Skip to main content

Hetzner CPX21

Ashburn, VA

Benchmark complete
Finished Oct 16, 2026
Geekbench 7 single-core
1,373
Geekbench 7 multi-core
Skipped
Disk 4K IOPS
55,905
fio random read and write, 50/50
Monthly price at test time
$37.49
Same price today

Server

  • Provider
    Hetzner
  • Plan
    CPX21
  • Location
    Ashburn, VA
  • vCPU
    3
  • RAM
    4 GB
  • Storage
    80 GB NVMe
  • Network
    Unknown
  • OS image
    ubuntu-24.04
  • Trigger
    Scheduled
  • Queued at
    Oct 16, 2026, 18:49:00 UTC
Visit Hetzner

Latest runs of this plan

Full history of this plan

CPU

Version Single-core Multi-core
Geekbench 7 1,373 Skipped Geekbench Browser

Multi-core was skipped because Geekbench 7 needs about 3.8 GB of free memory to run 3 threads, and this server didn't have that much. Single-core needs much less, so it still ran.

Scores from different Geekbench versions aren't comparable with each other. Charts only use Geekbench 7.

Geekbench 7 workloads

Workload Single-core Multi-core
File Compression
1,414
79th of 216
Single-core only
Navigation
1,418
75th of 216
Single-core only
HTML5 Browser
1,250
88th of 216
Single-core only
PDF Viewer
1,472
74th of 216
Single-core only
Photo Library
1,296
73rd of 216
Single-core only
Clang
1,159
92nd of 216
Single-core only
Text Processing
1,213
85th of 216
Single-core only
Asset Compression
1,380
87th of 216
Single-core only
HDR
1,318
82nd of 216
Single-core only
Photo Editor
1,632
63rd of 216
Single-core only
Ray Tracer
1,347
73rd of 216
Single-core only
Structure from Motion
1,677
56th of 216
Single-core only
Game Physics
1,262
91st of 216
Single-core only
Video Encoder
1,322
70th of 216
Single-core only
Audio Encoder
1,416
76th of 216
Single-core only
Video Player
1,500
70th of 216
Single-core only

Rank among the 216 runs with Geekbench 7 workload scores. Bars are relative to the best run. Geekbench doesn't run every workload multi-core, so some rows only have a single-core score.

Memory, TLS and scheduler

Memory bandwidth (STREAM)

Test Result Runs Spread
Copy higher is better 80.9 GB/s 3 22.9%
Scale higher is better 58.9 GB/s 3 12.8%
Add higher is better 63.6 GB/s 3 18.3%
Triad higher is better 66 GB/s 3 20.8%

Average of 3 runs. Spread is the standard deviation as a share of the average.

TLS and crypto (OpenSSL)

Test Result Runs Spread
RSA 4096 signs higher is better 633/s 3 0.657%
RSA 4096 verifies higher is better 41,891/s 3 0.269%
SHA-256 higher is better 2.37 GB/s 3 0.135%
AES-256-GCM higher is better 11.1 GB/s 3 0.65%
ChaCha20-Poly1305 higher is better 6.36 GB/s 3 0.501%

Average of 3 runs. Spread is the standard deviation as a share of the average.

Scheduler latency (schbench)

Test Result Runs Spread
Request latency, median lower is better 4.71 ms 3 0.196%
Requests per second, 90th percentile higher is better 620/s 3 0.186%
Request latency, 99.9th percentile lower is better 9.53 ms 3 12.6%
Requests per second, average higher is better 612/s 3 0.231%

Average of 3 runs. Spread is the standard deviation as a share of the average.

Disk

Block size Read Write Total
4K 109 MB/s (27,916 IOPS) 109.3 MB/s (27,989 IOPS) 218.4 MB/s (55,905 IOPS)
64K 1.06 GB/s (17,393 IOPS) 1.07 GB/s (17,484 IOPS) 2.13 GB/s (34,877 IOPS)
512K 1.63 GB/s (3,347 IOPS) 1.72 GB/s (3,525 IOPS) 3.36 GB/s (6,872 IOPS)
1M 1.9 GB/s (1,945 IOPS) 2.03 GB/s (2,074 IOPS) 3.93 GB/s (4,019 IOPS)

fio random read and write, 50/50, run by YABS against the boot disk.

