Best strategy to plot more 8tb disks

Have been on this to figure out max performance formula for 2 weeks plotting and deleting.

I am on PC with 15 x 8 TB Sata Seagate IronWolf drives. What is the best strategy to plot them. Go with big one file + 1 small or split each disk per 1 TB? What will give the best performance on mining?

As well I have i3 7100 process for rig and want to mine with Blaggo and feel it is to small for it. What processor on 1151 from your experience would be best for Blagoo mining?

Please help
Edit

3 Likes

I’ve had plot failures, power failures, etc. with the gpuPlotGenerator and lost the currently plotting file. Replotting stinks! For that reason I use files around 2GB-3GB in size so I don’t lose so much in the event of a catastrophe.

My current 8TB Seagates have three plot files per drive. I have 12 external drives and eight internal drives on two machines for a total of 119TB and my scans per block finish in the 22 to 25 second range. On paper one file should be better, but in the real world having more than one file still performs well.

1 Like

What gpu do you have? Perhaps gpu plotting/mining may be better than your cpu as that is only 2 cores 4 threads? But specifically as for better cpu for 1151, anything with more cores/threads, ie. i5 Or i7 series, as I believe blagos runs parralel threads though the assigned search plots.

1 Like

I have 1070 and 980 Ti on same machine but they are mining. I was planning better to go with Intel Core i7 7700. I dont know is it metter to take i7 or it is ok with i5 with 4 cores.

1 Like

For your information, I was able to plot 2 10TB ironwolf drives in 40 hours with a 980Ti.
I’m sure you could plot 4 drives (or even more) in the same amount of time.

1 Like

8 TB ploting with 1070 or 980 ti takes 18 hours and Optimization takes about same time. But still you need to make small file to fill rest space with CPu ploter but it is fast.

More question is Do Blago requires i7 ot it is ok to go with i5

1 Like

i currently mine 46.3 Tb’s with a i5-4460 (4 cores, 4 threads) and my read times with the newest miner (blago Cpu) is approx 20 seconds. i believe a i5-2500k is approx the same. This is with 6 - 8TB(7451) seagates connected by usb 3.0 and a little bit on some internals, i am also using a usb 3.0 expansion card with 4 additional controllers - http://sonnettech.com/product/allegroprousb3pcie.html so as not to bog down any controllers as i add more drives. Hopefully this provides some idea on if a i5 would work for you.

2 Likes

My own experience with GPU plotter was not great, but I don’t have a high-end graphics card either. I use XPlotter with 2 threads and as much RAM as I can allocate (8-10 GB), and I plot in 1 TB file sizes. These plots generally complete within 24 hours and are optimized.

For mining, I’ve always gotten much better performance from JMiner than anything else, including Blago, although apparently there is a new version of Blago coming out soon?

Hope this helps.

1 Like

Hi @freely,

From my experience its better to split into smaller size, and plot each using GPU plotter. But don’t just plot directly in to each drive. I, my self had 4xSeagate 8TB Ext HDD USB+HUB, which takes more than 100hrs to plot each (don’t know why but it took me over 5days to plot just 1 drive).

What I do to get around this is I bought 2x500GB SSD, plot a 500GB plot file on one of them (actually its some thing like 465GB) which took me around 1.5-2hrs to plot and then copy to my designated Ext HDD, which also took a little more than an hour to copy. And while copying, I also start to plot the second SSD(with the nonce number continue from the first plot file), since the GPU is now available. By the time the copying is finished, the second plotting is almost finish, I would deleted the plot file in the first SSD and prepare to plot the third plot file. When the second plot file is finished and start copy to the designated Ext.HDD, I start plotting the third plot file on the first SSD. I do this so on ans so on until all my Ext,HDD are full, all plotted.

One clear benefit of this method is you can start mining, using Blago’s miner, immediately right after you finish copying your first plot file. You don’t need to wait for the whole drive to be plot. Furthermore, having many plot files in the same HDD doesn’t slow down the mining, at least not noticeable.

One TIP on this method is when you copy the plot file, don’t copy directly to plot folder right away (ie. \burst\plots) just copy to some other place on the designated drive, after finished then “move” to the plot folder (or cut and paste as you may call). That way Blago’s miner wouldn’t show any error if it read to the plot file when it wasn’t finish copying.

