Plotting Speed

[Diamond242]

Hi everyone, I’m using HP DL380p Gen9, Cpu 2x E5-2620 v3 2.40 and 144GB ram (7x16 - 2133MHz) and a NVME Samsung 980 Evo 1TB that connected to motherboard via a PCI Express 3.1 Adapter and I’m using windows server 2019 Standard. I’m trying to plotting Chia using madmax and This is the best result I have ever had:

`C:\MadMax\chia_plot.exe -n 1 -r 20 -u 512 -t C:\MadMaxTemp\ -2 G:\ -d D:\MadMaxPlots\ -c ..... -f ......

PS C:\MadMax> C:\MadMax\Ploter.ps1 Multi-threaded pipelined Chia k32 plotter - c6664a0 Build 0.1.6-chives for Windows. Check for latest updates: https://stotiks.github.io/chia-plotter/

Network Port: 8444 Final Directory: D:\MadMaxPlots
Number of Plots: 1 Crafting plot 1 out of 1 Process ID: 2704 Number of Threads: 20 Number of Buckets P1: 2^9 (512) Number of Buckets P3+P4: 2^9 (512) Pool Puzzle Hash: ... Farmer Public Key: ... Working Directory: C:\MadMaxTemp
Working Directory 2: G:
Plot Name: ... [P1] Table 1 took 34.6496 sec [P1] Table 2 took 350.637 sec, found 4294995335 matches [P1] Table 3 took 419.068 sec, found 4294954620 matches [P1] Table 4 took 522.95 sec, found 4294934395 matches [P1] Table 5 took 513.955 sec, found 4294905714 matches [P1] Table 6 took 492.732 sec, found 4294912081 matches [P1] Table 7 took 368.36 sec, found 4294965738 matches Phase 1 took 2703.43 sec [P2] max_table_size = 4294995335 [P2] Table 7 scan took 63.4455 sec [P2] Table 7 rewrite took 112.093 sec, dropped 0 entries (0 %) [P2] Table 6 scan took 59.4219 sec [P2] Table 6 rewrite took 102.633 sec, dropped 581255466 entries (13.5336 %) [P2] Table 5 scan took 57.3512 sec [P2] Table 5 rewrite took 97.6528 sec, dropped 761870999 entries (17.7389 %) [P2] Table 4 scan took 55.7721 sec [P2] Table 4 rewrite took 98.4633 sec, dropped 828789279 entries (19.2969 %) [P2] Table 3 scan took 55.4877 sec [P2] Table 3 rewrite took 96.4939 sec, dropped 855029559 entries (19.9078 %) [P2] Table 2 scan took 56.0931 sec [P2] Table 2 rewrite took 99.2326 sec, dropped 865538961 entries (20.1523 %) Phase 2 took 963.739 sec Wrote plot header with 252 bytes [P3-1] Table 2 took 138.588 sec, wrote 3429456374 right entries [P3-2] Table 2 took 95.4325 sec, wrote 3429456374 left entries, 3429456374 final [P3-1] Table 3 took 178.706 sec, wrote 3439925061 right entries [P3-2] Table 3 took 103.488 sec, wrote 3439925061 left entries, 3439925061 final [P3-1] Table 4 took 192.625 sec, wrote 3466145116 right entries [P3-2] Table 4 took 109.546 sec, wrote 3466145116 left entries, 3466145116 final [P3-1] Table 5 took 194.436 sec, wrote 3533034715 right entries [P3-2] Table 5 took 107.309 sec, wrote 3533034715 left entries, 3533034715 final [P3-1] Table 6 took 224.568 sec, wrote 3713656615 right entries [P3-2] Table 6 took 101.594 sec, wrote 3713656615 left entries, 3713656615 final [P3-1] Table 7 took 130.576 sec, wrote 4294965738 right entries [P3-2] Table 7 took 126.082 sec, wrote 4294965738 left entries, 4294965738 final Phase 3 took 1707 sec, wrote 21877183619 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 160.426 sec, final plot size is 108835667321 bytes Total plot creation time was 5534.95 sec (92.2492 min) ` Also I created 110GB RamDisk using “OSFMount” and setted that to tmp2. but It takes about 100 minutes to generate each plot. What is wrong? please help me.

[itsme112358]

Nothing is wrong, that's round about what I would expect from that kind of hardware 😅 I'll compare it with my smaller plotter: i5-9400 --> you have around 50% more raw compute I would say 32GB --> you have 34GB left (not the bottleneck, since plotting doesn't really use that much RAM) 3x480Gb Seagate Nytro (RAID0) as temp 280GB Optane as temp2 I'm currently plotting burst/signum on this machine as well and farming it. So it slowed down a bit and takes around 110 min/plot. When I use it to plot exclusively, I get around 90-100min/plot I have two suggestions:

1. Don't use 7 Sticks of RAM, go with 8 and make sure they are in the correct slots. Your CPUs have four memory channels each. I'm really not sure what they do when you configure them with 7 sticks but I'm 100% certain they won't be using all of their bandwidth (max ~59 GB/s in quad channel mode). Worst case: They fall back to one channel, leaving you with less than 15GB/s.
2. Try a different SSD. The Evo has impressive "short term" performance. As long as you write only a couple of GB at a time (most benchmarks) you think "wow, that's one hell of a fast SSD". But once you exceed the capacity of your (virtual) SLC cache the SSD slows down by a lot. Are you sure it's a 980 Evo? I think there's only 980 and 980 Pro, there hasn't been a 980Evo released afaik

But as mentioned before don't expect any miracles, you are using old hardware, 12 cores may have been awesome 8 years ago but since IPC are way up and clock rates are far higher than on the Xeons you are using, a current gen ~300$ 8 Core Ryzen will run circles around these legacy Server Units: Use this as a reference: https://www.cpubenchmark.net/compare/%5BDual-CPU%5D-Intel-Xeon-E5-2620-v3-vs-Intel-i5-9400-vs-AMD-Ryzen-7-5800X/2418.2vs3414vs3869

[danarbraz]

You are way oversubscribing threads on that hardware. With 2x E5-2620, you have 12 physical cores at 24 threads with hyperthreading on. On very similar hardware, I get the best results setting setting initial threads (-r) to at least 2 (if not 4) lower than the number of physical cores. Then use the thread multiplier parameter (-K) to ramp threads by a factor of 2 for phase 2.

So, try creating a plot with this cmd and let us know how it goes: chia_plot.exe -n 1 -r 10 -K 2 -u 512 -t C:\MadMaxTemp\ -2 G:\ -d D:\MadMaxPlots\ -c ..... -f ......