The long-awaited Raspberry Pi 5 SBC is finally here with a powerful Broadcom BCM2712 quad-core Cortex-A76 processor clocked at 2.4 GHz, 4GB or 8GB memory, and the usual Raspberry Pi Model B form factor.
The Raspberry Pi 4 Model B was introduced in June 2019, so it took over four years to get a new model, and the Raspberry Pi 5 should deliver 2 to 4 times the performance, compete directly with Raspberry Pi-shaped Rockchip RK3588S SBCs such as the Radxa Rock 5A or Cool Pi 4 Model B and benefit from a larger community and better software support.
Raspberry Pi 5 specifications:
- SoC – Broadcom BCM2712
- CPU – Quad-core Arm Cortex-A76 processor @ 2.4 GHz with crypto extensions, 512KB per-core L2 caches, 2MB shared L3 cache
- GPU – VideoCore VII GPU @ 800 MHz with support for OpenGL ES 3.1, Vulkan 1.2, 4Kp60 HEVC decoder
- System Memory – 4GB or 8GB LPDDR4X-4267 SDRAM (at launch, but cheaper variants with 1GB and 2GB RAM should become available later on)
- Storage – MicroSD card slot with support for high-speed SDR104 mode (twice the performance over the Raspberry Pi 4)
- Video Output/Input
- 2x HDMI port up to 4Kp60 with HDR support
- 2x 4-lane MIPI camera/display transceivers
- Networking
- Gigabit Ethernet RJ45 port via Broadcom BCM54213PE transceiver; optional PoE+ support through Raspberry Pi PoE+ HAT
- Dual-band 802.11ac Wi-Fi 5
- Bluetooth 5.0 / Bluetooth Low Energy (BLE)
- USB – 2x USB 3.0 ports supporting simultaneous 5Gbps operation, 2x USB 2.0 ports
- Expansion
- 40-pin Raspberry Pi GPIO header
- PCIe 2.0 x1 interface (requires separate M.2 SSD HAT or other adapter)
- Misc – Real-time clock (RTC), powered from an external battery, power button, UART debug connector
- Power Supply – 5V/5A DC power via USB-C, with Power Delivery support
- Dimensions – 85 x 56 mm
Besides the new BCM2712 SoC with a much faster CPU (with Armv8 crypto extensions) and a new VideoCore VII GPU, the Raspberry Pi 5 gets twice the peak performance for the microSD card, HDR support, 4-lane MIPI interfaces (instead of 2-lane for the Pi 4), a PCIe 2.0 x1 FPC connector for compatible add-ons, a built-in RTC, a power button, a separate UART connector, and support for USB PD to power the board. Raspberry Pi OS is still the OS of choice.
The Raspberry Pi Model B form factor is starting to really show its limitations with the board featuring a non-standard PCIe connector, micro HDMI connectors, and the MIPI CSI/DSI connectors are shared, so I would not be surprised to see a Raspberry Pi 5 Model C in the future with a Pico-ITX form factor or similar, and potentially with extra features such as 2.5GbE networking. Alternatively, they may also leave this task to partners using a future Raspberry Pi Compute Module 5 (CM5).
This is also the first full-size SBC from the company using silicon designed by Raspberry Pi with the RP1 “Southbridge” providing the bulk of the I/O capabilities for Raspberry Pi 5, and delivering a step change in peripheral performance and functionality with over twice the USB bandwidth, two 1.5 Gbps MIPI transceivers, higher SD card performance, and a PCIe 2.0 x1 interface.
Pricing for the Raspberry Pi 5 is still very attractive, costing only $5 more than the Raspberry Pi 4 in equivalent configurations with the 4GB model going for $60 and the 8GB model for $80. Sales should start by the end of October, and the Raspberry Pi 5 will remain in production until at least January 2035, or about 11 years of guaranteed availability.
We’ve just received a complete kit with a Raspberry Pi 5 with 8GB RAM, an active cooler, a US keyboard, a mouse, and a power supply, so expect a review once I’m back from my holidays. In the meantime, more details may be found on the product page.
