Wow… is that true ? No need SSD VNMe, SATA…for plotting ? All using DDR3 RAM ? And how many HDD for store the Plot as it was created ? And what happened when all plot is full in the HDD, if it is overwrite the old plotted that was stored in the HDD, do we lost it forever, as it will be last for about nearly 5 years for anothet winning CHIA XCH next round ?
Ok. So… You will need 2 types of computers.
Plotter - for making the plot.
Farmer - for storage of the plot & farm chia
Plotters - as powerful as you can
Farmers - as electric efficient, as you can
Plotters - making the plots & put them on the farmers.
In my configuration that was ~1300$ Kit from Aliexpress + 120G SSD for a system, +1Tb SSD for the final plot, case, water cooling, and power supply.
the trick is that you don’t need to use SSD as temporary storage, which burns out your SSD quite quickly.
You make a virtual HDD in your RAM & making a ploy using your 1tb SSD only for 1-time writе. As Plotting is done - you move your plot to your Farm storage computer.
I’m using 24 HDD bay storage based on SuperMicro X10SLL
Do you need the same configuration? No.
But as I understand here (And I confirm that!) - an external drive is not a solution.
10 x external drives mean 10 power sources, which means a bunch of data/power cables that make you crazy in the first 2 months. So… try to build that solid from the very beginning.
Regards.
1 Like
I know of no one that has burned out their SSD.
I have been plotting on Samsung 980 Pro NVMe drives, nearly non-stop, for close to two years, and they work like they did when they were new. No performance degradation.
Perhaps “quickly”, when compared to non-chia use, is technically accurate, because you would never, in a lifetime, burn out a quality SSD. Perhaps plotting for 5 years, non-stop, will burn out a good SSD, whereas that SSD (if not used for Chia) would have lasted a lifetime. So, yes, 5 years is quick compared to a lifetime. But I have not seen any figures on how long quality SSDs will last with Chia processing.
If you run a utility that claims that the SSD is critical, that is likely the manufacturer looking to make new sales for perfectly working SSDs. You hit the terabytes written value, and you get a critical warning. Yet, the SSD keeps working at full speed for 2x, 3x, 10x that value.
Perhaps some 250 GB QLC SSD might burn out after a few petabytes, due to its limited writable storage space and the relatively cheap NAND fabric. But it is unlikely that anyone is using those SSDs, mainly because they would be slow – which, ironically, will make them last longer, because you would pump out fewer plots.
I read that BladeBit is faster than madmax-on-RAM-drive. I do not have the reference, but I did read it either in this forum or reddit. Others can chime in on BladeBit vs madmax when using RAM.
2 Likes
Thanks @GolDen for your good advise,
Could you please share details how to setup virtual HDD in my DDR 4 and making a Ploy using 1TB SSD ( in this case, can i use SSD, NVMe PCIe 4 M2 1TB ?)
Another question is 
Can we store Plot in HDD and then upload to Cloud ( ex: Google Cloud/GoogleDrive OneDrive) then, join the Pool for farming CHIA ?
If you have enough RAM, then BladeBit is probably best.
I have only 64 GB of RAM. So I can’t create a plot in RAM. But I read posts from others on doing so.
Also, I believe that with BladeBit, you do not have to create a RAM drive (one less step). As far as I know, BladeBit is coded to run from RAM, just by using it. If you do not have enough RAM, BladeBit will refuse to run (I know this, because I tried).
I am not against @GolDen
1 Like
Ah, I’d got the problem, the main issues is RAM. I will try to use 3x 32 GB RAM to check, but the problem is how to setup and configure it, do tou have any link to show details how to setup up ? Many thanks yoi and @GolDen
Depends how you define burned out.
Id expect that drive is probably useless now for long term storage.
Itll be fine while pc is on for short term memory ( chia plotting ), but useless for storing data long term if you want it to last after turning the power off.
At least thats what ive gathered to be the case.
Not personally tested it though.
1 Like
I’d recommend 128G minimum!
Why? that give you a chance to use MadMax (I’m not familiar with BladeBit)
And… about “I’d buy a plots” - check your internet connection first. It’s 101Gb file.
Even in my local network (1Gb) its takes ~ 20 min for transfer.
If your internet connection not fiber optics → Looking for local “plotter guy” & order a packed HDD with post delivery.
If you are lookin for “passive” investments - I’d recommend you order everything like service. Just find someone who can make everything for you. If you never build a custom PC, install a linux, etc - think twice. From other hand - that’s a VERY interesting quest. If you can spent a few months to teach that - you will get a lot of fun & knowledge.
2 Likes
May be, my friend. May be…
I never tested that by myself. But, as I understand, you are using SSD in the wrong way.
HDD/SSD means storage of data for long. RAM - for save the data until PC is working.
When we plotting - we use SSD for short period of time. Until it will be transferred to the farm system.
So… from the beginning I decide to use RAM for that.
