Hey!
So here is a node itself:
And here is a VM:
So as you can see there is a difference in speed in 4k blocks and geekbench showing less multicore "power".
Current setup for CPU: 1 socket, 32 cores, all others on default.
Disks: VirtIO-block / Write back.
My question is if I do correctly assign all cores? Why then there is such multicore difference?
And the second one is there a possibility to improve 4k blocks disk speed?
Cheers!
So here is a node itself:
Basic System Information:
---------------------------------
Processor : AMD EPYC 7313 16-Core Processor
CPU cores : 32 @ 1500.000 MHz
AES-NI : ✔ Enabled
VM-x/AMD-V : ✔ Enabled
RAM : 62.8 GiB
Swap : 0.0 KiB
Disk : 1.7 TiB
fio Disk Speed Tests (Mixed R/W 50/50):
---------------------------------
Block Size | 4k (IOPS) | 64k (IOPS)
------ | --- ---- | ---- ----
Read | 973.22 MB/s (243.3k) | 1.48 GB/s (23.1k)
Write | 975.79 MB/s (243.9k) | 1.49 GB/s (23.3k)
Total | 1.94 GB/s (487.2k) | 2.97 GB/s (46.5k)
| |
Block Size | 512k (IOPS) | 1m (IOPS)
------ | --- ---- | ---- ----
Read | 1.87 GB/s (3.6k) | 1.92 GB/s (1.8k)
Write | 1.97 GB/s (3.8k) | 2.05 GB/s (2.0k)
Total | 3.85 GB/s (7.5k) | 3.98 GB/s (3.8k)
iperf3 Network Speed Tests (IPv4):
---------------------------------
Provider | Location (Link) | Send Speed | Recv Speed
| | |
Clouvider | London, UK (10G) | 974 Mbits/sec | 8.18 Gbits/sec
Online.net | Paris, FR (10G) | 976 Mbits/sec | 8.20 Gbits/sec
Clouvider | NYC, NY, US (10G) | 921 Mbits/sec | 2.10 Gbits/sec
Geekbench 4 Benchmark Test:
---------------------------------
Test | Value
|
Single Core | 6090
Multi Core | 70778
Full Test | https://browser.geekbench.com/v4/cpu/16526240
Geekbench 5 Benchmark Test:
---------------------------------
Test | Value
|
Single Core | 1387
Multi Core | 18104
Full Test | https://browser.geekbench.com/v5/cpu/14129066
And here is a VM:
Basic System Information:
---------------------------------
Processor : AMD EPYC 7313 16-Core Processor
CPU cores : 32 @ 3000.006 MHz
AES-NI : ✔ Enabled
VM-x/AMD-V : ✔ Enabled
RAM : 15.6 GiB
Swap : 0.0 KiB
Disk : 20.0 GiB
fio Disk Speed Tests (Mixed R/W 50/50):
---------------------------------
Block Size | 4k (IOPS) | 64k (IOPS)
------ | --- ---- | ---- ----
Read | 552.17 MB/s (138.0k) | 5.10 GB/s (79.7k)
Write | 553.62 MB/s (138.4k) | 5.12 GB/s (80.1k)
Total | 1.10 GB/s (276.4k) | 10.23 GB/s (159.8k)
| |
Block Size | 512k (IOPS) | 1m (IOPS)
------ | --- ---- | ---- ----
Read | 5.15 GB/s (10.0k) | 5.47 GB/s (5.3k)
Write | 5.43 GB/s (10.6k) | 5.83 GB/s (5.6k)
Total | 10.59 GB/s (20.6k) | 11.30 GB/s (11.0k)
iperf3 Network Speed Tests (IPv4):
---------------------------------
Provider | Location (Link) | Send Speed | Recv Speed
| | |
Clouvider | London, UK (10G) | 974 Mbits/sec | 8.71 Gbits/sec
Online.net | Paris, FR (10G) | 976 Mbits/sec | 7.53 Gbits/sec
Clouvider | NYC, NY, US (10G) | 915 Mbits/sec | 1.91 Gbits/sec
Geekbench 4 Benchmark Test:
---------------------------------
Test | Value
|
Single Core | 5971
Multi Core | 60857
Full Test | https://browser.geekbench.com/v4/cpu/16526228
Geekbench 5 Benchmark Test:
---------------------------------
Test | Value
|
Single Core | 1402
Multi Core | 15634
Full Test | https://browser.geekbench.com/v5/cpu/14128941
So as you can see there is a difference in speed in 4k blocks and geekbench showing less multicore "power".
Current setup for CPU: 1 socket, 32 cores, all others on default.
Disks: VirtIO-block / Write back.
My question is if I do correctly assign all cores? Why then there is such multicore difference?
And the second one is there a possibility to improve 4k blocks disk speed?
Cheers!