Great and less great SSDs for Ethereum nodes

HallOfBlame.md

Overview

Syncing an Ethereum node is largely reliant on latency and IOPS, I/O Per Second, of the storage. Budget SSDs will struggle to an extent, and some won't be able to sync at all. For simplicity, this page treats IOPS as a proxy for/predictor of latency.

This document aims to snapshot some known good and known bad models.

The drive lists are ordered by interface and then by capacity and alphabetically by vendor name, not by preference. The lists are not exhaustive at all. @mwpastore linked a filterable spreadsheet in comments that has a far greater variety of drives and their characteristics. Filter it by DRAM yes, NAND Type TLC, Form Factor M.2, and desired capacity.

For size, 4TB comes recommended as of mid 2024. The smaller 2TB drive should last an Ethereum full node until early 2025 or thereabouts, with crystal ball uncertainty. The Portal team aim to make 2TB last forever with EIP-4444 by late 2024. Remy wrote a migration guide to 4TB.

High-level, QLC and DRAMless are far slower than "mainstream" SSDs. QLC has lower endurance as well. Any savings will be gone when the drive fails early and needs to be replaced.

Other than a slow SSD model, these are things that can slow IOPS down:

• Heat. Check with smartctl -x; the SSD should be below 50C so it does not throttle.
• TRIM not being allowed. This can happen with some hardware RAID controllers, as well as on macOS with non-Apple SSDs
• ZFS
• RAID5/6 - write amplification is no joke
• On SATA, the controller in UEFI/BIOS set to anything other than AHCI. Set it to AHCI for good performance.

If you haven't already, do turn off atime on your DB volume, it'll increase SSD lifetime and speed things up a little bit.

Some users have reported that NUC instability with certain drives can be cured by adding nvme_core.default_ps_max_latency_us=0 pcie_aspm=off to their GRUB_CMDLINE_LINUX_DEFAULT kernel parameters via sudo nano /etc/default/grub and sudo update-grub. This keeps the drive from entering powersave states by itself.

The Good

"Mainstream" and "Performance" drive models that can sync mainnet execution layer clients in a reasonable amount of time.

• Higher endurance (TBW) than most: Seagate Firecuda 530, WD Red SN700
• Lowest power draw: SK Hynix P31 Gold - was a great choice for Rock5 B and other low-power devices, but 2TB only

We've started crowd-sourcing some IOPS numbers. If you want to join the fun, run fio --randrepeat=1 --ioengine=libaio --direct=1 --gtod_reduce=1 --name=test --filename=test --bs=4k --iodepth=64 --size=150G --readwrite=randrw --rwmixread=75; rm test and give us the read and write IOPS.

If you have room for it and need an excellent heatsink, consider the "Rocket NVMe Heatsink". It is quite high however, and may not fit in some miniPC cases.

Hardware

M.2 NVMe "Mainstream" - TLC, DRAM, PCIe 3, 4TB drives

• Any data center/enterprise NVMe SSD
• Teamgroup MP34, between 94k/31k and 118k/39k r/w IOPS
• WD Red SN700, 141k/47k r/w IOPS

M.2 NVMe "Performance" - TLC, DRAM, PCIe 4 or 5, 4TB drives

• Any data center/enterprise NVMe SSD
• Acer GM7000 "Predator", 125k/41k r/w IOPS
• ADATA XPG Gammix S70, 272k/91k r/w IOPS
• Corsair Force MP600 Pro and variants (but not "MP600 Core XT"), 138k/46k r/w IOPS
• Crucial T700, 215k/71k r/w IOPS
• Kingston KC3000, 377k/126k r/w IOPS
• Kingston Fury Renegade, 211k/70k r/w IOPS
• Mushkin Redline Vortex (but not LX)
• Sabrent Rocket 4 Plus, 149k/49k r/w IOPS. @SnoepNFTs reports the Rocket NVMe Heatsink keeps it very cool.
• Samsung 990 Pro, 124k/41k r/w IOPS - there are reports of 990 Pro rapidly losing health. A firmware update to 1B2QJXD7 is meant to stop the rapid degradation, but won't reverse any that happened on earlier firmware.
• Seagate Firecuda 530, 218k/73k r/w IOPS
• Teamgroup MP44, 105k/35k r/w IOPS - caution that this is DRAMless and uses a Host Memory Buffer (HMB), yet appears to perform fine.
• Transcend 250s, 127k/42k r/w IOPS. @SnoepNFTs reports it gets very hot, you'd want to add a good heatsink to it.
• WD Black SN850X, 101k/33k r/w IOPS