Network

IPv4 transfer speed

Loading...
Target Protocol Send Receive
Clouvider, London, UK (10G) IPv4 2.44 Gbps 3 Gbps
Eranium, Amsterdam, NL (100G) IPv4 2.23 Gbps 2.84 Gbps
Leaseweb, Singapore, SG (10G) IPv4 756 Mbps 937 Mbps
Clouvider, Los Angeles, CA, US (10G) IPv4 3.14 Gbps 3.95 Gbps
Leaseweb, NYC, NY, US (10G) IPv4 9.36 Gbps 8.55 Gbps

Public iperf3 servers vary in capacity and availability. Compare a target only with itself over time.

Datastores

PostgreSQL

fsync speed (pg_test_fsync)

Method Operations per second Time per operation
One 8 kB write
open_datasync 3,804 0.263 ms
fdatasync 3,754 0.266 ms
fsync 2,179 0.459 ms
fsync_writethrough Not supported here n/a
open_sync 2,234 0.448 ms
Two 8 kB writes
open_datasync 1,898 0.527 ms
fdatasync 3,513 0.285 ms
fsync 1,846 0.542 ms
fsync_writethrough Not supported here n/a
open_sync 997 1 ms
open_sync at different write sizes
1 * 16kB open_sync write 2,193 0.456 ms
2 * 8kB open_sync writes 1,076 0.929 ms
4 * 4kB open_sync writes 549 1.82 ms
8 * 2kB open_sync writes 247 4.04 ms
16 * 1kB open_sync writes 140 7.16 ms
fsync on a non-write descriptor
write, fsync, close 1,932 0.518 ms
write, close, fsync 2,086 0.479 ms
Non-synced 8 kB writes
write 1,343,468 0.001 ms

pg_test_fsync against the Docker volume Postgres keeps its data on. fdatasync is Linux's default wal_sync_method.

pgbench at scale 10

Workload Clients Transactions/s Median p95 p99 p99.9 Failed
Read/write (TPC-B like) 1 1,239 0.763 ms 1.07 ms 1.59 ms 2.7 ms 0
Read/write (TPC-B like) 8 5,549 1.28 ms 2.54 ms 3.82 ms 7.24 ms 0
Read/write (TPC-B like) 32 5,823 4.2 ms 13.5 ms 22.1 ms 38.6 ms 0
Read only 1 15,794 0.058 ms 0.093 ms 0.124 ms 0.195 ms 0
Read only 8 78,866 0.091 ms 0.174 ms 0.232 ms 0.379 ms 0
Read only 32 90,570 0.306 ms 0.527 ms 0.703 ms 1.07 ms 0

Scale 10 (1,000,000 accounts), loaded in 1.03 s. 60 s per run, prepared statements. Latency comes from sampled transaction logs, and p99.9 is left out when fewer than 1,000 were sampled.

pgbench at scale 100

Workload Clients Transactions/s Median p95 p99 p99.9 Failed
Read/write (TPC-B like) 1 1,062 0.845 ms 1.53 ms 2.15 ms 4.25 ms 0
Read/write (TPC-B like) 8 5,233 1.38 ms 2.62 ms 4.34 ms 9.19 ms 0
Read/write (TPC-B like) 32 7,876 3.72 ms 6.87 ms 10.1 ms 17.4 ms 0
Read only 1 13,034 0.071 ms 0.11 ms 0.153 ms 0.446 ms 0
Read only 8 60,424 0.122 ms 0.221 ms 0.287 ms 0.547 ms 0
Read only 32 73,645 0.39 ms 0.678 ms 0.913 ms 1.4 ms 0

Scale 100 (10,000,000 accounts), loaded in 11.1 s. 60 s per run, prepared statements. Latency comes from sampled transaction logs, and p99.9 is left out when fewer than 1,000 were sampled.