Jean-Luc started CNX Software in 2010 as a part-time endeavor, before quitting his job as a software engineering manager, and starting to write daily news, and reviews full time later in 2011.
Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress
Seems it is already at a disadvantage against the rockchip models, it is missing a eMMC ad M2 slot.
Yes, agreed. Their advantages now is only their support and community. For pricing, I’d rather go with OPi5. RPi5 $60 vs OPi5 $66. Just with +$6, you get +4 core, M.2 + eMMC slot, full HDMI port + type-c with DP. It’s no-brainer for me.
Where is the OPi5 for $66?
Same places where you’ll find RPi5 for $60 a few hours after launch, nowhere.
https://www.aliexpress.com/store/group/OPi-5-4G/1553371_40000004098085.html
You could get 8GB with $85. Just search there.
I just got OPicm4 1.8ghz 8gb 64gb EMMC for 60 with base board, m.2 and 2x MIPI-CSI. same size as RPi5. and has 1 top NPU. RPI5 has competitions and it’s growing
Yes. RPi as of now and in the future will have many competitors.
Its is slightly inferior in most respects but comes close but also has many missing feature and 1x lane of pcie2.0 is sort of mweh in use.
Its a Raspberry with the same GPIO pin mux and all the support Pi’s get so that bridges the gap.
I have one on pre order but thinking PiStockGate has ended Pi dominance but Raspberry still wears a crown over a lesser empire.
5V/5A… Who though that was a good idea? That’s not part of the PD spec at all.
Also, this form factor needs to be killed with fire.
Otherwise a big step up in terms of the hardware, if only the Videocore wasn’t so limited.
“not part of the PD spec at all.”
See rev 3.1 of the spec from the USB implementers forum, specifically table 10-2 and associated notes.
It appears you are misinformed, as more than 5V at 3A *is* permitted – it’s just a matter of time until third party power supplies add 5V/5A support.
The issue is, that you need a marked cable too to deliver 5A, which is not really common.
Designing a low end PSU for 5V 5A is quite a feat, even my 100W USB-C PSUs cannot supply 5A at 5V.
USB-PD also requires actual negotiation for anything above 5V@3A.
A 9V 3A supply requirement would be a lot easier to comply with.
> The issue is, that you need a marked cable too to deliver 5A
And since these cables are more expensive than RPi Trading Ltd’s new $12 USB-C 5V/5A power brick everyone right in his mind will buy the latter or simply doesn’t care at all if it’s not about powering USB consumers.
IMO the more important question is the power path: are the 5V at the USB receptacles stable 5V or suffer from the usual voltage drop when consumption goes up?
The Raspberry Pi people bought some Radxa PD 30W to test.
https://radxa.com/power-pd-30w
What is the highest amperage your power brick is able to negotiate at 5V?
5A@5V(5.2V actually).
Can we stop with stupid things like this and just go 9 or 15V please? You guys as board makers are responsible for following standards instead of making up random shit as you go along.
If you look into the real world problem or study a little bit, you will not think this is stupid.
Radxa PD 30W supports PD 3.0 with 9V / 12V in first priority, but if the negotiation fails or the target device doesn’t’ support PD, the Radxa PD 30W can still provide 25W power at 5V for the max capability.
To be honest I prefer the Rock5b buck and just plug 12v into it so that I have a 12v rail.
PD is great for mobile devices, but its a bit of a head scratch for devices that are likely statically powered.
PS anyone does PD have an efficiency standard as my PD is much less efficient than the Pi 5v USB-C but guess the loading is much less with a 120watt PD?
Whats the Radxa 5v efficiency like Tom as I prob should not of got a 120watt as only power tiny things.
