I also turned on my I7-2600. I bought a NVMe 1 TB and put it on a pcie to NVMe card.
I run 4 plots simultaneously with 2 hours interval between the next start.
Tried yesterday with 3 plots but I believe 4 is possible to. With 3 plots a got 6 to 7 plots per day. Average time with 3 plots 34.000 seconds per plot.
Later this day I’ll update my reply with the average of the first 3 plots while running 4 simultaneously…
FWIW, I fired up my i5-2500K for sh*ts and giggles. It actually fared surprisingly well.
With a 1TB Rocket4 NVMe, it can do one plot at -b 3390 in 25k seconds.
3 parallel plots average 36k seconds each.
Swapping out the NVMe for a SATA SSD lets me do 2-by averaging 32k each. I know it’s not apples to apples, but this suggests CPU saturation rather than IO bottleneck.
Finally, for fun, I tried making a bunch of 1-by plots using a 7200rpm spinner only: 38k. Not terrible all considering.
Good to hear that many other are having good performance with these old CPUs!
I finally received today the extra RAM I bought, so will update this in a few days!
My plan is to start by plotting 4 in parallel (1 plot on each HDD).
Let’s see if it holds the 40k sec. per plot after a few days… if it holds will add 2 more.
Temp size wise the 1TB NVMe can do 4 plots simultaneously. Average load is about 20% and max used space about 75%
So maybe a fifth plot could also work. However the CPU is the bottleneck here. Amount of threads are set to 3 but the first and third phase are taking way longer when doing 3 plots simultaneously…
First plot is about to finish after 11 hours… Tomorrow I have more data from the other 3 plots and even more…
I changed the config back to 3 jobs simultaneously… 2 jobs at the same time in phase 1 my CPU couldn’t handle. Times went from 10000 seconds to 18000 seconds.
The same goes for phase 3 so eventually my jobs were running 13 or 14 hours in stead of 10.
The output is bassicly the same or even worse.
I use a plot manager in Windows. Settings are:
-3 jobs simultaneously
-3500 ram
-4 threats
-Max 1 job in phase 1
-Early start for next job when 1 job is entering phase 4
Bro can you post your logs plz?
My setup :
8gb ram (2400)
I7 7700 hq
Ssd 512go
2t hdd
With one plot about 12 hours
With 2 plots paralel 23 hours to finish them all
How you can mange to do that ? And how czn you manage to parallel plot on 250ssd?
Are you plotting on the SSD? Which model it is?
If you are plotting 2 on parallel on the HDD, that times are completely normal. The key for plotting on HDDs is not running more than one plot simultaneously.
I’m not using the SSD to plot. I’m using the HDDs to plot, 1 plot on each drive at once.
I use 2 threads, 3400MB RAM per plot.
You can see my plotman stats after 10 plots, running 2 in parallel:
And here are my stats so far running 4 plots in parallel (1 per HDD):
Am using the cli on windows and i cant even get 2 plots a day not even on solo plotting
My temp ssd 512gb called hikvision n200 external with usb c
I tried 4 threads and 4000 ram for each plot in parallel and i got 2 plots per 24 hours thats insane
In whiche face you start the second plot ? I do it at 36%?
My pc is so bizarre !!!
I have a recycled iBuyPower laptop:
3rd gen i7 which clocks at 3.5ghz in turbo
16gb DDR3 mem
512GB ssd for os and node
1x1.5tb 5400rpm in bay 2
1x1.5tb 5400rpm in cdrom bay
3TB External
Win 10 x64
I am using this as a farmer - is the processor overkill or if I were to run a single plot with 4 threads and 8GB memory, the idle 4 threads and 8GB memory would be able to successfully farm and run any corresponding processing in the event of a successful hit, while the plotting was still going on…or is the system entirely a potato altogether? Please assist…
For plotting I am using another recycled Hp laptop with 6th gen i7, 32GB RAM on PopOS - running two in parallel with 4 threads each and 12gb and 16gb resource allotment on 1tb 2.0NVME - takes 7hrs to finish the two and I transfer them over to the farmer…
I am in the third world so 1) we don’t have access to reasonably priced computer hardware, nor the variety 2)the electricity here is unreliable, laptops prove to be relatively durable against small surges, are less power consuming, and have batteries so built in ups in event of power outages which are frequent here…my idea is not to hoard as many hard drives but to max out available drives around the house, pick up a few reasonable ones in recycling to max out the ports and leave the farmer alone for a couple of years. If I hit, I commit 50% of XCH value in hardware again and hold onto the XCH… If this plan pays consistently I can scale my investment just as frequently to invest close to 75%…with this plan I have a data center in my dreams, in my dreams that is…
I could’t find many info about that SSD, but I guess it is you bottleneck…
I would try plotting just one plot on the HDD to see the times and have something to compare.
