Dual CPU, 32gb of ram, 2TB nvme. madmax still at 2.5 hours for one plot?

Using a Mac Pro. Mac OS 10.14.6 with 8C 32GB RAM and 2*Kingston KC1000 NVME i get this:

[P1] Table 1 took 71.5531 sec
[P1] Table 2 took 302.718 sec

Just tried it an stopped after that, remember the Mac Pro provides PCI-E 2.0 only for the NVME thus limiting the speed.

If I was willing to spend money on my Mac pro for stuff I can repurpose, I’d get a PCIe x16 to dual NVMe and at least 1x 1TB Inland (or similar) NVMe (just 1 NVME to start). I will try tonight after work, if I have time, using a USB 3 dock and the Adata 500GB SSD.

I am aware that the Mac Pro maxes out at SATA 2 (early spec) and PCIe 2 (early spec).

Is it even worth it to use the Mac Pro for plotting?

I ran another test with my setup. 150 minutes for temp and final drive the inland NVME. I set it to 16 threads.

I have used this system with the original chia plotter and would run 5-6 plots in parallel and could get around 10 plots a day

I have a plotter with the same CPUs same amount of ram and same inland 2t. in a hp z800

Windows times are 53 min ~
ubuntu times are 47 min per plot stable for weeks

I wouldn’t used that SSD as second… That inland 2tb is awesome. I have it in 3 plotters that use those and they don’t give up. One has generated 208TB worth of plots single handedly.

you might be using that ssd as second temp not first. just use the nvme as temp

Are you saying use the inland for Temp and Temp 2? Final location is a HDD. how many threads do you use? Id love to get 53 minutes.

I ran another setup with the nvme as temp and temp 2. Final directory was the nvme as well, just so i didnt have to wait for transfer time. 174 MIN? wft? im not sure where im going wrong? I tried 36 threads for some reason but i usually run it at 16 with about the same results.

Any help would be great, im banging my head against the wall here.

jjs, Are you doing anything differently than i am? besides having a third of my plot time lol
Here is my log…
Keep in mind the D drive is the nvme


Multi-threaded pipelined Chia k32 plotter - 82f92a9
Build 0.1.1 for Windows. Check for latest updates: Multi-threaded pipelined Chia k32 plotter | Build for Windows

Final Directory: D:
Number of Plots: 4
Crafting plot 1 out of 4
Process ID: 5784
Number of Threads: 36
Number of Buckets P1: 2^8 (256)
Number of Buckets P3+P4: 2^8 (256)
Pool Puzzle Hash:
Farmer Public Key:
Working Directory: D:
Working Directory 2: D:
Plot Name: plot-k32-2021-08-03-02-01-de70b68626042f4cbc0964a9856f1464d9ca29840dd6b6f214b9ad19f497b4f4
[P1] Table 1 took 130.068 sec
[P1] Table 2 took 639.168 sec, found 4294919836 matches
[P1] Table 3 took 911.92 sec, found 4294888406 matches
[P1] Table 4 took 1110.46 sec, found 4294641980 matches
[P1] Table 5 took 1083.81 sec, found 4294223625 matches
[P1] Table 6 took 1019.63 sec, found 4293498300 matches
[P1] Table 7 took 702.218 sec, found 4291996410 matches
Phase 1 took 5597.58 sec
[P2] max_table_size = 4294967296
[P2] Table 7 scan took 58.3338 sec
[P2] Table 7 rewrite took 153.702 sec, dropped 0 entries (0 %)
[P2] Table 6 scan took 164.372 sec
[P2] Table 6 rewrite took 304.606 sec, dropped 581448514 entries (13.5425 %)
[P2] Table 5 scan took 163.286 sec
[P2] Table 5 rewrite took 288.129 sec, dropped 762163890 entries (17.7486 %)
[P2] Table 4 scan took 148.496 sec
[P2] Table 4 rewrite took 254.921 sec, dropped 829046012 entries (19.3042 %)
[P2] Table 3 scan took 149.46 sec
[P2] Table 3 rewrite took 255.946 sec, dropped 855207151 entries (19.9122 %)
[P2] Table 2 scan took 146.663 sec
[P2] Table 2 rewrite took 252.052 sec, dropped 865593266 entries (20.1539 %)
Phase 2 took 2353.14 sec
Wrote plot header with 252 bytes
[P3-1] Table 2 took 221.267 sec, wrote 3429326570 right entries
[P3-2] Table 2 took 140.835 sec, wrote 3429326570 left entries, 3429326570 final
[P3-1] Table 3 took 220.009 sec, wrote 3439681255 right entries
[P3-2] Table 3 took 152.795 sec, wrote 3439681255 left entries, 3439681255 final
[P3-1] Table 4 took 214.662 sec, wrote 3465595968 right entries
[P3-2] Table 4 took 149.66 sec, wrote 3465595968 left entries, 3465595968 final
[P3-1] Table 5 took 221.73 sec, wrote 3532059735 right entries
[P3-2] Table 5 took 154.361 sec, wrote 3532059735 left entries, 3532059735 final
[P3-1] Table 6 took 229.818 sec, wrote 3712049786 right entries
[P3-2] Table 6 took 163.099 sec, wrote 3712049786 left entries, 3712049786 final
[P3-1] Table 7 took 245.635 sec, wrote 4291996410 right entries
[P3-2] Table 7 took 197.282 sec, wrote 4291996410 left entries, 4291996410 final
Phase 3 took 2314.92 sec, wrote 21870709724 entries to final plot
[P4] Starting to write C1 and C3 tables
[P4] Finished writing C1 and C3 tables
[P4] Writing C2 table
[P4] Finished writing C2 table
Phase 4 took 187.009 sec, final plot size is 108796175840 bytes
Total plot creation time was 10452.8 sec (174.214 min)