Second TIP, Its also a good idea to keep a log file, an Excel sheet of you will, to log the plot file name, size, first and last nonce number. That way if any plot file shows too many Bad DL, hence plot file maybe corrupt, you can just simply delete and re-plot only that plot file.

Good luck and Happy Bursting!

3 Likes

It can go even lower CPU. Blago doesn’t require high performance CPU at all, however having many core does help when you have many HDD. currently, I tried mining using Blago(the previous version, but the latest one wouldn’t be much of a difference) on many different machines. What I found is, a Pentium G4400 can mine one 8TB HDD at almost the same speed as an i5 mining the same HDD(with few MB/s difference which is likely due to MB and other component). However, when I have more HDD drives than the number of CPU core, I did fine some slow down. I mine 8x 8TB, 4 Int and 4 Ext USB HDD, on a Celeron G3900 (2core 2 thread) is significantly slower than i5 7600K (4core 4thread)

Good luck and Happy Bursting!

P.S. Sorry for the multiple editing, English is not my first language. So, there’re many spelling and gamma mistake.

1 Like

Thanks all for suggestions. I have decided to plot whole disk in one file + 1 small to full the disk. Plotting takes 18 hours with GPU and 18 more hours to optimize.

Lets see how CPu acts after I will turn on whole 15 disks.

1 Like

Why are you plotting in buffer + optimization instead of doing direct mode?

1 Like

I stareted to ask this question to my self as well. Have to try what how much time it takes

1 Like

Ok direct mode fails, Was plotting big file. GPU writes full disk in 12hours and then starts to count nonces it from 30 000 n/s drops to 3500. from 12 hours lasted to 5 days.

Buffer + optimize takes 36 hours for 8 tb

1 Like

sounds like the hard drive is going to sleep at 12 hours maybe

1 Like

I see that disk is spinning on 99% on TaskM and system is taking whole 16 gb ram.

1 Like

What is your stagger size?

1 Like

I am using one file: 30408703 nonces, stager 16384
Other file 112496 nonces and 14062 stager

My device configuration for 16 GB RAM with 1070 GTX mini is 1024 384 8192

1 Like

To me, the stager number is way off. Assuming you are using GPU plotter to plot, the stager should be the same number as hashesNumber. Furthermore, your globalWorkSize and localWorkSize is very odd. Since you have 16GB of RAM, you should be able to use larger globalWorkSize and the localWorkSize should be able to divide globalWorkSize and get integer value. And all this numbers should be a power of 2. Please have a look at the README file in the plotter’s folder.

Here’s my start.bat file to plot on a 500GB HDD
gpuPlotGenerator generate direct d:\1793389570451629XXXX_400000001_1900544_8192
My Device confiure
0 0 8192 64 8192
My plotting system is a Ryzen 1800x 16GB RAM, GPU RX480 8GB.

Another possible issue is, since you are using Nvidia card, are you using the CUDA version of GPU plotter? If you are using the OpenGL version, of cause its going to be very slow. This also had been mentioned in the README file.

Good luck and Happy Bursting!

2 Likes

@freely Here are my devices.txt values for a GTX 1060 : 0 0 2048 128 2048

I could not use the recommended values, probably since I also use the GPU for my graphics.

The direct mode operation sounds normal: one fast full write, followed by a much slower series of writes. These look alarming at first, as they will be very slow. They do get successively faster though, as do the time estimates to complete.

I mentioned before I did NOT use one big file…I had unexplained lockups or failures on a really big file. I now plot three or four 2TB to 3TB files to fill a drive. You might want to try this.

To help with settings, I wrote a small piece of software that lets a user input the free drive space, a lower and upper RAM range, and the number of desired plot files. The application then chooses file sizes and a stagger value so as to minimize leftover disk space while using the largest stagger value in the range allowed.

Below is a screenshot for an 8TB drive and three plot files. Note the equal sized plot files, the small amount of unused drive space and that the stagger value divides evenly into the plot file size while fitting within the RAM window.

NOTE: Do NOT use these values as the calculations are based on YOUR exact free byte count!

2 Likes