Are you using your PC for something else while plotting?
And I think it’s not needed 4 threads and 4000 RAM, you won’t notice any significant improvement. I would keep using default 2 threads and 3400mb.
I’m not staggering the plots on this machine as I only use one drive per plot and I have enough threads and RAM.
I guess you can also use this machine for plotting, just plot with 2 threads and 3400 RAM, then Just check the debug.log file to see if farming is affected in any way, but I doubt it, I think you have enough resources to plot and farm on this machine.
Imho you are wasting resources on this machine. Maybe you can plot 4 in parallel on that NVME drive; with 2 threads, 3400 RAM and 2h stagger between each plot. But I would recommend you trying first only with three though.
Sure, go for it! I think the idea behind this project is using resources we already have available (at least at the beginnings ).
Thanks man, I just started a plot on the farmer with 4gb ram and 2 threads…
When I do simultaneous Plots the two take 10060s to end stage 1, so if I plot 3 in parallel then plot two I start at time delay 10060s and plot three at time delay 20120s or something like that with 2 threads…
With the current configuration where I am dedicating 4 threads simultaneously and churning out the two together, is there any physical damage occuring to the processor? I have observed the resource monitor and its never that the 8 threads are always occupied, the 100% activity marker jumps from core to core at any given time 2 threads get occupied 100% simultaneously but again not in a rush to fill the drives up, doing two at a time with higher resource allocation, does it improve the consistency of the finished plot?
(Imagine if you were operating a video rendering software and gave a job to a better resourced pc and the same job to a lower powered one, the final results have visible differences, I understand the analogy does not fit but you get the idea I am getting at here regarding the quality of the finished plots…if at all that may even be a thing)
@guillem
Hi, I tried plotting on HDD drives and I stumbled upon a problem that I cannot resolve. When I ran 1 plot on hdd as temp drive, it finished in ~13h, which I consider good. Then I added second hdd drive to be temp drive also. I ran one plot on each hdd drive, thinking that times would be same because there is one plot per hdd, no parallel plots on one hdd. But times got bad, it took ~25h FOR EACH plot to finish. Each plot was ran with 4 threads and 4GB and another test with 2 threads and 4GB ram. Processor was utilized about 30%, there was enough free ram, etc. I tried different sata ports for those hdd drives, but it was the same. Final plot file was written to third HDD that is dedicated for plots.
This are the pc specs:
AMD FX-8350
Asus sabertooth 990fx r2.0
2x8GB DDR3 Kingston hyperx beast
2x500GB WD Green for temp drive
1x4TB Seagate as final plot destination drive
Does anyone know why this happens or did anyone had similar experience while temp plotting on hdd? It’s odd to me that when I do one HDD plot time is 13h but when I do 2 hdd plot time is 25h.
Also, I tried rearranging sata ports between controllers (motherboard has 8 ports, 6 for 1 controller and 2 for another). But that did not make any difference.
Remember cpus did get “faster” by using parallelization, not so much clock speed. So them old processors still run at 3.4Ghz, and it so happens that a single Chia process does not seem to be doing much per clock, so high clocks == good.
well, as you can see in the reports i posted, my overall time per plot also decreases when adding more plots in parallel. I assume that’s completely normal, cause you add more load to computer resources.
however, in my case is not as significant as yours. it went from 11hours or 38k sec. (with 2 plots) to 13h or 46k sec. (with 4 plots).
have you tried staggering the plots? which OS are you using? how are the temp drives formatted?
I did not stagger plots, just as you because each plot was on different drive. Drives are formatted as NTFS and have indexing disabled on them. I am currently using windows. I’ll try with Linux tomorrow and with pci to sata adapter (just to see if there is a problem with sata controller).
).