Gigahorse GPU Plotter, GPU Farmer, Plot Sink: Boost Your Chia Farm by 47.2%! Plot in <2 Minutes

I’m guessing your drives are formatted NTFS, this will cause slow write speeds in linux.


Indeed and not by a little bit either.

I’m plotting in Linux but farming on windows. I’m now using ExFat that has good copy times in Linux. But honestly I have no idea if there might be a reason not to use Exfat for farming

Didn’t the drive manufactures set the external drive to be formated with exfat so you could plug the drive into a windows or Apple machine right away.

1 Like

There is 1 problem :

Diff. between ext4 and NTFS.

Hosting plots on hdd ( 6TB, 8 TB , 10 TB, 18 TB ) – is better in NTFS.

example: my 18 TB harddrive can host 165 plots - ntfs. If i format it in ext4 – 163 or 164. I lose 1.
plots bring us money…

Transfer from NVME ( ext4 ) to sata 3 drives ( ntfs ) can be between 45 - 220 MB/sec…

Not true in every case…

$ mount | grep /mnt/hdd2
/dev/sde on /mnt/hdd2 type ext4 (ro,relatime,seclabel)
$ ls -l /mnt/hdd2/*.plot | wc -l
$ sudo fdisk -l /dev/sde | grep model
Disk model: TOSHIBA MG09ACA1

Just 2c,

Update: Just checked, every my HDD (same specs as above) has been hosting 165 plots.

1 Like

Did you disable reserved space in ext4? sudo tune2fs -m 0 /dev/sdX


I will try this tomorrow

Thank you. I did it on nvme ext4 and i got extra 86 Gib.

I need help/advice.

I plotted Compression C8 - 108 hdd of 8TB ( refurbished) for testing Gigahorse/flexfarmer

After 6-7 days of farming - I already have 3 hdds dead.

I measured all 24 hard drives on the system and it had 221-223W/h meaning 9.21W/hard. = full load every minute.

At the same time, the other hard drives plotted no compression ( 38 Harversters + 1 farmer full note) have a consumption of 4.7W/hard.

What I am doing wrong ? I dont want after 6-7 days another 3-4 hard dead…



I noticed in the table above, 3060ti can farm 1.78Pib, if it is K33, is it still 1.78Pib, or half.

Can anyone else confirm if apx dbl elec usage is correct?

Did you set difficulty to 18+?

1 Like

I had 50% of refurbs die on me, it’s all a big gamble.

~9W does not mean full load, it depends how old the drive is, if it’s 7200 or 5400 RPM, etc etc

he did

1 Like

What is the login link that is needed to change the difficulty?

You will need to specify the Launcher ID and use the Login Link from the Chia client to modify the difficulty.

Chia client?

I would assume that @BlackEgo got all of his drives more or less of the same type. I would also assume that he would mention that the drives with higher power draw are different type, thus maybe such difference could be explained. Although, those are my guesses.

Still, @madMAx43v3r the question is whether assuming that both sets are exactly the same, where one set has no compression, and the other C8, will there be a big difference in disk access, or it will correlate with compression level (e.g., 100% more disk access for 50% compression - basically, disk access matches the number of plots per disk). If the disk access is correlated with the number of plots, we should not see much power draw difference, as the disk idle time will dominate disk power usage.

I doubt it. That 5W means that the disk is mostly idle (8TB disk has ~70 plots, assuming 512:1 filter, it will be hit once every 70 secs or so for few KBs). Assuming that we have 50% compression, it will be hit every 35 secs (2x plots, but potentially less read per plot), what also means zilch. Also, some Seagate drives can be set to go to idle_b mode very fast (say 10 secs), so after each lookup the drive can reduce power draw right away.

Actually, here is the datasheet for 8TB Barracuda - Doing 100% time writes, that disk draws just 5.3 W, and while idling (no stepper, board goes to power saver) it drops to 3.4 W. Neither of those is anywhere close to 10W. To get somewhere closer to that 10W, those drives should be 15k RPM SAS drives, and I don’t think that those drives go down to 5W at any point (my guess).

1 Like

Chia client is what your gui / cli talks to.

In the gui, login links are found on the pooling tab on each nft. ( iirc 3 little dots on the nft ).

Not sure how that all works with comp plots and max’s interface though.

They are different type, 8 TB vs 18 TB for the other farm.

I don’t believe those numbers. Also in practice you have more than just HDDs, like a JBOD or some other type of enclosure with fans, power supply losses, etc.

I have 24 hdd on system with 3060. My hdd consume 9.25W /hdd - total 222W only hards + 88W sistem with gpu when my old 20 hdd ( 18 TB) system consume 127W all togher.

So we have 6.35 W / HDD for the newer 18 TB uncompressed and 9.25 W / HDD for the older 8 TB HDDs with C8. It would be helpful to measure power without any load to see if additional power draw is from higher idle draw or from additional farming load.

We also dont know how @BlackEgo is powering his drives, for example using a dedicated power supply for every USB HDD is very inefficient.

Exactly, he has 104 C8 plots on the 8 TB HDDs, which is a lookup every ~50 sec on average. I can’t believe that would cause any increase in power draw over idle.

Now if those drives are powering down in-between, that would cause a lot more power draw, because spinning up again pulls a lot of power.


Thank you. That was my question. So, whatever is going on in his setup has nothing to do with compressed plots.


I use flexfarmer and the plot nft is from a wallet i don’t have “warm”

Why is it bad for both at the same time?