I thought I would post my success story with this older SAS PCI-E P812 smart array controller card, which I picked up on Ebay for $12.50, including shipping!
It came configured with 1GB onboard cache memory and the cache backup capacitors, so I don't need to worry about replacing aging batteries. Due to this, I have turned off write caching in Windows for arrays and drives connected to this controller. In the event of a power failure or a system hang/reboot, the unwritten cache is stored in NVRAM using the power from the capacitors, and when the system is restarted, the cached writes are retrieved from NVRAM and then posted to the connected hard drives as the card boots up. Pretty slick stuff to have in a workstation....
I have it installed in an HP Z640 workstation (basically the same MB as the Z440, but with a daughterboard header for a second processor.daughterboard). I am running Windows 10 Pro x-64 build 20H2 (2020 fall feature update) with all updates applied.
I am using it for a pair of internal 4TB SAS drives configured in Raid 1, and also have a cable hooked up to one of the external ports that has 5 sata connectors that I use for hot-swapping various drives around for various machines I build for family and others.
Had to disable card's option rom in BIOS due to bootloop:
I had to disable the option rom in the workstation's BIOS because the card would cause a bootloop when it tried to run the card's option rom program. Strangely enough, the same card in an ASUS (non HP) motherboard didn't cause a boot-loop, and I could access the option rom's configuration utility in that machine.
So, since I can't use the option rom's configuration on this workstation, I use HP's windows based "Array Configuration Utility" to configure the card. It is available for download on HPs website.
Important warning regarding firmware versions:
The firmware on the card is 5.14.
Important warning -- I have read elsewhere and in this thread that the later versions of the firmware for the card disable the back ports for SATA drives, and also have read elsewhere that later firmware may not run correctly if not installed in an HP server, so I have NOT updated the firmware. V5.14 seems to be running fine.
I am not booting from a drive connected to the P812 smart array card, so I can't confirm that it will boot from it in a Z640 workstation. Instead I boot from an Nvme SSD connected to a M.2. Nvme PCI-e adapter card. BTW, this was an AMAZING upgrade from a SATA based SSD hooked up to one of the motherboard's SATA connectors.
The card runs hot:
The one somewhat negative comment I would make about this card is that it runs fairly hot. The large heat sink on the card gets too hot to comfortably touch, so I positioned a spare fan internally to blow air across the heatsink to help keep the temps down. I could not find any power specs for the card, which I found a bit odd, but I suspect it draws quite a bit of power from the PCI-E bus.
The card is large:
The card is quite long (31.1cm (12.3 in)) -- but it fit into an X8 slot in this workstation with one minor modification -- these workstations have a spacer attached to the opening side of the case which I suspect is intended to help hold longer half-height cards firmly in their slots and that interfered with closing up the side of the case. It is easily removed, and that solved the interference issue.
However, it would NOT fit into a Mini-Tower format PC I had previously because the internal hard drive bay rack got in the way, so please pay attention to the card's length and your PC's particular case configuration when considering using the card.
Update:
The card is still running well both with internal SAS drives and external SATA drives. However, I have not been able to figure out a way to run HP's smartarray drive diagnostics in Windows 10. The HP utilities seem to require that they be run on one of HPs proliant servers in order for them to installed. The necessary management drivers won't install natively on Windows 10 Pro. If someone has a workaround for this, I would appreciating getting contacted with the details.....
2nd Update (on 2022/04/28):
The card is still running fine. The only issue I have with it is that once-in-a-great-while the controller doesn't complete initialization before windows boots, and I have to do a reboot to see the arrays. I have added a 5 second boot delay in the bios settings, but it still happens maybe once every other month or so -- so it is a minor annoyance.
I bought an open box external 5 bay SAS/SATA drive rack on Amazon Warehouse Deals for $84 (
ISTAR BPN-DE350HD-RED Trayless 3X 5.25 to 5X 3.5 12Gb/s HDD Hot-swap Rack : Electronics) and am running another raid-1 with a pair of 4TB recycled/recertified HGST datacenter SATA drives in it. The rack had some mixed reviews, but I took a chance and disassembled and did some minor internal cabling routing and insulation modifications and reassembled it to correct the internal cabling issues that had caused some of the failures that had been noted in the reviews. I bought a 4 port 4 wire molex power bracket (I mounted it in front of the unused PCI slot) to supply the 12v and 5v to the rack from the workstation power supply, and then got a couple of molex to sata power cables to run power from that bracket to the rack. It has been stable and reliable since.
Two different types of SAS SFF 8088 to SATA cables:
I did learn one thing that I hadn't realized was an issue. It turns out that there are two different kinds of "Mini SAS 26P SFF 8088 Male to 4 SATA" cables out there. One has wires crossed and I believe is used to bridge between two backplanes/racks of drives, but this type won't work between the controller and drives, and the other doesn't have crossed wires and will connect from the controller directly to the drives. The cables look exactly the same, and have the same connectors on both ends! I learned this the hard way. I got the wrong one on Ebay, and had figure out what was wrong and wait for the 2nd (and correct) cable to show up. The cable that worked was called a "Mini SAS 26P SFF 8088 Male to 4 SATA 4Pin Female 1M 3.3FT Cable with Latch, Mini SAS Host/Controller to 4 SATA Target/Backplane", and I found it on Amazon here:
https://smile.amazon.com/gp/product/B07M8F1MV7. Hopefully this will save another SAS controller novice from making this mistake.