[ quality vendor power supplies are possibly ~86-90% efficient for PD modes, slightly lower % for QC modes, mentioned within a testing setup were Apple(~96W), Hyperjuice, Omega or a Kovol (these 100-240V, ~120W, ~86%_5V_15-20Woutput, ~90%_20V_~98Woutput and ~$80-90). A Kovol hottest part was ~158°F/70°C. ]
This ^
The important thing to explain to those reading this without being aware of the problem is that the voltage drop across the cable directly depends on the current, and that the lower the actual voltage, the higher the current and further higher the relative drop. E.g. dropping 0.4V at 5V leaves with with 4.6V only, while for the same power, under 10V only 0.2V would be lost. And while 5V is fed directly into some components (e.g. USB ports) and thus has to be extremely stable, higher voltages are always converted via an onboard regulator which is basically insensitive to… Read more »
> That’s why 5V input at large amperage is not good at all.
Sure. Maybe the new Renesas DA9091 PMIC contains a step-up converter to provide stable 5.1V on USB and HDMI ports even with input voltage low as hell? 🙂
BTW: the integrated ADC in the RP1 chip provides 4 voltage and 1 temperature reading.
If it does include step up converters to do that, then they’re doubly fools as it would have been more efficient and more compatable to negotiate a higher input voltage and regulate that *down* to the voltages you need.
Sure, but they already sold millions of 5.2V USB-C power bricks…
Care to share that table, as I can’t find any such info for 5V. Yes, 5A is used from 20 V and up, but not below. And permitted, nice, but no-one is going to make it, as it’s a stupid inclusion that is super inefficient. My brand new PD 3.1 140 W doesn’t support the mode and I don’t know of a single charger that does. I guess we’ll have to wait and see if anyone supports this “permitted” mode, I very much doubt it.
While there may be a technicality that 5V/5A is permitted (probably under the PPS spec), the USB-PD spec strongly discourages nonstandard voltages/currents in supplies for the same of compatability and interoperability. You want to be able to say “this needs a 20W supply” and not “try to find a supply that does 15V with at least 1.33A” because you know the 20W tier requires 15V/1.33A.
5V/5A is not a technicality, and it’s not non-standard – it’s right there, in the PD spec. The only thing you can say about it right now is that it’s not very common, because the majority of devices haven’t needed it (opting for higher voltages instead) but it’s very likely that this is about to change.
Did you miss the previous table 10.1? Here’s some important highlights: Simple to identify capabilityHigher power Sources are a superset of smaller onesUnambiguous Source definitionsMaximize 3A cable utilization A 5V/5A supply violates these four considerations. Please understand that the USB-PD spec is very complex and you have to take it in as a totality. You can’t just pick and choose little parts of it. IMHO, that last one is the most damning–as it’s an electrical engineering concern and I’m an engineer. Voltage drops over resistances are proportional to current (Ohm’s law). Power is the product of current and voltage. The… Read more »
This^
None of which actually precludes or prevents 5V/5A. Look, I’m not arguing that it’s a good idea, but it’s pretty clear there is nothing in the USB PD standard which actually prevents it. But arguing that it’s non-standard and violates the specification is total bullshit – it’s simply not a very common combination rather than actually being non-standard (although that is almost certainly going to change based on recent correspondence from Chinese PSU factories that are interested in supporting this VI combination). Right or wrong, the standard explicitly permits 5V/5A – the only caveat is that the correct 5A cable… Read more »
Care to share that table, as I can’t find any such info for 5V. @TLS USB Power Delivery, rev 3.1, version 1.8, page 805, table 10-2 (“SPR Normative Voltages and Minimum Currents”), note 2: 2. The Fixed PDOs Maximum Current field Shall advertise at least 3A, but May advertise up to RoundUp (PDP/Voltage) to the nearest 10mA. Requires a 5A cable if over 3A is advertised. So, there you go – 5V/5A is within the PD spec – just don’t use a piece of wet string as a cable. As it happens, Chinese PSU/charger factories are already tooling up to… Read more »
“Not illustrated are that currents higher than 3A are allowed to be offered up to a limit of 5A given that a 5A cable is detected by the Source and the Voltage times current remains within the Source PDP Rating.”
So it won’t become a thing, as the eMarker cables cost roughly twice that of non eMarker cables.
Also, you’re aware this would mean new USB-PD controllers and chips as well, since it’s not the norm, which makes it even less likely to happen.
