Espressif just dropped the ESP32-E22

[ali_asadzadeh]

Hey everyone,Big news from Espressif today (well, announced mid-January but still fresh): they've officially unveiled the ESP32-E22, their first chip with Wi-Fi 6E and tri-band support (2.4/5/6 GHz).
It's not your classic standalone ESP32 MCU – this one's positioned as a Radio Co-Processor (RCP), meaning it offloads all the Wi-Fi + Bluetooth stack to itself via PCIe 2.1 or SDIO 3.0, so your main host CPU (could be another ESP32, Raspberry Pi CM, some NPU board, x86, whatever) can focus purely on the app.Key highlights from the announcement and product page:Tri-band Wi-Fi 6E (2.4/5/6 GHz), 160 MHz channels, 2×2 MU-MIMO, beamforming – they claim up to ~2.4 Gbps peak (iperf tested around 2.1 Gbps PHY)
Dual-mode Bluetooth: Classic (BR/EDR – yay, A2DP audio possible again!) + BLE 5.4
Dual-core proprietary Espressif RISC-V @ up to 500 MHz (with 1 MB SRAM, no PSRAM mentioned)
Host interfaces: PCIe 2.1, SDIO 3.0 (also USB FS OTG, lots of GPIOs if used standalone?)
Designed for high-throughput, low-latency stuff in crowded RF environments (6 GHz clean spectrum ftw)

Engineering samples are apparently available now (vetting required probably), so it's getting close to real-world testing.Official announcement:
https://www.espressif.com/en/news/ESP32_E22_AnnouncementProduct page:
https://www.espressif.com/en/products/socs/esp32-e22This feels like a big step up from the S/C series for anyone needing proper high-speed Wi-Fi without killing the main processor.Questions for the brains trust here:What projects would you throw this at? Video streaming IoT cams, wireless VR peripherals, high-res mesh networks, industrial high-bandwidth sensors...?
How do you plan to use it – as a pure co-processor with a beefy host (like RK3588 / Jetson / PC), or standalone for something?
Any early thoughts on rivals? Murata/Type 2 modules with Qualcomm/Intel Wi-Fi 6E chips? Other co-proc options like the new-ish NXP IW612 or something Broadcom? How does this stack up on power/cost/performance when it hits volume?
Anyone already got samples or inside info on dev tools / SDK / Linux driver status?

Very keen to hear what people think – is this the Wi-Fi upgrade we've been waiting for in the ESP ecosystem, or niche for now?Cheers!

[hans]

Wait, is this a MCU or a coprocessor? It says coprocessor on the tin. Why is it part of their ESP32 series then? Isn't the 32 synonymous to all other vendors naming their 32-bit processors FOO32?
If my assumptions is correct, then I've no idea what they were thinking. Like why would Microchip call ethernet controllers like the ENC424J600 a PIC32.

Regardless, Wifi 6E.. I've no idea what I would this chip for.
I'm already not very impressed by Espressif silicon and ESP-IDF in the first place. I tried to lookup how I2c slave works on the ESP32-C6. Arduino variant: broken. ESP-IDF: complete backwards thinking of how a I2c slave works, every I2c instance state made static, not expansion friendly at all. RTL level: a few dozen registers with each a dozen bits to set everything up. Ugh.
Meanwhile code for any generic ARM microcontroller made by ST, Atmel/Microchip, heck even Raspberry Pi: probably 100LOC.
 
Can't imagine what kind of a shitshow this would be inside the Linux ecosystem. Maybe its better if "it just works" when properly integrated inside the Linux kernel. But then that's more a case of being as jellybean as possible, not because it excites me with new application possibilities.
Anyhow, rant over.

[az1]

I'm already not very impressed by Espressif silicon and ESP-IDF in the first place. I tried to lookup how I2c slave works on the ESP32-C6. Arduino variant: broken. ESP-IDF: complete backwards thinking of how a I2c slave works, every I2c instance state made static, not expansion friendly at all. RTL level: a few dozen registers with each a dozen bits to set everything up. Ugh.

I wouldn't be so quick to judge based on the quality of the HAL, *especially* not with the Arduino HAL on not-AVR where they consistently make a mess out of perfectly serviceable chips.  The real question is: does I2C work reliably after you set all the knobs?

[hans]

Nope, Espressif example had unreliable data responses too. It was a simple ping-pong kind of slave and it would corrupt about 1 in 5 messages of 3 bytes back.
But I digress, lets talk about the E22.

[ftg]

There was already SDIO firmware for select ESP32's to use them as wifi cards in embedded systems via SDIO.
This looks like a continuation of it, with WiFi 6E bands now included.
When I first read of the E22 I got exited about tri-band wifi and PCIe.
But now that it looks more like an SDIO WiFi card and not a microcontroller, I'm less excited.

[BadeBhaiya]

Seems like they're really trying to establish themselves in the WiFi/BT/RF coprocessor space. I'm really looking forward to the data rates these chips can do. For too long I have struggled with Realtek Bluetooth/WiFi dongle Linux support. It has been getting worse especially for their newer chips. I wish Espressif could give them some competition there.

[MikeK]

The proper word is "released" or "announced", not "dropped".

[mikerj]

The proper word is "released" or "announced", not "dropped".

I was assuming they'd discontinued one of their products from the title.

[MikeK]

The proper word is "released" or "announced", not "dropped".

I was assuming they'd discontinued one of their products from the title.

Re-read the first sentence of the first post.

[mikerj]

The proper word is "released" or "announced", not "dropped".

I was assuming they'd discontinued one of their products from the title.

Re-read the first sentence of the first post.

Re-read my post.  From the TITLE

[PCB.Wiz]

The proper word is "released" or "announced", not "dropped".

Yes, “announced” is a more correct title, as it is not yet really available.

That said, a quick dictionary check has at #5 :
To Cease or Abandon: To stop including, pursuing, or doing something

but making an appearance at #9 is media slang
To Release (Media/Products): To make something publicly available, such as a new album or product launch.

[az1]

Nope, Espressif example had unreliable data responses too. It was a simple ping-pong kind of slave and it would corrupt about 1 in 5 messages of 3 bytes back.
But I digress, lets talk about the E22.

I meant more along the lines of: could you take the documentation, bang on the registers, and make something work reliably.  Software development is not one of Espressif's core competencies like most other manufacturers.  Vendor provided HALs are all going to be varying degrees of bad.

[__george__]

It seems that they want to get some market share from Realtek or similar companies which have these PCIE cards used in PC and laptops. Sounds quite boring but the naming is awful for sure. Sure, it is a dual core RISCV so technically it has 32 bit cores but I am guessing that they will not publish any info about it and thus it will not be possible to program it yourself. So mixing the names with your own microcontrollers sounds terrible 

[Geoff-AU]

The proper word is "released" or "announced", not "dropped".

Agreed.  Having clicked the link I can only assume "dropped" is intended in the same vein as "mic drop" or "bass drop" either of which is the opposite of common parlance in this industry.

"Dropped" means discontinued in electronics.  Always has, always will, and it's too late to change it.