M.2 NVMe "Mainstream" - TLC, DRAM, PCIe 3, 2TB drives

• Any data center/enterprise NVMe SSD
• AData XPG Gammix S11/SX8200 Pro. Several hardware revisions. It's slower than some QLC drives. 68k/22k r/w IOPS
• AData XPG Gammix S50 Lite
• HP EX950
• Mushkin Pilot-E
• Samsung 970 EVO Plus 2TB, pre-rework (firmware 2B2QEXM7). 140k/46k r/w IOPS
• Samsung 970 EVO Plus 2TB, post-rework (firmware 3B2QEXM7 or 4B2QEXM7). In testing this syncs just as quickly as the pre-rework drive
• SK Hynix P31 Gold
• WD Black SN750 (but not SN750 SE)

M.2 NVMe "Performance" - TLC, DRAM, PCIe 4 or 5, 2TB drives

• Any data center/enterprise NVMe SSD
• Crucial P5 Plus
• Kingston KC2000
• Samsung 980 Pro (not 980) - a firmware update to 5B2QGXA7 is necessary to keep them from dying, if they are firmware 3B2QGXA7. Samsung's boot Linux is a bit broken, you may want to flash from your own Linux.
• SK Hynix P41 Platinum / Solidigm P44 Pro, 99k/33k r/w IOPS
• WD Black SN850

Cloud

• Any baremetal/dedicated server service
• AWS i3en.(2)xlarge or is4gen.xlarge
• AWS gp3 w/ >=10k IOPS provisioned and an m7i/a.xlarge

The Bad

These "Budget" drive models are reportedly too slow to sync (all) mainnet execution layer clients.

Hardware

• AData S40G/SX8100 4TB, QLC - the 2TB model is TLC and should be fine; 4TB is reportedly too slow
• Crucial P1, QLC - users report it can't sync Nethermind
• Crucial P2 and P3 (Plus), QLC and DRAMless - users report it can't sync Nethermind, 27k/9k r/w IOPS
• Kingston NV1 - probably QLC and DRAMless and thus too slow on 2TB, but could be "anything" as Kingston do not guarantee specific components.
• Kingston NV2 - like NV1 no guaranteed components
• WD Green SN350, QLC and DRAMless
• Anything both QLC and DRAMless will likely not be able to sync at all or not be able to consistently keep up with "chain head"
• Crucial BX500 SATA, HP S650 SATA, probably most SATA budget drives
• Samsung 980, DRAMless - unsure, this may belong in "Ugly". If you have one and can say for sure, please come to ethstaker Discord.
• Samsung T7 USB, even with current firmware

The Ugly

"Budget" drive models that reportedly can sync mainnet execution layer clients, if slowly.

Note that QLC drives usually have a markedly lower TBW than TLC, and will fail earlier.

Hardware

• Corsair MP400, QLC
• Inland Professional 3D NAND, QLC
• Intel 660p, QLC. It's faster than some "mainstream" drives. 98k/33k r/w IOPS
• Seagata Barracuda Q5, QLC
• WD Black SN770, DRAMless
• Samsung 870 QVO SATA, QLC

2.5" SATA "Mainstream" - TLC, DRAM

• These have been moved to "ugly" because there are user reports that only Nimbus/Geth will now sync on SATA, and even that takes 3 days. It looks like after Dencun, NVMe is squarely the way to go.
• Any data center/enterprise SATA SSD
• Crucial MX500 SATA, 46k/15k r/w IOPS
• Samsung 860 EVO SATA, 55k/18k r/w IOPS
• Samsung 870 EVO SATA, 63k/20k r/w IOPS
• WD Blue 3D NAND SATA

Cloud

• Netcup RS G11 Servers. Impressively fast; but it still depends on your neighbors in the service.
• Contabo SSD - reportedly able to sync Geth 1.13.0 and Nethermind, if slowly
• Netcup VPS Servers - reportedly able to sync Geth 1.13.0 and Nethermind, if slowly
• Contabo NVMe - fast enough but not enough space. 800 GiB is not sufficient.

