Review Raspberry Pi 4 Review: The New Gold Standard for Single-Board Computing

[AllanGH]

Why not a NUC? ...unless it'd cost too much. AFAIK, they include a VESA mounting bracket.

Mainly because I have a drawer with a mix of the Model 3 variant boards in it...probably still have 6 of the genuine RPF-made boards left...along with a smattering of Pi-like boards and something just under a hundred different Arduino and other Atmel and ARM Cortex-based boards.

All of them are good for quick solutions to various problems, really, so it's nice to just have them on hand.

Mounting isn't an issue. Any case I have stashed away will be better-off after a visit with the drill press.

 

[g-unit1111]

Do it.


About running Windows on it, here's what their FAQ says:



https://www.raspberrypi.org/documentation/faqs/#pi-software

Very interesting, I have never heard of the IoT version of Windows before.

 

[shivansps]

Windows IoT is petty much useless, for any use.
Windows 10 ARM in other hand runs the full Windows 10 desktop with the support to run x86 .exe files on it.
On the RPI3 was very unusable, tje I/O was not enoght, runing Windows on microsd ends bad no matter the rest of the hardware, and USB 2 was not enoght to run Windows 10 decently.

Now those problems are gone, BUT i guess we back to square 1 in terms of driver support, meaning, it probably dosent work at all, and has no driver for the USB and every other device.

 

[bit_user]

Very interesting, I have never heard of the IoT version of Windows before.

In the AMA, Dr. Upton addresses the issues of getting full-blown Windows to run on it.

https://forums.tomshardware.com/thr...-ask-your-questions-now.3492239/post-21115209

Sounds like it could happen, someday.

 

[AllanGH]

Sounds like it could happen, someday.

LOL...I can't say that it would be an improvement.

 

[bit_user]

LOL...I can't say that it would be an improvement.

Well, I wouldn't use it. But I can see the appeal for some.

Plus, some businesses require their employees to use Windows. So, that would make the Pi 4 the cheapest out-of-the box Win 10 system you could buy (new).

 

[bit_user]

The 'Everything We Know about the Pi 4" article (from ~4 months ago) didn't age well

"Release Date: Not Coming in 2019"
https://www.tomshardware.com/news/raspberry-pi-4-everything-we-know,38539.html

As I was reviewing the original post on their site, I came to the FAQ section:

Raspberry Pi 4 FAQs
...

Wait, is it 2020 yet?

In the past, we’ve indicated 2020 as a likely introduction date for Raspberry Pi 4. We budgeted time for four silicon revisions of BCM2711 (A0, B0, C0, and C1); in comparison, we ship BCM2835C2 (the fifth revision of that design) on Raspberry Pi 1 and Zero.

Fortunately, 2711B0 has turned out to be production-ready, which has taken roughly 9–12 months out of the schedule.

So, the answer is that they were able to skip two of the four respins of the SoC that they anticipated.

Source: https://www.raspberrypi.org/blog/raspberry-pi-4-on-sale-now-from-35/

 

[in_the_loop]

"it also provides out-of-order execution so it’s not waiting for the output of one process to start on another "

Well then, Raspberry Pi 4 welcomes Specter and Meltdown then?
Since the "in-order" execution of the former Pi:s prevented these type of exploits (and several other newer out-of-order/branch prediction cache exploits).
https://www.raspberrypi.org/blog/why-raspberry-pi-isnt-vulnerable-to-spectre-or-meltdown/

With that, if patched, also a possible decrease in performance, probably more likely than on normal PC:s since as a IOT-device or media server (as in many installations) it will use the network part much more, which makes it much more likely to see the performance hit from the exploits (which kicks in more when the network part is accessed).

 

[Dianne.S]

Any plans on comparing the atomic PI to the raspberry?
It's based on x86 so it would be much more compatible with what people need to run.
https://dlidirect.com/products/atomic-pi

The Atomic Pi is really just a rebranded AAEON MF-001. It's this: https://industrialbid.com/m/lot-details/index/catalog/1604/lot/285151/2000---High-Speed-PCBAs

That explains why it's being sold below cost (and why it's unlikely to be a long-term product.)

