Community
cancel
Showing results for 
Search instead for 
Did you mean: 
idata
Community Manager
3,219 Views

cluster size other that 4KB on Intel SSD

<!-- /* Style Definitions */ p.MsoNormal, li.MsoNormal, div.MsoNormal {mso-style-unhide:no; mso-style-qformat:yes; mso-style-parent:""; margin:0cm; margin-bottom:.0001pt; mso-pagination:widow-orphan; font-size:16.0pt; mso-bidi-font-size:12.0pt; font-family:"Arial","sans-serif"; mso-fareast-font-family:"Times New Roman";} .MsoChpDefault {mso-style-type:export-only; mso-default-props:yes; font-size:10.0pt; mso-ansi-font-size:10.0pt; mso-bidi-font-size:10.0pt;} @page WordSection1 {size:612.0pt 792.0pt; margin:70.85pt 70.85pt 70.85pt 70.85pt; mso-header-margin:36.0pt; mso-footer-margin:36.0pt; mso-paper-source:0;} div.WordSection1 {page:WordSection1;} -->

Most program and data files are much larger than they were when 4KB became the default value for cluster size.

I would like to know whether there is any inconvenience (other than wasting some space) in having 64KB clusters.

And what performance improvements could I expect while loading today's programs and data files.

The information found on the Internet is rather conflicting.

But Intel certainly has solid measured results.

10 Replies
RGiff
Honored Contributor I
715 Views

Goggle Paragon Tech. and see what they have to say , while you are there check out the Alignment tool .

idata
Community Manager
715 Views

Sorry, but can't you be any more precise with your reference?

RGiff
Honored Contributor I
715 Views

Read post by James Walker about PAT

idata
Community Manager
715 Views

I read the post about PAT.

Alignment seems to be important for SSD performance, so I will use PAT if someone can provide it (the free offer does not exist any more).

BTW, I have tried to do the right thing by using the align= parameter in diskpart.

But my original question about cluster size other than 4KB is still unanswered.

RGiff
Honored Contributor I
715 Views

I read a Intel pdf about cluster size ,it seams that the reason for the smaller size is for higher IOPS, see pdf , download.intel.com/it/Solid_State_drives_in Enterprize.pdf .

idata
Community Manager
715 Views

The referenced pdf reports of tests with workloads of various "block" sizes, but it does not compare various "cluster" sizes in e.g. NTFS.

RGiff
Honored Contributor I
715 Views

I think after reading the pdf , the jist of it is do you want high IOPS use 4kb , If you are doing video editing use 64kb , It a choice over speed or file size.

idata
Community Manager
715 Views

There is a section (in the pdf link below) on how block sizes can influence performance, which basically confers with what the white paper and Robert are saying.

http://www.snia-europe.org/Document.nsf/A53AA71753BDD64E8825772D003E4CD3/$FILE/Performance_Benchmark...

idata
Community Manager
715 Views

The motivation for my original question is that I want to maximize the responsiveness of my W7 laptop.

So I think I have to optimize it for large program and data files being loaded, small ones will load quickly anyway.

I therefore generated my data and backup partitions to NTFS with 64KB clusters (unit=64k in diskpart parlance).

Despite many efforts, I did not succeed in changing the system partition's cluster size to anything greater than 4KB, neither before installation (diskpart), nor after installation (Partition Wizard Pro) which provides a "change cluster size" function.

The Partition Wizard Pro's error message indicated that my system contains compressed files, files which can only be held in 4KB clusters.

When I will do the next clean install, I will create a separate partition for the program files, partition which I can create with 64KB clusters.

Thank you for informing me about the alignment issue.

It turned out that my boot + system partition was not correctly aligned, even though I had used the diskpart align=1024 parameter.

I finally succeded by omitting the align= parameter, in other words, W7 does align properly without any help from the user.

idata
Community Manager
715 Views

Hello,

 

Almost two years later, I finally was able to satisfy my curiosity.

 

I got a new laptop, with to different SSDs installed (not Intel though).

 

On five identical size partitions, except for the cluster sizes 4KB, 8KB, 16KB, 32KB and 64KB respectively, I ran CrystalDiskMark Seq and 512K random performance tests.

 

Result:

All indicated performance figures were within 1.5%, a difference which I consider to be smaller than the measurement accuracy.

And no trend in one way or another!

 

Conclusion: I formatted all my partitions with 4KB cluster sizes.

Reply