code itself).
Once I have a stable version (really soon Grin), I will work on this particular problem.

Well, I've kind of figured out what the problem is... I have a 3GB XFX 280X (R9-280X-TDBD) but when I run the new gpuminer v2 and run "list devices" it shows my card with a max global memory of 2048 instead of 3072.  Maybe they forgot to put in a third of my vRAM? Cheesy

Any idea why the discrepancy?  Not that it REALLY matters anymore... I'm almost done plotting, lol.
member
Activity: 60
Merit: 10
One per plot. So, you'd have to start multiple plotters to use multiple GPUs
Thanks. So does this mean that basically the only way to use multiple GPUs in one rig efficiently (= simultaneously) is to build plots on multiple HDDs at the same time with 1 gpu per 1 HDD? Although I suppose it's possible to divide 1 HDD into several partitions and fill each partition of the same disk simultaneously, assigning 1 gpu per partition? If that is possible, then we'll run into maximum disk write speed limit pretty quickly, cause ordinary HDDs won't like it when huge chunks of data are written to different partitions at the same time..

The multi-GPU support is in the roadmap, be patient Wink
legendary
Activity: 3248
Merit: 1070
full member
Activity: 154
Merit: 100
Add me on Twitter! @AnonOnAMoose
Aww, no one gave me + trust for overseeing the BURST multipool fund =(
legendary
Activity: 1059
Merit: 1000
One question: but the 1024m is used to indicate the ram memory, right?
If you increase this value in multiples,  gain more? Since I have 128GB of ram. Wink

java -Xmx1024m -cp pocminer.jar;lib/*;lib/akka/*;lib/jetty/* pocminer.POCMiner mine "http://burst-pool.cryptoport.io"


Which is the better of them?

http://burst-pool.cryptoport.io/

or

http://burst-pool2.cryptoport.io/
newbie
Activity: 55
Merit: 0
Error: Failed to submit nonce, retrying...
Error: Failed to submit nonce

 Huh Huh Huh
full member
Activity: 294
Merit: 101
The Future of Security Tokens
So much help from LivsUA, thank you dude, you are the best Cheesy

now im creating my second file with gpu (7970) instead of cpu and i´m 3 times faster.
Actual plotting 9300nonces/min.
I think i could go a bit faster, but thats ok for now.

i go for jackpot version and try solomine with 6TB, if i dont find anything the next days, im gonna join a pool.
Just want to test my luck.

Maybe my chances getting struck by a lightning are better.
hero member
Activity: 518
Merit: 500
Messed around with the 2.0.1 plotter kernel a bit, and got it running a bit faster.

https://mega.co.nz/#!vsommLbQ!C9v0TshBm0JY8DMBC8DsM9wCtZvvGjsVDNl9wGr0DKU
Replace the kernel folder with the one provided.

This is about 35% slower on a Titan Black.

Giga 280x, parameters 0 8000 4000 128 4000
original: 11162,8 nonces/min
mod: 17777,8 nonces/min

