RAID and SSD Benchmarks

[jen4950]

I figured it's been a while since I contributed something meaningful other than the Soapbox, so here are some stats.

So- I've been busy on the hardware front recently. I run my own business and handle all of the IT, so naturally I need to test out platforms

Both machines are clean installs of Windows 7 64 bit. I've got 14GB of RAM at work, 12GB DDR3 at home; Dual Xeons at work, single Xeon at home. I'm big into virtualization, so it works out well. Both are running Windows 7 64 bit.

I've got the following sub-systems that I've benchmarked for your entertainment:

Work:

Dell Precision 690 - almost 3 years old
14 GB DDR2
Dual Xeon E5335 @ 2.0 Ghz
Quadro FX3450 w/ a Dell WFP2709

OCZ 240GB Vertex 2 - 240GB SSD
4 x 750GB Seagates on 3Ware controller (9650SE-8LPML) in RAID10 - 1.3TB

Home:

Dell Precision T7500 - 2 weeks old
12 GB DDR3
Single Xeon X5680 @ 3.33 Ghz
Quadro FX4800 with one Dell WFP3008 (soon to be two)

2 x 15k Fujitsu 300GB SAS drives in RAID0 on Dell PERC 6/i - 300GB
6 x 2TB RE4 Western Digital Drives in RAID5 on Adaptec 51645 - Almost 10TB

The only benchmark I don't have is the 15k SAS drives; I'll put those up once I get HD Tach up and running again. (Ran out of trial period- oops! Went ahead and bought it, waiting for serial numbers..)

Dell Precision 690 - Work

SDD Performance - TRIM ENABLED

RAID10 Performance (3 year old technology)

Dell Precision T7500 - Home

RAID 0 PERC 6/i with 15k 300GB drives

[TBD]

Various RAID arrays with 6 2TB RE4's on the Adaptec RAID controller. With 6 drives, you can try out all of the possible configurations. The final configuration is RAID5 @ 9.985TB - DOH!

RAID0

RAID1E

RAID5

RAID5 During build Inspect - I ran the benchmark while it was rebuilding the array.

RAID 5EE

RAID6

RAID10

RAID50

Let me know what other benchmarks you would like to see on these two systems and I'll see what I can do.

 

[sub.mesa]

Those are only sequential read benchmarks; try some real benchmarks with these tools:

CrystalDiskMark
AS SSD

Those are real benchmarks; and would show an extremely large difference between SSD and HDD arrays since it benchmarks also random I/O. There you'll see how slow HDDs really are compared to SSDs. The most important benchmark of these two programs is the 4K random read benchmark. That's the most common on system disks and your most important I/O stat in most circumstances.

 

[jen4950]

Those are only sequential read benchmarks; try some real benchmarks with these tools:

CrystalDiskMark
AS SSD

Those are real benchmarks; and would show an extremely large difference between SSD and HDD arrays since it benchmarks also random I/O. There you'll see how slow HDDs really are compared to SSDs. The most important benchmark of these two programs is the 4K random read benchmark. That's the most common on system disks and your most important I/O stat in most circumstances.

Yeah- I've ran some of those. Didn't want the thread to get too long on the first post. I'll put them up tonight.

 

[jen4950]

Some more benchmarks.

On the 690 (SSD):

SSD

RAID5 Array

On the T7500 (RAID5):

 

[sub.mesa]

Pretty interesting, some notes:

- the SSD AS SSD benchmark has too low 4K-64 scores; you sure AHCI/NCQ is working? You should see 200MB/s+ here
- due to your controller (not ahci) you likely lack TRIM capability
- benchmarks done on your RAID5 HDD array show very low random I/O scores; which is normal. You'll see the scores letting lower (closer to the actual scores) as you increase the filesize from 50MB to something larger than 4GB.
- Since your controller has enabled write-back, you can see 950MB/s write values when writing only 50MB; since you will be writing to DRAM memory chip directly instead of HDD. Note that this extremely high score disappears as you increase the filesize.

 

[jen4950]

I'll look into the AHCI/NCQ- its on the SAS controller-

No matter what, I am quite pleased- it's a huge improvement over my old machines.