Crucial MX500 2TB 3D NAND SATA, P/N CT2000MX500SSD1:

fio --randrepeat=1 --ioengine=libaio --direct=1 --gtod_reduce=1 --name=test --filename=test --bs=4k --iodepth=64 --size=150G --readwrite=randrw --rwmixread=75; rm test
test: (g=0): rw=randrw, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=libaio, iodepth=64
fio-3.36
Starting 1 process
test: Laying out IO file (1 file / 153600MiB)
Jobs: 1 (f=1): [m(1)][100.0%][r=170MiB/s,w=55.8MiB/s][r=43.5k,w=14.3k IOPS][eta 00m:00s]
test: (groupid=0, jobs=1): err= 0: pid=58013: Tue Aug  6 20:41:22 2024
  read: IOPS=40.1k, BW=157MiB/s (164MB/s)(113GiB/735227msec)
   bw (  KiB/s): min=56120, max=178656, per=100.00%, avg=160549.40, stdev=16317.69, samples=1470
   iops        : min=14030, max=44664, avg=40137.29, stdev=4079.44, samples=1470
  write: IOPS=13.4k, BW=52.2MiB/s (54.8MB/s)(37.5GiB/735227msec); 0 zone resets
   bw (  KiB/s): min=18936, max=59840, per=100.00%, avg=53507.94, stdev=5464.86, samples=1470
   iops        : min= 4734, max=14960, avg=13376.91, stdev=1366.21, samples=1470
  cpu          : usr=11.35%, sys=37.07%, ctx=20912690, majf=0, minf=8
  IO depths    : 1=0.1%, 2=0.1%, 4=0.1%, 8=0.1%, 16=0.1%, 32=0.1%, >=64=100.0%
     submit    : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
     complete  : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.1%, >=64=0.0%
     issued rwts: total=29492326,9829274,0,0 short=0,0,0,0 dropped=0,0,0,0
     latency   : target=0, window=0, percentile=100.00%, depth=64

Run status group 0 (all jobs):
   READ: bw=157MiB/s (164MB/s), 157MiB/s-157MiB/s (164MB/s-164MB/s), io=113GiB (121GB), run=735227-735227msec
  WRITE: bw=52.2MiB/s (54.8MB/s), 52.2MiB/s-52.2MiB/s (54.8MB/s-54.8MB/s), io=37.5GiB (40.3GB), run=735227-735227msec

Disk stats (read/write):
    dm-0: ios=29488516/9828823, sectors=235908176/78630008, merge=0/0, ticks=33612141/11641385, in_queue=45253526, util=100.00%, aggrios=29483985/9828644, aggsectors=235938760/78640656, aggrmerge=8354/1510, aggrticks=33412161/11652208, aggrin_queue=45069154, aggrutil=100.00%
  sda: ios=29483985/9828644, sectors=235938760/78640656, merge=8354/1510, ticks=33412161/11652208, in_queue=45069154, util=100.00%

