P-series RAID 0: Is it worth it?

I haven't found any benchmarks for RAID 0 in these laptops.

Here are my benchmarks for a P-7805u with 1 Hitachi 7K320 using HD Tune 2.55 and block size set to 64KB:
Minimum: 27.4 MB/s (low spike seems irrelevant)
Maximum: 74.7 MB/s
Average: 57.2 MB/s
Access Time: 16.8 ms
Burst Rate: 62.3 MB/s
CPU Usage: 11.2%

Actual performance would be better than this because I have the hard drive configured so that all the used files are on the outer, faster part of the disk. The test had to use the inner part to do the testing. Also for some reason the transfer rate went consistently down over the duration of the test. Likewise, the access time went up over time. Perhaps the test starts with the outer most part that it can use and works its way into the center.

I could definitely see RAID 0 being a big improvement if the 1 hard drive were nearly full and programs were being stored on the inside of the disk.

Plz post benchmarks using other standard P-series hard drives because this will be useful for other people and will be useful to compare RAID 0 in general with these laptops.

Note: I discovered reading an older post that RAID 1 is poorly implemented in these laptops (a P-7805u) and don't increase the read speed by much. I have yet to find someone using a RAID 1 system where the files on the two disks are read independently, making the read faster than RAID 0. According to Wikipedia such a RAID 1 exists.

 

I have a Raid0 setup in my 6860 with two wd scorpion 5400rpm drives. I'm using a 64 kb stripe size. Here are my HD tune scores.

Transfer Rate
Min-44.6mb/s
max-125.3mb/s
avg-97.5mb/s

access time-16.8 ms
burst rate-76.7mb/s
cpu usage-5.7%

A screen shot attached.

 

Those are surprisingly good. What CPU do you have? RAID 0 supposedly taxes the CPU more.

I got these specs from Tom's Hardware charts for the Hitachi 7k320. http://www.tomshardware.com/charts/...s/h2benchw-3.12-Max-Read-Throughput,1109.html

Avg Read/Write both about = to 66MB/s
Write Access Time = 8.00ms
Read Access Time = 15.90ms

There are a lot of other specs but I don't understand what they mean or they are irrelevant to this comparison.

 

I have a t9300 in there right now. I've been considering an Extreme chip but the cost is not worth it IMO.

As for cpu load, I am running a very striped down version on 64bit vista. I've got 26 processes at idle using <500mb of ram. That being said, perhaps less load is reflected by the slimmer OS.

 

Damn under 500mb ram! I don't think I've ever gone below a gig. What exactly did you cut out and how did you do it?

 

Alot of it was done just through disabling various services and start up items. Granted, I've disabled services that have led to loss of functionality of different parts of windows, but I've done this in a way to only remove the features I don't use, or don't see as beneficial.

Here's a screen shot of the process tree at idle.

 

I have two seagate 320GB 5400rpm drives in Raid 0 configured for high performance on my stock CPU p6860fx, here's the test data:

HD Tune: Intel Raid 0 Volume Benchmark

Transfer Rate Minimum : 46.0 MB/sec
Transfer Rate Maximum : 122.3 MB/sec
Transfer Rate Average : 91.7 MB/sec
Access Time : 25.5 ms
Burst Rate : 488.1 MB/sec
CPU Usage : 13.1%

 

Nice results, they look similar to mine, though your burst rate is significantly higher than mine is.

Do you remember the stripe size you are using? I've been considering a 128kb stripe size instead of my current 64kb.

 

Those results are really good (I'm especially surprised by the low CPU usage), but the access time is shockingly high. Oh wow that burst rate seems unrealistically high.

I am wondering how RAID 0 would hold up if the block size of the benchmarker were lowered, and what the average file size for is in gaming. I've read some old articles (I think around 2005) that stated that there was basically no improvement in game load times, only a minimal improvement in boot times, with really the only high gains in video/audio encoding or playback. I'm sure this is less the case now, but it would be nice if there was a way to get real world benchmarks.

 

I noticed something similar in way of game load times. Not much difference or any really. I did notice a big difference when working with photographs in adobe lightroom and photoshop. Save times for huge file sizes dropped by more than half.

 

The high burst rate comes from enabling write caching and advanced performance in the policy property of the drive. I'm using the default stripe size.

In practice, I also don't notice much difference in application performance. However, there's a huge difference in large file transfer performance. This comes into play most for me when doing system backups. I just bought a 500GB 7k external notebook HDD with a USB 3.0 enclosure and an expresscard with USB 3.0 ports. The RAID 0 array speed matches to my new drive quite well; I average almost 90 MB/sec to the external drive.

 

Wow USB 3.0. I haven't looked into it, but I bet eSATA is still better, because I remember with USB 2.0 the real world data transfer rate was really only about 200Mbps and USB supposedly uses a lot more of the CPU because it has to do some converting. Is USB 3.0 able to power 3.5" hard drives? Any way, once flash drives become fast and support 3.0, it will then probably be worth the investment.

 

Interesting, I'm unable to change that on my drive policy tab.

 

As far as speed is concerned, I dont think there's much practical difference between USB 3.0 and eSATA. However, I think USB 3.0 will come out on top due to the addtional power (which means no additional power cable needed for 7k rpm drives) and the fact that it's backwards compatible with USB 2.0. I don't know off hand if USB 3.0 will power 3.5" drives, but I suspect not; certainly not through a USB 3.0 expresscard which is still limited by power availability from the expresscard interface.

 

I think you need to do it through an app that you're running as administrator. It should work if you try modifying the properties using storage manager.

 

Ahh excellent, thank you for pointing me in the right direction. I needed to install the Intel Storage Matrix Manager. From there I could enable write caching.

I ran hdtune again with caching enabled with advanced performance enabled too. I got some interesting results for transfer rate. I had huge drops during the test as you can see in the first attachment.

I disabled advanced performance but left write caching enabled, this leveled out the transfer rates.
As you can see, write caching drove the burst rate through the roof, 80ish mb/sec without caching 700mb/sec with caching. Everything else stayed pretty much the same.

Thanks for the tip Starcub!

 

np.

I re-ran the same tests on my machine. There was very little difference between the non-adv and adv results. I did notice that my plots were significantly more variable than yours. I suspect the difference between yours and my plots is the drives we are using. Would probably need to do a more controlled test to determine if there were any real difference between the two modes. I'm not that interested due to the fact that both modes warn you that you could lose data if you don't have power backup, and both provide fairly equal performance.

 

I'm probably going to wait for SSD prices come down and buy two of those and put them in RAID. I've read that they improve much more from it than hard drives, not to mention that one SSD would probably beat 2 3.5" hard drives in RAID 0. Hopefully our laptop's RAID 0 controller will work well for SSDs.

Anyone know if SATA II drives work in these laptops?

 

They certainly do, also I have two of the 500gig travelstar 7200rpm in raid 0.
Synthetics look great but its meh. Depends on your application.

http://forum.notebookreview.com/showpost.php?p=5888314&postcount=336
[edit, added link to previous benches]