> which makes it even less likely to happen
Chinese PSU factories are already lining up to support 5V/5A, based on their interest which has been communicated in the last few weeks. I would wager they have a better idea of the likely demand and market opportunity than typical consumers.
I would not be surprised if 5V/5A is supported by at least some if not most Chinese PD PSUs within the next 6-9 months.
Wonderful, a flood of non-standard power supplies of questionable quality. Exactly what the USB forum was trying to prevent with table 10.1–which you conveniently keep ignoring. “You want to ruin a specification? That’s how you ruin a specification.”
Again, what’s with the “non-standard” FUD? It’s explicitly permitted in the spec! Table 10-1 in no way prevents 5V/5A.
I believe you’ve read that table wrong, nowhere does it say 5A there.
Furthermore, there are no 5A modes below 20V anywhere else.
I’m sorry if you can’t do basic math, I can’t help you. PDP/Voltage.
There is no PDP profile for more than 3A below 20V though…
Just because 5A is allowed for USB-PD, doesn’t mean it’s allowed at all Voltage levels.
So I can’t help you if you can’t read.
I guess you’re proof that while everyone can read a spec only some people can actually comprehend it. There’s 2 table notes that you’re ignoring which explicitly confirm that >3A on 5V is permitted. I can’t help it if you’re ignoring what’s right in front of you – maybe English isn’t your native language? May/Shall confusing you? I dunno. Anyway, this is pointless – how you or I interpret this specification is totally irellevant. If you’ve got a concern, contact the RPi Foundation and/or usb.org!
Aren’t you nice and friendly…
Ok, I asked someone I know that was involved in writing the specs and yes, you’re correct, it’s is “allowed” but he also said it makes no sense, so no-one would implement it, since this is the only device out there with those kind of power requirements. However, 5.1 V is not within spec, unless the RPi 5 uses PPS mode, which is unclear.
> Aren’t you nice and friendly… Yes, I am normally! However stating and re-stating what is clearly written in the spec does become a little frustrating. 🙂 > and yes, you’re correct Thanks, it’s actually decent of you to acknowledge that, as most would not! > However, 5.1 V is not within spec, unless the RPi 5 uses PPS mode, which is unclear. The Source output voltage max/min is +/- 5%, so at 5V an output of 5.1V would be within spec (up to 5.25V, actually). Since the Pi5 PSU is designed to output 5.1V then presumably they are regulating… Read more »
The RPi5 PD 5.1V/5A PSU is based on Rubycon and Panasonic caps with GaN switching, so I would not be surprised if it has very good regulation (ie. better than 5%, which would mean it is safely within the PD spec) even though it only costs $12. With additional support for the more common 9V/12V/15V PD modes (up to 27W) it is a very cheap, versatile and high quality PSU with (almost certainly) excellent voltage regulation (which is very distinct from Chargers that tend to focus more on current regulation). The Radxa 30W PSU also implements the 5V/5A PD mode… Read more »
A few years ago I bought a 5.2V/2.4A power brick from energizer that I was very happy with because this extra voltage would cover for voltage drop across cables and connectors. It has long been the one giving me the most stable experience with various SBCs. This just requires a good regulation to avoid overshoots, but basically all 5V-compatible regulators found in field, and even very old TTL chips designed for 5V supported up to 5.5 in the spec. And 30 years ago, the Pentium 66 required a modified ATX PSU producing 5.15V (actually PSUs found in field would usually… Read more »
that’s minimum requirements for sources, so RPI power supply is USB-PD compliant and implements optional features. The problem is on the sink-side (RPI5 board)
Let’s see how it really performs (especially the DRAM bandwidth which is often one of their bottlenecks), but at first glance they seem to have finally listened to their users who were hacking the board to try to get PCIe, and all those who’ve been complaining for more than a decade regarding the lack of crypto extensions. Sure there’s probably still a lot to say regarding the form factor and lack of on-board storage, but maybe after the RPi4 that I considered “decent”, this one will finally be “good enough” for most use cases ? It’s important that RPi focuses… Read more »
> maybe after the RPi4 that I considered “decent”, this one will finally be “good enough” for most use cases ?
c’t magazine measured 3.25W in idle. No thanks, x86 exists…
Sounds high, maybe it will get better later, we’ll see. Anyway the pricing remains below decent x86 boards of comparable size. This does count for a number of users.