You want the slower nvme on temp 1, faster on temp 2.

I would switch to ubuntu and use only one nvme for temp
Im running ubuntu not windows.

show what you are using in the powershell…cover the keys.

I use -r 12 and -K 2 on the plotter everything else default.

No, it is not worth it. The CPU generation used is a tad bit slow and the limit of the NVME speed does not allow for adequate # of plots/day.

whats not worth it? Id be happy with 53 minute plots

.\chia_plot.exe -n 6 -r 20 -u 256 -t D:\ -2 D:\ -d D:\ -c xxxxx -f xxxxxx | tee ‘%LOG_FILE%’"

I guess i will try “-r 12” but i thought it was threads and I have 24 available threads. hmmm

Please, check that each CPU has his ram modules installed in dual mode. Other cause would be that the ram modules doesn’t have good health. I have experience getting ECC corrections erros. I saw It in my /var/log file of my Ubuntu system.
The errors were like this: EDAC MC1: 1 CE memory read error on CPU_SrcID#1_Ha#0_Chan#0_DIMM#0.

I replaced the ram modules and the system is fast now, 50 min madmax plot. (Dual xenon e5-2640 V3, 64gb ram). I use SSD and nvme for temp dir as you.

Please check your ram modules health and the situation on each cpu. Ensure you are working in dual mode if supoorted by your motherboard.

Regards.

Which Inland NVME model do you have, Premium or Platinum? Platinum quickly slows down to SATA SSD speed for sustained write. Check if you have hyper-threading enabled, under Task Manager, Performance → CPU - right click - Change Graph to - Logical Processors. Do all 24 cores show up?

My secondary plotter is a 2009 MacPro with similar dual Xeons as yours only slightly faster. The best I can do under MadMax is about 66 minutes.
image

1 Like

I meant that it’s not worth for me since i am plotting on a Win10 Workstation and getting about 32min plots

According to CPU-Z it is in single mode. Since my mobo has two processors i put one stick of 16gb ram on cpu 0 slot 0 and another 16gb ram on cpu 1 slot 0. Should I set it up differently? How would i get it into dual mode?

Ram is tested and fine.

I have the Platinum Inland NVME (damnit, lol) I am also using a PCIe 3.0 x4 adapter.

All 24 cores show up in the task manager.

Thanks for everyone’s help btw

Should I change my memory configuration and get a different NVME?

I dont really need a 2gb since I’m not plotting in parallels, what would you recommend?

If you want to work in dual mode, you need to add two 16gb new modules. The actual setup of ram is correct, but for work in dual mode you need 2 ram stick for each cpu (if your mobo support It).

How much of a difference is dual mode going to make? is it worth my buying another 16gb x2 of ram

Most likely. Not your system, but was watching a YouTube of benchmark comparison of single channel vs dual channel memory setups on a Ryzen APU and the benchmark was like half for single stick. Depending on your setup it can matter. Not so expensive really, so why not …and if cost is a concern, put 2x 16GB on one CPU and get 2x8GB for the other and compare for fun.

1 Like

Should I look into a RAM disk?