My Intel DC P3700 1.6TB drive has an endurance rating of 30 PB, considerably more that 2 to 3 times a consumer drive 
Even my 200GB S3700 drives have an endurance rating of 3.65PB, again considerably more than 2 to 3 times a similar size consumer driver.
3 PBW for a 2 TB drive is definitely not the norm for enterprise SSDs. I also have an intel DC P3XXX model drive with a 22 PBW rating. The LX3030 model you mention does not have the kind of sustained write speeds required for continuous use like an enterprise drive as well.
Yes, electricity cost as well as your W/TB are the primary factors in your operating expenses, as was stated later in the post. You can use the spreadsheet linked afterwards to see for yourself how the numbers work out.
With low electricity costs, and before the plot filter reduction, C15 can be the best choice for profitability. W/TiBe, the metric I was originally commenting on for efficiency, will still be better for lower C levels however.
Wow. I just found out that you can only use 1 gpu at a time to plot, I had assumed that you could use one per instance when plotting as long as you had enough ram. I donāt understand why you canāt specify which gpu you want to use to plot with the -x switch. But after the first instance you only use more cpu resources. Sighā¦ā¦ā¦. If you set up a virtual machine for each instance could you use a different gpu for each then?
My 5950x with 128 GB is pretty much a tossup with my dual E5-2680 v2 with 128 GB - both are HD write speed limited doing CPU plotting -c15 at about 7.5-8 minutes a plot.
Limit is the HD write speed, though putting 512GB in the dual Xeon workstation and using 384GB of the RAM as a ramdrive to plot to did NOT speed it up, even when I used itās GTX1070ti to plot with instead of the CPUs.
I donāt use a SSD, and refuse to wear one out or waste the money on a high-TBW drive like the PNY Chia drive or even the Digital Spaceport things, having 5 machines that can plot at the same time is plenty fast enough than trying to set up ONE super-fast plotter-specific machine and burning a SSD to make it about the same speed as the entire group of machines instead.
Your setup is probably different.
Plot via NFS mounts works.
Better yet, plot local then plow to NFS mounted drives via 2.5GB or faster LAN (2.5 has gotten CHEAP) will match local plotting speeds to one HD as the LAN is a little faster than the HD write speed.
You have to use a GPU to FARM -c10 or higher.
My primary plotting machine doesnāt even have a CPU in it (1u HP rackmount server) but plots -c15 just fine.
The real issue was the total lack of any -d entries telling the plotter WHERE to plot to.
Great plotting machine with no CPU, lovely 
How do you creating your ram disk? can you show exact command?
Im plotting nossd on my Dell T620 in 4m single plotter instance. But Iāve plotted in 2 plotter instances simultaneously(second instance plotter takes 5m30sec),
Ramdisk + GPU
There may be no processor in the plotting machineā¦
The most important thing is to have a central unit or central computing unit, sometimes called a central processing unitā¦
Although various miracles happen in lifeā¦
You definitely can plot using multiple GPUs across multiple instances. The -x switch can specify which GPU to use based on a couple different things, the easiest is to use a portion of the GPUās serial number; something like: ā-x 0abcā.
I used your instruction from the early days, sir.
Created a ram disk and it could access. However read write speed was barely 1GB/sec.
Both CPU and GPU plot on ram disk were over 8 minutes a plot.
I went back using a 10 disk SAS10K 600GB RAID 0 mechanical 2.5 HDD. RAID0 was faster than RAM disk. I could do both CPU (dual 2680V2) and GPU (1080) in parallel. CPU plot was a little over 7 minutes. GPU plot was near 8 minutes. So I am getting <4 minutes a plot output on average.
oddly slow ram performance.
After you created ramdisk on /dev/ram0 you would create a file system right? With mkfs.ext4 after that you would mount it on /mnt/ramdisk after that did you sudo chown $(whoami):$(whoami) /mnt/ramdisk?
as for ram, did you changed ram mode in BIOS for your dell server?
And on what clock speed its running?
From my experience its greatly affects performance.
Do you use supervisors like proxmox or native OS installation?
Command line
sudo mkfs.ext4 /dev/ram0
then
sudo mount /dev/ram0 /mnt/ramdisk
yes?
Thanks! I am not near the Dell R720.
sudo mount ...
and then
sudo chown -R $(whoami):$(whoami) /mnt/ramdisk
When you will had a chance try to look at dellās bios. In my case Iāve got from ~10m to 4m plotting time decrease as well as ram IO bandwith increased from 1Gb/s to 3Gb/s
Of course you are right. I had to use the --no-gpu-mining option otherwise I would run out of video ram on my following instances and the 2nd 3rd and 4th gpus would fail to initialize. I hadnāt realized they were not initializing. With the 4 Tesla P4 gpus I am using I am managing to spit out a plot every 2.75 minutes. I am happy now.
Is anyone able to use a mergerFS filesystem as destination for nossd plotting?
