I tried SWAR’s plot manager as well. A bit sketchy compared with plotman but that works too.
I am thinking of TBW on consumer drives. Journaling will result in extra writes and I/O
I agree in regards to async discards if you have enterprise SSD
Yeah, it is. I’ll share the plotman analyze output when I get home.
I really think that it is more bound by the cpu and that with proper staggering the drives are not hit all that hard.
I had a hackintosh nuc, where I could monitor read/write speeds over time, doing 5, maybe 6 in parallel I think. But it just sat around 350Mb/s.
1 Like
My machine completely freezes!!! Got unify b550 mobo, 128 GB RAM, 4 x 970 EVO PLUS, 850W corsair PSU. Got latest bios installed.
Same thing here. Got 5950 crashing on me every night.
For some reason it’s significantly worse on a 5950X, but I have also experienced regular crashes on a 5900X since posting that.
Are you using your SSDs in software RAID 0? For me that turned out to be the culprit.
2 Likes
hi quindor, i have the same cpu and would appreciate your help in setting up the system
Hello 
I got my 5950X with 64ram 4tb nvmssd and so on.
I’m using swar plot manager.
Can anyone post a example job config.
Im bit confused only phase 1 is using more than one core should I asign something like 8 threats per plot and call him only take 4 phase ones in action ?
Of I asign only 2 at each plot and only 2 plots are allowed on phase 1 wouldnt that be a waste?
Maybe somebody can help me with a start config.
Regards shizzle
I will give that a try. I think it might have to do with the motherboard. Because I also have a tuf b550 motherboard without any issue running with 5900x. Only the msi board are giving me these issues.
1 Like
Hi parsifal295,
I have
- X570 Aorus Ultra
- overclocked 5950x
- 64GB 3200Mhz
- 4x Evo 970 Pro in RAID 0
and tried copying your settings, but it seems jobs take close to 7 hours.
Are you still doing 55 plots a day?
Yes. it varies between 53-55.
- Pcie 3.0 x4 raid 0 has half bandwidth compared to pcie 4.0.
- OS optimization (if you’re NOT on Clear Linux).
- Can you share your overclock settings?
If I understand correctly you are using just one slot pcie 4.0 x16 which offers 32GB/s total
I am using 4 slots of PCI 3.0 x4 with each 4GB/s = 16GB/s total.
My SSDs are only rated at 3.5/3.3GB/s read/write, so the slots don’t limit me, correct?
I use Ubuntu 20.04.2 LTS with hourly trim enabled.
I did a very quick overclock:
Global C-state Control = Enabled
Power Supply Idle Control = Low Current Idle
CPPC = Enabled
CPPC Preferred Cores = Enabled
PPT = 270
TDC = 220
EDC = 220
Tested very briefly in Windows with Cinebench R23 and did around 28500, but unfortunately lost the screenshots I made of HWINFO64, so can’t tell you what my effective clocks are. Do you know how to check in Ubuntu?
Just added a staging drive. Here is what plotman looks right now:
To check your CPU freq,
- download cpufrequtils
sudo apt install cpufrequtils
- run script to get CPU frequency.
./getCPUfreq.sh
There’s a reason I’m using Clear Linux over Ubuntu.
3 Likes
The benchmarks look good. This article (https://www.phoronix.com/scan.php?page=article&item=ubuntu-2104-clear&num=1) says:
Out of nearly 100 tests, Ubuntu 21.04 was about 4% faster than Ubuntu 20.10 overall while Intel’s Clear Linux had a 8% advantage over the new Ubuntu 21.04 “Hirsute Hippo” release.
Have you tried disabling the Spectre/Meltdown mitigations?
What folder do you run getCPUfreq.sh
I can’t find that file anywhere on the system after installing cpufrequtils
Impressive. What is utilisation time for your nvme drives? Or better post whole iostat -x -k 1 output if you don’t mind.
Coding, i’m a beginner in this game.
But i have a linux for my OS and I can’t find a throttlestop for this OS, can you recommend any throttlestop that will work for linux?
Can you recommend me the best throttlestop for linux? Currently using focalfossa…
Hey Guys 
I built a new system with a 5950x and have the problem that my temp drives always “peak” above 1GB R/W and then partially drop back to 0-10 MB/s R/W for a few seconds. As a result, I have massive CPU I/O waits, extremely high CPU idles.
What I want to do now is to test it without raid or ext4 as filesystem.
Has anyone had the same problem before or any ideas on how else to address the problem?
Specs and Settings used:
OS: Ubuntu Server 20.04.
CPU: 5950x
RAM: 32GB x 2 3600 MHz CL16
M2: 3 x 2TB Firecuda 520 (RAID 0 with Chunksize 64 and filesystem xfs, mounted with -t xfs -o discard)
Using SWAR for plotting with 15 plots parallel, staggered 20 min with 4 threads (I think the main problem is the inconsistent read write speeds of the SDDs)