Speed comparison - show your speeds! #452
Replies: 33 comments 107 replies
-
|
New settings: 2x 500GB NVME WD BLACK SN750, RAID 0, CHUNKSIZE 128k 26 minutes per plot
|
Beta Was this translation helpful? Give feedback.
-
Thanks, so far I have only mastered downloading ubuntu from the market and installing it in windows 10, after that I can enable ubuntu and get it in the screenshot, then I do not understand what to do, I know that I have to download the plotter here and configure it for my configuration, which by the way xeonn e5 2690 v3 32 ram nvme 2t. With windows 10 through the "cmd" it gives 1 plot in 1 hour. Only I lack the knowledge to load the plotter into the ubuntu environment and configure it for my pc configuration. |
Beta Was this translation helpful? Give feedback.
-
|
Start ubuntu.
./chia_plot -n 1 -r 14 -u 128 -p HERE_PUNLIC_KEY -f HERE_FARMER_KEY -t /mnt/d/ -d /mnt/e/ under /mnt/ your drive are listed. if your temp drive is d: you enter -t /mnt/d/ |
Beta Was this translation helpful? Give feedback.
-
|
|
Beta Was this translation helpful? Give feedback.
-
|
Spacebar After cd |
Beta Was this translation helpful? Give feedback.
-
|
I started Plotter with wsl ubuntu 20.04 on windows. But it works slower than from Windows OS Why is that so? |
Beta Was this translation helpful? Give feedback.
-
|
OS: Windows 10 -u 64 -r 16
|
Beta Was this translation helpful? Give feedback.
-
|
Hello. |
Beta Was this translation helpful? Give feedback.
-
|
I am not using the windows 10 version, i am using wsl ubuntu 20.04 on windows. But just try ubuntu on windows first. |
Beta Was this translation helpful? Give feedback.
-
|
Mmm... Im lamer in ubuntu))) if you have a small instructions))) |
Beta Was this translation helpful? Give feedback.
-
|
Ive install wsl have worst result. Then install clear ubuntu and processor load is below 25% while plotting. 11700k 16 threads 128 buckets(((( please help |
Beta Was this translation helpful? Give feedback.
-
|
ok i think the main difference here is SSD speed. I am using 2x nvme in raid 0. You are using 1 single nvme. |
Beta Was this translation helpful? Give feedback.
-
|
@Isking007 I'm having the same problem. I tried with docker and with ubuntu WSL but I'm not able to use all the processor. Did you had any luck? |
Beta Was this translation helpful? Give feedback.
-
|
Ryzen 9 5900x Running with all 24 threads. In my case 512 buckets is always faster. I dont have enough RAM for 64 buckets, but the first phase (that ran) was taking longer. All times with second plot (first plot is always faster) - ubuntu and clear with server versions:
I didn`t knew about WSL, currently I'm running with a dedicated pc just for plotting and another with windows for farming. |
Beta Was this translation helpful? Give feedback.
-
|
Hi, just an update, I just turned on PBO and overclocked my memory to 3.8GHz, I`m now getting near 1300s (second run) with clear linux. I wasnt able to install clearlinux on WSL, I tried to install on docker but the times where way too high. |
Beta Was this translation helpful? Give feedback.
-
|
No , the raid is made in the bios, Intel raid controller on my motherboard |
Beta Was this translation helpful? Give feedback.
-
|
Oh, ok, but how about inside WSL? It is formatted as NTFS? |
Beta Was this translation helpful? Give feedback.
-
|
Yess ntfs |
Beta Was this translation helpful? Give feedback.
-
|
CPU:Ryzen 9 3900X Single run (lab-time): 1600s In my production environment, this gives 5.2TiB/d - all done and fine. |
Beta Was this translation helpful? Give feedback.
-
|
I have the same system and only 1 ssd disk m2, only I using wsl ubuntu 20.04 on windows 10, and my result is 2150-2200 seconds. Have you done raid0 or just uses 2 ssd as working directories 1 and 2? |
Beta Was this translation helpful? Give feedback.
-
|
intressting thats a great result for 3900X rig. can you post your Madmax timing? |
Beta Was this translation helpful? Give feedback.
-
|
Wsl ubuntu 20.04 on windows 10. Results 38 minute. r24, u512, 1 ssd nvme format on Cluster 256Kb |
Beta Was this translation helpful? Give feedback.
-
Beta Was this translation helpful? Give feedback.
-
|
No worries, sir |
Beta Was this translation helpful? Give feedback.
