Sort of low raid-0 crystalmark results...thoughts?

[Trix teh Pony]

I bought two Samsung F4 hard-drives for a raid 0 setup. I got them installed and decided to run a benchmark to see where my performance was at. I was expecting more, to be honest. People in the Newegg listing I linked are reporting numbers much higher than me. I have the stripe configured at 64k blocks.

I have two suspicions...

The drives do not have jumpers set. Both are currently jumperless (slave/slave), but I read that in SATA raid, with each drive on a different channel, this shouldn't matter.

Another potential problem is that My board is kind of out of date, being that Nvidia doesn't provide Nforce drivers for Windows 7 64-bit.

 

[SickBeast]

From what I've read, RAID-0 will give you a 20% boost when gaming, a larger boost when working with media, and a negligible boost when it comes to you OS and general usage.

Your numbers look off. They look as though they would for a single drive. Are you sure you set up your array properly? You need to format the array as RAID-0 and choose a cluster size.

 

[Trix teh Pony]

I'm pretty sure I did. I set my system's sata controller to raid and then created the stripe in the raid controller's bios. My raid's name in my system bios is named "raidlol", whereas before I created the stripe it just showed the two hard-drives. I then provided the raid drivers to the windows installation via floppy and the OS installed.

I'm going to try again and see what happens. Good thing I didn't format my old drive, lol, because the semester isn't over yet and I need a working system.

This is a screenshot with just one hard drive...

 

[Zap]

Your numbers look off. They look as though they would for a single drive.

Maybe it was set up as RAID1?

 

[SickBeast]

My guess is that a larger stripe size block could improve things. What a PITA to try, though.

In a strange way it looks "capped" at 150mb/sec, but it can't be because he's on SATA2, right?

 

[spinejam]

2 F3's in RAID0 (128 kb stripe & write cache enabled):


Uploaded with ImageShack.us

 

[corkyg]

Maybe it was set up as RAID1?

That's what it looks like. If it were really RAID 0, how could one drive be measured by Crystalmark or anything else? A partial RAID 0 = nada.

 

[dorion]

My guess is that a larger stripe size block could improve things. What a PITA to try, though.

In a strange way it looks "capped" at 150mb/sec, but it can't be because he's on SATA2, right?

Well his board has 2 SATA2 ports and 2 SATA1 ports...

Trix teh Pony are you plugged into the red and orange SATA2 connectors or the black Sata1 connectors?

 

[tweakboy]

Also cluster size can make a difference.

 

[Trix teh Pony]

My write caching is enabled. My stripe was 64k and I changed it to 128k (no difference), and I'm plugged into the correct sata-2 ports. I made sure my that my raid is raid-0 and not raid-1. I'm also using sata-2 cables, and I swapped them out with different ones just incase if they were faulty. I also disabled every channel in my bios except for sata-2 (ide and sata-1 all disabled). That didn't help. I also tried finding drivers or anything that I haven't installed. My only guess is the chipset drivers, because there aren't any on Nvidia's site for Windows 7 64bit. My motherboard is basically kind of obsolete, and I'm considering buying a raid controller card to get the raid off the motherboard because of this.

My raid controller is an on-board Jmicron, and in my research I've read that they're pretty much crap. Not sure if this is the reason. Both drives are currently jumperless, but since they're each on their own port, this shouldn't matter, right?

I ran a different benchmarking software, but I am getting the same results. I thought about unplugging one of the drives to see if the raid would break, but looking the device manager tells me that it probably would.

 

[sub.mesa]

Do you have another controller you can connect your drives to, than JMicron?

JMicron is only good for connecting an occasional 'lost disk' and comprises the absolute bottom of low-quality RAID solutions on Windows. If your chipset does not provide RAID, then you have no "real" onboard RAID functionality. Real onboard RAID would be Intel ICHxR southbridge or AMD SB8xx southbridge or another chipset with RAID functions enabled. Intel has the best onboard RAID drivers available. Remember that onboard RAID is 100% software RAID; everything is done by drivers and nothing by the hardware. But as you may already assume, the Intel RAID drivers will only function on Intel RAID controller; not other controllers or Intel non-RAID controller.

 

[Emulex]

jmicron is below software raid.

lol.

 

[Rubycon]

Use HD Tach or HD Tune benchmark (and its file benchmark) to test STR. ATTO also works. Crystal and AS-SSD are more i/o oriented for solid state disks.

 

[Trix teh Pony]

I decided to just do all three benchmarks for consistency.

 

[Rubycon]

Yeah you're bottlenecked for sure.

LOLRAID?

 

[Trix teh Pony]

I keep reading that my onboard raid sucks, and have also read about performance issues with it in raid0. I guess that's that then. I'll be upgrading my mobo when the x68 chipsets are released, which won't be for a few months. Lol, damn...

Thanks for all your help guys, I guess I'll just ride out the bottleneck until then. Atleast I'm getting some extra performance with two drives as opposed to one. When I actually do upgrade, I'll report back with a before-and-after.

 

[Trix teh Pony]

Posting an update. I've basically built a new system and I'm still using the same Samsung F4 hard drives. Everything is performing well, including the RAID.

For reference, here's my upgrade path..

AMD X2 6400+ cooled by Corsair H70
ASROCK Alivedual Esata2
4gb OCZ Reaper
500w powersupply
EVGA Nvidia 8800gts 512 g92

Intel i5 2500k cooled by Corsair H70
Gigabyte GA-Z68X-UD4-B3
8gb DDR3 gskill ripper-x 1600
850w power supply
EVGA Nvidia gtx 570

 

[tweakboy]

I tell you the problem. The F4 drive , you need to run it at 512k

you will get much better benchmark results instead of 4096kb.

I would get 154 read from my F4 non raid and 155 write, but since the drives have kinda went bunk,,, it shows in speedfan SMART the performance is half way the bar. instead of full.... both drives are busted they have errors but HD Pro says drives are
used to say 115 in details now its 101 Ill put a screenie up soon. Now I get 100 on one drive and 115 on other, its 512k but these drives are picky,, things are slow I can RMA both of them perhaps. I hate these F4 drives also I have to defrag Trace.fx the readyboost prefetch I have to defrag and wat not sometimes apps delay on open,, or if I defrag with defraggler it opens winamp in 0 seconds instant. But if I reboot try again its gonna take couple seconds like 3. soo I have to defrag again and what not.