I guess the high idle consumption is due to the I/O situation. The 40nm RP1 chip next to the SoC is attached via PCIe Gen2 x4 providing USB3, Ethernet and the single Gen2 PCIe lane to the outside. RPi folks set ASPM to performance and not default or powersave (most probably for a reason).
BTW: memory bandwidth/latency is not that bad.
The single Gen2 PCIe is not from the RP1 PCH if you look at the board. It is directly from BCM2712 SoC which makes sense as the pcb layout would be extremely hard given the position of the PCH and the slot. The same reason applies for the SDIO which also from the SoC not the PCH.
> The single Gen2 PCIe is not from the RP1 PCH if you look at the board
Yep, and this PCIe lane can also be forced to Gen3 speeds though then signaling problems then are to be expected: https://github.com/geerlingguy/sbc-reviews/issues/21#issuecomment-1738733390
> BTW: memory bandwidth/latency is not that bad.
Indeed!
But there seems to be something seriously wrong (maybe with Geekbench, maybe with the hardware): https://github.com/geerlingguy/sbc-reviews/issues/21#issuecomment-1739842852
Hmm but isn’t this geekbench joke seriously dependent on starting programs ? If so, given that the storage here is just an SD card, it’s seriously impaired compared to a device with a properly working eMMC.
I retested on my Rock 5B with Geekbench 6.2 and the installation lives on one of the worst SD cards I own.
Then I tried to start the Phoronix crap to let it run over night. But that filled the 8 GB card completely and the experiment ended with ‘no space left on device’.
> Then I tried to start the Phoronix crap to let it run over night. But that filled the 8 GB card completely and the experiment ended with ‘no space left on device’.
🙂
OrangePi5-Plus has M.2 with PCIe 3.0 x4
RPi5 has PCIe 2.0 x1 interface and no M.2
OPI5+ has a theoretical bandwidth of 4GB/s.
RPi5 has a theoretical bandwidth of 500MB/s.
I’m really disappointed that it has no M.2 under the board.
> RPi5 has a theoretical bandwidth of 500MB/s.
Nope, since you can force it to Gen3 speeds, see Jeff Geerling’s findings (setting ‘dtparam=pciex1_gen=3’ in the ThreadX config file instructs the kernel to switch to twice the throughput).
Asides that: who on earth relies on storage bandwidth? Which use case does this represent other than silly benchmarks? Booting from an NVMe SSD is great due to its low latency.
I was hoping to use an M.2 board with PCIe to 5 SATA (poor man’s NAS). If the Gen3 speeds are available, that is awesome. At 1GB/s you can read 2 SATA in parallel: one for streaming, one for other stuff.
Wait for the SATA HAT for Raspberry Pi 5 from Radxa.
Ooh… I know someone who would love to test that 😉
Even if you saturate the gigabit ethernet link with streaming you’re only using 25% of that PCIE-2 link speed. PCIE-3 isn’t going to offer much.
The man which sold them told me that those of PCIe line was design as 3.0 but run as 2.0 due to FPC line or the connector.
Prob Radxa had a whole load of probs with FPC and cross talk and likely will be the same.
Guess it all depends how they have split gnd and signal lines and on this version you may be stuck with 2.0
> Asides that: who on earth relies on storage bandwidth? Which use case does this represent other than silly benchmarks?
