Is there enough ribbon to raise it up another 8.5mm with a gpio header extender and some m2.5*8+6 standoffs? I think the extra space would allow for a better cooling system for both the board and the hat. I really have to watch the cooling because the ambient room temperature is rarely below 80f here and I am wanting an always on solution. I want to send files at high speed to ssd’s and let the system passively distribute them to a backup and a main storage in the background afterwards.
Penta SATA HAT is now available
I would think that the case design should making the fans blown from the side instead of the top in the original Pi4 quad SATA nas… it would work better in a dual fan setup… would help with thermals for all 3 main components of the setup… will draw a diagram later tonight, forgot to bring my pen out today…
Good luck with such a setup while being tortured with storage traffic. 
I hope you’re using a ‘checksummed’ filesystem like btrfs or ZFS since this will show data corruption due to overheating way easier/earlier than relying on concepts of the past where most of this stuff gets undetected anyway (mdraid10/ext4).
Not that much luck, just well designed custom software.
The extra 8.5mm would allow for the inclusion of heat shielding between the main board and the hat.
A redesign of the mainboard cooling would make it much more efficient and a well placed side blower on a customized case could then blow air up and out working in conjunction with a top fan to alleviate that heat build up on the hat that it shows in the thermal image.
There are some very efficient blowers designed for 3D printers with a 10mm clearance that could easily be adapted to 8.5mm with a simple 90 degree curved 3D printed plastic piece. That way it could be side mounted on the exterior of a case.
As for the software I made; it runs on NTFS but it uses checksums from the original files to verify all the transfers.
It also maintains directory structure and uses Google Translate on any unflagged file names that do not match the System Default Language. Software folders are automatically flagged, media files and folders as well as various other documents that are not in software folders are not flagged.
Also, file names without extensions get the headers scraped in an attempt to identify them.
Failed transfers result in a drive check and automated repair if needed and then the transfer is reattempted.
So far, it works, but I am running it directly on my main system and not passively on a secondary system such as a rpi5. I believe that transferring and verifying files onto a setup such as this running SSD’s and letting the rpi5 handle the rest would significantly reduce the time my main system is tied up with file transfers.
ZFS seems to work well, though at RAIDZ1 (or any other level that needs parity data), the Pi 5 seems to struggle to maintain 1 Gbps write even for smaller file sizes (which is a little disappointing, tbh—I thought it could at least hit 115 MB/sec or so writes, and maintain that speed if the drives can keep up).
I recently tested IPFire on the RPi 5 (yeah, not supported but if you’re able to combine bootloader+kernel with whatever userland – former Armbian/OMV dev here – then it works ofc). When testing NAT throughput and allowing the A76s to clock lower than ~2000 MHz cpu0 was always the bottleneck since all the IRQs ended up there and througput started to suffer.
Have you had a look at this (/proc/interrupts and watching atop/htop output)? And maybe know a way to use IRQ affinity with RPi kernels to get rid of this stupid cpu0 bottleneck by moving the IRQs in question to cpu1 – cpu3?
Or more generally speaking… did you have a look at what the bottleneck is in these situations? 
BTW: As mentioned in some Github issue where we met: RPi folks don’t use io_is_busy=1 while relying on ondemand cpufreq governor and in certain situations (though usually not benchmarks) this will result in the CPU cores not ramping up clockspeeds when needed.
I didn’t observe atop in my testing last week, but I’ll fire it up soon. It’s annoying/odd that it only seems to affect writes over the network (samba more strongly than NFS)—any ideas why wouldn’t it also affect read performance? (Where I can saturate a 2.5 Gbps connection easily, reading 100 GB from a 4-drive array.)
In general I would better look than speculate and that’s what atop and /proc/interrupts are for
IRQ affinity with RPi kernels is broken and as such cpu0 might be the bottleneck. That combined with ondemand w/o io_is_busy=1 and cpu0 while serving all interrupts might not even clock at 2400 MHz. And writes may generate twice as much IRQs compared to reads. But that’s speculation and it’s really easy to check for this with the aforementioned methods.
Also switching to performance governor might result in better throughput numbers since this masquerades parts of the problem (clockspeeds not ramping up quickly enough).
It doesn’t seem like that’s the problem (testing with atop, monitoring cpu freq with vcgencmd, and also setting force_turbo—see my ongoing testing at https://github.com/geerlingguy/raspberry-pi-pcie-devices/issues/615).
I have an extra Top Board if you want it. I watch your videos all the time and would be happy to send it along as a means of support.
The official case for the ROCK 4 works really well because the CPUs on those boards are on the undersides and the bottom of the case is essentially a big aluminum heatsink. Mine didn’t quite touch, so I bought some copper shims of various sizes to make up the difference. As a result, I have never seen the temperature of my CPU go above the high 30s.
Matt
It seems the resellers are all out of stock of the Penta SATA HAT for Raspberry Pi 5.
Is there any ETA on stock?
Are there resellers in Australia?
1 Like
@setq
Is there a way to get only the FPC PCI cable for the Pi?
I just got mine hat in, and Arace sent me the ROCK version, assuming these boards are the same…
Thanks I have ordered.
1 Like
Hi @seang,
You can get cable from here https://arace.tech/products/m-2-e-key-to-ipex-board
It is currently out of stock and should be in stock soon.
Hi @setq would like to check on whether it is possible to set HDD sleep via firmware like the previous quad SATA hat model, or the only way to set sleep is via HDPARM command? Asking this because i’m still using spinning disk for long term backup…
Hi! Looks like the hat is out of stock on Arace.tech (much cheaper than Allnet) and Amazon, does anyone have an idea on future availability (@hipboi @setq)? Thank you!
Edit: Why not also ask for any expected date on the OLED/fan extension and case
1 Like
I can’t get my SSD’s to show up under ubuntu server 24.04, I added the parameter dtparam=pciex1, rebooted… no luck. Anyone has any idea what might help me?
Wrong topic, watch the headline please.
Seems like it is out of stock everywhere. Any ideas, when it will come back?