MySQL

sysbench OLTP

Workload Threads Transactions/s Queries/s Median p95 p99 Errors
Read/write 1 237 4,736 3.82 ms 6.55 ms 10.7 ms 0
Read/write 8 1,004 20,080 7.7 ms 12.3 ms 17 ms 0
Read/write 32 1,331 26,616 23.5 ms 38.2 ms 46.6 ms 0
Read only 1 665 10,645 1.42 ms 2.03 ms 2.52 ms 0
Read only 8 2,627 42,035 2.86 ms 4.82 ms 5.77 ms 0
Read only 32 2,725 43,607 11.2 ms 16.1 ms 19 ms 0

10 tables of 100,000 rows, loaded in 5.7 s. 60 s per run, one connection per thread.

Redis

memtier

Mix Operations/s Hits/s Median p99 p99.9
1:10 set:get 101,528 92,297 0.607 ms 1.23 ms 1.92 ms
1:1 set:get 102,818 51,409 0.599 ms 1.26 ms 2.05 ms

1,000,000 keys of 32 bytes. 64 connections over 2 threads, 60 s per ratio.

Web frameworks

Framework Fortunes req/s Single query req/s JSON req/s Fortunes p99 at 256 connections
Next.js nextjs 987 3,436 4,874 410 ms
Express express-postgresjs 11,421 15,955 47,635 41.5 ms
Fastify fastify-postgres 11,548 16,649 53,762 43.2 ms
NestJS nestjs 7,164 7,822 25,467 65.9 ms
Nuxt nuxt 10,188 15,544 52,251 58 ms
FastAPI fastapi 5,620 6,602 27,019 211 ms
Django django-postgresql 2,709 3,053 7,133 192 ms
Rails rails 4,435 6,258 22,902 77.6 ms
Laravel laravel 1,165 1,241 2,248 245 ms
Symfony symfony 1,508 1,651 4,476 189 ms

Peak requests per second across the 64 and 256 connection levels, measured with wrk. Each test has to pass the TechEmpower response checks first, or it's marked invalid.

Web apps

WooCommerce

Page mix

Step Requests/s Median p95 p99 Error rate
8 virtual users
Home page 3.41 461 ms 578 ms 617 ms 0%
Shop page 3.41 647 ms 757 ms 810 ms 0%
Category page 3.41 391 ms 603 ms 722 ms 0%
Product page 3.41 443 ms 542 ms 572 ms 0%
Product search 3.41 378 ms 493 ms 542 ms 0%
32 virtual users
Home page 3.27 1.8 s 2.05 s 2.16 s 0%
Shop page 3.27 2.41 s 2.71 s 2.77 s 0%
Category page 3.27 1.99 s 2.5 s 2.64 s 0%
Product page 3.27 1.8 s 1.99 s 2.02 s 0%
Product search 3.27 1.65 s 1.82 s 1.87 s 0%

Cart and checkout

Step Requests/s Median p95 p99 Error rate
8 virtual users
Add to cart 6.55 499 ms 622 ms 649 ms 0%
Cart 6.55 411 ms 498 ms 526 ms 0%
Checkout 6.55 318 ms 412 ms 436 ms 0%
32 virtual users
Add to cart 6.37 1.79 s 2.1 s 2.29 s 0%
Cart 6.37 1.72 s 1.97 s 2.14 s 0%
Checkout 6.37 1.43 s 1.72 s 1.92 s 0%

k6 with a fixed number of virtual users, 61 to 68 s per level. The error rate counts responses that weren't a 200.

Discourse

Signed in

Step Requests/s Median p95 p99 Error rate
1 virtual user
Topic page 2.08 208 ms 340 ms 452 ms 0%
Topic JSON 1.04 104 ms 146 ms 196 ms 0%
Home page 1.04 187 ms 283 ms 413 ms 0%
Categories page 0.529 218 ms 242 ms 254 ms 0%
User activity page 0.512 134 ms 180 ms 273 ms 0%
4 virtual users
Topic page 6.13 292 ms 392 ms 585 ms 0%
Topic JSON 3.06 150 ms 200 ms 249 ms 0%
Home page 3.08 262 ms 350 ms 569 ms 0%
Categories page 1.54 300 ms 429 ms 625 ms 0%
User activity page 1.54 190 ms 321 ms 481 ms 0%
16 virtual users
Topic page 5.9 1.1 s 1.33 s 1.67 s 0%
Topic JSON 2.94 920 ms 1.08 s 1.31 s 0%
Home page 2.97 1.06 s 1.3 s 1.59 s 0%
Categories page 1.49 1.1 s 1.37 s 1.55 s 0%
User activity page 1.49 950 ms 1.15 s 1.41 s 0%

