Perfect Disk - SSD/HDD optimization

This thread is for sharing information and results in optimizing SSD and HDD with Perfect Disk software... enjoy

 

That's great, thank you for creating this thread @QuillP !!

To start us off here is some info from the previous thread that brought us here:

Just before your post, in the same thread:

And, previously on NBR, a thread @Phoenix started, but now closed:

What do you think of this optimization? Perfect Disk
http://forum.notebookreview.com/threads/what-do-you-think-of-this-optimization-perfect-disk.771762/

And, a link to PerfectDisk Trial:
http://download.raxco.com/perfectdi...perfect disk&gclid=CPqJwqTV5LsCFStBQgodblMAhA

And, PerfectDisk Pro product page:
http://www.raxco.com/home/products/perfectdisk-pro

 

Hey @hmscott , quien you force to HDD with SmartPlacement, select Classic, Performance, Performance aggressive or ...?

 

Yes, it's in the options per device, poke around, you'll find it

 

Of all these, which ones do you use? Classic?

 

Please post/show me a Snipping Tool cut of the dialog with a circle around what you are referring to so I can see what you are seeing.

There should be a clear option for SSD or HDD.

 

here...

 

How to solve the problem of 4k scaling of Perfect Disk in Windows 10 ...

Just put the " PerfectDisk.exe.manifest" file in the Perfect Disk home directory: C:\Program Files\Raxco\PerfectDisk\

Then, tell windows to prefer an external manifest file.

1. Press Windows Button + R, type “regedit”, and then click OK.
2. Navigate to the following registry subkey:
   • HKEY_LOCAL_MACHINE > SOFTWARE > Microsoft > Windows > CurrentVersion > SideBySide
3. Right-click, select NEW > DWORD (32 bit) Value
4. Type PreferExternalManifest, and then press ENTER.
5. Right-click PreferExternalManifest, and then click Modify.
6. Enter Value Data 1 and select Decimal.
7. Click OK. Exit Registry Editor.

Done!

 

Ah, I thought you were talking about the forcing of HDD over SSD first, but it looks like you've already done it, and now want to know about the level of optimization options.

It's interesting, I'm sure there are reasons for each, but when I've tried them they haven't made much of a difference.

If you hover over(?) the options (or read the detailed help) they might have more descriptive reasons for each. Raxco is good about helping clients with particular application needs find better optimizations, and then Raxco put's them in the product.

I'd start with Classic (default), and run tests, and then try the others...

Here's an example from an old RAID0 4x512 test:

2TB M600 x4 first run - no updates - perfectdisk smartplacement + ssd optimize + writeback cache in RST.PNG

2TB M600 x4 512GB - updates done - perfectdisk smartplacement + writeback cache in RST


Those are from a new RAID0 volume, the optimizations won't kick in under you've used it for a while, and the optimizations might change over time as you complete installing apps and move on to using them.

Please post your findings if you find a good model for optimization

 

I think you used the same CrystalDiskMark capture...

 

This is the performance with 1) Performance aggressive + 2) Ssd optimized

 

No, they are different, they all look the same given the same snipping area. I changed to another one in my post below as an example, and it has different numbers, but looks the same. I have a steady snipping hand

 

Is that better?

 

Only a little... not much. Now works as the first day.

BEFORE OPTIMIZATION



AFTER OPTIMIZATION

 

Yup, that's about right, it restores it back to the performance of a freshly created volume.

Doing the Optimizations on a new install after updates is also a good idea, and then let PD automation and Write Optimization keep things optimized moving forward.

That's where those CrystalDiskMark runs are from, after doing a new OS install, showing the progressing of optimization and results.

 

Why not use the window-snipping option?


Sent from my iPhone using Tapatalk

 

Those images are from 2015, I haven't found a need in all that time to do anything else except use PrtScn and Snipping tool.

It's a small amount of time, not frequently enough for me to worry about finding or perfecting another way.

I know people that need to do it so much they have much better tools and techniques, but I haven't found the need. I've worked with them doing manuals and documentation.

You know how it is, once the method rolls off your fingers so quickly, it's not worth changing the motions.

Besides, I like the raw feel from capturing other behind the dialog info, and side by side dialogs and Windows all in one shot.

 

I'm going to try the automation and Write Optimization options.

I have another question: is anyone trying to create a raid0 with two different ssd pcie models? I have 1 960 Pro x1tb (MLC) and 1 pm961 (it's like 960 EVO) x1tb (TCL). They have a similar performance in reading, the difference is in writing. With this raid configuration, I lose or gain performance?

