Choosing the right SSD...

Hi,
Yeah bad word to use
Dynamic disks aren't a good choice

geneo said:

All SSD have garbage collection.

Click to expand...

Yes, but on SandForce SSDs that could support TRIM, if TRIM either was not supported by the SATA controller or was not supported by the firmware thereof (as was originally the case with Marvell controllers until finally, at very long last, a firmware update came out to fix that...), the garbage collection wouldn't kick in until the last moment, after the SSD had already become substantially slower. So then you had to secure erase the SSD or use "Tony Trim". Whereas the C300 the garbage collection effectively prevented the SDD from gradually becoming slower over time.

geneo said:

Garbage collection consolidates partially filled blocks into filled ones and generates free unused pre-erased blocks for new writes. The only difference between the gargabe collection on the C300 without TRIM and an SSD with TRIM is that the garbage collection with TRIM has the information of deleted blocks.
So it was not as good as an SSD using TRIM or we wouldn't have TRIM.

Click to expand...

See above.

geneo said:

Also, you could and should have over-provisioned the C300 operating without TRIM.

Click to expand...

Yes, not letting the used space exceed 70% of total capacity was standard practice.

P.S., back at the time, another thing that you could do to prevent your SSD from becoming slow over time was to use SSDkeeper (from the original company that also made Diskeeper) that later became the HyperFast feature that is part of Condusiv Diskeeper (the predecessor of DymaxIO).

geneo said:

What does "dynamical" over-provisioned mean?

Click to expand...

As shown in the graphic illustration that @NavyLCDR posted, it means that both the un-partitioned space (if any) and the available free space on each partition gets automagically added to the pool of unallocated cells [to which the over-provisioning is being applied].

This recent information from micron about use of their storage executive might be of interest.

hdmi said:

Yes, but on SandForce SSDs that could support TRIM, if TRIM either was not supported by the SATA controller or was not supported by the firmware thereof (as was originally the case with Marvell controllers until finally, at very long last, a firmware update came out to fix that...), the garbage collection wouldn't kick in until the last moment, after the SSD had already become substantially slower. So then you had to secure erase the SSD or use "Tony Trim". Whereas the C300 the garbage collection effectively prevented the SDD from gradually becoming slower over time.

See above.

Yes, not letting the used space exceed 70% of total capacity was standard practice.

Click to expand...

If you read what I said, my point was the C300 ran garbage collection in the background just like all SSD do now (your point being other did not until they ran low on space), but with out TRIM they were not efficient no matter what - it was no replacement for TRIM. And they would also slow down as they filled.

hdmi said:

As shown in the graphic illustration that @NavyLCDR posted, it means that both the un-partitioned space (if any) and the available free space on each partition gets automagically added to the pool of unallocated cells.

Click to expand...

The available free space on partitions that are not fully utilized is not over-provisioned space - it is just that, free space. Inventing terms like "dynamical over provisioning" is kind of confusing. Free space on partitions is not the same as over provisioned space which is fixed and guaranteed to be avaialble to the SSD. As far as the SSD is concerned though, it is the same as over provisioned space, but it is not guaranteed to be available (if the partitions fill). There is no automagic addition - it's is just a bunch of unused blocks to the SSD as I have said earlier.

geneo said:

The available free space on partitions that are not fully utilized is not over-provisioned space - it is just that, free space. Inventing terms like "dynamical over provisioning" is kind of confusing. Free space on partitions is not the same as over provisioned space which is fixed and guaranteed to be avaialble to the SSD. As far as the SSD is concerned though, it is the same as over provisioned space, but it is not guaranteed to be available (if the partitions fill). There is no automagic addition - it's is just a bunch of unused blocks to the SSD as I have said earlier.

Click to expand...

Well, you need to talk to Seagate then. The term dynamic over-provisioning was "invented" by them, not us. As well as Techspot:

Quote:
The SSD uses non-user space in the flash memory (over provisioning or OP) to improve performance and longevity of the SSD. In addition, any user space not consumed by the user becomes what we call dynamic over provisioning – dynamic because it changes as the amount of stored data changes.

And EDN:

Figure 4 shows the effect of the TRIM command on over-provisioning. For a “marketed” percentage of over-provisioning (28 percent in this example), the amount effectively increases after performing a TRIM operation. Note how the capacity originally designated as Free Space remains consumed as Presumed Valid Data by the SSD after being deleted by the operating system or the user until a TRIM command is received. In effect, the TRIM operation provides dynamic over-provisioning because it increases the resulting over-provisioning after completion.


Figure 4. The effect of the TRIM command on over-provisioning percentage

BunnyJ said:

The title says this "Windows 11 might seriously slow down your SSD",, key word.. might. You left that out. Oh and one of the sources is Redit not the best one on the internet.

Click to expand...

Also from the story
"The issue isn’t universal, and at least some drives seem to be performing at their full speed with the adoption of newer drivers. Less drastic reductions, in the 10 to 25 percent range, were also reported on some drives."

Seems?? Sounds hit and miss at best.

geneo said:

, but with out TRIM they were not efficient no matter what - it was no replacement for TRIM.