Anonymous

Step Requests/s Median p95 p99 Error rate
16 virtual users
Topic page 129 32.3 ms 73.3 ms 103 ms 0%
Topic JSON 64.5 31 ms 75.3 ms 105 ms 0%
Home page 64.5 32.3 ms 74.2 ms 101 ms 0%
Categories page 32.3 31.3 ms 69.2 ms 101 ms 0%
User activity page 32.3 151 ms 213 ms 343 ms 0%

k6 with a fixed number of virtual users, 60 to 61 s per level, after a 21 s warmup. The error rate counts responses that weren't a 200. Discourse ran with 6 app workers and 1024 MB of Postgres shared buffers, what its installer picks for a 4 GB box. It added a 2 GB swapfile first, the way its installer does, and didn't need it. The container peaked at 2,757 MB.

Ghost

Readers

Step Requests/s Median p95 p99 Error rate
1 virtual user
Home page 9.56 29.1 ms 38.5 ms 47.1 ms 0%
Post page 23.9 15.7 ms 20.6 ms 28.3 ms 0%
Tag page 4.78 19.8 ms 25.7 ms 29.5 ms 0%
Author page 4.78 20.1 ms 25.2 ms 31.3 ms 0%
RSS feed 4.78 24.7 ms 31.2 ms 38.7 ms 0%
4 virtual users
Home page 11.6 90.2 ms 116 ms 132 ms 0%
Post page 29.1 64 ms 91.1 ms 104 ms 0%
Tag page 5.81 57.8 ms 78.7 ms 91.6 ms 0%
Author page 5.81 57.1 ms 78 ms 88.5 ms 0%
RSS feed 5.79 63.5 ms 86.9 ms 98.7 ms 0%
16 virtual users
Home page 11.5 351 ms 417 ms 458 ms 0%
Post page 28.7 296 ms 347 ms 383 ms 0%
Tag page 5.71 193 ms 238 ms 256 ms 0%
Author page 5.76 190 ms 234 ms 250 ms 0%
RSS feed 5.75 198 ms 240 ms 258 ms 0%

Staff in Ghost Admin

Step Requests/s Median p95 p99 Error rate
1 virtual user
Post list 5.03 29.2 ms 43.2 ms 49.5 ms 0%
Post editor 15.1 11.1 ms 15.4 ms 20.2 ms 0%
Saving a draft 15.1 29.2 ms 38.1 ms 49.6 ms 0%
Member list 5.03 35 ms 48.7 ms 55.5 ms 0%
4 virtual users
Post list 6.61 36.3 ms 45.6 ms 52.4 ms 0%
Post editor 19.9 18.1 ms 25.1 ms 32.3 ms 0%
Saving a draft 19.9 153 ms 190 ms 208 ms 0%
Member list 6.64 46.3 ms 59.5 ms 75.4 ms 0%

k6 with a fixed number of virtual users, 60 s per level, after a 20 s warmup. The error rate counts responses that weren't a 200. Ghost ran the way its ghost-docker setup runs it: one Node process, and MySQL with its stock 128 MB buffer pool. Memory peaked with Ghost at 313 MB and MySQL at 625 MB.