LLMs. When you run tests on many of them and you have to load a model as large as the whole RAM from the SSD, you appreciate it when it’s fast. I’ve seen myself complaining when loading a 44G file at 1 GB/s because it took forever 🙂
> I’ve seen myself complaining when loading a 44G file at 1 GB/s because it took forever
Fortunately on RPi 5B the RAM is limited to 8GB 😉
(at least for now, maybe BCM2712 is capable of addressing more so RPi Trading Ltd. might present a 16GB board in the future once 128Gb LPDDR4x modules become affordable)
From playing with Llama and whisper its pretty pointless going above the 7b param models that weigh in at just 4gb https://github.com/ggerganov/llama.cpp#quantization
Should be good as bench rk3588(s) vs Pi5 though as the code is well optimised at newer Arm achitecture such as v8.2 and onwards.
https://github.com/ggerganov/whisper.cpp also
Its sort of pointless loading a model so large its so slow it could be considered unusable 8gb is prob a sweetspot on a A76
7B models with reasonable quantization (Q5_M) take about 4.8GB of storage plus half a gig of context at runtime. I agree that for interactive processing on A76 a 7B model is fine. However for offline processing (I’m using them to review patches) 7B is not accurate enough, even Mistral which is already exceptional for a 7B one, and others like openorca+platypus-13B give me excellent results for about 10G of RAM. The processing time is long but I don’t care, as the goal is to batch them. Even if it runs for one hour each day it’s OK.
Phoronix already posted some benchmarks against the Orange Pi 5 (RK3588S).
The performance is good, too bad the form factor sucks and no nvme or even emmc.
I have thought Eben Upton said that next RPi will be in the next year… But it seems they are even now, after release, behind the market.
Upton/RPT always misdirect about future plans, and sometimes something goes right behind the scenes and months are shaved off the release date.
Some choices are questionable, but it seems to hold its own against RK3588S. It will sell big as usual. I am off the Pi train though.
How many uart ports on the GPIO pin? There are 4 uarts on pi 4.
So what’s wrong with the GPIOs? Why all the GPIO is 2*2. It looks terrible.
Ordered. Wait for early november to come
The wifi antenna on the 4 (and Zero 2W) has been very very directive in my experience. At least “opposite side” still somewhat worked but now with the active cooling, that seems very unlikely to work well. Any chance we will see some benchmark for wifi (signal quality, latency, throughput and latency during high throughput) with various orientation and with or without the active cooler? I haven’t seen any such benchmark but in two places where the wifi “path” is not a straight line, turning the board has been absolutely required to get decent speeds. I was seeing < 2Mbps,… Read more »
I can see they put a lot of effort in cost down to keep the pricing as low as possible. Although it is no longer the darn cheap $35 SBC that it used to be, the engineering decision clearly focused on cost more than performance. Now let’s hope we can get these boards with its listing price which is lower than its RK3588 based counterparts. I expect the software support would be much better than rockchip but it also has significantly less hardware subsystems that rockchip provides like video encoder and the npu.
They mentioned the RP1 (Southbridge) is a cousin of the rp2040 – and it includes a PIO and a Cortext M3. Sounds like some interesting things could be done with that combination. (PIO is very handy on the RP2040, so having this on a SBC connected via PCIe unlocks a lot fun ideas at least for me)
Hardly a cousin, just designed by the same team, but according to the article on the RPi site, they’ve been working on it since 2016…
Natural evolution of rpi4, selectively adding key stuff but considering tradeoffs and cost. Pcie .. But no m.2 plug to stick a drive into? Likely to continue to disappoint people hoping for every improvement they could list. As ever the rpi software and community give it a big leg up but it’s important to be clear eyed you’re buying that instead of the latest hardware or even the best board design. In part, I am not a fan of micro HDMI, I know the board is small but it’s not THAT small, this is the end of the lineup prioritizing… Read more »
look like an overclocked rp4… 5A is crazy.
What ??? no audio jack ??? they thought that the missing microphone wasnt crazy enough ?
Some people can get by without the audio jack, for example most HTPC users. I’ll be interested to see who must have the audio jack, and if it’s going to be added right back by expensive HATs and cases.