-
|
IBM x3550 M4: Number of Threads: 16
|
Beta Was this translation helpful? Give feedback.
-
|
how many cores for these servers per cpu ? |
Beta Was this translation helpful? Give feedback.
-
|
write model cpu plz |
Beta Was this translation helpful? Give feedback.
-
|
Hi, post updated with the requested info |
Beta Was this translation helpful? Give feedback.
-
|
2x Xeon(R) Gold 6149 CPU @ 3.10GHz |
Beta Was this translation helpful? Give feedback.
-
|
My status Ryzen 5 5600x 12 Thread -r 12 -u 7 Time for plot 5145.53 sec. or 85 min. I do not understand, why is it time for create plot is long. Thanks |
Beta Was this translation helpful? Give feedback.
-
|
yes set it to the highest. |
Beta Was this translation helpful? Give feedback.
-
@vctrb |
Beta Was this translation helpful? Give feedback.
-
No Windows is on other ssd. The raid is only the plott Drive. |
Beta Was this translation helpful? Give feedback.
-
By reading the manual and googling your motherboard and pci lane routing. |
Beta Was this translation helpful? Give feedback.
-
|
motherboard specification |
Beta Was this translation helpful? Give feedback.
-
|
Some Tests ... Dual Socket 7F32 CPU with 8 cores and 512 GB Memory. r4 dual ramdisk *@vision-amd:[/media/plots/chia-plotter/build]: ./chia_plot -n 1 -r 4 -u 128 -t /media/ramdisk/ -2 /media/ramdisk/ -d /nfs/chia/ -p * -f *Multi-threaded pipelined Chia k32 plotter - f495816 Next: r8 dual ramdisk *@vision-amd:[/media/plots/chia-plotter/build]: ./chia_plot -n 1 -r 8 -u 128 -t /media/ramdisk/ -2 /media/ramdisk/ -d /nfs/chia/ -p *-f *Multi-threaded pipelined Chia k32 plotter - f495816 Next r 16 dual ramdisk *@vision-amd:[/media/plots/chia-plotter/build]: ./chia_plot -n 1 -r 16 -u 128 -t /media/ramdisk/ -2 /media/ramdisk/ -d /nfs/chia/ -p *-f *Multi-threaded pipelined Chia k32 plotter - f495816 Next r16 with ramdisk + ssd *@vision-amd:[/media/plots/chia-plotter/build]: ./chia_plot -n 1 -r 16 -u 64 -t /media/chia-temp/ -2 /media/ramdisk/ -d /nfs/chia/ -p *-f *Multi-threaded pipelined Chia k32 plotter - f495816 now ... setting mitigation off, will add them later |
Beta Was this translation helpful? Give feedback.
-
New single CPU record: 22.32 mins per plotServer Mount RAM disk Plotter command Code Results |
Beta Was this translation helpful? Give feedback.
-
|
Hi @cvilasis69, you should try clearlinux and use all your 32 threads, you probably will do 10 to 15min. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for posting this. I am going to install clear linux next. |
Beta Was this translation helpful? Give feedback.
-
|
Cisco UCS C220 M3 server Process ID: 2377135 |
Beta Was this translation helpful? Give feedback.
-
|
Still dialing in settings but so far only seeing sub 1 minute improvements. Dell R730XD /media/chia/4tbS07/plot$ /home/chia/chia-plotter/build/chia_plot -r 86 -n 1 -u 512 -d /media/chia/4tbS07/plot/ -p pubkey1 -f pubkey2 -t /media/chia/1tbS12/temp/ -2 /media/chia/ram/ |
Beta Was this translation helpful? Give feedback.
-
|
32 mins 80 cores 256 buckets 36.8 mins 40 cores 256 buckets parallel #1 37.7 mins 42 cores 512 buckets parallel #2 I feel my slower clock speed may be holding me back; will update when the new hardware arrives. |
Beta Was this translation helpful? Give feedback.
-
|
System: Multi-threaded pipelined Chia k32 plotter - a2ef17b I've tried less buckets and haven't seen much of any change... Maybe a few minutes up or down. Even tried temp 2 using a RAM disk and it was actually slightly slower, but by almost nothing. It never fully saturates the NVMe bandwidth. Also tried parallel plotting 4 at a time and it just ended up equaling the same time. Same with changing threads to the full 36 (18 cores x 2)... No real speed up. About 35 minutes seems to be where my system is at. |
Beta Was this translation helpful? Give feedback.
