Seaberry carrier board turns a Raspberry Pi into a desktop PC with 11 PCIe slots Fancy a Raspberry Pi 4 in a desktop ITX form factor with 11 PCIe slots? The new Seaberry carrier board may make your wish come true – but for a fairly hefty price. As we mentioned earlier this month, the forthcoming Linux kernel 5.16 will sport better support for the Raspberry Pi 4 Compute Module. This tiny daughterboard plugs …
My basement dwelling server cost the same as a RPi 4 - came with 8 sas drive slots, 24gb of ram, 4x 1 gigabit ethernet, 2 x 8 gigabit fibre cards, 2 x 4 core CPUs, 2 psu's... Admittedly it's cost me about £250+ for sas hard disks and a second 8 drive caddy (plus second raid array to manage it).
Still cheaper than one of these efforts...
Whole life cost over several years normally reduces the upfront cost element to incidental. I'd hazard a guess that the annual leccy cost for that server isn't far short of the assembled cost given that the 2x 4-core CPUs will be pulling north of 100w & (16?) sas drives maybe 10w each.
I'm not knocking the setup, A few hours to setup and its probably going to out last the decade if you've a spare hdd or two.
It normally idles around 88watts which is surprisingly small when it's not being hammered. Put a heavy load on it and it'll hit about 3-400 Watts. It does have 2x 750 watt platinum rated Psu's it can go quiet high but since I'm not running a graphics/AI/mining card it sips it's juice normally.
Agreed. Many uses for the Pi (not all but many) could be solved equally well or better with a £10-20 thin client from eBay. Used of course but those things tend not to break. With the "optional" extras (case, power supply, storage) supplied as part of the deal rather than on top.
The Pi has its place but it isn't a universal solution. It's too expensive, big or power hungry in some applications, it's underpowered in others. I get tired of all those comments here and elsewhere for any other SBC that amount to little more than "it's not a Pi" as if that is a bad thing.
I am struggling to see the use case for this board, the price point rules it out for a lot of stuff. It's up against desktop boards (both ARM and PC based) for far less and much better I/O throughput in particular. There doesn't even appear to be GPIO facilities for control systems. I can't really see where it fits unless Pi compatibility is a specific need.
Probably mostly useful for test. If I want to develop Pi compute + some random cards, I might want to test on a bench with existing cards before I build an integrated solution.
Or as an industrial control/capture. If I need 10 pciE cards to capture N inputs but don't need much processing, this might easier/cheaper/more reliable than a rack of 3-4 PCs with a few cards in each.
Thinking back to VME crates with a single CPU card and a dozen data I/O cards
well if I wanted a NAS solution I could use a USB 3 external drive, right?
FreeNAS is based on FreeBSD which (last I checked) supports RPi pretty well. (I actually contributed some code for that and tested on a 3B+for 12, and I think 13 has support for RPi 4).
So a simpler and cheaper NAS: just use a USB drive, run the OS from the SD card and use ZFS on the USB drive with FreeNAS or FreeBSD itself. No problem!! Low cost, too. And if you add a couple of USB 3 "thumb drives" with large storage capacity, even better.
Still having a bunch of PCIe interfaces available for an RPi is an interesting toy. Beyond that [for a real solution] not so much. The USB 3 ports on the RPi 4 should provide all of that and more.
The Pi has its place but it isn't a universal solution. It's too expensive, big or power hungry in some applications, it's underpowered in others. I get tired of all those comments here and elsewhere for any other SBC that amount to little more than "it's not a Pi" as if that is a bad thing.
Does a "universal solution" exist in any form? Nobody ever claimed a Pi was a universal solution.
The difference between Pi and other upstart SBCs that attempt to out-spec it, is the Pi creators stand behind it and offer real backing. Instead of just chucking it out to market with half baked OS and drivers and just hoping an extra feature will sell it. So many have tried and fizzled out.
40 million shipped and counting.
Only Jetson Nano has really offered any kind of competition.
Anyone who is thinking about running Pi CM4 in a mini-itx form might want to consider waiting for Turing Pi 2.
The engineers are not making a flying rocket car for Joe Public, it's a tricked out Raspberry PI. I've long figured out that a PC running Proxmox is a better option than loads of RPI modules. However this is a whole 'scene' in itself now. All over the world people are doing interesting things with the PI. There are loads of computing tasks which only require a little computer and the PI fits well. The ability to extend the PI massively with this new board is very popular. I'm going to enjoy seeing what Jeff Gearling and others do with this.
This is more or less exactly what I want to build a home NAS. I want something that makes a Pi fit in a standard case (tick), that gives it plenty of onboard storage (tick - although I’d settle for Sata ports, I don’t need PCIe), and that doesn’t chew through the power like a junky on a binge (which is why that old PC or Mac just won’t cut the mustard)
That being said, I don’t want it at that price. At £120 or less, including the compute board, I’d buy. More than that? No thank you.
For that I would want something that allows easy swapping of disks. Preferably hot-swapping even, but a quick shutdown/slide out/slide in/start up is still way better than opening its case or even extracting this board from it, removing one of the upper mPCIe sticks to get at the lower one (of course it's one of those that's gone wonky as Murphy dictates), putting it all back together with a fresh SSD fitted and booting up again.