Realistically, you are going to need a new case, power supply and active cooling. Although definitely better, as one would expect, it feels rather underwhelming overall once you add up the real price, although I think that many will still go for it because of the hype that surrounds it. That PCIE 2.0 interface needs to be shown to be of good use too (more cost for an adapter) as the storage options don’t exactly shine out of the box. It feels like a partial catch up on the competition and would (if precedent is followed) have to sustain what… Read more »
Ebon Upton will tell you that they haven’t even started thinking obout RPi5 while holding an RPi5 in his hand.
They have been working for 7 years (since 2016) at the RP1 chip used by RPi5. You can be 100% sure that the hardware for RPi6 is ready (but no software) and they are now working on RPi7.
The RP1 was originally intended for a much older Pi but they were still tweaking the design to get it to work well… all the way through to finally including it on Pi 5.
Is Mr. Upton still also employed at BroadCom’s STB division?
Anyway, he should know better than most others how BroadCom’s roadmap wrt I/O capabilities of their VideoCore SoCs looks like. As such RPi Trading Ltd. knew for a long time they had to design their own I/O chip and based on pictures in early 2022 the C0 stepping was ready…
What might be interesting with the Broadcom chip is does it OC?
I think somebody, probably Jeff, said the CPU is good for 2.6 GHz (+8.3%), but not stable at 2.8 GHz. That could change down the line with firmware updates and hardware revisions.
tomshardware.com/news/overclocking-raspberry-pi-5
Here’s Tom’s with 3.0 GHz CPU, 1.1 GHz GPU. GPU overclock didn’t help but they also didn’t test it much. 25% CPU overclock led to a couple of 25% improvements, 8% and 22% in two other tests.
I see people complaining about the lack of eMMC or M.2. All those copycats (I do not favor RPi or OPi or all the others) had to put something out there to be distinguished from RPi. I never used eMMC because their compatibility is rather poor compared to a microSD card and they usually cost a lot more. Since it has UHS-1 support, RPi now has doubled the storage speed. One can still use nVME through a good USB enclosure. On one of the videos about RPi 5 some are complaining about the fact they will have to use cooling… Read more »
The SD card and USB port speeds still suck though.
> The SD card and USB port speeds still suck though
The equipment used by Tom’s Hardware is more likely to blame since Jeff measured +90 MB/s sequential transfer speeds with a good TF card and +300 MB/s at the USB ports (where I suspect his USB SSD is to blame and +400 MB/s should be possible as well as already measured with other SoCs relying on DesignWare 3 USB3 IP)
https://github.com/geerlingguy/sbc-reviews/issues/21
Indeed; I did not use my faster USB to NVMe drive as it was employed in another adventure when I ran the USB tests. That was with a flash drive with an older SSD controller.
I don’t know what hardware Tom’s hardware used, but it was quite inadequate!
No av1 decoder… *sigh*
And there is no dsp for h264 either, so for media use this will be rather bad & short lived…
You should read what LibreELEC had to say first: libreelec.tv/2023/09/28/rpi5-support/ BCM2712 supports HEVC 4K60 hardware decoding. It no longer supports H264 in hardware. This might sound odd but it removes the RPi4’s 1080p restriction on H264 decoding and the 4K H264 test media we have has played. The big increase in performance from the Quad-Core A76 chip means RPi5 can software decode AV1, H264, VC1, VP9, and more at 1080p with ease. In our testing with YouTube and inputstream.adaptive a surprising amount of 4K media also plays. Optimised (lower refresh-rate and bitrate) 4K30 VP9 is generally fine while more demanding… Read more »
PS those who got in on the 1st day with a preorder, if you would give a shout out here on reciept as likely then we can judge when they are shipping.
I ordered one as an SBC geek and got an itch to get my hands on it 🙂
Thanks
Rpi4 was my 2nd board I was trying after ROCKPro64 as light desktop replacement. Software support was much better and liked that even for browsing internet and retrogaming. Also I was enjoying silence but I was still missing a bit more power so I got rid of both.After having dispute with AliExpress so I am no more their customer i was unable to buy Orange Pi 5 for reasonable price and here you are Rpi5 so I placed order and hoping for best.
Anyone got thiers as mine has turned up wondering if anyone is doing any 3rd party testing and evaluation of the released boards?