Click to expand...

Real-world testing in the long run proved beyond questioning that they [C300] remained equally as efficient as when they were still new and unused. The same could not be said about SandForce SSDs as pratically every gamer/enthusiast who owned one kept secure erasing it the whole time, all simply because leaving the PC sit idle in BIOS for hours was not a viable solution to them.

geneo said:

And they would also slow down as they filled.

Click to expand...

That's just because the over-provisioning was skimped on, as the NAND was so highly expensive that you really had no other choice but to keep the size of your user data below 70% of total capacity─excepting of course if you were able to afford an SSD from Intel that was hyper expensive during those days.

NavyLCDR said:

Well, you need to talk to Seagate then. The term dynamic over-provisioning was "invented" by them, not us. As well as Techspot:

Understanding SSDs: the need for TRIM, overprovisioning, and more

A number of years ago, the storage industry got together and developed a solution between the OS and the SSD by creating a new SATA command called...

www.techspot.com

Quote:
The SSD uses non-user space in the flash memory (over provisioning or OP) to improve performance and longevity of the SSD. In addition, any user space not consumed by the user becomes what we call dynamic over provisioning – dynamic because it changes as the amount of stored data changes.

And EDN:

https://www.edn.com/understanding-ssd-over-provisioning/

Figure 4 shows the effect of the TRIM command on over-provisioning. For a “marketed” percentage of over-provisioning (28 percent in this example), the amount effectively increases after performing a TRIM operation. Note how the capacity originally designated as Free Space remains consumed as Presumed Valid Data by the SSD after being deleted by the operating system or the user until a TRIM command is received. In effect, the TRIM operation provides dynamic over-provisioning because it increases the resulting over-provisioning after completion.

View attachment 15920
Figure 4. The effect of the TRIM command on over-provisioning percentage

Click to expand...

Yeah in all my years of working with storage I never saw that term used.

hdmi said:

Real-world testing in the long run proved beyond questioning that they [C300] remained equally as efficient as when they were still new and unused. The same could not be said about SandForce SSDs as pratically every gamer/enthusiast who owned one kept secure erasing it the whole time, all simply because leaving the PC sit idle in BIOS for hours was not a viable solution to them.

That's just because the over-provisioning was skimped on, as the NAND was so highly expensive that you really had no other choice but to keep the size of your user data below 70% of total capacity─excepting of course if you were able to afford an SSD from Intel that was hyper expensive during those days.

Click to expand...

Well the Sandforce was different. It compressed data and had issues because of that. Comparing them is a bit of apples to oranges. They skimped on over provisioning with the excuse that because of the compression of data it was effectively over provisioned. I think part of the sandforce issue was the compression itself. That wasn't a great success.

May be off again. Issue was not the modem apparently.

geneo said:

Bree said:

...Your Samsung 840 has 3-bit MLC...

Click to expand...

MLC is almost exclusively used to refer to 2 bit/cell NAND....

Click to expand...

Hardly 'exclusive'. I see MLC in common use for all cases where 2 or more bits are stored per cell. The terminology I used was quoted direct from Samsung's own specs.

STORAGE MEMORY
Samsung V-NAND 4bit MLC

Click to expand...

STORAGE MEMORY
Samsung V-NAND 3bit MLC

Click to expand...

Crucial also use MLC to refer to their 3-bit devices.

Memory Type

Micron® 3D MLC NAND

Click to expand...

Bree said:

Hardly 'exclusive'. I see MLC in common use for all cases where 2 or more bits are stored per cell. The terminology I used was quoted direct from Samsung's own specs.

Samsung 870 QVO SATA SSD | Samsung Semiconductor Global

Samsung 870 QVO offers up to 8TB of storage capacity. Experience the highest-capacity computing with both speed and reliability-enhanced internal SSD.

www.samsung.com

Samsung 870 EVO SATA SSD | Samsung Semiconductor Global

Samsung 870 EVO achieves the maximum speed of SATA interface. Intelligent TurboWrite maintains long-term high performance. Available from 250GB to 4TB.

www.samsung.com

Crucial also use MLC to refer to their 3-bit devices.

Choose your BX300 SSD

uk.crucial.com

Click to expand...

Yes, it is is supposed to mean more than one level, but has fallen to use of meaning 2 level as that was the first multi-level SSD. Mystic even used it in this manner in his first post ( SLC > MLC > TLC -> QLC), so..

You will see TLC and QLC SSD often listed only as MLC in their specs as a subtle marketing trick - it either makes you think 2-bit/cell or at the most tells you nothing at all. Their specs should say TLC if they are 3 level. QLC if they are 4. Often they don't even include the bit level in the spec if it is a QLC (e.g. like WD blues). These SSD drive vendors can be very unscrupulous andlaunch an SSD with pretty good performance that are sent to reviewers and shipped to initial customers, then switch out parts for lower performance ones without changing the model name or anything, like the Western Digital switch from TLC to QLC on the blues:


They are not the only ones. And they launch them as MLC and let the reviewers uncover they are TLC (and later switch the parts to QLC!)


