cancel
Showing results for 
Search instead for 
Did you mean: 

intel P3608

SGant
New Contributor

MB: SUPERMICRO X10DRH-iT

Proc: 2x INTEL Xeon E5-2620V3 2.4GHz 15MB 6C/12T

I have a question regarding Intel P3608.

i don't know if we get all the performance from the NVMe Intel P3608.

is there any other way to test the performance?

thx

CrystalDiskMark 5.2.1 x64 (C) 2007-2017 hiyohiyo

Crystal Dew World : http://crystalmark.info/

-----------------------------------------------------------------------

* MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]

* KB = 1000 bytes, KiB = 1024 bytes

Sequential Read (Q= 32,T= 1) : 5395.325 MB/s

Sequential Write (Q= 32,T= 1) : 2015.354 MB/s

Random Read 4KiB (Q= 32,T= 1) : 462.633 MB/s [112947.5 IOPS]

Random Write 4KiB (Q= 32,T= 1) : 436.565 MB/s [106583.3 IOPS]

Sequential Read (T= 1) : 2259.450 MB/s

Sequential Write (T= 1) : 1899.885 MB/s

Random Read 4KiB (Q= 1,T= 1) : 37.511 MB/s [ 9158.0 IOPS]

Random Write 4KiB (Q= 1,T= 1) : 210.011 MB/s [ 51272.2 IOPS]

Test : 4096 MiB [X: 0.0% (0.2/1490.4 GiB)] (x5) [Interval=5 sec]

Date : 2017/01/15 16:32:53

OS : Windows Server 2016 Server Standard (full installation) [10.0 Build 14393] (x64)

37 REPLIES 37

idata
Esteemed Contributor III

@Hello S.G.

The way that IOMeter* works, you will need to run two separate tests in order to get both read and write results. As it stands, it does seem like your SSD is under performing, but this could be a test configuration issue.We can recommend running the tests in the following manner:First Test ̶ Sequential Writes Test:The first workload is the sequential writes test. Instead of using the whole drive, it only uses 1 worker at 32 outstanding I/Os per target over a span of 8GB. This is done by defining the Maximum Disk SizeFor screenshots, refer the number on the section title to the pictures attached.Setting the target Disk (1.jpg)
  1. Click on the Disk Target tab.
  2. Click on Worker 1.
  3. Click on the target drive
  4. Set Maximum Disk Size to 16777216 sectors.
  5. Leave Starting Disk Sector at 0.
  6. Set # of Outstanding I/Os to 32 per target.

Setting Access Specifications (2.jpg)

  1. Go to the Access Specifications Tab
  2. Under Global Access Specifications, click on New (A new window will open) (3a.jpg)
  3. Specify a name, ie. 128 KB Secuential Writes
  4. Set Transfer Request Size to 128 Kilobytes
  5. Set Percent/Sequential Distribution to 100% Sequential
  6. Set Percent Read/Write Distribution to 100% Write
  7. Press OK

Setting Up Test Duration (4.jpg)

  1. Click the Test Setup tab.
  2. Set for 20 minutes.

Running the Test (5.jpg)

  1. Click the Results Display tab.
  2. Set Last Update and Update Frequency to 1 second.
  3. Click the green flag to run the workload.
  4. Name and save the output file.
  5. Note the result under Total MBs per Second

--------------------------------------------------------------------------------------------

Second Test ̶ Sequential Reads Test:The second workload is the sequential read workload. Just like the sequential writes workload, it only uses 1worker at 32 outstanding I/Os per target, spans over 8GB, and is run for 20 minutes.Setting the target Disk (1.jpg)
  1. Click on the Disk Target tab.
  2. Click on Worker 1.
  3. Click on the target drive
  4. Set Maximum Disk Size to 16777216 sectors.
  5. Leave Starting Disk Sector at 0.
  6. Set # of Outstanding I/Os to 32 per target.

Setting Access Specifications (2.jpg)

  1. Go to the Access Specifications Tab
  2. Under Global Access Specifications, click on New (A new window will open) (3b.jpg)
  3. Specify a name, ie. 128 KB Secuential Reads
  4. Set Transfer Request Size to 128 Kilobytes
  5. Set Percent/Sequential Distribution to 100% Sequential
  6. Set Percent Read/Write Distribution to 100% Read
  7. Press OK

Setting Up Test Duration (4.jpg)

  1. Click the Test Setup tab.
  2. Set for 20 minutes.

Running the Test (5.jpg)

  1. Click the Results Display tab.
  2. Set Last Update and Update Frequency to 1 second.
  3. Click the green flag to run the workload.
  4. Name and save the output file.
  5. Note the result under Total MBs per Second

Please let us know the results.

Best regards,

Carlos A.

SGant
New Contributor

here are my results

idata
Esteemed Contributor III

Hello S.G.,

I've reviewed the reports you've provided us, and I'm glad to inform you that your SSD is performing very well.

SSD performance is advertised based on results from tests ran on new/empty drives, under ideal conditions. Your overall hardware configuration, working temperature, amount of data currently stored on the drive, etc, will have an effect on your performance results. Results close to the 80% mark are considered normal.

Here are your results:

- Sequential Read: 5126 MB/s (Better than advertised for all capacities: 102% if 1.6TB or 4TB units, 113% if 3.2TB SSD).

- Sequential Write: 1952 MB/s (96% if 1.6TB SSD, 75% if 3.2TB unit).

As a point of comparison, you may review the following link:

- http://ark.intel.com/compare/88731,88734,88735 http://ark.intel.com/compare/88731,88734,88735

Please let us know if you have any further questions or concerns, as we'll be more than glad to continue assisting you.

Best regards,

Carlos A.

SGant
New Contributor

I am glad to hear that ...

what about random reads and writes.

Aren't random reads and writes more important for SQL server

What kind of test should i do just to be sure?

thanks

idata
Esteemed Contributor III

S.G.,

For this, we would need to run the tests again, but modifying only one step:

Under the "Setting Access Specifications" portion of the instructions, step 5 tells you to set the "Percent Random/Sequential Distribution". To test your random read/write performance, you'll simply need to make sure this option is set to 100% Random on each of the two tests. All other settings will remain the same.

When you get the results, instead of looking under the "Read/Write MBps (Decimal)" columns for Worker 1, you will need to look for the "Read/Write IOps" columns.

I've attached an example of what to look for in your results. I added color to some rows in order to make the file easier to read, but that's of course not necessary.

If you'd like for us to review and interpret these results for your SSD, feel free to post them on this thread.

Best regards,

Carlos A.