 

[bit_user]

"it also provides out-of-order execution so it’s not waiting for the output of one process to start on another "

Yeah, that's pretty sketchy wording. "process" would usually refer to a UNIX Process, but OoO execution is orders of magnitude more fine-grained, and cannot cross process boundaries if there's no SMT/hyperthreading. The best way to describe it is by simply saying the CPU can reorder the instructions in a more optimal (but still correct) pattern.

Well then, Raspberry Pi 4 welcomes Specter and Meltdown then?
Since the "in-order" execution of the former Pi:s prevented these type of exploits (and several other newer out-of-order/branch prediction cache exploits).
https://www.raspberrypi.org/blog/why-raspberry-pi-isnt-vulnerable-to-spectre-or-meltdown/

With that, if patched, also a possible decrease in performance, probably more likely than on normal PC:s since as a IOT-device or media server (as in many installations) it will use the network part much more, which makes it much more likely to see the performance hit from the exploits (which kicks in more when the network part is accessed).

It's a good catch. However, there's no way those mitigations would entirely nullify the benefit of moving to an OoO core.

And the Pi is still somewhat less vulnerable to side-channel attacks by virtue of no SMT/hyperthreading.

 

[rattle99]

Does the read speeds for an external HDD over USB3.0 depend over how much RAM the device has?

 

[Piguy490]

Can the Pi4 Boot from an external SSD located on the USB 3.0 port? I have the Pi3B+ and never was able to get it to boot from my external 256 Gig USB drive at all, I had to boot from the Micro sd, although I could save and grab data from the SSD at the USB 2.0 speeds. I know the 3B+ was capable of booting from certain compatible USB 2.0 drives, I just hope that the Pi4 can boot directly from my SSD. Appreciate any information on this topic!!! Thanks Much Appreciated! If you have a link to best article or steps for setting it up with OS installed on SSD and booting from the SSD would be appreciated!

For the previous message, I don't think the amount of ram would really matter when looking at USB 3.0 actual transfer speeds, but I may be wrong.

Jeff

 

[bit_user]

Can the Pi4 Boot from an external SSD located on the USB 3.0 port?

They said that initially, it cannot. However, that is planned to be fixed in the near future (don't know the timeframe).

In the meantime, you can install most of the OS on a USB 3-connected drive, so long as /boot/ resides on a SD card. That way, you get 99% of the performance benefits of USB 3, although it means you still need a SD card.

 

[in_the_loop]

Yeah, that's pretty sketchy wording. "process" would usually refer to a UNIX Process, but OoO execution is orders of magnitude more fine-grained, and cannot cross process boundaries if there's no SMT/hyperthreading. The best way to describe it is by simply saying the CPU can reorder the instructions in a more optimal (but still correct) pattern.


It's a good catch. However, there's no way those mitigations would entirely nullify the benefit of moving to an OoO core.

And the Pi is still somewhat less vulnerable to side-channel attacks by virtue of no SMT/hyperthreading.

Still, there is also a possibility some of these PI4:s will end up in custom IOT-devices (maybe even sold as part of another solution) and left unpatched in later exploits.
And, for that sake, other IOT products that will incorporate this ARM and later models that probably will have an out-of-order execution type of architecture.
It's often devices meant to solve a specific thing and there is already a problem with many of them not being updated (either due to the person owning them no knowing/caring about it or that there are no updates made possible by manufacturers.
They will often just "sit" there.
It is often best to have these type of devices running on its own little subnet, if that is possible to set up on the router.

 

[bit_user]

Still, there is also a possibility some of these PI4:s will end up in custom IOT-devices (maybe even sold as part of another solution) and left unpatched in later exploits.

The patches to the original spectre/meltdown vulnerabilities that affected ARM should already be in the Raspbian kernel used by the Pi v4. Perhaps the Pi 4 even used updates of the A72 that has them fixed in hardware.

As for other attacks that are discovered in future, the issue of vendors & users failing to patch their devices is currently true of all IoT platforms. As for the A72 being OoO, we can hope that, by now, any side-channel vulneabilities in it have been found - it is basically a 4-year-old core that's probably already in hundreds of millions of devices. But there are no guarantees.