-
|
Actually when I tried 256 buckets it gave me an error: So at least on my system I see set at 128 or lower, unless I'm doing something wrong. |
Beta Was this translation helpful? Give feedback.
-
|
This plotter maxxes out everything. So in your case mybe it is ram latancy of RAM speed or some other sord of controler. When i overclock my RAM to 4666 (unstable After 60 min 🥺) And set temp 2 to second raid 0 Drive i get 19 min per Plot |
Beta Was this translation helpful? Give feedback.
-
|
Could be... Can't really overclock on this system or do any more deep tweaks, so the NVMe drives will be how I'll go forward. May as well keep my RAM free considering the TBW on these SSDs. Would like to see this plotter going on k33s as well as I use the to fill up drives. This is all rad either way! |
Beta Was this translation helpful? Give feedback.
-
|
I have 3 machines: 5950x with 110G ramdisk + 3x 970 Evo Plus 2TB in Raid 5900x with 64GB ram + 970 Pro 1TB as Temp1 and 980 Pro 1TB as Temp2 9900k clocked at 5Ghz all cores + 16GB ram + 970 Evo Plus 2TB as Temp 1 and SN750 2TB as Temp 2 |
Beta Was this translation helpful? Give feedback.
-
|
That 9900k ist somehow bottlenekt by sonething |
Beta Was this translation helpful? Give feedback.
-
Is anyone getting much better results with the same cpu? |
Beta Was this translation helpful? Give feedback.
-
|
At 5ghz it should be at least doing 30-35 min |
Beta Was this translation helpful? Give feedback.
-
I have nobody to compare to, so not really sure it can do that well. |
Beta Was this translation helpful? Give feedback.
-
|
How about speed and cost / longevity. Surely lots of fast-ish RAM is important? I just picked up an R630 with 256GB DDR4, added 8xSATA Intel 200gb SSD (RAID0 on the PERC H730) as the tmpdir1. Total cost approx £1100 GBP. I think it was less than the cost of the RAM alone would have been. I have an additional 32GB ram coming in a couple of days which should allow me to use the full 256GB for tmpdir1 & tmpdir2. So zero SSD usage or wear - until we have to move to k33. Don't forget that the value of this coin is dropping. This is becoming a fun nerd-race that actually costs money. We are doing nerd drag racing. Racing is an expensive hobby / sport. But it's cool ;-)
|
Beta Was this translation helpful? Give feedback.
-
2TB MP600 Corsair , Bucket 256 , I only gave 90 threads for this. over it will just be slower. |
Beta Was this translation helpful? Give feedback.
-
Try 512 buckets. Seems to be faster for everyone. |
Beta Was this translation helpful? Give feedback.
-
|
Thank you , I will do. Is anyone getting decent speeds when doing in parallel 2 madmax instances ? |
Beta Was this translation helpful? Give feedback.
-
|
How can you use 90 threads if you only have 256GB RAM? I can't use even 64 threads with b=256 due to lack of RAM. I also have 256GB. With b=512 I can use r=64 but the time is not better than with r=32 (TR 3970X, 32 cores). |
Beta Was this translation helpful? Give feedback.
-
I find that running 2 madmax instances and using 'taskset' to avoid NUMA spanning; my plot time goes from 21 minutes up to 30 -45 mins. And that that is with separate resources for each instance. I/O contention? When I maxed the RAM the speed drops down from 2166 to 1833 so lost some bandwidth there. |
Beta Was this translation helpful? Give feedback.
-
|
Dell R720 2x E5-2690v2 64GB Ram. Array 1 = 4x SAS SSD (HUSSL4010BSS600) raid0, Array2 = 4x SAS SSD (HUSSL4010BSS600) raid0. |
Beta Was this translation helpful? Give feedback.
-
So no RAM disk, plotting on raids only. how many threads allocated ? |
Beta Was this translation helpful? Give feedback.
-
|
Nope no ramdisk. plotting on 2 arrays of 4x sas ssd Raid 0. Plotting 2 in parallel with 20 threads each. Each plot gets its own raid0 array. |
Beta Was this translation helpful? Give feedback.
-
|
finishe my testing 1 X Intance 1590-1600 S = 54 Plots per day 24tr 2 X intances 3100S = 56 Plots per day 12Tr each not a big difference 2 |
Beta Was this translation helpful? Give feedback.
-
Multi-threaded pipelined Chia k32 plotter - 9e7902c Working Directory: /mnt/chiatemp/ |
Beta Was this translation helpful? Give feedback.
