The world of the overclocker contains many arcane tweaks to squeeze the last drops of performance from a computer, many of which require expert knowledge to understand. Happily for Raspberry Pi 5 owners the Pi engineers have come up with a set of tweaks you don’t have to be an overclocker to benefit from, working on the DRAM timings to extract a healthy speed boost. Serial Pi hacker [Jeff Geerling] has tested them and thinks they should be good for as much as 20% boost on a stock board. When overclocked to 3.2 GHz, he found an unbelievable 32% increase in performance.
We’re not DRAM experts here at Hackaday, but as we understand it they have been using timings from the Micron data sheets designed to play it safe. In consultation with Micron engineers they were able to use settings designed to be much faster, we gather by monitoring RAM temperature to ensure the chips stay within their parameters. Best of all, there’s no need to get down and dirty with the settings, and they can be available to all with a firmware update. It’s claimed this will help Pi 4 owners to some extent as well as those with a Pi 5, so even slightly older boards get some love. So if you have a Pi 5, don’t wait for the Pi 6, upgrade today, for free!
Having written microcode for RAM, it really pays to know your hardware. There are a lot of clever people who come up with tricks that still comply with timings.
Opteron?
Or possibly IBM POWER10’s OMI.
Or who knows, maybe Samsung or some other manufacturer’s DDR really has gotten so advanced that it needs firmware to manage it.
Meanwhile, I just grabbed in my closet and got out a raspberry pi 3 :-p
A pi3 is perfectly capable, depending on what you do with it. Not everything needs to be speedy, and there are good savings when it comes to power usage. I still use a pi3 as an Internet radio. It even has a sexy touch screen attached.
It would be really nice if raspi went back to releasing a board that was cheap and power efficient. I guess that’s kinda what the zero is, but it’s definitely lacking in ports unlike the old raspi 2.
The last several generations of pi have just been an upward march towards price and performance nobody asked for.
The professional/commercial embedded hardware market asked for higher performance, and they get what they want because they make up at least 80% of the Pi market.
The university R&D market benefits by getting access at lower prices than the commercial embedded market is willing/able to pay, by RPi Trading guaranteeing that the education customers are able to buy onesie/twosies at launch for cost plus 10%.
The latter is plowed back into development of and support for new products, until commercial demand is being satisfied by ramping of products to max production/yields.
i definitely agree with and appreciate that different markets have different demands and the pi 5 is definitely targetting a specific demand that surely must exist.
but i am not impressed with what i have seen of the commercial hardware developers that use pis. i genuinely don’t know the extent of this market, but the product that landed on my workbench is a helium miner (a crypto scam with a radio component to try to deceive people about the crypto fundamentals). i believe it cost about $300 in 2020. it is an aluminum box with a raspberry pi 4 and a lora daughter-board inside of it. it died after 2 years because it used the stock raspberry pi os configuration of mounting a cheap microsd card read-write and logging to it. it couldn’t be restored because the vendor had put crypto private keys on the obliterated microsd card, which are now lost forever.
there’s so much wrong with this picture: the product did not require the power of a pi4, but they thoughtlessly used one. the product was sold for too high a price, and the high price was maintained through the scarcity enforced by it and all its competitors relying on unobtanium pi4 in 2020. the fly-by-night vendor didn’t resolve the well known and easily-avoidable problem of constantly writing to a microsd card. the fly-by-night vendor didn’t resolve the problem of replacing the microsd card.
i don’t know, were scams and extremely low-quality products 10% of the reason that pi4 was unobtainable in 2020? or 90% of it? i don’t know. but i do know it’s a non-zero fraction. personally, i believe that a major component of pi’s sales model is poorly-thought-out products that don’t need to exist. this is a huge problem in all of technology, so i’m sorry to single out pi but we do keep hearing about their institutional customers and i just wanted to gesture in the direction of the cretin behind the curtain. no wizard there.
Greg just because some scam products used a Pi poorly doesn’t mean they are not in a huge number of excellent products that are using them for good reason and selling at reasonable prices. A scam can contain anybody’s parts, they are just going to use Pi’s because they are the lowest effort solution to a cobbled together janky product – doesn’t matter the SBC is expensive overkill when you are going to charge a small fortune for your ‘audiophile gold plated TOS link’ level scam…
I suspect you just haven’t noticed the Pi’s in commercial products much because so many products use one and don’t bother to mention it anywhere. Rather par for the course now if not entirely traditional to have bare minimum technical details listed on the adverts and with hard to find or no detailed tech specs.
Though most of the really quality products I’ve seen that use a Pi use a compute module version (but that may just be a bias based on where I’m looking). Which for most makes sense only the IO you actually need, in the form factor you want but still getting to piggyback upon the work done keeping a Pi product updated and functional for probably longer than your products expected lifespan… But there have been plenty of useful for your task things I’ve noticed using a regular Pi too over the years they have been in operation.
One of my favourites that I think I’ve seen done about 4 times now with varied generations of Pi is the Din rail computer/GPIO possibly with beefy relay combo meant for giving you flexible powerful compute and IO for your construction virtually anywhere you’d need it in a rugged and commonly used (industrially) system. It might still get used for tasks it is overkill for. But deploying 30 identical nodes that are capable of every task you need is preferable to a little bit of this one, and that one, and the other that are too specifically tailored to only the current task – lost all flexibility to reconfigure the space easily and have to stock spares of all these different nodes.
All the older boards are still in production, if you want a Pi2, you can have one. And loads of people continually ask for more power. The Pi5 and CM five are huge sellers.
I wish they would put a damn dsi connector on a zero/zero2. Hobbiest portables would take off.
As long as they keep selling, I guess many buyers are glad to pay the price for performance. If someone is satisfied with less performance, the earlier models are still available. I would argue that “somebody” must have asked for higher performance.
