Raid controler determines speed?

Computer Hardware
#1

Maybe this is common knoledge and it is just new to me but I though I might share it anyway. Not to long ago I discovered that my siig raid card performed substantially faster than my promise 4 channel card (even though the promise is a much more expensive card). Thats using the same exact hard drives. The siig was much faster with 2 drives than the promise was with 2 or 3 drives. Well I was helping a friend setup raid last night. He has sata 1 raid and sata 2 raid on his motherboard. One uses a silicon image controler and the other uses an nvidia controler. I sugested that he try both and benchmark them. the sata 1 controler using the nvidia raid controler performed resonably faster using the same drives. I have seen lots of comparisons on the speed of diffrent drives for raid buy have yet to have ran across any comparisons of diffrent cards with the same dirves, and how their speed compares. It would seem that the controler can make a huge diffrence on the speed, reguardless of what drives are used.

I just though some of you might find this interesting.

#2

Aye, I knew this. I upgraded a mobo from Gigabyte … with another Gigabyte board + Faster CPU + Faster Ram + Slower Raid controller :a

Needless to say, I ended up just imaging the raid partitions & restoring to straight single IDE drives, as the performance of the two single drives was much better than striped (raid 0).

I found SATA raid with the Sil3114/Sil3112 controller was painfully slow & gave that up as well. Of course, Nforce2 chipset was renowned for issues with the Sil311X series.

Soooo … which one is faster then?

The Silicon Image controller or the Nvidia Chipset Controller?

#3

Beware of “benchmarks”. They are not consistant enough to draw conclusions unless the differences are pretty large. Even then, there’s lots of variables that have nothing to do with the controller. You really can’t compare different installations, nor new installs vs old installs. Benchmarks are just that, and should really only be used to compare speeds on the same drives, same install, same data, same fragmentation, etc., etc.

Likewise, many benchmarks don’t tell you much to start with. Some are little more than burst rates, others use only one size file, etc. Sustained transfer rate tests also tell you nothing about how fast a drive/array will “feel”.

One big difference in controllers is the method and type of cache they use, if any. So often you may be comparing a controller with properly implemented cache with one that has none. If the files in the test are small, the difference can be huge.

Another variable is CPU load. Controllers can vary a lot in this area and a loaded CPU will affect the drive speeds.

Also be aware that most, if not all, boards with 2 controllers have one on the PCI bus and the other on the chipset. Chipset based controllers should always be faster by quite a bit, and have much lower CPU loads.

#4

Aye … but when the CPU is 30% faster, ram is faster(333->400), the new HD’s have higher densities, same cache size, spindle rotation, higher speed interface & similar seek times, You should expect it to be comparable, rather than the slower :wink:

That said, after you lose everything on your Raid0 array because one HD has an issue, or windows has an issue, Raid0 will never again be an option, except for non-essential stuff, like OS, program files or temp files :wink:

#5

That said, after you lose everything on your Raid0 array because one HD has an issue, or windows has an issue,

Which is also true of any drive or configuration, just as much as RAID-0. There’s no excuse for losing any data, period. There’s certainly lots of HD’s around that don’t benefit from RAID-0, but the risk of data loss is no reason to avoid using it. I’ve run on RAID almost exclusively for years, and never lost a thing, nor will I ever lose anything unless I take a direct lightning hit and the house burns down and floods all at the same time. Even then, I have backups in a water proof fire vault, so I still might be alright. Anyway, don’t blame RAID if you lose data, blame yourself.

#6

raid performance in level 0 should just be icing on the cake. what’s really important should be performance in redundancy modes, especially 5. my cheap highpoint raid5 controller blows any consumer integrated raid5 solution out of the water; course at level 0 it suffers from serious pci bandwidth issues.

#7

Lol, I didn’t technically lose anything, because I did have backups of everything important.
It was just interesting the way that it worked perfectly up until the moment, when simply restarting, windowsXP managed to unrecoverably commit suicide and corrupt all the partitions. An interesting experience indeed.

#8

The nvidia (chipset based I assume) that controled the sata 1 ports was faster than the silicon image controler by about 15% overall if I recall right. He only told me the results over the phone but the silicon image gave a sandra drive index in the highe 70’s where the nvidia controler gave a drive intex in the low 90’s. The diffrence I saw on my raid cards was much more substantial (both are pci raid cards). The promise card gave a drive index of about 60 with 2 drives, around 75 with three drives, where the siig card gave a drive index of about 93 with 2 drives. while all the diffrent tests were higher with the siig card, it was defanatlly write speeds where the siig seriously outperformed the promise card. Some speeds were about double.
Fyi I know that benchmarks don’t really directly relate to real world performance with programs and sandra is not the best benchmark tool out there for hard drives, but I figured it would be good enough for a comparison of the two diffrent controlers. These speeds are all with 7200rpm maxtors by the way. Mine are 60 gig (have 4 of them) though I am going to start using two 160 gig seagates I recently got soon, and his are 200 gig maxtors. I wish there were some reviews that gave comparisons of diffrent raid controlers with the same drives (I havent found any yet), not that I am going to buy another raid c ard soon. I have the 2 pci cards and one of my motherboards has onboard raid so I figure thats enought for now.