-
|
wow, cannot believe it by this so old configure. Impressive ! |
Beta Was this translation helpful? Give feedback.
-
|
Very nice dude! Man the 990x extreme was my favorite proc, good to see the old girls got some life in her. May have to pull mine out of the tech bin.
Very nice dude! Man the 990x extreme was my favorite proc, good to see the old girls got some life in her. May have to pull mine out of the tech bin. |
Beta Was this translation helpful? Give feedback.
-
|
The x58 900 series i7s are absolute beasts. They overclock like a DREAM. I buy and use them for my ETH miners because they're so stable. |
Beta Was this translation helpful? Give feedback.
-
|
System 1: Threadripper 3970X with 256GB 3200MHz RAM. Using b=256 and r=32 I get 15m:30s per plot. Additional 50s are required to move the final plot form ramdisk to SSD from where the plots are distributed to HDDs. -> 88 plots / day System 2: Ryzen 3950X with 48GB 3000MHz RAM. Using b=128 and r=16 I get around 36m per plot. tmpdisk is on RAID0 consisting of 4x400GB Intel enterprise SATA SSDs (RAID0 continuous reads/writes about 2GB/s). -> 40 plots / day |
Beta Was this translation helpful? Give feedback.
-
|
I'm getting ~35 minutes per plot in Ubuntu with i9-10850k (stock, no power limit), 32gb 3000MHz CL14, single 2TB 970 EVO Plus for temp1/2, b=default, r=20. Times stay consistent over hundreds of plots, even without doing trim. Maybe your SSD array is limiting the 3950X? The 10850k has higher IPC and clock speed, but a 3950X still has 12 more threads. |
Beta Was this translation helpful? Give feedback.
-
|
that is very slow for 32 cores imo, my 10980xe do one plot with 110G ram disk under 20 mins |
Beta Was this translation helpful? Give feedback.
-
|
Ryzen 9 5900X Vermer 12 Ubuntu 5.11.0-27-generic #29~20.04.1-Ubuntu mount -t tmpfs -o size=105G tmpfs /mnt/ram/ Started copy to /mnt/farm/007/plot-k32-2021-08-25-15-09-XXXXXXXXXXXXXXXXXx.plot More buckets and threads don't seem to help trying with more is slower and fewer no better. |
Beta Was this translation helpful? Give feedback.
-
|
try switching to clear linux (but kernel 5.12, not 5.13), my times with almost the same configuration - only nvme is 970 pro 512gb are barely above 18 minutes. Total plot creation time was 1083.27 sec (18.0545 min) |
Beta Was this translation helpful? Give feedback.
-
|
Dell PowerEdge R820 sudo mount -t tmpfs -o size=330G tmpfs /mnt/ram/ Number of Threads: 32 |
Beta Was this translation helpful? Give feedback.
-
|
48 minutes - I'm impressed, it's a machine I knocked together from old gaming equipment. ./chia-plotter/build/chia_plot -n -1 -r 16 -u 256 -v 512 -t /mnt/stage/ -2 /mnt/plot/ -d /media/andrew/Farm_15/ -c x -f x Network Port: 8444 |
Beta Was this translation helpful? Give feedback.
-
|
10850K cpu at 4.9 all cores ASUS 590M Plus MB WIN 10 , madmax running is powershell 3200MHZ 64GB ram XMPII temp drive 1 = WD Black 2TB NVME (gen3) temp drive 2 = Corsair P2 2TB NVME (x2) Raid 0 (gen3) running 18 threads (seems to be the sweet spot) Total plot creation time was 1900.76 sec (31.6793 min)Number of Plots: infinite |
Beta Was this translation helpful? Give feedback.
-
|
not sure why people wasting time plotting in Windows when there's 20-30% free speedup with just using linux (especially clear linux) |
Beta Was this translation helpful? Give feedback.
-
|
I did not even get 1 minute , ubuntu, clear linux vs windows 10. with ryzen 5900x, ryzen 5600x, Epyc 7713 ... was not worth the tests and time wasting |
Beta Was this translation helpful? Give feedback.
-
im am plotting 23 min.on windows. dont spread false information. it is round about the same if you have fast hardware. |
Beta Was this translation helpful? Give feedback.
-
|
it's not false information, i've tested (almost) the same hardware on different OS-es and found quite big differences. Even switch from Ubuntu 20.04 to Clear linux gave me ~10% faster plotting. However, when CL automatically upgrades kernel to 5.13 i got drop in speed, something like on Ubuntu and i had to switch back manually (to 5.12). So plotting speed is very OS dependent, and very File system dependent (APFS and HFS are big slow down on Macs) |
Beta Was this translation helpful? Give feedback.
-
And which best time you got with 5900x? |
Beta Was this translation helpful? Give feedback.
-
|
Debian 10 ./chia_plot -r 18 -n 1 -t /mnt/tempdir/ -2 /mnt/ram/ -d /mnt/final/ Total plot creation time was 3071.42 sec (51.1904 min) |
Beta Was this translation helpful? Give feedback.
-
|
Clarification: Raid 10 consists of 4 SAS HDDs |
Beta Was this translation helpful? Give feedback.
-
|
I think with Raid 0 you should be faster. I mean why Raid 10? You need speed not safty. |
Beta Was this translation helpful? Give feedback.
-
maybe he wants to keep data safe and also double writting speeds with extra HDD ;-) |
Beta Was this translation helpful? Give feedback.
-
|
I recall from normal Chia plotting that any kind of RAID was slower than using parallel ploting on each NVMe sepparate.but MadMax doesn't utilize more drives |
Beta Was this translation helpful? Give feedback.
-
|
I made a RAID0 test afterwards but the difference was negligible. Actually it was only a test for interessted users for plotting without SSDs. :) |
Beta Was this translation helpful? Give feedback.
-
|
Used to make 10 plots parallel every 10 hours in Ubuntu with Plotman. Once, MadMax became official Chia plotter...I do not bother anymore. MP600 has got sustainable 1500MBs...only WD SN850 is faster but with a half life span. I have been researching "cheapest" option, and didn't find anything better yet. Of course, if someone tends to plot 32EB...it is another challenge. |
Beta Was this translation helpful? Give feedback.
-
|
RAMdisk in Win10 ~37min, Clear LInux ~25min with G.Skill RipJaw DDR-4 3200 CL14 128GB |
Beta Was this translation helpful? Give feedback.
-
|
dedicated Clear Linux rig
|
Beta Was this translation helpful? Give feedback.
-
|
56W/plot |
Beta Was this translation helpful? Give feedback.
-
|
53.5W/plot |
Beta Was this translation helpful? Give feedback.
-
|
Later runs while my super inefficient final copy is taking place to a sshfs mount slow down to a consistent ~2400s. No doubt room for optimization, but it'd be small gains. First up would be moving to a low overhead copy. This is mostly a curiosity benchmark for me and serves as a great upgrade over my FX-6300 16GiB ram system that took 8200+ seconds. Update: Compiling with clang instead of gcc on Ubuntu 20.04 shaves 3% off the plot time. 2053 seconds. This is very close to the Clear Linux gcc 11 build's speed (2020 seconds). See #981 for what is needed to compile using clang. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks. XFS indeed helps over ext4, it shaves 60 seconds off the above. Clear Linux: I tried it. If I run the binary I compiled on ubuntu 20.04's g++ 9 timing doesn't change significantly (their kernel tweaks don't appear to matter much here). But if I recompile the plotter with Clear's g++ 11 [does that default to nicer -march= flags?] another 80 seconds are shaved off. Recompiling the plotter with a better compiler & flags is doable regardless of underlying distro. AMD's AOCC clang/llvm fork might be interesting to try next. |
Beta Was this translation helpful? Give feedback.
-
|
I used git hub pull. never tried recompile. how did you recompile, please? what flags? my guess is your limit is CPU and NVMe...do not know Silicon Power. Fastest poor men NVMe on market is WD SN850, followed by Corsair MP600. They have sustainable writing speeds 1200-1700MBs. No other normal NVMe can do it.. RAM speed and timing would be another limit. I have got G.Skill Rip Jaw 128GB 3200MHz...they are pushed to the limit, and cannot be clocked to 3800MHz for max. performance. Fast memory cost fortune...sadly. It would pay off only for hard core plotters. |
Beta Was this translation helpful? Give feedback.
-
|
how the hell did you manage -r 15 cores on 8 core CPU? :-D or do you have a divider for threads only? |
Beta Was this translation helpful? Give feedback.
-
|
I haven't try using my own flags, I just used the make_devel.sh as the README describes. Agreed that a gen4 ssd and faster ram should make a difference. No need to spend that money. My ram runs at DDR-3066 CL16 despite claiming to be rated for 3200 via its XMP profile because it couldn't pass a prolonged memtest86+ at higher speeds in this system (and cost a lot less than any that might). It already feels shameful to build a system without ECC ram in it, but that kind of costly reliability wasn't the point of this toy. :) |
Beta Was this translation helpful? Give feedback.
