AdaptecRAID 71605 Storage Controller 2274400-R

The short:Looking at the spec sheet, this controller has high potential. 16 6Gb/s SAS channels on PCIe3*8, 1GB cache (though the option for more would be nice), power saving features, optional flash memory plus supercapacitor based cache backup, support for numerous popular operating systems, support for easily available drives and backplanes.More recent firmware/driver that still works for me: BIOS/Firmware/Boot 7.5-0 (32084); Linux driver 1.2-1 (52033). Other combinations besides this and the ones I mention elsewhere in the review have not done so well for me on my 3.4.6 kernel based Xeon server. This combination of versions also works well for the 78165 and the 72405 models for me.My initial review rated it 1 star because there were many problems, and Adaptec's support appeared to have brushed me off at that point and I had spent a while just trying to get it to boot reliably. Then, someone at Adaptec has pointed me to the (then) single 'obsolete' firmware version that worked, and I have been able to hack the drivers into working under Linux, but there were many problems. More recently, a round of updates to firmware and drivers and utilities has resolved a lot of my issues.If you get one of these, I strongly advise ignoring the disc that ships with it and downloading firmware 30612, along with the appropriate drivers (Linux = 30300, others = 30617) and utilities 20618. These seem to be more stable, and the performance still seems to be good. One can hope that any versions that follow these will also work well.The long:I want to make clear that I am using the firmware, drivers, utilities releases I mentioned above, which have resolved many of the problems mentioned here (exceptions are noted).I am using an SGPIO backplane, which the firmware claims to support. The oldest firmware releases do not seem to want to send disc activity indication or fault indication to my backplane. The newer versions (at least 30040 and later) consistently can send fault indications, but still seem to be unreliable about sending activity indications. Since my backplane can use the drive-provided activity indication, I am using that instead of the one provided by the array controller, so I do not know whether the newest version handles activity indications correctly via SGPIO, but it does seem to reliably provide fault information via SGPIO.Only two versions of the firmware work reliably with 'power saving' mode, or with 'delayed spin' on power-up: 30040 and 30612. All of the other versions will lose drives if they do not come full ready within about 5 seconds, even from 'spindle stopped and heads unloaded' state, and I have met no mechanical drives that can accomplish this feat reliably -- most take 15 seconds or longer. If a drive fails to come ready in time, the controller will mark it as failed; if a hotspare is present, the controller will attempt to rebuild to the hotspare, but the hotspare will be failed as well because it will not spin up in time, as well as the other drives in the array, thus causing the array to completely fail (maybe you can recover your data) and any participating hotspare(s) to be marked as failed. This also affects 'delayed spin' during power on -- if drives are not ready, they will be marked failed. In this case, I often saw around eight drives come up okay and then the remainder failed.A bus scan takes a long time. I guess as a long-time SCSI user, I should be accustomed to absurdly long bus scan times, but I had hoped that since SAS is point-to-point (at least with no expanders) the scan times would be shorter on a fully populated host (I have 16 drives, one on each of the 16 SAS links, and no expanders, but a device scan still takes twenty or thirty seconds). The exact length of time has varied between firmware versions, but never less than 15 seconds, even with all drives ready and awaiting commands. I dread what it would be if I had expanders and maxed out the number of drives.The chassis and cables and drives I am using are listed in the so-called 'compatibility' list -- 16 Hitachi Ultrastar drives of the same model (directly connected without expanders), Adaptec cables, and Chenbro backplane modules in a Chenbro chassis, powered by a power converter that provides at least 100% headroom over the nominal peak load of the system on all rails but +5V (where it can provide only 50% over the nominal peak load) and carefully planned wiring that involves no power splitters or extensions and provides two separate power connections to each backplane module. Despite all this, only two firmware versions (30040 and 30612) have proven to work consistently.The documentation lists features that are simply not there. Such features include, for example, a way to set whether activity indications will be provided by the controller to the backplane.I gave up on the graphical configuration software, so I can not comment on how well it works in the recent versions. I'm moving my comments about it to a comment since I am no longer considering it in my review.The command line software for configuration works fairly well, but the documentation is inconsistent and could use a little more information about what particular features mean and how the various settings affect things -- some things are thoroughly documented, while others are just mentioned, as if in passing.The support people will ask for a diagnostic dump when getting information about your case. Stupidly, that dump includes all user names and passwords that have been used in setting up and in accessing the software that controls the device. I think the diagnostic dump is a good idea, but there is no reason why it should include (in cleartext as it does -- or even scrambled or encrypted) every user name and password ever tried against the tools. If you do intend to send them a support dump, edit it or change your passwords afterward. Hopefully the newer versions omit this; I have not had need to check lately.The Linux driver binaries they include work only on the specific distributions they list. Trying to install directly to an array on the 71605 seems to be a bad idea, since a lot of installers will load the module Adaptec provides but they do not later install it to the initrd for the installed system, rendering it unable to access the drives after installation has finished (this particular item is almost certainly not Adaptec's bug, but it is something of which you need to be aware). If you want to build your own kernel (for optimisation, hardening, or other reasons), you need to install DKMS support and then maybe the DKMS based source version will work with your kernel. Good luck; I went through several installs and numerous iterations with DKMS and various kernel source trees. I finally got it to work by installing to a SATA drive on the system board's connectors; once I got the driver build working, I moved my boot and other volumes to arrays on this card. All of this struggle was under Linux releases explicitly mentioned in Adaptec's compatibility list. While it is still tedious to deal with certain kernel changes, once the DKMS package is installed and the DKMS driver source is in place and building correctly, things settle down and work as long as you remember to refresh the DKMS based drivers whenever making changes to the kernel or the initrd.I will mention the Windows driver binaries do seem to work, and at least there Adaptec seems to have tested things. Since the server on which I have this card runs Linux, I have not had an opportunity to test the more recent Windows drivers. Previous ones I tried while initially investigating this card seemed good, but even in Windows you probably want to avoid the older firmware variants.The 'bootable' DVD included with the card, which supposedly provides diagnostic and configuration tools, as well as some other features, is not bootable by any system I have that handles DVDs. I tried on machines with firmware and hardware vintage between 1999 and 2013, with SCSI and PATA and SATA connected DVD burners and BD burners (I have no non-burner DVD or BD drives); while all of them could read the disc and access files from it, not one of them saw the disc as bootable. I was able, in contrast, to boot a Windows7 install disc, several Linux install discs, and a MemTest86 disc in all of them; though not all of them would run the things I tried, they at least saw them as bootable and tried, but they all said the disc included with this controller was not bootable.I have given up on UEFI, and gone back to the more stable 'legacy' firmware interface, so I moved my comments about that to a comment instead of the main review.The firmware setup tool is lacking what I consider to be important features. The most glaring one is the lack of a way to set a hotspare -- you can build arrays, manage arrays, destroy arrays, but not add spares. It also has a very strange interface for setting the order of volumes -- instead of moving them around, or even the funky selection sort most system BIOSes offer for drive boot order, you can only swap a volume with the first one in the boot order, and this only works if both the first one in the boot order and the one being swapped are in 'optimal' state (so you can't do this until they have finished building). On the positive side, you can do what you need for initial setup with the firmware setup tool, and make tweaks later with the setup tools that run under the OS.Conclusion:I had promised to update my review if Adaptec provided drivers and firmware that worked reliably for my configuration. It looks like they have done so, and I like the results so far. If it all works well for a while (until I can afford more drives), I will probably buy another of these (or maybe the 24-link version) to run the next stack of drives I attach to this server.

Overheating is a major concern with this card. Make sure you have enough airflow in your case to properly cool it (when it gets too hot an alarm/buzzer will go off).

No cables but correct

Bought this to use in my home media server. My on-board raid could not keep up with the amount of HD files I had when trying to stream even 2 at the same time. And forget about adding movies while a stream is running. Now I can have multiple HD streams running, and transfer video files and still have a +115 transfer rate.

States ESXi 5.5 and 6 support but have not been able to create a good build with the .tgz files as the CLI Tool wants a .VIB driver. Controller also has heat issues. Had to install a fan to keep it cool.

Works great in my home server

Worked fine in my old Intel system that I sold off the parts from but does not work right in my new AMD system on the X570 chipset. (I also tried it on a server I setup using the x470 with Ryzen 3600 and it has the exact same issues.) I updated the motherboards and RAID controller to the most recent BIOS and drivers but also tried older ones. Unable to access the card's setup on boot (using Ctrl-A causes system lockup) and can only access it through the MaxView software. The card cannot handle RAID 5 without randomly locking up and losing the array by corrupting the allocation tables. Any attempts to read/write from more than one drive on the controller causes intermittent system freezes of 3 to 5 seconds. Unable to access the card via MaxView while doing any kind of file copies between drives on the controller and between the motherboard SATA drives and controller drives. The only other card in either my system or the server is an nVidia graphics card. I've since tried the LSI cards and they work great. Avoid Adaptec/Microsemi cards if you use AMD systems.