That's about 59% increase in speed..... Shocked
full member
Activity: 224
Merit: 100
He дo кoнцa пoнял пpo пpoвepкy пepвыx бaйтoв пocлe гeнepaции плoтcoв нa пpoцeccope и видeoкapтe. Moжeт ктo нибyдь пoяcнить ?
coздaй фaйл c пoмoщью пpoцeccopa 48712486872462871628746 1 1 1000 5
A пoтoм coздaй фaйл c тaкими жe xapaктepиcтикaми c пoмoщью гpaфичecкoй кapты и cpaвни вec oбoиx фaйлoв в пaпкax нaзнaчeния.Ecли вec в бaйтax coвпaдaeт тo вcё в шoкoлaдe Wink
вec мoжeт и coвпacть, cмoтpeть нaдo coдepжимoe. в кoммaнднoй cтpoкe нaбиpaeшь  fc /b имяфaйлaЦпy имяфaйлaГпy     и oн тeбe пoкaжeт oтличия внyтpи фaйлoв, ecли нeт oтличий - тo вce нopм. фaйлы дoлжны быть c oдинaкoвыми пapaмeтpaми, пoлoжи иx в paзныe пaпки.
G:\plots>FC /b cpu gpu
Cpaвнeниe фaйлoв CPU и GPU
FC: paзличия нe нaйдeны
this what i see...
full member
Activity: 164
Merit: 100
Also, by partitioning a drive, you are not doing yourself any favors. Sure the drive shows up as 2 drive letters... but it is still the same physical disk, so it will still be confined to the same max write speed as if it was un-partitioned.
Well, that's exactly what I meant when saying "we'll run into maximum disk write speed limit pretty quickly", since the maximum write speed is for the whole drive, so when writing to multiple partitions simultaneously, you'll be limited by 'max_write_speed/number_of_partitions' at best. The more partitions you write at once, the bigger penalty you'll have. If we assume the average max write speed of around 100-150 MB/sec, then it'll be 50-75 MB/sec at most when writing to 2 partitions simultaneously. I'm using 7970 cards right now, and getting 15-16 k per minute, that's what, 60-65 MB/sec.. It might actually make sense to divide each drive in 2 partitions and use 1 GPU per partition simultaneously. Would be 1.5-2x times faster to fill one drive with 2 partitions and 2 gpus, rather than to use 1 gpu for plotting on 1 big partition.
...partitioning the drive doesn't make it faster. No idea how you're coming to that conclusion.
Using 2 GPUs to plot to the same folder on a single partition is going to be the same speed as plotting with 2 GPUs to separate partitions on the same drive.
member
Activity: 67
Merit: 10
Also, by partitioning a drive, you are not doing yourself any favors. Sure the drive shows up as 2 drive letters... but it is still the same physical disk, so it will still be confined to the same max write speed as if it was un-partitioned.
Well, that's exactly what I meant when saying "we'll run into maximum disk write speed limit pretty quickly", since the maximum write speed is for the whole drive, so when writing to multiple partitions simultaneously, you'll be limited by 'max_write_speed/number_of_partitions' at best. The more partitions you write at once, the bigger penalty you'll have. If we assume the average max write speed of around 100-150 MB/sec, then it'll be 50-75 MB/sec at most when writing to 2 partitions simultaneously. I'm using 7970 cards right now, and getting 15-16 k per minute, that's what, 60-65 MB/sec.. It might actually make sense to divide each drive in 2 partitions and use 1 GPU per partition simultaneously. Would be 1.5-2x times faster to fill one drive with 2 partitions and 2 gpus, rather than to use 1 gpu for plotting on 1 big partition.

No idea how you came to this conclusion, assuming identical gpus it would take the same amount of time to write one partition as two as three as four etc. since the number of gpus = number of partitions.  Still assuming of course that 'you'll be limited by 'max_write_speed/number_of_partitions at best'
newbie
Activity: 18
Merit: 0
Guys, I have a problem with mining. I'm running two mine.bat and the process java.exe is using all my RAM. I have 4GB of RAM. When I start the mining applications it uses about 3GB but after some time it uses all RAM and computer becomes SO SLOW that I literally cannot do anything, while CPU usage is about 10%.
I have 2x1TB of plots.
Any ideas? Do I just need more ram or is there a way to solve this?
I updated java but it's still same.
legendary
Activity: 1106
Merit: 1014
Also, by partitioning a drive, you are not doing yourself any favors. Sure the drive shows up as 2 drive letters... but it is still the same physical disk, so it will still be confined to the same max write speed as if it was un-partitioned.
Well, that's exactly what I meant when saying "we'll run into maximum disk write speed limit pretty quickly", since the maximum write speed is for the whole drive, so when writing to multiple partitions simultaneously, you'll be limited by 'max_write_speed/number_of_partitions' at best. The more partitions you write at once, the bigger penalty you'll have. If we assume the average max write speed of around 100-150 MB/sec, then it'll be 50-75 MB/sec at most when writing to 2 partitions simultaneously. I'm using 7970 cards right now, and getting 15-16 k per minute, that's what, 60-65 MB/sec.. It might actually make sense to divide each drive in 2 partitions and use 1 GPU per partition simultaneously. Would be 1.5-2x times faster to fill one drive with 2 partitions and 2 gpus, rather than to use 1 gpu for plotting on 1 big partition.
full member
Activity: 164
Merit: 100
One per plot. So, you'd have to start multiple plotters to use multiple GPUs
Thanks. So does this mean that basically the only way to use multiple GPUs in one rig efficiently (= simultaneously) is to build plots on multiple HDDs at the same time with 1 gpu per 1 HDD? Although I suppose it's possible to divide 1 HDD into several partitions and fill each partition of the same disk simultaneously, assigning 1 gpu per partition? If that is possible, then we'll run into maximum disk write speed limit pretty quickly, cause ordinary HDDs won't like it when huge chunks of data are written to different partitions at the same time..
I'm not sure of the throughput required for GPU Plotting, so I'm not sure if plotting to the same drive would cause slowdowns. Someone would need to test.. Wink
Also, by partitioning a drive, you are not doing yourself any favors. Sure the drive shows up as 2 drive letters... but it is still the same physical disk, so it will still be confined to the same max write speed as if it was a single partition.
sr. member
Activity: 416
Merit: 250
Messed around with the 2.0.1 plotter kernel a bit, and got it running a bit faster.