Wagtail

Visitors

Step Requests/s Median p95 p99 Error rate
1 virtual user
Home page 1.56 83.5 ms 106 ms 116 ms 0%
Blog index 1.56 274 ms 322 ms 371 ms 0%
Blog post 4.67 28.1 ms 45 ms 50.2 ms 0%
Bread page 4.67 31 ms 40.9 ms 56.4 ms 0%
Location page 1.56 29.4 ms 40.8 ms 52.4 ms 0%
Site search 1.55 49.2 ms 76.8 ms 80.6 ms 0%
4 virtual users
Home page 2.92 199 ms 295 ms 347 ms 0%
Blog index 2.92 683 ms 938 ms 1 s 0%
Blog post 8.77 52.5 ms 95.9 ms 138 ms 0%
Bread page 8.77 58.3 ms 95.9 ms 112 ms 0%
Location page 2.94 51.5 ms 88.4 ms 95 ms 0%
Site search 2.92 80 ms 174 ms 196 ms 0%
16 virtual users
Home page 3.1 604 ms 756 ms 869 ms 0%
Blog index 3.09 1.6 s 1.9 s 2.02 s 0%
Blog post 9.21 344 ms 539 ms 619 ms 0%
Bread page 9.21 344 ms 512 ms 655 ms 0%
Location page 3.1 337 ms 517 ms 642 ms 0%
Site search 3.05 404 ms 612 ms 658 ms 0%

Editors in the Wagtail admin

Step Requests/s Median p95 p99 Error rate
1 virtual user
Admin dashboard 0.694 87.1 ms 99 ms 132 ms 0%
Page explorer 1.4 338 ms 394 ms 491 ms 0%
Page editor 2.1 159 ms 191 ms 295 ms 0%
Image library 0.694 64.3 ms 76.7 ms 588 ms 0%
Admin page search 0.694 39.8 ms 75.1 ms 80.6 ms 0%
4 virtual users
Admin dashboard 1.06 288 ms 548 ms 639 ms 0%
Page explorer 2.16 800 ms 1.86 s 2.71 s 0%
Page editor 3.23 394 ms 759 ms 848 ms 0%
Image library 1.08 199 ms 485 ms 577 ms 0%
Admin page search 1.06 107 ms 266 ms 434 ms 0%

k6 with a fixed number of virtual users, 60 to 61 s per level, after a 20 s warmup. The error rate counts responses that weren't a 200. Wagtail ran the way bakerydemo ships it: uWSGI with 2 workers of 4 threads each, whatever the box, and Postgres and Redis on stock settings. Memory peaked with Wagtail at 298 MB, Postgres at 189 MB, and Redis at 26 MB.

Advertised vs. measured

Advertised Measured
vCPU 3 3
CPU Shared vCPU 0% of busy CPU time stolen
RAM 4 GB 3.7 GiB
Storage 80 GB NVMe 75 GiB
Architecture x86_64 x86_64
Virtualization KVM KVM

Guests usually report slightly less RAM and disk than advertised because of kernel reservations and filesystem overhead.

Detected hardware

CPU model
AMD EPYC-Rome Processor
CPU cores
3
CPU frequency
2445 MHz
AES-NI
Yes
VM-x/AMD-V
No
RAM reported
3.7 GiB
Swap reported
0 MiB
Disk reported
75 GiB
Virtualization reported
KVM
CPU steal
0% of busy CPU time stolen
Distro
Ubuntu 24.04.4 LTS
Kernel
6.8.0-138-generic
Architecture
x86_64
IPv4 reachable
Yes
IPv6 reachable
Yes

Advertised at test time

Plan
CPX21
Product line
regular_purpose
Location
Ashburn, VA
vCPU
3 shared vCPU
Architecture
x86_64
RAM
4 GB
Storage
80 GB NVMe
Virtualization
KVM
Network
Unknown
Transfer
2.2 TB
Port 25
On request
Monthly price at test time
$37.49
Hourly price at test time
$0.060
YABS version
v2026-09-20
Ran at
Oct 16, 2026, 18:49:39 UTC
Parser version
5

Run timeline

Queued
Oct 16, 2026, 18:49:00 UTC
Server created
Oct 16, 2026, 18:49:34 UTC
First update from the server
Oct 16, 2026, 18:49:34 UTC
Completed
Oct 16, 2026, 20:17:39 UTC
Server deleted
Oct 16, 2026, 20:17:39 UTC
Cost of this run
$0.089
Watch party 0 watching

This party closed at . It's read only now.

Watch party

Loading the party