 

I use Greenshot ... it's very useful. http://getgreenshot.org/

 

You'll need to set it up and let us know to be sure

Usually it's not a good idea to have different units, they should all be the same, as the algorithms used to run the RAID may expect performance very close to the same for each unit.

When the write speed of one unit is much slower than the write speed of the other unit, I've seen errors reported - like a waiting for response timeout error - or a silent failure.

It's possible things have improved since I've tried that way back when, so please give it a shot and let us know how it works out, and what the performance drop is for writes.

Given no failure for the RAID0, at least the slower writes will slow down the overall speed of the RAID0.

It could be as slow as the slowest device if the software decides to pick the slowest device throughput for the timing - which is about the only way I can think it would be made to work - either that or ignore the "timeout" of the slower write completion.

 

I will try and then publish the result. With 2x 960 Pro in Raid0, did not "perceive" a performance difference vs 1 single 960 pro.

 

Thanks for the tip

 

Depends on how your PCIE lanes are wired, you might be bandwidth limited, you might not be seeing the full potential performance of your 2x RAID0. Ignoring human perceptions, what did the benchmarks show?

Also, you've discovered a little known fact. PCIE NVME drives are a waste of money for the most part, and it's better to invest in SATA storage for a personal desktop or laptop.

As you said there wasn't a perceived speed improvement when logically there should be.

Most operations only require a small amount of disk IO, everything else takes up a lot more time. Improving the time spent on disk IO by 1/2, won't show up as a 2x improvement overall, that improvement is drown out by the much longer time spent doing the rest of the task.

When going from a 50MB/sec HDD to a 500MB/sec SATA SSD you save a huge chunk of time off the wall clock, a large enough amount of time to be "perceptible" even within the mix of doing other things to complete a task.

But, going from 500MB/sec to 1500MB/sec won't. Even though logically it's slicing off 2/3 the time to do a transfer, the time saved will be minuscule next to the time of other operations, and the wall time includes all operations.

There are some things that a PCIE NVME SSD shine for, but most people will only have fleeting experience with them. If your specialized workflow needs this speed up, then you must have it, but most people don't need it.

It's better to pay 1/2 the cost for storage and get SATA SSD's, than spend a premium for high speed storage that most people will never need.

 

Gotcha. To clarify, I'm talking about the window snip "mode" of the snipping too you're already using. You can do a free form snip, a rectangular snip (the default), or a window snip (which is a rectangular snip that automatically grabs the full app window).


Sent from my iPhone using Tapatalk

 

It comes down to the sterile effect of capturing a dialog or window isolated from the environment it occupies space with.

I like the raw feel from capturing other behind the dialog info, and side by side dialogs and Windows all in one shot.

Having done *lots* of sterile capture of dialogs and Windows for documentation over the years, I associate one with the other.

By grabbing a shot hovering over the area of interest, while capturing some of the environment, helps me convey the context along with the focus of the capture.

It's been a long time, but you've reminded me exactly why I don't use the isolated capture, it breaks when trying to get popups over dialogs or Windows, as @QuillP needed to capture.

There were other such state information that was isolated and grabbed instead of the overall image I needed to grab; I recall fighting with it at one point.

Like I said, it's been a while, but that is how I recall why I ended up doing it the way I do it now. Even Snipping Tool has such limitations, and I use PrtScn with MSPaint that I then use Snipping Tool to grab the area I want to show.

So far it's worked well, and have never had a single complaint that what I provided has been confusing or inadequate to convey the information or direction I am giving.

 

@QuillP: Did you try Samsung NVMe driver in AHCI mode instead of RAID?

 

Yes, I prefer mod AHCI because I can check the ssd status and use Samsung Magician to update the firmware. Now, i converted the block size to 64k (from the Windows primary partition) to test the performance.

 

Did your RAID0 performance scale 2x from single SSD throughput? Or, was it limited by the chipset to something less than 2x?

A simultaneous transfer test against both SSD's would be interesting too, to see if the throughput is limited through the chipset.

 

Yes, I tried them. 2x 960 pro. 3.0x4. The result is impressive in the benchmark, but no difference in real usage. Only increases writing performance. The reading performance is equal to that of 1 960 Pro (I think that because the reading performance of the 960 pro already reaches the limit of pcie 3.0x4). The only real gain is that you have a single 2tb drive.