I am not making this up you know, I have a lot of experience in enterprise storage and follow the consumer market fairly closely.

geneo said:

But over-provisioning is simply creating unallocated space on the drive. It is not available to the OS, and it takes away space that would otherwise be available to partitions, so doesn't help there.

It is like a guardband the helps guarantee a level of performance and longevity at the cost of some space. But not necessary if you maintain equivalent free space on your partitions.

I should add that I beleive most SSD have some level of overprovisioning from the factory that is burned in (extra, un-provisionable capacity ).

Click to expand...

When I say that OverProvisioning (OP) is available to the OS, I meant that the unallocated space can be made use of by the OS (or the user), if required, by creating a partition using that space.

I will give you an example:

I have a T5 with 10 partitions, of which 8 are in APFS+/ HFS+ format, 1 in NTFS (which has Windows installed) and 1 in exFAT.

When I connect the T5 to my Windows machine, it only sees the NTFS and exFAT partitions. So if these partitions are near full, file operations will get hampered even if there is (plenty of) free space in the other 8 partitions. Efficiency of File operations is dependent on 2 factors: What the OS has access to and what the onboard micro-controller in the SSD has access to.

In this case, if I had made use of OP, I can reclaim that space (temporarily) to do what I want without hampering performance. This is something not possible if I don't have OP and I have all that free space allocated to partitions that the OS can't see.

OK. I obviously am not explaining this adequately. So let me try and go through the whole process. I may not have some things exactly right, but I think the gist is correct.

When you first initialize and SSD, all blocks on it are free and unused and are on a free list. The operating system knows how many logical blocks are on the drive. These blocks are not the same as the SSD blocks btw. The logical blocks are 512 Bytes in size while the blocks on the SSD are typically in the MB range. The SSD blocks contain atomic units called pages that are close to the size of logical blocks.

When you create partitions on the disk, SSD blocks are created to contain the partition table, and these blocks are allocated from the free list and are mapped to OS logical blocks. The partition table contains information on the partitions, including the logical block address (LBA) range. So far these blocks are the only ones allocated on the SSD, all the remaining blocks are free and not associated with any partition.

As data is written to a partition, SSD blocks are allocated from the free list for them and mapped to the OS LBAs.

At this point the only blocks that have been written are associated with logical blocks (and indirectly to the partition). All other blocks are unallocated and free for any partition to use - when something gets written to the partition, SSD blocks are allocated from the free list and mapped to the approiate LBA in that partition.

So say you have written only 10 SSD blocks of data to the HSF partition and nothing to any other partition you have created, then you dismount the disk and mount it on a Windows system. All blocks but those allocated to the partition table and the 10 blocks in HSF partitions are on the free list and are unallocated and not associated with any LBA or partitions.

When you mount the disk back in the Apple machine and the partition is mounted and you want to read back those blocks, the OS requests to read the LBA from the SSD, the SSD looks up the LBA in its mapping of LBA to SSD pages, and returns the data.

The SSD only knows about free blocks and allocated blocks and their mappings to OS LBA. When you create a partition, the operating system allocates a range of LBA for it, but the SSD does not allocate any blocks for the partition. This is not that much different than hard disk drives - when you create a partition for a HDD, the HDD doesn't actually allocate anything except space for the partition table entry. The difference between HDD and SSD is that the mapping between LBA and physical blocks on a hard drive is fixed and managed solely by the OS, the mapping between LBA and physical blocks on an SSD is dynamic and managed by the SSD - an extra level of indirection.

partition table:

ThrashZone said:

Hi,
I have mx100 that all are still at 99%
These are nearly 7 years old

Click to expand...

Then why are you so negative about QLC disks?

NavyLCDR said:

No OS can use unallocated space. I don't know where you got that idea. The OS can only access space in partitions.

Click to expand...

I meant to say that the unallocated space can be made available to the OS, and this where OP has an advantage over allocating free space to partitions using a filesystem that the OS can't read.

The-Hive said:

@TheMystic I fitted the Evo 970 2TB SSD in the enclosure and it is really fast, not up to thunderbolt speed but fast enough not to notice. Still it's not the same as an internal drive. What shall I do with all this free space
View attachment 15859

Click to expand...

Any special affection for SSDs?

Why did you buy one when you had so much free space sitting unused?

ThrashZone said:

Hi,
Partitions just overly complicate disks and you eventually realize they need resizing.
I just use a simple folder scheme pretty much why gpt has never been needed.

Click to expand...

They can be helpful too at times, especially during times of disk failures.

The-Hive said:

I am with the @TheMystic I love Samsung SSD's, in fact I like pretty much all Samsung stuff. especially their phones, I had an old J3, fantastic but I had to replace it with an S10 as the J3 wasn't compatible with my car and S10 works great with the My Phone app and a better one the Alienware Connect app.

Click to expand...

Unless you try out the competition, how would you know if you are missing out on something not available on Samsung products?

Samsung phones don't have one thing that is extremely important to me: Built-in Firewall!