K33 @Ram, 2 Servers difference of timing plot

Hi guys

I have 2 server plots all on Ubuntu. The GPU is the same for both.
I use Gigahorse farming and plot.

Dell 2x Xeon 12cores 512Gb 2133mhz
Time plot: 6min

Other Bundle
MZ32-AR0 (rev. 1.0) and EPYC 7551P 32cores 512Gb 2400mhz
Time plot: Starting with 18min and after 12 hours 33min per plot and farming.

What do I miss here?? There Is any different command line to EPYC?
When is farming together significantly increase the time to plot?
Gigabyte there is a poor performance at Linux?

I had similar problems (reported in another thread). I originally had Ubuntu 22.04 installed, but after few hours the OS was giving up and plotting started to grind. I was banging on every corner to no avail. Actually, plot times were kind of fine, but moving plots was grinding, so the plotter was constantly waiting for a free space on NVMe.

The end result was that I did a fresh install of 20.04. Also, I let cuda_plotter only dump to NVMe, and used shell scripts (basically mv cmd) to move plots from NVMe to final dest. I think that even right now, when I use cuda_plotter (or plot-sink) to move plots to final dest. it craps out after few hours.

So, I would rather not try to troubleshoot your box, but would do a fresh install of 20.04 (LTS) and made sure that your GPU drivers are good to go right after install (nvidia-smi).

By the way, just looking at the console is not helping much. Run some resource monitor, as well as monitor your temps to check what is potentially the issue. In my case, RAM temps were really high, so I had to add extra fans to blow on RAM.

The copy is not my problem. I fixed it coping for 10 hdd different, and the SSD never is full at the Gigabyte.
With the memory being too much hot: , I can agree with this can be an option, and I will up de fans.
But is my big problem.

The problem is: Why a new EPYC with 32 cores is slower to plot than a dual 12 cores???
The GPU, memory, SSD, and SAS control are all the same…

I can’t understand it.

yeah the CPU speed should have little impact anyway, using a GPU to plot.

Is it possible that there is some problem with the Pci-e slot speed?

It may be because you’re farming at the same time? That was a frequent issue for USB drive farmers back in the days.

1 Like

Yes, of course. The CPU has little impact, but the difference in performance is enormous.

No is that, I stopped farming to test and I don’t have USB drives.

I am back here to let you all know, I solve the problem after long tests.

I needed to reinstall the Ubuntu to solve the problem.
Now is plot a K33 in 7min for more the 12hous. Along Chia update together
.

But I backed all stuff to Dell, to make all tests.

I used here
Dell Precision rack 7910, 512Gb
RTX 3070
2x SSD Dell 1.9tb SAS in Raid 0.

With this GPU I can’t plot and farm. But before reinstalling, It spend 35min to plot a K33.

If you ask me: what can do this? My answer is: I really, I don’t know.

Now I will reinstall the Ubuntu on EPYC and make a test. But first I will buy this cable to connect the SSD directly at de motherboard.
PCIe 4.0 SlimSAS SFF-8654 4i to SFF-8639 U.2 Cable

Does anyone know if this port Slim can of the Gigabyte Motherboard can do a Raid 0?

1 Like