• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

First play with SAS

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

mackerel

Member
Joined
Mar 7, 2008
To recap I salvaged a couple of 146GB drives from an old server that was being thrown out. Nothing else of potential use was left although with hindsight I could have taken the heatsinks. Anyway, I got the drives home and went to wipe them and... found they're SAS. I didn't have anything to connect them to. After seeing the price of new controllers, or mobos with built in support, I put them to one side and forgot about them.

Later on, while doing the SSD test I did stumble across various affordable SAS controllers. These were from the likes of HP and Dell. They're relatively old server kit, but at least they're PCIe 8x. I ruled out the lower cost Dells as they didn't run driver support beyond Server 2003. The HPs did go to Server 2008 R2, which is equivalent to Windows 7. I found a bundle with cables and got it.

It was soon apparent I made a mistake, as the cables came with the same connector on both ends. I didn't check this detail in the listing, only that it had cables at all. After another trip to ebay I had the correct cable which splits it to 4 drives, delivering power via molex.

Ok, I plugged the card into the my various systems, and here is where the pain started. Most systems didn't recognise the existence of the card at all. One Haswell/Win7 system did, but I couldn't install the 2008R2 drivers on it, either with setup utility, or after extracting it and manually selecting the files. I'm sure I had seen it working somewhere before, but I couldn't remember where. It was a Win10 system... and I only built 3 of those so far. It definitely wasn't my main system or VR rig, leaving another old Haswell system. I put the card in and... nothing.

Maybe it was the test install of Win10 I made on one of the SSDs. I tried that in two more Win7 Skylake machines, no luck. Then I remembered, it was an old Dell SFF desktop that recognised it. I put it in and sure enough, Win10 was on it and installed a driver for me.

Now to put the cable in and power the drives. As it was a compact Dell desktop, there were no spare power connectors as there is nothing else you could add in that needed power in there. After debating with using a pico-PSU, instead I just unhooked the SSD on another desktop and took the power cables off that. I could hear a faint whine and feel a vibration. It's alive!

The drives weren't visible in Windows though. A quick google revealed a HP configuration utility, which despite its awful dialog boxes installed fine and allowed me to set up an array. I wanted to test the drives separately and the controller didn't offer any simple connection mode, so each drive ended up being a raid 0 by itself.

One drive popped up but not the other. A trip to disk management showed the 2nd wasn't online as its signature conflicted with the 1st. I forced it and then it appeared too. Looking at the little contents on there, it appears they were used in raid 1 hence the conflict. This is a little surprising as one was a 10k drive, the other 15k.

I wiped the drives and created fresh partitions on both, and run CrystalDiskMark with the following results.

Code:
   Sequential Read (Q= 32,T= 1) :   141.263 MB/s
  Sequential Write (Q= 32,T= 1) :   114.858 MB/s
  Random Read 4KiB (Q= 32,T= 1) :     3.450 MB/s [   842.3 IOPS]
 Random Write 4KiB (Q= 32,T= 1) :     2.623 MB/s [   640.4 IOPS]
         Sequential Read (T= 1) :   141.556 MB/s
        Sequential Write (T= 1) :    77.603 MB/s
   Random Read 4KiB (Q= 1,T= 1) :     1.001 MB/s [   244.4 IOPS]
  Random Write 4KiB (Q= 1,T= 1) :     0.906 MB/s [   221.2 IOPS]

   Sequential Read (Q= 32,T= 1) :   163.773 MB/s
  Sequential Write (Q= 32,T= 1) :   112.441 MB/s
  Random Read 4KiB (Q= 32,T= 1) :     3.303 MB/s [   806.4 IOPS]
 Random Write 4KiB (Q= 32,T= 1) :     2.221 MB/s [   542.2 IOPS]
         Sequential Read (T= 1) :   161.711 MB/s
        Sequential Write (T= 1) :    99.414 MB/s
   Random Read 4KiB (Q= 1,T= 1) :     1.329 MB/s [   324.5 IOPS]
  Random Write 4KiB (Q= 1,T= 1) :     1.068 MB/s [   260.7 IOPS]

The first set of results are for the 10k drive, and the 2nd set for 15k. The latter has a slight boost in sequential. I thought it would have a clear advantage over the 10k model for random, which isn't the case.

The controller card has 256MB cache which was enabled. As I understand it, without the optional battery pack it wont cache writes, so it should only be caching reads. Still, the random write performance is above any consumer hard disk I've seen.

I'd also comment, the drives got really hot. The 10k drive was hot to the touch. Uncomfortable but not painful. The 15k drive was painful and I couldn't keep contact for more than a few seconds. It can't be good for the drive either so these really need forced air cooling.

With that, my initial play is over. I'm not sure what, if anything, I'll do with them longer term. Maybe stick the lot on ebay for someone else to play with. I've done the hard work and know everything is there and works at least.
 

mokrunka

Member
Joined
Dec 21, 2009
Cool write up dude. I messed around with some 15k 73 Gb SAS drives a while back. Your post reminded me of all the difficulties I had along the way. I basically did all of the same things you did. Those cables tricked me as well! The server I took apart though had an LSI MegaRAID SAS controller though, which worked fantastically well, so I didn't have the configuration problems that you did.

thanks for the nostalgia!