In any case, IoT has less exposure to side-channel attacks, since those require executing malicious code. The main way that most people execute malicious code is through their web browser, which most IoT devices don't have. The other way is for cloud-based servers to have a malicious VM running alongside its target, on the same physical machine - also not an issue for IoT. So, I think your concern is slightly misplaced.

there is already a problem with many of them not being updated (either due to the person owning them no knowing/caring about it or that there are no updates made possible by manufacturers.
They will often just "sit" there. It is often best to have these type of devices running on its own little subnet, if that is possible to set up on the router.

This is an ongoing issue, and it's not helped by the fact that many IoT devices need to access the cloud as part of their function. I agree with segregating them on their own subnet, but we both know that most people won't do this.

What's needed is for some kind of certification program where manufacturers not only undergo 3rd party testing, but also set aside funds to provide fixes and firmware updates for some pre-determined amount of time. There could also be a standardized way of updating the devices, so that either they could self-update or you could automatically scan your network and update all the devices on it.

 

[dfarrell]

I'm having an issue with my Logitech wireless keyboard/mouse. I'm plugged into one of the USB 2.0 ports. 2GB RPi4, plenty of power, 5GHz WiFi. Keyboard typing can be non responsive for seconds at a time. I did not notice this with early Pis. This seems to randomly happen.

 

[bit_user]

I'm having an issue with my Logitech wireless keyboard/mouse.

Thanks for reporting that, but hopefully this isn't the only place you're posting about it. I'd imagine there are much better places to go, if you're looking for solutions. There don't seem to be a lot of Pi users on here, and even fewer will already have the Pi 4.

 

[redtailboas]

Very strange, the issues with video playback. 4 Cortex A-72 should make it no problem at all. No hardware decoding?

 

[Giroro]

I see the review was updated to include the thermals for the unreleased alpha firmware.
I don't really think that's fair, considering some users were seeing their USB 3.0 speeds absolutely decimated by the update.

Maybe it's jumping the gun to assume that firmware tweak is going to reach production with the thermal improvements intact, as the issue seems related to their implementation of USB power management.

https://www.raspberrypi.org/forums/viewtopic.php?f=28&t=244421

 

[bit_user]

Very strange, the issues with video playback. 4 Cortex A-72 should make it no problem at all. No hardware decoding?

From what I've read, hardware-accelerated decoding is dependent on which player app you use. I don't have a link handy, but I recommend trying to find out what others are having success with.

 

[secretxax]

So which OS did you test this on? You didn't mention. Mentioned no supporting OSes.

 

[bit_user]

So which OS did you test this on? You didn't mention. Mentioned no supporting OSes.

The only OS supported at launch was surely Raspbian - Raspberry Pi's offshoot of the Debian Linux distro.

One thing to know about Raspbian is that it's 32-bit, even though this hardware is capable of 64-bit. The Raspberry Pi foundation seems in no hurry to switch to 64-bit, instead preferring to keep binary compatibility between the different generations.

Other ARM boards, such as ODROID's N2, can run some mainstream distros, like Ubuntu. I don't know if the same is true of the Pi, but I wouldn't be surprised if it unofficially works.

 

[StewartHH]

I got one of these Raspberry PI 4 with loaded 4G of ram, primary because I interested in mentoring programming to 18 year Aspie. But I spent only a 30 minutes or so before I decided to say it was not worth the time. In know way this device can replace a real pc - even one with the Core M processor. I did notice that Microsoft has Visual Studio that can be used with it - I even had it hook up to my LG ultra wide but I found it awkward to used.

I think the best way to write application is to figured out how to cross compile to run applications and write the code on real machine then this basically toy. The good news it was only $99. I am curious that Microsoft could come out with Windows for ARM on than they may charge $1000 for like the Surface Pro X. But it would be cool to pay only $99 for same thing as Surface Pro X.

 

[bit_user]

I spent only a 30 minutes or so before I decided to say it was not worth the time. In know way this device can replace a real pc - even one with the Core M processor.

What did you use for storage?

I would recommend a SATA SSD with UB3 enclosure/adapter. I think that's the only way to get PC-level I/O performance out of it, which should greatly affect the user experience.