RAID 10 Benchmarks Large Differential between 2 sets

I'm trying to benchmark 2 different RAID 10 arrays run by hardware RAID cards on a single machine. One has the OS, the other has data.

CrystalDiskMark3.0.1 shows a huge delta though between the two arrays, and I'm trying to figure out why.

The OS array reports 1400MB/s and 1500MB/s for SeqRead/Write respectively.

But the 'data' array reports 370MB/s and 150MB/s SeqRead/Write respectively.

I'm assuming it has something to do with CrystalDiskMark is actually reading how fast the OS communicates with the Array, and not how fast the array is actually communicating with the motherboard (whereas since CrystalDiskMark is sending the test data from the OS... it registers the 'true' speed on it's own Drive.)

In order to test that though, I ran the program from the 'data' array, and got the same result. Leaving me a bit confused.

Any ideas?
7 answers Last reply
More about raid benchmarks large differential sets
  1. What hard disks and how many per RAID 10? What controller and options?
  2. RAID Array OS: OCZ SSDx4 120GB (LSi 9260 Controller) 128kb units
    RAID Array DATA: Intel SSDx4 80GB (Adaptec 2405 Controller) 128kb units
  3. Are these the controllers taht you have?

    You are comparing two RAID controllers. All benchmarks show that the LSI is much faster when using SSDs. The LSI has a larger cache, faster transfer rates, is more expensive, etc.
  4. Public access computer... posted extra by mistake...
  5. That's a good point - I knew they were a bit different (the 2405 is listed by many as a 'budget' RAID card (despite the very high price I paid for it 3 years ago))... I hadn't expected THAT significant a difference - I had never even bothered to benchmark it since it was performing at what was use-wise great speeds.

    1. Still, I've read benchmarks in SSDs all over the place - from 200 to 800MB/s on the 2405, but they weren't all RAID 10 (lot of RAID 0 which I assume would account for the delta).

    2. Another factor to consider (for anyone else that reads this) is that the TEMPERATURE of the card, a notoriously hot-running one, is apparently also related to a decrease in performance.

    As I haven't done any after-market cooling on the card, it's possible the extremely hot temperature (130F singe your hand hot) of the card could be downgrading the performance - I'll try to squeeze a cooler on its heatsink... if I manage to I'll post a performance gain - I haven't seen any information on temperature vs performance on this (or any) RAID cards to date.
  6. ...
  7. In theory RAID 0 should be twice as fast as RAID 10 with the same number of SSDs. The OCZ SSDs may also be faster than the Intel SSDs. Cooling the Adaptec controller probably won't make it faster, but it's worth a try.
Ask a new question

Read More

NAS / RAID Hardware Benchmark Storage Product