-
|
every RAM is officially 1066MHz according to JEDEC standard. Everything else is binning and OC of the best silicon. If you want reliability, I doubt there is a ECC DDR4-3200 or 3600MHz RAM :o) most importantly, most crap is just marketing hype. I have official RAM modules DDR4-3600 that are slower than DDR4-3200 because they forget to tell you that the remaining 30+ timing parameters are slow ROFL...PC tuners love first 4 primary parameters because that is "all" it matters. gen 3 or 4 makes little difference in normal customers market...there is no NVMe that can utilize full PCIe lanes bandwidth anyway ;-) That is why hardcore plotters use enterprise/data center SSDs that can do more speed/IOs |
Beta Was this translation helpful? Give feedback.
-
|
12900K CPU @ All cores Gigabyte Z690 Gaming X WIN 11 and madmax 2x DDR5 5000Mhz 16GB = 32GB temp drive = Seagate FireCuda 530 @ PCIe 4.0 x4 [2022/02/13 23:08:53] ./chia_ploter/chia_plot.exe -n 1 -r 16 -K 16 -u 256 -v 256 -t C:\tempChia\ -d G:\ -w [2022/02/13 23:09:02] [P1] Table 1 took 8.74995 sec 22min |
Beta Was this translation helpful? Give feedback.
-
try Clear Linux, I bet it will be even faster. I am getting <25min with old HW |
Beta Was this translation helpful? Give feedback.
-
|
as RAM and clock speed are very important, why you guys don't show it? here is my rigs 1, [email protected] AVX offset 0, RAM gloway 32G 2666 cl19 x4, RAM disk, disk: PX04(HHHL) 1.6T, system: Ubuntu 20.04; => 21.8 min |
Beta Was this translation helpful? Give feedback.
-
and the 7960X can even do a 19.6 min job @ 4.7G AVX offset 0, but it eats 400w power, that gonna to be too crazy, so I just set it to 4.4G and offset 0, acceptable 250w |
Beta Was this translation helpful? Give feedback.
-
|
24min @ 150-170W |
Beta Was this translation helpful? Give feedback.
-
|
I built pure plotting machine from scrap lying around. Clear Linux is high performance oriented...if you wanna play games, use modern GPU...it needs lots of time for development. It also does not follow usual Linux configuration....it is classical "I do my own distro, why make it easy" ROFL |
Beta Was this translation helpful? Give feedback.
-
finally I tested Clear Linux, 30.5 on Ubuntu 21.04 and 27.3 on ClearLinux, 7960X, 900P, SN550，that's about 11.7% improvement, I spent a night on the configuration, and will test it on a 128G system tonight, and I noticed that in phase one, Ubuntu will not make all cores full load, most cores are at max frequency, but still some cores are not, while ClearLinux force all cores run at max frequency, that should be the reason, of cause, more power consumed |
Beta Was this translation helpful? Give feedback.
-
|
most general-purpose Linux distros have issues with CPU scheduling for years. I recall from old days of chia classic plotter, you had to force CPU scheduling with numactl or nice and so on. Plotman resolved this issue automatically, either by manipulating MSR or numactl. IO scheduling also crap....so much wasted cycles. It has been not designed for minority of power users. Otherwise, processes jump from one core to another, wasting at least 10% of time by pointless scheduling...I spent hours to figure out this crap...I couldn`t fully load all cores. Clear Linux is performance and simplicity focused and results are "Clearly" visible |
Beta Was this translation helpful? Give feedback.
-
|
2 x 6242 Gold, 384GB Ram, No Disk Drives, Running Ubunto (22) Process ID: 284881 Next step is to try Clear Linux |
Beta Was this translation helpful? Give feedback.
and others
-
Hello,
It would be very interesting if we all compare our speeds.
System: 10900k @ 5,2 GHZ Core 4,8ghz cache, 32gb 4000mhz ram @ cl 16
nvme 2x 500gb SN750 @ RAID 0 RAID CHUNK SIZE: 64k
Settings 14 threads 128 buckets
System windows 10 running with wsl ubuntu 20.04
Speed: 31 minutes per plot
UPDATE:
i9 12900k @5,1 P-Core 4,0 E-Core
32GB DDR4 3600 CL14
Software Raid in WIndows
4x WDBLACK SDD RAID0 64k Chunk Size
24 minutes
Beta Was this translation helpful? Give feedback.
All reactions