I use the extra bays for cloning drives when I help family and friends do upgrades. This way I don't have to open my z460's case to hook up drives to do cloning -- a real time saver. The only downside is that I have to provision each drive I insert into the rack via the HP "Array Configuration Utility" in Windows in order to access it, but it's only a minor annoyance.
I am going to see if it is possible to upgrade the array in the rack from raid-1 to raid-5 by adding another 4TB SATA drive -- which would double my raid array from 4TB to 8TB. We'll see if the software is smart enough to do the in-place raid-1 to raid-5 array rebuilding. I'll try to update this post with my experience attempting this....
BTW, These recertified HGST 4TB drives are selling for around $37 on Amazon these days, are guaranteed to have zero bad sectors and they have a 5 year warranty from the refurbisher. (Note, some listings offer the 5 year warranty and some don't, but the price is the same.) Since these data center drives have about a 2 million hour MTBF rating, I think there's plenty of life left in them -- much more so than the current consumer drives out there, and at a small fraction of their cost.
I got the idea to do the conversion from raid-1 to raid-5 from having done this on my TerraMaster NAS. I was initially running two 12TB shucked Western Digital external drives in a raid-1 array, and doing periodic full drive backups to a 3rd 12TB unshucked external USB 3 drive, but then realized that I could convert the raid-1 to raid-5 by shucking and using the 3rd drive inside the NAS as the 3rd drive in a raid-5 array. The NAS software was smart enough to do the in-place upgrade (which took 2 days to complete). This conversion doubled the available size of my array to 24TB and I still have redundancy (the raid can be rebuilt without data loss should one of the drives fail). I may do an upgrade to raid-6 if I can find another 12tb drive on this year's black friday sales -- which would give me double redundancy (up to two drives could fail and the array could be rebuilt).
BTW, I have read that these large capacity external drives are actually helium filled datacenter drives that have the firmware modified to run at a slower speed (about 5200RPM), so they should last a good long while -- probably longer than the expensive "red" drives that are sold as being "made for NAS".
I hope this information is helpful to some considering using one of these controller cards...
Philip
Last update (3/24/2024):
I did successfully upgrade the raid-1 array to raid-5 by adding another drive, thus doubling my capacity.
The issue of this board sometimes not initializing properly on a boot seems to have gone away. Not sure exactly why, unless there was a windows 10 update that fixed something. I did not upgrade the firmware on the controller.
Since then I have added another of the 5 bay external drive enclosures I mentioned above, and currently have 2 raid-5 arrays working well. One is with 4 6tb sas enterprise class drives, for a usable 18tb of storage, the other is with 4 8tb sas enterprise class drives for a usable 24tb of storage. I have an extra of each drive size set up as a "hot spare" so that if one of the drives fails, the array will automatically rebuild using the spare.
One thing I had to correct when I expanded the arrays was that I had to modify the power going to the external drive enclosures. Previously I had supplied power to them from only one of the internal drive power cables. That worked okay for up to 5 drives, but when I added more, I started getting intermittent drive failure messages that randomly changed from drive to drive. When I removed the 6th drive, the errors went away, which lead me to believe that I had a power issue as all the drives had tested out fine individually.
So I split the power up between two of the internal drive power cables, and the issues went away. What must have been happening is that the 18 gauge conductors in the internal drive power cables were not capable of carrying the necessary amps for more than 5 drives, so by splitting them up, more power became available to the drives. Just to be clear -- what I mean by "internal drive power cable" is the ENTIRE drive power cable from where it originates at the power supply to the last drive connector on that string of drive power connectors. Most power supplies have two or more of these separate drive power cables.
This required my modifying the 4 molex power external bracket, which had a single molex power connector on the inside as the power source, to now have two molex power connectors for the input power -- one to the 1st two external molex connectors (we'll call those A and B), and the other to the other two external molex connectors (we'll call them C and D). I already had enough of the 24 inch long molex to sata power cables I used to connect from the molex power external bracket to the drive enclosures, I just made sure that those 24 inch power cables were now connected to the external molex connector A, and the other to external molex connector C on the bracket, and that solved the power issue.
I will be retiring the p812 controller soon, replacing it with a newer HP Smart Array P822 2gb cache controller. I am doing this because:
1. this newer controller will use less power,
2. It will still give me both internal and external ports,
3. It will double the cache memory size,
4. It still uses supercap for cache power failure recovery.
5. It has a much smaller footprint than the p812 controller, which should allow for better airflow inside the workstation case.
What is still unknown is whether the p822 can support 4kn advanced format drives. The P812 can't support the 4kn advanced format drives, but it does at least support the 512e advanced format drives. I'd like to get 4kn format support as I understand there is a throughput increase when the drive doesn't have to do conversions from the native formatting on the drive to the 512e reported format the controller understands.