The Wirerustee and Radxa boards mentioned look a lot better for such an application, even though they only offer SATA and not PCIe, but you're bandwidth-constrained anyway so that's a moot point. Mounting one of those in a small case using a four- or six-slot drive cradle that mounts as a single 5.25" bay is not exactly rocket surgery.
The biggest problem with home built NAS are that they mainly just support Raid 5, i.e. ALL drives HAVE to be the same size or the raid uses the smallest drive size as the std.
For a home user who has a lot of various size drives it just does not make sense. That is why i went with a synology. If you are happy purchasing your drives by the dozen then go ahead.
Never mind that as far as I'm concerned Synology can go fsck themselves with a splintery broomstick, for the price difference between a DIY NAS and a commercial one you can easily buy one or two disks replacing the smallest drives you would want to use with ones whose size is more in line with the others.
Also, the boards I mentioned take four and five drives respectively, not a dozen.
"The biggest problem with home built NAS are that they mainly just support Raid 5, i.e. ALL drives HAVE to be the same size or the raid uses the smallest drive size as the std."
That's the RAID standard and is true for all RAID setups. Smallest drive sets the size for all. You can replace each drive in turn and upscale the storage over time tho. My home lab QNAP has RAID 6 and 9 x 1TB SSD where it started with 512GB drives.
Or just use btrfs and use pretty much whatever disks you have to hand. Don't try to use RAID5 or RAID6, though - it is buggy on btrfs and you will lose data.
Being able to use different RAID levels for data and metadata can be very useful (as can using RAID1C3 or higher for metadata).
Personally, I use a variety of disks, stick LUKS and LVM over them, and run btrfs over various combinations of LVs for different storage requirements (live data, long term backups, snapshots, etc).
That's well within budget for the serious DIY crew - unless they're pricing in USD.
It seems many PI users are looking for an easy way to get at the PCIe, otherwise it wouldn't have sold out so fast. Throw it in an old case slap in a few bits from the spares box and you've a near ideal 'get it working with a PI' rig.
For industrial controllers this could be just what everyone was waiting for.
Yes, A PCIE slot and that's perfect for small systems, this allows a very complex control system to be boxed on one board.
Industrial plant monitoring systems can easily have hundreds/thoudands of low* bandwidth inputs and these days via a mix of protocols, this could easily replace half a dozen single slot carrier boards each running a few dozen lines (max) of a specific type. Reducing the number of controllers by a large % is always something to think about. Adding a card to an existing controller board is almost always easier that adding a whole new controller.
* from traditional serial into the low single Mb/s
Plus, you have 11 PCIe slots, but it's going to be "just a tad" bandwidth restricted -- the Pi CM4 PCIe slot is a 1x gen 2 slot, 400MB/sec. That's a respectable amount of bandwidth, but not really if you're expecting to actually give all 11 cards a workout -- that's 35MB/sec per slot with no overhead (would not be surprised to find it's closer to 30MB/sec, PCIe is usually point-to-point so divvying up a PCIe lane 11 ways probably does incur some overhead.)
I could see doing an ARM desktop though. I ran a Acer Chromebook with a ARM (Tegra K1), Chrubuntu ran sweet off it. The Chrubuntu is regular Ubuntu with some tweaks for the chromebook (power management stuff to handle the big/little CPU setup -- it had a low power core, which it used very successfully (22 hour battery life), more or less if it was going to run 1 core under about 800-900mhz, it'd kick over to the low-power core. OpenGL, OpenGLES, and CUDA all worked which was sweet. All the typical Ubuntu packages are available native; I had qemu-user-x86 and qemu-user-x86-64 installed, could set the package system to allow x86/x86-64 packages to be installed, it'd successfully run x86 and x86-64 binaries (.. as long as they weren't multithreaded.) I ran a Samsung color printer "binary blob" x86 printer driver under it, it ran at about 1/4 native speed but still could spit out pages and send them to the printer over wifi faster than it could print them.
So the software is there even for an ARM desktop. This one's too pricey for me though 8-).
It depends what you use if for. I've got a machine that I use for AI research* and most of the stuff just gets dumped on the GPU and then the PCIe is effectively ignored for three or four hours. Setup and Results/teardown would be faster on a faster machine but overall it would cost me a fortune for a small % of total run time.
* OK pissing about but then isn't gaming?
I've played with my Pi3B+ and found it good for prototyping use cases for 'better' (more capable kit)
Just bought a Fujitsu TX120 s3 for £40 on fleabay. 3pcie slots, Pci, 4 hot-swap drive bays (came with raid card in one PCIe slot) upto 32GB ram possible, iLO equivalent
Downer is the Celeron cpu (I have a 'spare' E3-1200 series cpu). Works fine as TVheadend+MPD+Snapcast server for 5 endpoints ("TVs" or monitor+thin client)
Point is that much of testing/config was with Pi but in no way could it do the 'full' job of simultaneously collecting broadcasts+storage+serving output to endpoints. Then again, I would not have invested in the TX120 if I couldnt make setup work
One of my Pi3's is my day-to-day Work computer, has been for a number of years.
Does everything I need to do work-wise (read pdfs, check webmail/ wikipedia/ el Reg, libre office for when I do something) with no complaints from either of us- Pi or me.
Leaves plenty of desk space to play with things.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