And… return to the “performance degradation”, I’d be glad to know, what happened, when plotter wright on the “bad” sector. It should be controlled by CRC code etc… but… I can’t imagine how that will work in the plotting process.
Not really. You do realize that plot is a collection of proofs and for k32 it is about 100GB. On the other hand, one bad sector is just few KB. So, if a bad sector will happens, as long as it will not hit the header portion, it will destroy just a very tiny fraction of proofs. So, one can argue that such plot is still in a very good shape, i.e., drives that cannot be used for anything else may still be good enough for plotting, as long as the degradation doesn’t progress rapidly (platter damage).
Also, for SSDs, the bad sector problem is rather less frequent issue. What is happening is that gates are getting structural damage, and when SSD is powered off they discharge (much) faster. So, as long as you plot, your SSD may work well beyond their expected lifetime. However, when you turn your box off, after some time (days), all the data will be lost. Although, once powered on, you may still be able to format it, and continue plotting. This is the main reason to not use boot drive for plotting, or rather use plotting SSD only as a scratch drive.
1 Like
m-m-m… I mean the other scenario.
The plotter writes the plot on SSD.
SSD writes it block by block (correct?), counts CRC for each block, and writes it. Ok.
A few moments later some bit is burned & where’s “1” became “0”. But!!! SSD knew nothing about that (as I understand!) because, as far as I know, it does not check CRC every time without I/O.
Ok.so… we have a situation when our data is corrupted & we can’t restore that.
Now it’s time to transfer the plot to the farm. As I understand, it’s read data by block and checks CRC If Ok - put it into the buffer.
But what happens, if CRC isn’t ok?
Does some error message & plotting process stop? So we will lose the time.
That was kind of answered.
There is really no need for CRC on the plot file, as a single cluster damage or a handful is not really a big deal (as long as the header structure is intact). The header structure can potentially be checked by chia plot test. In case it is not checked, during one of proof lookups that plot will be flagged as bad. So, again not a big deal to ignore CRC processing. You can also look at that that if such corruption will happen, then instead of having a proof like 123654 you will get something like 120054 (what will be a garbage proof, so will never show up in lookups).
The SSD is trying to average writes, so gate destruction is rather similar for all cells, and most likely it will just refuse to cooperate rather than throw a fit. So, there is really no direct comparison between HD and SSD degradation.
I am not trying to advocate to plot on a compromised SSDs, just trying to say that in case some bad things happen the damage to the plot could be less costly (as far as potential win value) than the cost of electricity to replot such plot. So, as long as the system is happy with such SSD and chia is not complaining about such plot there is no point to worry about it too much.
1 Like
I do not know what you mean.
I purchased my SSD specifically for producing Chia plots. How is that wrong?
I suspect that the vast majority of people plotting are doing so via an SSD, and not via RAM.
They could all be doing it wrong. But it is more likely that most folks do not have 128 GB of RAM.
For K33 plots, you need just north of 521 GB of RAM (plus RAM for the OS). That narrows down the field of folks that could use RAM for those plots.
For K34 plots, you need over 1 TB of RAM (plus RAM for the OS). The eliminates just about everyone that could use RAM for those plots.
SSDs are for manipulating data at high speed (not as fast as RAM, and not for the same purpose as RAM, but fast, nonetheless).
It is why virtually every PC is built with the OS on an SSD. And the OS is forever reading and writing to the SSD.
The plot will likely be invalid, and will not count towards your total farm space.
$ chia plots check -g plotname.plot will very likely identify bad plots, as well as the GUI will very likely show invalid plots (it shows duplicate plots, so it would probably show invalid plots, too).
If after years of plotting, and my SSDs have no more trim space, and become unreliable, then that SSD will be removed and dumped.
2 Likes
Actually, you have got the same opinion with me, i’d used to be thinka bout to hire someone as consultant, support me everything from the scratch as setup computer configiration and setup everything, install software, maximize the current computer configuration, join thr pool…to have a plot and earn CHIA,
His/her income can be discussed or negotiation depend on the result that can bring !
Bte, if you know someone, please introfuce to me to support this many thanks
So, wjat happened if we stored plot in the HDD and for about 3-4 years the HDD is invalid/errors to operate, in this case ? Do we lost all of our data/i mean ,lost all of our Plot there ?
In this case ? Can we use Cloud ( Google Drive or One Drive or anything elase as backup solution ?)
you will buy a new HDD. Plot again & fill the HDD with a new plots. Doesn’t matter how “new” your plot is. Your reward depends on how many plots you keep totally.
& (as far as I know!) there is no “cloud” solution for chia (in terms of cost efficiency).
Access speed to the plot is matter. So… big enough cloud with quick access to the plot will cost you more, then reward you get.
2 Likes
There is 10+ years payback (more than hardware lifespan cycle). If you want to try in cloud, you can attach storage to server on AWS for example. It was profitable in first months of farming.
2 Likes
I recommend this site, they plotted 154 plots for me and I had no problems.
I hired them because I didn’t want to kill my NVME
1 Like
Doubt you would have for 154 plots.
1 Like