https://mega.co.nz/#!vsommLbQ!C9v0TshBm0JY8DMBC8DsM9wCtZvvGjsVDNl9wGr0DKU
Replace the kernel folder with the one provided.

This is about 35% slower on a Titan Black.

Giga 280x, parameters 0 8000 4000 128 4000
original: 11162,8 nonces/min
mod: 17777,8 nonces/min
legendary
Activity: 1106
Merit: 1014
One per plot. So, you'd have to start multiple plotters to use multiple GPUs
Thanks. So does this mean that basically the only way to use multiple GPUs in one rig efficiently (= simultaneously) is to build plots on multiple HDDs at the same time with 1 gpu per 1 HDD? Although I suppose it's possible to divide 1 HDD into several partitions and fill each partition of the same disk simultaneously, assigning 1 gpu per partition? If that is possible, then we'll run into maximum disk write speed limit pretty quickly, cause ordinary HDDs won't like it when huge chunks of data are written to different partitions at the same time..
hero member
Activity: 1036
Merit: 531
and as long as mine are correct i don't care about others payout lol

That's the spirit.
If you get hit by a truck, "others" won't care, either.

in that case i won't care also lmao

Road Pizza en Francais LOL

you're french?
hero member
Activity: 1036
Merit: 531
Block explorer down?
full member
Activity: 164
Merit: 100
I guess I have a problem. I've been generating plots all night, currently got 170,000 nonces and still none of my shares have been valid and I have 0 mined balance.
I know that 170,000 isn't that much, but seriously - nothing? I guess I'm doing something wrong.


EDIT: Also, I always get the "error reading file message" - any idea what's wrong?
I said you that you are overlapping! Delete the plots that start from 17001 and begin a new plot that start from 600001..
This is wrong information... the 2nd number is the number of nonces to create, NOT the stopping nonce.
If starting nonce is 17000, and it is set to create 600000 nonces the next file would NOT start at nonce 600001. Doing so would cause an overlap of 17000 nonces.

Formula is {Starting_Nonce} + {Number_of_nonce_to_create} = {next starting nonce}

So, if the current file is plotted with 17000 600000, then the 2nd plot would be 617000 600000
full member
Activity: 164
Merit: 100
The readme for gpuplotter 2.0.1 says:

Quote
Multiple graphic cards (preferred way)
--
In this mode, all the non-bound graphic cards can be used to generate plots without any timeout issue.
The typical parameter will be 8160.
Is it possible to use multiple gpus at once to build one plot, or is it only "one card per one plot" at the same time? Seems like it's required to specify gpu id in the command-line (after the platform id), so is there a way to specify 2 or 3 gpus there somehow?
One per plot. So, you'd have to start multiple plotters to use multiple GPUs
Jump to: