On Thu, Jan 29, 2015 at 8:47 PM, David Christensen < dpchr...@holgerdanske.com> wrote:
> On 01/29/2015 06:23 PM, Kevin O'Gorman wrote: > >> There are lots of choices where the info in directed at Windows users, but >> precious info available if you want to be sure the card will work on >> Linux. >> >>> From Amazon, i tried the HighPoint 640L because I know someone who had a >>> >> good experience. Mine was not so much. I've tried two of the cards but >> got connected to my drives only briefly once. I have no idea why. Now >> I'm >> looking for a solution that will let me connect to my 4TB SATA-3 drives >> internally, without paying more for it than I paid for the mainboard. >> Details: >> - running Xubuntui 14.04 >> - I have PCIe x1 and x16 slots available >> - I even have an old PCI slot available, but it may be too slow. >> - using mdadm, I have no interest in RAID capabilities, just JBOD. >> - The drives are internal, it would be awkward to set them up otherwise. >> - I need to connect 2 drives, may want 2 more later, but it can wait. >> - Drives are 4TB SATA 3 with GPT partitions >> - I can set the mainboard to use AHCI or IDE compatibility >> - there is not trace of Windows, so it needs to be configurable with just >> Linux. >> > > The key is getting a card that has chip(s) that are well supported by > Linux. Deducing what chips are on a given card, and whether or not they > are supported by Linux, for any given piece of hardware is something that > I've yet to have much success with. > > > So, I asked a similar question in July 2013: > > https://lists.debian.org/debian-user/2013/07/msg00310.html > > > Basically, there were two choices: > > 1. 2 @ SATA2 host bus adapter ($20). > > 2. 4 @ SAS/SATA3 RAID controller ($200). > > > I ended up going with the first option -- Syba SD-SA2PEX-2IR: > > http://www.sybausa.com/productInfo.php?iid=536 > > I'm not using it now, but don't recall any problems with Debian GNU/Linux, > Windows XP, and the various free DOS versions used with Norton Ghost and > hard drive manufacturer bootable diagnostic discs. > > > I prefer PCIe x1 over PCI because PCIe should have faster access to the > CPU/ chipset/ memory and because newer ATX/ microATX motherboards usually > have at least one PCIe x1 slot; some don't even have PCI. > > > As for SATA 1/2/3 speed differences on my SOHO network with LUKS encrypted > drives, the bottleneck is usually the CPU (non-AES-NI) or the drive > itself. At one point, I had a ZFS on Linux mirror on two 3 TB Seagate > ST3000DM001's with LUKS on an Intel DQ67SW motherboard and Core i7-2600S > processor, with one drive on a motherboard SATA3 port and the other on a > motherboard SATA2 port. I recall that the drive on the SATA2 port had > higher busy time than the SATA3 port, but the bottleneck was seeks. > > Thanks. I looked up Syba and found that board is now legacy, and I'm using only SATA III drives these days, so I found the current similar boards. I chose SI-PEX40062 which was the best fit for my drive needs and available PCIe slots (4 internal drives, 2x PCIe). Amazon had a board with the same model number and specs, but its brand name is IO Crest. I ordered it and it arrived today. It's working 50%. Ugh. Two of the four ports work fine. It happens that's all I _need_ to keep working, and I've started an RMA and order for a replacement, so that's okay -- it's time that's most important to me right now. At least there was no fooling around figuring it out. I can hot-swap the cables so long as no partition is mounted, so it was quick to determine that two ports work and two don't. My wife has Amazon Prime anyway, so the shipping is free and the replacement is coming even before the return has been sent (I'm using it in the meantime). I even get deliveries on Sunday because the USPS is trying harder. BTW, I'm interested in bus speeds this much because the drives are going to be doing huge sequential file accesses, and seeks are not an issue. I'm a lot happier than I was a week ago. Thanks again. -- Kevin O'Gorman #define QUESTION ((bb) || (!bb)) /* Shakespeare */ Please consider the environment before printing this email.
