Just sharing the results of my USB speed tests. I am using s3 at 160Mhz. I found that the fastest speed possible was about 500K Bytes per second. That's in the area that I have been seeing form other peoples posts. To get that speed the RTOS spends about 40% of its time running the USB code, so not enough left for my particular application, maybe enough for yours. 100K Bytes per second was where the USB code started to have an impact on how much processing time was left for other things. I would say a good Rule Of Thumb would be, if your going faster than 100K Bytes per second, keep an eye on how much time is spent doing the USB. If you are thinking "the spec says it can go to 12Mbps" no, it can't.
Enjoy
USB speed test
-
- Posts: 1688
- Joined: Mon Oct 17, 2022 7:38 pm
- Location: Europe, Germany
Re: USB speed test
What happens when you run the CPU at its full 240MHz?
Re: USB speed test
https://community.st.com/t5/mems-sensor ... d-p/523996
Now you have to add flash read/write operations time, which is another overhead.
Couple years ago I performed USB MSC tests on S3, you can find results somewhere on the forum.
As far as I remember the read speed could reach even 900kB/s.
I am suggesting using 240MHz clock, which seems to have nice boost on DMA operations.
Now you have to add flash read/write operations time, which is another overhead.
Couple years ago I performed USB MSC tests on S3, you can find results somewhere on the forum.
As far as I remember the read speed could reach even 900kB/s.
I am suggesting using 240MHz clock, which seems to have nice boost on DMA operations.
Re: USB speed test
I should have mentioned that this was transmitting only. The product I am designing will be transmitting continuous sensor data which is the critical path. It will only be receiving a small amount of configuration data at initialization so no simultaneous transmit and receive.
I should have also mentioned I am using CDC_ACM.
I should have also mentioned I am using CDC_ACM.
Re: USB speed test
https://developerhelp.microchip.com/xwi ... ull-speed/
In theory CDC ACM should reach transfer around 1MB/s, not sure if esp32 can reach such transfer speed, but it only theory.
To get highest throughput you have to make sure that:
- esp32 is sending data in tight loop without congestion
- your client app on PC can transfer data that fast
So, can you confirm somehow your client/test app can go that fast?
In theory CDC ACM should reach transfer around 1MB/s, not sure if esp32 can reach such transfer speed, but it only theory.
To get highest throughput you have to make sure that:
- esp32 is sending data in tight loop without congestion
- your client app on PC can transfer data that fast
So, can you confirm somehow your client/test app can go that fast?
Re: USB speed test
The only thing the ESP32 was doing was sending USB, so yes, it was a very tight loop.
The PC application that was doing the receiving was written by my colleague. He has a lot of experience with USB communication, especially in high speed industrial control applications. We were able to verify that the application was indeed fast enough.
The PC application that was doing the receiving was written by my colleague. He has a lot of experience with USB communication, especially in high speed industrial control applications. We were able to verify that the application was indeed fast enough.
Re: USB speed test
I dont have good tools to make performance test on CDC-ACM yet, but this test should be good enough.
It is CDC-NCM doing iperf test from linux PC, DUT is this devkit
https://docs.espressif.com/projects/esp ... guide.html
results
It is not 1MB/s, but close enough.
Hope this helps you a bit.
It is CDC-NCM doing iperf test from linux PC, DUT is this devkit
https://docs.espressif.com/projects/esp ... guide.html
results
Code: Select all
------------------------------------------------------------
Client connecting to 192.168.4.1, TCP port 5001
TCP window size: 16.0 KByte (default)
------------------------------------------------------------
[ 1] local 192.168.4.2 port 40978 connected with 192.168.4.1 port 5001 (icwnd/mss/irtt=14/1440/0)
[ ID] Interval Transfer Bandwidth
[ 1] 0.0000-3.0000 sec 2.12 MBytes 5.94 Mbits/sec
[ 1] 3.0000-6.0000 sec 2.25 MBytes 6.29 Mbits/sec
[ 1] 6.0000-9.0000 sec 2.12 MBytes 5.94 Mbits/sec
[ 1] 9.0000-12.0000 sec 2.25 MBytes 6.29 Mbits/sec
[ 1] 12.0000-15.0000 sec 2.12 MBytes 5.94 Mbits/sec
[ 1] 15.0000-18.0000 sec 2.25 MBytes 6.29 Mbits/sec
[ 1] 18.0000-21.0000 sec 2.12 MBytes 5.94 Mbits/sec
[ 1] 21.0000-24.0000 sec 2.25 MBytes 6.29 Mbits/sec
[ 1] 24.0000-27.0000 sec 2.25 MBytes 6.29 Mbits/sec
write failed: Broken pipe
shutdown failed: Transport endpoint is not connected
[ 1] 27.0000-30.0000 sec 2.37 MBytes 6.64 Mbits/sec
[ 1] 0.0000-30.0066 sec 22.3 MBytes 6.24 Mbits/sec
Code: Select all
iperf> iperf -s -i 3
I (10007) IPERF: mode=tcp-server sip=localhost:5001, dip=0.0.0.0:5001, interval=3, time=30
I (10017) iperf: Socket created
iperf> I (13887) iperf: accept: 192.168.4.2,40978
Interval Bandwidth
0.0- 3.0 sec 5.76 Mbits/sec
3.0- 6.0 sec 5.77 Mbits/sec
6.0- 9.0 sec 5.91 Mbits/sec
9.0-12.0 sec 5.89 Mbits/sec
12.0-15.0 sec 5.81 Mbits/sec
15.0-18.0 sec 5.83 Mbits/sec
18.0-21.0 sec 5.86 Mbits/sec
21.0-24.0 sec 5.91 Mbits/sec
24.0-27.0 sec 6.06 Mbits/sec
27.0-30.0 sec 6.17 Mbits/sec
0.0-30.0 sec 5.90 Mbits/sec
I (43887) iperf: TCP Socket server is closed.
I (43887) iperf: iperf exit
Hope this helps you a bit.
Who is online
Users browsing this forum: AG_OHP and 67 guests