4x SAMSUNG MZQL27T6HBLA-00A07 8TB

on LVM RAID0

test: (g=0): rw=randrw, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=libaio, iodepth=64
fio-3.28
Starting 1 process
test: Laying out IO file (1 file / 153600MiB)
Jobs: 1 (f=1): [m(1)][100.0%][r=418MiB/s,w=138MiB/s][r=107k,w=35.2k IOPS][eta 00m:00s]
test: (groupid=0, jobs=1): err= 0: pid=1791624: Fri Sep 20 06:50:44 2024
  read: IOPS=106k, BW=415MiB/s (435MB/s)(113GiB/277427msec)
   bw (  KiB/s): min=412601, max=433600, per=100.00%, avg=425912.77, stdev=2952.66, samples=554
   iops        : min=103150, max=108400, avg=106478.01, stdev=738.21, samples=554
  write: IOPS=35.4k, BW=138MiB/s (145MB/s)(37.5GiB/277427msec); 0 zone resets
   bw (  KiB/s): min=137098, max=145432, per=100.00%, avg=141948.63, stdev=1246.79, samples=554
   iops        : min=34274, max=36358, avg=35486.98, stdev=311.72, samples=554
  cpu          : usr=14.71%, sys=73.15%, ctx=8764196, majf=0, minf=10
  IO depths    : 1=0.1%, 2=0.1%, 4=0.1%, 8=0.1%, 16=0.1%, 32=0.1%, >=64=100.0%
     submit    : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
     complete  : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.1%, >=64=0.0%
     issued rwts: total=29492326,9829274,0,0 short=0,0,0,0 dropped=0,0,0,0
     latency   : target=0, window=0, percentile=100.00%, depth=64

Run status group 0 (all jobs):
   READ: bw=415MiB/s (435MB/s), 415MiB/s-415MiB/s (435MB/s-435MB/s), io=113GiB (121GB), run=277427-277427msec
  WRITE: bw=138MiB/s (145MB/s), 138MiB/s-138MiB/s (145MB/s-145MB/s), io=37.5GiB (40.3GB), run=277427-277427msec

Disk stats (read/write):
    dm-6: ios=29470675/9822052, merge=0/0, ticks=2172516/114900, in_queue=2287416, util=100.00%, aggrios=7373081/2457323, aggrmerge=0/0, aggrticks=537561/26680, aggrin_queue=564241, aggrutil=100.00%
    dm-4: ios=7374797/2455609, merge=0/0, ticks=536436/26820, in_queue=563256, util=100.00%, aggrios=7375991/2455677, aggrmerge=431/31, aggrticks=538438/28509, aggrin_queue=566947, aggrutil=100.00%
  nvme2n1: ios=7375991/2455677, merge=431/31, ticks=538438/28509, in_queue=566947, util=100.00%
    dm-2: ios=7372459/2457946, merge=0/0, ticks=538468/27240, in_queue=565708, util=100.00%, aggrios=7431982/2460076, aggrmerge=450/55, aggrticks=546460/28955, aggrin_queue=575442, aggrutil=100.00%
  nvme0n1: ios=7431982/2460076, merge=450/55, ticks=546460/28955, in_queue=575442, util=100.00%
    dm-5: ios=7373249/2457155, merge=0/0, ticks=538392/26592, in_queue=564984, util=100.00%, aggrios=7374415/2457222, aggrmerge=421/35, aggrticks=538730/28707, aggrin_queue=567438, aggrutil=100.00%
  nvme3n1: ios=7374415/2457222, merge=421/35, ticks=538730/28707, in_queue=567438, util=100.00%
    dm-3: ios=7371821/2458585, merge=0/0, ticks=536948/26068, in_queue=563016, util=100.00%, aggrios=7372991/2458649, aggrmerge=454/28, aggrticks=538861/28691, aggrin_queue=567552, aggrutil=100.00%
  nvme1n1: ios=7372991/2458649, merge=454/28, ticks=538861/28691, in_queue=567552, util=100.00%

I also have a Samsung 990 Pro w/ Heatsink on my laptop, running the same test was same results as SnoepNFT https://gist.github.com/yorickdowne/f3a3e79a573bf35767cd002cc977b038?permalink_comment_id=4958391#gistcomment-4958391

Solidigm P44 Pro 2TB

I get quite a bit better results on my P44 Pro 2TB (250k/83k; zero tuning):

