Hi to all,
I am having my gigahorse plotter fail after some 10-15 plots. It first starts to slow down from 5-7 min to 12-15 min and after a while, it quits. Sometimes it freezes the whole PC. I am plotting on DELL 7810 WS with 64 RAM and 4x980Pro on an ASUS HYPER m.2 to PCE-E x 16 adapter, using MSI 3070 GPU. I have checked everything and it seems all is fine but… Temps are fine, the WS have been user for rendering before and works fine, and just do not know what is the problem here. Can anyone help?
Someone said that ubuntu 22.10 is the problem and you should try 22.04 but you didn’t say what you use?
You could just run in batches…plot 8 plots in a loop until you work it out
I also have Dell T5810 with a ASUS Hyper card fille with 4 Samsung EVO 970 1TB units. I also have 256gb system memory in my box and using Win 10 with no issues making plots for MMX. You have your 4 980 spanned for one drive?
If you don’t plot and just let it sit, and go back 30 minutes later whats the status of the box?
Are you running harvester or chia_recompute_server while plotting? if so, that’s the problem. You will need a 12gb GPU to run all at once. At least on my experience.
1 Like
I’m using 22.2.0.4 and confirm I also have that issue. The only thing I can figure it to be is the 64GB of RAM is not enough. I can plot 10-12 before it slows down and kills itself. I ended up creating a script to loop every 5 minutes and check if the process was running or restart it.
22.04. And I have already done the “batch plotting” with 10, 8 and 5 per batch. The same thing, it quits plotting with no reason and error code. I sometimes find the whole terminal window closed or just the plotter frozen.
I am not, I just do plotting on this machine.
Maybe. In order to test the GPU and Asus Hyper M.2 to PCIE x 16 adapter I inserted them in another PC, with AMD 570 mobo and Ryzen 5900X, 64 RAM and started plotting. One freeze in several hours, now at the 17th plot and doing OK.
1 Like
Will have to test that, I have not tried that yet, I usually start the machine for plotting. And for the record, I have done a LOT of CPU ploting on this machine and it worked great.
You could save some memory by removing the desktop
Hmm. I case I find the memory is not enough, it might be a kind of solution. This way, I need the desktop for remote access and other things.
Btw, I have tested the GPU+AsusHyper and I have the same problems on another 64 GB, Ubuntu 22.04 machine.
256gb or bust!!! Will save you headaches real quick.
Use - w (w8 for plot to be copied) and see if it degrades over time.
Very likely your mobo is @ max at the saturation of pci lanes (bandwith), in > out in data terms.
Ram @ 2667 or default
Comeback with the results
I am searching for a memory upgrade on eBay since it can not be found here locally, but everything I find takes weeks for delivery. Plus I don’t want to buy memory for a PC that is having problems plotting, do I? 
I have tested the GPU+NVMEs on another mobo and have had very similar results. Quitting after a while. 
Gen3 forced on pci lanes? X570 have 20+4 lanes to be shared.
Really that behaviour is consistent with bandwith limitation/Ram not stable / nvme overheating (low cache) in short, u are producing more data than your system can spit out. So check the chipset of the mobos and how they work. Understanding that can help how to setup a stable plotter.
Dell T7810 with a E5-2683v3 can handle 40 pci lanes. So… Nvmes overheating or ram
(gen3 here so no need for limitation)
I have that same asus nvme and usually the stay around 50ºc but with proper ventilation on the case.
My best 2cents. Hope it helps.
MemTest86 - Official Site of the x86 and ARM Memory Testing Tool
Use this to test your memory, I have been using these versions since 1995
The temps are fine, I checked them so many times. As for lanes of X570, I am aware of that but I am using this PC to find if/what is wrong with my setup on 7810. It is just that the problems are similar on both, so in my opinion, it is the GPU+ASUS_HYPER or the 64RAM limit I have on both machines.
I will test plotting on another XEON machine later, with another GPU and NVMEs if I can find some around.
1 Like
I have the T5810 and a T7810 both with 256gb system memory.