I now have to try a raid0 with 2 Samsung pm961 x 1tb. (Is like 960 Evo). This weekend, I'll try to create a raid with these.

I found a benchmark that I did to a Raid0 of 960 pro 1tb x2

 

AOMEI Partition Assistant it is very useful. I always use it for everything else.

 

Yup, I had the same problem with the 4x SATA RAID0, the read speed topped out at just under 1700MB/sec due to the chipset, instead of the 2200MB/sec theoretical unlimited Read speed.

Hopefully this throughput limitation for 2x-3x RAID0 PCIEx4 is going away in the 3xx series motherboard chipset, waiting for more details to emerge as they get closer to release.

 

In my case, I prefer to use only 1 960 pro x1tb (ahci) instead of a Raid, because the actual performance of single 960 is almost the same as with Raid0. I think the Raid consumes a little more cpu (do not confirm it, but I think it is). Also, I do not need the benefit of a 2tb partition. I think I can settle with 3500mb of reading and 2200 of writing hehehe. That's enough for now.

 

Anyone know what is the fastest and easiest way to change from ahci to raid in Windows 10? I started in safe mode and changed to raid. The last time I tried it did not work, I think something changed in the last update of Windows.

 

I agree, practically speaking it's better to not RAID the NVME SSD's, as there are quickly diminishing returns due to the chipset limitations.

Before we found out about this limitation we were hoping for 6GB/sec 2x RAID0's... sigh.

What about 12GB/sec 4x RAID0... it could happen

 

So, what's wrong with Default Windows 4k alignment? I mean with 64k block size you'd be wasting lot of space for smaller files though.

 

nothing bad. I was just testing the ssd performance. In raid0 with strip size 128k and block size 64k increases reading and writing performance. I wanted to know if the ssd in Ahci was the same result...

 

There are some applications where tuning the stripe size to the access and data flow makes a big enough difference to put in the effort to test and tune - especially if you are rolling out 10's, 100's, and more storage devices

For laptops and desktops with widely varying demands, I usually tune for the storage size vs cluster size vs stripe size.

But even then it's usually a small gain, but fun to do

 

Sure, the gain is always small in home desktop or laptop. At my job, we use big servers with storage devices where fine tunning is necessary and makes a big difference.

 

Exactly my point, just syncing up for future reference

 

Hey @hmscott, OptiWrite and auto-optimization options actually do something? I meand Do you see any long or short term benefit?

 

It'll reuse existing cells for writing new data after TRIMming. So, new cells are not used thereby increasing lifetime of the drive.

 

I thought it was no longer a problem in the new ssd. For example, 960 pro has 800tbw of lifetime, I think my grandchildren could continue to use it.

 

Well your grand children might have an 800TB drive PCIe 5.0 for downloading torrents, steam games, downloading Win 10 every day and nuking it, etc etc....

 

Maybe you can help me. Do you know a way for Windows 10 to boot properly after switching on the Bios from Ahci to Raid? Before I started Safe mode to take the change but now it does not work...

 

See this thread
Uninstall existing RAID driver in device manager and use the cmd as mentioned in the thread.

 

This method no longer works. Something to change with Creators Update ....

 

I don't notice it working, but I also don't see my benchmarks decrease over time like I did before optimizing and then enabling the automatic write optimization.

You have completed the optimization, you can check for performance drops by running the benchmarks occasionally, for me the performance hardly varied over normal run variance differences.

Silent, but efficient

 

Are you sure? Since this method is suggested by MSFT TechNet.
I tried it when I got my PC and I tried on Anniversary update which is RS1 and aside from not allowing Unsigned Driver Install on RS2 it must be same.
I'll see. I use Dual Boot Dual SSDs for Linux and Windows. It took me a month to get Linux boot perfectly.

 

Hi, this weekend I tried to create a raid 0 of 2x 960 Pro 1tb. These are the Pro and Cons:

(Hardware detail Alienware 17 R4 7820hk, 32 RAM, 2 ports pcie 3.0 x4).

Pro:
- duplicate the space x1.96 tb
- 25% max writing speed (Raid0 3.000mb/s vs. 2.000mb/s 960 Pro)

Cons
- more CPU consumption, + 10%
- higher temperature +10 c (When the Raid is used)
- In normal use, the increase in writing performance is not perceived.

The reading values are practically equal to those of a 960 Pro 1tb. In my case, only Raid 0 is justified to duplicate the space...