$ fio --randrepeat=1 --ioengine=libaio --direct=1 --gtod_reduce=1 --name=test --filename=test --bs=4k --iodepth=64 --size=150G --readwrite=randrw --rwmixread=75
test: (g=0): rw=randrw, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=libaio, iodepth=64
fio-3.36
Starting 1 process
test: Laying out IO file (1 file / 153600MiB)
Jobs: 1 (f=1): [m(1)][100.0%][r=838MiB/s,w=279MiB/s][r=215k,w=71.4k IOPS][eta 00m:00s]
test: (groupid=0, jobs=1): err= 0: pid=12922: Sun Sep 29 08:48:05 2024
  read: IOPS=250k, BW=977MiB/s (1025MB/s)(113GiB/117867msec)
   bw (  KiB/s): min=809536, max=1110872, per=100.00%, avg=1001964.77, stdev=99876.67, samples=235
   iops        : min=202384, max=277718, avg=250491.18, stdev=24969.18, samples=235
  write: IOPS=83.4k, BW=326MiB/s (342MB/s)(37.5GiB/117867msec); 0 zone resets
   bw (  KiB/s): min=271720, max=371000, per=100.00%, avg=333941.07, stdev=33374.23, samples=235
   iops        : min=67930, max=92750, avg=83485.27, stdev=8343.57, samples=235
  cpu          : usr=15.19%, sys=69.35%, ctx=8336094, majf=0, minf=8
  IO depths    : 1=0.1%, 2=0.1%, 4=0.1%, 8=0.1%, 16=0.1%, 32=0.1%, >=64=100.0%
     submit    : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
     complete  : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.1%, >=64=0.0%
     issued rwts: total=29492326,9829274,0,0 short=0,0,0,0 dropped=0,0,0,0
     latency   : target=0, window=0, percentile=100.00%, depth=64

Run status group 0 (all jobs):
   READ: bw=977MiB/s (1025MB/s), 977MiB/s-977MiB/s (1025MB/s-1025MB/s), io=113GiB (121GB), run=117867-117867msec
  WRITE: bw=326MiB/s (342MB/s), 326MiB/s-326MiB/s (342MB/s-342MB/s), io=37.5GiB (40.3GB), run=117867-117867msec

Disk stats (read/write):
  nvme0n1: ios=29461506/9818976, sectors=235692056/78552008, merge=0/25, ticks=3013987/57806, in_queue=3071817, util=71.84%

.

ADATA XPG GAMMIX S70 BLADE 4TB (AGAMMIXS70B-4T-CS)
301k / 100k

Got it from Amazon USA in october 2024
Tested on framework laptop 13 (11 gen Intel), on empty ex4 drive, booted from usb

DRAM chip was a bit hot, so looks like heatsink and a cooler might be a good idea.

fio --randrepeat=1 --ioengine=libaio --direct=1 --gtod_reduce=1 --name=test --filename=test --bs=4k --iodepth=64 --size=150G --readwrite=randrw --rwmixread=75; rm test
test: (g=0): rw=randrw, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=libaio, iodepth=64
fio-3.28
Starting 1 process
test: Laying out IO file (1 file / 153600MiB)
Jobs: 1 (f=1): [m(1)][100.0%][r=1176MiB/s,w=394MiB/s][r=301k,w=101k IOPS][eta 00m:00s] 
test: (groupid=0, jobs=1): err= 0: pid=3884: Mon Oct 21 07:46:53 2024
  read: IOPS=301k, BW=1176MiB/s (1234MB/s)(113GiB/97929msec)
   bw (  MiB/s): min=    1, max= 1307, per=100.00%, avg=1177.29, stdev=110.59, samples=195
   iops        : min=  389, max=334608, avg=301386.99, stdev=28309.82, samples=195
  write: IOPS=100k, BW=392MiB/s (411MB/s)(37.5GiB/97929msec); 0 zone resets
   bw (  KiB/s): min=97520, max=449912, per=100.00%, avg=403870.42, stdev=24508.67, samples=194
   iops        : min=24380, max=112478, avg=100967.57, stdev=6127.17, samples=194
  cpu          : usr=18.97%, sys=65.46%, ctx=8422129, majf=0, minf=6
  IO depths    : 1=0.1%, 2=0.1%, 4=0.1%, 8=0.1%, 16=0.1%, 32=0.1%, >=64=100.0%
     submit    : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
     complete  : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.1%, >=64=0.0%
     issued rwts: total=29492326,9829274,0,0 short=0,0,0,0 dropped=0,0,0,0
     latency   : target=0, window=0, percentile=100.00%, depth=64