Not a kernel patch but a firmware update, to the Intel-Management-Engine-adjacent VPU core. You’re not in control of your Pi, because the raspi is Not Open Source.
The closest anyone ever came to writing open firmware for the VPU which is [1] which is too incomplete to be useful in any meaningful sense. It boots, but storage is unreliable and USB, Ethernet, DMA, video, etc. don’t work. This also required a massive reverse-engineering effort because raspi don’t provide any support whatsoever.
[1] https://github.com/christinaa/rpi-open-firmware
don’t let perfect be the enemy of good, this sort of idealistic BS is how good things die. The Pi is fantastic value and far better supported than almost anything else out there.
Value is becoming more debatable with things like the cheap N100 boards out there. And x86 will have better support for at least awhile.
this is what i came to say as well. there are a lot of upsides and downsides. i don’t mean to malign pi. but it is essential to understand the limits of this closed source hack. the forum post linked in the article implicitly displays the weakness that upsets me the most, but doesn’t directly discuss it.
the whole userland, kernel, and firmware are bundled together and upgraded together and several of the people commenting on the thread appear to be struggling to get the right versions. for some use cases, this doesn’t matter at all. for others, it’s a fatal flaw. ymmv.
personally, i hate it. i understand that there will be closed-source firmware but i want it to at least have a stable, generalized interface so that i am not forced to upgrade my open source kernel in lock step with the firmware. it destroys a lot of the advantages of the open source components, if they are forced to march to the beat of a changing interface to closed components. it’s not a hypothetical problem for me — the one and only project i have done on pi was needlessly difficult because of the bad design of the closed component and then was entirely 100% obsoleted by literally the very next firmware update.
if you’re an open source guy who wants to hack the kernel, this kind of garbage approach to closed firmware is a show-stopper. if you aren’t, you can afford to not care. but keep in mind, you won’t be benefitting from the work of people like me. the community of open source hackers around the pi is really hampered by forcing kernel hackers to bifurcate into two camps: people like me who hate it and avoid it versus insiders who have privileged access to the closed components.
Great, now I just need two things: [1] Readily available RPi-5 stock. [2] Affordable RPi-5 prices. Remember when high-end RPi computers used to be available everywhere for truly affordable prices? Yeah, those were the good old days – sigh…
Pull the other one, every new Pi has sold out quickly since the very first one.
If you want a cheap one buy a zero, don’t bitch about the highest performance model costing a bit more than the one released 4 years ago as if inflation doesn’t exist.
The highest performance model now costs more than an x86 box (usually small form factor with power supply ram and storage). While the Pi does not come with storage (M.2) or power supplies.
I was a huge proponent of Pi and own every model from the very first one to come out through the Pi4 in multiple variants.
I will not buy a Pi 5. It simply makes no sense, and what Pi was originally targeting is not what they are targeting now.
I guess avaliablility depends on where you live (plenty avalaible in the UK). The prices look pretty good to me, compared to the price of anything similar.
Until you look at all the peripherals required for a Pi5. Then once you have coolers, cases, power supplies, maybe an M.2 and so on, you might as well buy a small form factor, ready to go n100 box which is far more capable and supported by every linux distro.
I see these same comments every time the Pi is mentioned on Hackaday. I’m not saying I don’t believe you, but Hackaday is all about sharing information.
Do you have any links to your favourite “ready to go n100 box”? Are they available in ARM flavours or only Intel? If they are x86-only, what about AMD? Are there any similar products that have a port analogous to the Pi’s GPIO header? Are they available globally or just in your own country? Will they be supported for at least a few years? If I’m basing something around it, whether it’s a commercial product or just something to put in my house, I like the idea of being able to buy spare parts or download software updates for many years to come.
I know the Pi isn’t perfect*, but generally the comments disparaging the Pi or suggesting a faster/cheaper/better alternative are very light on detail.
In the spirit of contributing, one example is the SD card storage. Nicely mitigated in recent Pis that have the option of booting from USB or even SATA. They are picky about power supply, requiring 5V at quite a high current. The power issue I think is explained by trying to keep the price down. For commercial products the designer already has a PSU in mind. For personal projects I’d wager most hackers want to save money by repurposing an old phone charger and a few capacitors. For people who don’t want to do either of these things, there is an official Pi power supply. Rest assured that if the Pi’s power supply *was changed, there would be hordes of people on here complaining that is was too expensive, or not fit for their particular niche application :-)
**OK, while we’re on niche applications, I am disappointed that the PiStorm Amiga accelerator isn’t compatible with the Pi5. The speedup would be awesome! But then again, the Pi4 is still available and will be for years to come. Now try that with some random cheapo little computer.
You do realise that the old models and zeros are still available and relatively low cost still. Also the relative performance to a ‘real’ computer has massively shifted – the Pi 1 could run a very lightweight desktop, badly to the point its barely useable, the Pi 3 leaps up to being pretty close to fully useable for most folks and comparable enough to a low end more conventional computer – the netbook and the like. But now the 4 and 5 have leapt up far enough that they could really be your only computer for anybody that isn’t addicted to videogames or doing a seriously compute heavy hobby/job, though with Mr Geerling and friends getting AMD GPU working on the 5…
The availability of Pi’s is generally not bad, it went to shit over the pandemic same as everything (though they probably did better than most tech) and they have usually underestimated the initial demand a bit for a rough month or two around launch. But on the whole you can buy ’em more readily than most electronics project boards.
Is this an official update? When it’s releasing?
I’ll give this a go on my RPi 400, as it has become very trying to use as a desktop anymore. Even a single Web browser (and tab) seems to be enough to make it stumble all over itself. It wasn’t like this, just a year ago, and I’d hate to relegate it to other, e.g. headless, duties so soon.