Cold-cache Sequential IO Benchmark

@angstrom Excellent work!
Link: https://ccsiobench.com/s/nncyQ

 

v0.9.7 - STANDARD RAMdisk Test, STANDARD SSD Test

The SSD WRITE #s are clearly in CyberSpace somewhere (that device is not that fast ) during the BUFFERED operations. They were all over the place until the RAW DATA (unbuffered?) tests started, then they started to look normal at about 420mB/s.

@angstrom I like it when you work at the speed of light to solve problems... thanks!

 

> The SSD WRITE #s are clearly in CyberSpace somewhere

Yeah, I hear you, noticed that too. The thing is that even disabling Caching in device properties appears not to have any effect on this. I'm going to exclude buffered writes from the default test. Stand by for an update

 

I always disable Caching on my SSDs... that's what made it seem so strange when I saw those buffered numbers for the WRITEs.

Good luck in chasing this one down... I'm sure MicroSloth is helping you with this one (caching in buffered mode but not at the disk level)

 

@WildByDesign , how much RAM do you have in that NetBook?

 

@angstrom , also, on the SSD, when running the QUICK TEST rather than the STANDARD, the WRITE numbers are even more outrageous (about half the speed of the RAMdisk... more caching buffers available due to faster test?)... FYI!

 

I've got it maxed out at 8GB. Despite beating this netbook to death with multiple Hyper-V virtual machines, in addition to having Chromium and Thunderbird running from a 2GB RAM Disk, this netbook is hanging on relatively well. Although I will be looking into getting a new netbook or ultrabook in the next few months.

 

FWIW, after testing CCSIO on different types of storage elements, the BUFFERED modes really make the devices look different than what they really are, especially in terms of their specified speed (the PEAK detector). The Windows buffering (not device write caching necessarily) really plays havoc with your speed detection.

I did some testing on USB2 and USB3 devices and the results are really weird... USB2 devices reading at 50mB/s+ (I've never seen that with a USB2 device), USB3 devices closer to reality. But the RAW testing modes seem to come in very close to expected target on just about everything from USB2 to RAMdisks.

For a minute there, I thought I had some really speedy CHEAP USB2 storage devices

 

Version 0.9.8 is out.

1. Hammered those write tests in place by excluding "buffered sequential" mode from test write speed testing. This drops the total number of tests per benchmark to 285 (used to be ~350).

2. Added an option for switching between normal kilo/mega units (1024-based) and the hipster version (1000-based):



Test report shows them both as does the raw measurement data in the log.

https://ccsiobench.com/get

@TheRollbackFrog - yeah, I too had a 5-year old USB stick that was showing some overly optimistic numbers

 

@angstrom - the download link above (and the one on the web site) always returns a ZERO length file, any help greatly appreciated.

 

Whoops. Try again please.

 

Thank you, sir!

 

Here is Quick test with RAM Disk with 0.9.8:
https://ccsiobench.com/s/n2lle

 

Yep, with v0.9.8, all the tested devices (USB2 & 3 incl.) now look very normal for what they are.

 

Thanks, guys. Much appreciate the input and all the help.

 

There is a newer version available:
Cold-Cache Sequential IO Benchmark -- https://ccsiobench.com

v0.9.9 - Released Feb 01, 2017​

 

Don't think I need any RamDisks.. My SSD is fast enough for all operations I need..

Neat app,thanks.

https://ccsiobench.com/s/Js0k1

 

Version 1.0 is officially out.

A couple of changes related to the detection of NVMe drives and an option for uploading test results on exit.

PS. Don't forget to hit Alt to get the menu where you can select between Quick, Standard and Extended test modes and switch between binary and decimal Kilo/Mega/Gigabytes.

 

@angstrom Thank you for sharing this great tool with us. Congrats on reaching stable release!

 

Great job, Alex... congrats!

 

Thank you. Nice tool

 

Thanks, glad you like it. Don't forget to tell the others

 

Update's out. Adds full test profile customization as well as support for testing random-access IO (the Access Pattern setting).



Sample results (Samsung NVMe):

https://ccsiobench.com/s/biBuUN - sequential IO rates
https://ccsiobench.com/s/biBuHT - random-access IO rates

 

Alex, where may I find access to the function that gives me the window above (either I'm blind <that's probably it> or it ain't there)?

 

Click on blue "Change" in Configuration line in main window. Drop the list down and select Customize...