Performance loss plotting on multiple HDDs in parallel

Plot in HD is an output not to spoil SSD.

Why are you going to make 4 parallel plots on a very expensive, life-limited SSD when you can make 4 parallel plots on 4 cheap HDDs?

11,15h is good time. What is your hardware?

I plot on an NVMe because the expense of NVMe and SSD replacement is trivial compared to the price of storage for your plots and because I cannot produce plots nearly as quickly on an HDD plotting setup for which my machine is not suited anyways.

I am warranted for over 1200 TBW on my 2TB EVO 970 Plus NVMe which is over 2000 K32 plots. I should get closer to 5 or 6,000 K32s plots from my $440 NVMe based on the data we have been collecting.

Many NVMes and SSDs that have reported show TBW already at 2-300% over their warranted TBW and still going strong. Very few report dying early or even at expected warranted TBW.

if you ever hope to win a coin, how quickly plots are produced is much more important than an under 15 cent per plot NVMe replacement cost. You save nothing if you are plotting slowly.

BTW After using every ounce of muscle in my weak plotter I plot 3 K33s every 17.5 hours, equivalent to 8 K32s every 24 hours. This is as fast as I can plot, as my bottleneck is in the CPU, not at the plotting drive. If I was plotting using HDD I would have a huge bottleneck at the HDDs and produce far fewer plots.

Most of the money I have spent is for storage, and it will take me two months to finish filling what I have. Producing more slowly would make me sad.

I’m running the following:
Dell T5610 with dual E5-2680v2 and128GB RAM
Rocketraid 2840A 16 channel controller with 8 x HP 2.5" 10K 300GB HDD
I have a few 10TB 3.5 HDD as final drive currently hook into the Mobo SATA ports. (I have room for 10 more final drives - 8 from the controller and 2 on the mobo)

PrimoCache settings (as you recommended) is 100% write, Level 1 Cache at 64gb, 16KB block size, deferred write at 160 seconds, Mode: intelligent.

Testing 8 parallel plots now on all 8 drives. Will post results (hopefully in 12 hours).

Thanks for your recommendation Ninja!

1 Like

Good work! Keep me informed of your tests. I’m always trying to gain speed on HDD.

You should check out Sloth’s youtube video on plotting with HDD. I just used his idea.
I got 8x HP Enterprise 2.5" 300GB HDD for literal $10.00 each including shipping. (You can still get them for $12 to $14 today). $80 Total
I got Rocketraid 2840A 16 channel controller for $150. Two breakout cables for $50. $200 total
Total of $280 for this set up. These drives should not burn out and longevity is suppose to be “unlimited”.
Thanks to NinjaDummie, with PrimoCache, I just plotted 4 plots with 4 drives in 11.15 hours. I am currently testing 8 plots with all 8 drives. IF this works, I hope to get 8 plots every 11 hours, or 16 plots every 22 hours. And you are right, this is not speed olympics. If I can plot 16 plots everyday for x numbers of days, I think it might work.

I don’t have the time, space, or inclination to fix something that aint broke.

NVMe or other high speed SSD is obviously the way to go for plotting fast - no question about it.

I’ve been convinced though to set up a machine I was planning to use a secondary harvester (otherwise was just waiting to be put together when required), put all the disks that are waiting to be farming disks into it, use those as tmp disks. and then any plots it makes are basically free - I did something similar initially on my main farmer, but I’m a bit too fanatical about challenge response on that machine to want to let anything else run on it.

So yeah, any machine that isn’t otherwise working could probably be producing > 0 plots, and even it took 24 hrs a plot - that’s better than 0 plots.

1 Like

Great news! All 8 plots completed around 11.78 hours to 12.1 hours.
FYI - I had six disk in JBOD plotting separately and two disk in Raid 0 (plotting two plots) - The six JBOD disk finished six plots under 12 hours. The two plots in RAID 0 finished above 12 hours. All good!

1 Like