Run status group 0 (all jobs):
   READ: bw=1176MiB/s (1234MB/s), 1176MiB/s-1176MiB/s (1234MB/s-1234MB/s), io=113GiB (121GB), run=97929-97929msec
  WRITE: bw=392MiB/s (411MB/s), 392MiB/s-392MiB/s (411MB/s-411MB/s), io=37.5GiB (40.3GB), run=97929-97929msec

Disk stats (read/write):
  nvme0n1: ios=29458080/9817976, merge=0/34, ticks=1449679/142800, in_queue=1592483, util=99.51%

Tested SSD:

• Emtec X400-10 Power pro

Managed to get the following results:

Emtec X400-10 Power pro

fio --randrepeat=1 --ioengine=libaio --direct=1 --gtod_reduce=1 --name=test --filename=test --bs=4k --iodepth=64 --size=150G --readwrite=randrw --rwmixread=75
test: (g=0): rw=randrw, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=libaio, iodepth=64
fio-3.36
Starting 1 process
test: Laying out IO file (1 file / 153600MiB)
Jobs: 1 (f=1): [m(1)][99.8%][r=324MiB/s,w=107MiB/s][r=82.9k,w=27.4k IOPS][eta 00m:01s] 
test: (groupid=0, jobs=1): err= 0: pid=3257: Wed Nov  6 20:46:37 2024
  read: IOPS=60.0k, BW=234MiB/s (246MB/s)(113GiB/491712msec)
   bw (  KiB/s): min=165432, max=366584, per=99.97%, avg=239835.00, stdev=15147.15, samples=982
   iops        : min=41358, max=91646, avg=59958.75, stdev=3786.78, samples=982
  write: IOPS=20.0k, BW=78.1MiB/s (81.9MB/s)(37.5GiB/491712msec); 0 zone resets
   bw (  KiB/s): min=53960, max=120912, per=99.97%, avg=79933.04, stdev=5104.43, samples=982
   iops        : min=13490, max=30228, avg=19983.25, stdev=1276.11, samples=982
  cpu          : usr=18.32%, sys=41.47%, ctx=8479778, majf=0, minf=9
  IO depths    : 1=0.1%, 2=0.1%, 4=0.1%, 8=0.1%, 16=0.1%, 32=0.1%, >=64=100.0%
     submit    : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
     complete  : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.1%, >=64=0.0%
     issued rwts: total=29492326,9829274,0,0 short=0,0,0,0 dropped=0,0,0,0
     latency   : target=0, window=0, percentile=100.00%, depth=64

Run status group 0 (all jobs):
   READ: bw=234MiB/s (246MB/s), 234MiB/s-234MiB/s (246MB/s-246MB/s), io=113GiB (121GB), run=491712-491712msec
  WRITE: bw=78.1MiB/s (81.9MB/s), 78.1MiB/s-78.1MiB/s (81.9MB/s-81.9MB/s), io=37.5GiB (40.3GB), run=491712-491712msec

Disk stats (read/write):
  nvme0n1: ios=29465216/9820893, sectors=235723592/78573680, merge=0/595, ticks=27622647/606658, in_queue=28233556, util=63.66%

Additional Notes:

• A non-mainstream drive and therefore a cheap SSD (especially second hand)
• Gets burning hot under load

Kingston FURY Renegade 4TB, PCIe 3.0

Heat

The temperature under load, as displayed by nvme smart-log /dev/nvme0, was reaching 80°C without heatsink, I bought a basic low profile heatsink for $8 (Thermalright M.2) and now the temperature does not exceed 50°C under heavy load. It's worth it to eliminate the chance of damage or throttling.

btrfs

I am getting 30k read / 10k write IOPS with btrfs. fio seems to be mostly slowed by 100% 1-core CPU utilization of my Intel Core i5 9500T. It could be due to my btrfs setup (defaults, I tested with or without zstd:1 compression which did not make much difference) or the way fio measures performance.

ext4

With ext4, I am now getting 180k read / 60k write (x6 compared to btrfs).

Additional Notes

phoronix also benchmarked btrfs significantly slower than f2fs, ext4 and xfs in some of their Aug'24 tests. Not necessarily a fundamental btrfs issue, it may simply require more tuning.

Conclusion

Just to say that the filesystem and its configuration, and not only noatime, matter.