Total host writes > Total NAND writes why?

[RTXOutOfStockEdition]

CrystalDiskInfo

Total host writes 3518 GB

Total NAND writes 2035GB

 

Drive Intel 545S 512GB

1 partition 48% full, boot drive

 

Why is NAND writes lower than host writes?

[AbydosOne]

12 minutes ago, RTXOutOfStockEdition said:

Why is NAND writes lower than host writes?

What comes to mind is "compression", but I'd be surprised if it managed to compress almost half of what was written to it.

[keskparane]

Does CrystalDiskInfo have a help file?

[Wooden Law - Black]

This is caused by the write amplification (WA) and because the flash must be erased in order to write.

[RTXOutOfStockEdition]

-_-

[RTXOutOfStockEdition]

38 minutes ago, AbydosOne said:

What comes to mind is "compression", but I'd be surprised if it managed to compress almost half of what was written to it.

I looked it up and someone mentioned compression as well, but this is wild

 

21 minutes ago, Wooden Law - Black said:

This is caused by the write amplification (WA) and because the flash must be erased in order to write.

 

this is like reverse write amplification

[Wooden Law - Black]

1 hour ago, RTXOutOfStockEdition said:

I looked it up and someone mentioned compression as well, but this is wild

 

 

this is like reverse write amplification

Why? 
 

Can you post the CrystalDiskInfo screen?

[NewMaxx]

5 hours ago, RTXOutOfStockEdition said:

CrystalDiskInfo

Total host writes 3518 GB

Total NAND writes 2035GB

 

Drive Intel 545S 512GB

1 partition 48% full, boot drive

 

Why is NAND writes lower than host writes?

This just means you have a write amplification factor (WAF) of less than 1.0. As AbydosOne states, one cause of this is compression, although it's been a long time since we've seen that on consumer drives - SF-2281 based drives, specifically.

 

However, in your case, it's because that Intel 545s utilizes static SLC and only static SLC. Static SLC is in OP space, outside of user space, and is dedicated for the lifetime of the device. As such it has a separate garbage collection (GC) and wear-leveling zone from the native flash, which is different than how dynamic SLC operates. The wear on the drive is therefore the worse of two factors: the static SLC zone, and the native TLC zone. Writes will go to SLC first and if they do not end up being committed to TLC there is no effective wear. So it's possible to have more host than TLC NAND writes for a total WAF of <1.0.

 

If you use an appropriate program it should show SLC NAND writes as distinct from TLC NAND writes for this reason. The static SLC can handle far more program/erase cycles (factor of 10). So, for example, the drive endurance could be 30K P/E for SLC and 3K P/E for TLC, whichever comes first - however programs may be looking at the TLC writes which will be less than hosts in normal usage (since everything goes through SLC first, excepting sustained writes).

 

For the record this also applies to other drives such as those based on the Marvell 88SS1074 (WD Blue 3D, SanDisk Ultra 3D, et cetera) that have static-only SLC. Many cheaper drives using SMI controllers with Samsung's 64L TLC (overstock) also do this, possibly to improve endurance on that older flash. It's more common today to have dynamic or hybrid (static + dynamic) and dynamic flash shares a GC/wear zone with native TLC/QLC which actually can increase wear. If you want more information you can check out my subreddit, discord, etc.

[RTXOutOfStockEdition]

22 minutes ago, NewMaxx said:

This just means you have a write amplification factor (WAF) of less than 1.0. As AbydosOne states, one cause of this is compression, although it's been a long time since we've seen that on consumer drives - SF-2281 based drives, specifically.

 

However, in your case, it's because that Intel 545s utilizes static SLC and only static SLC. Static SLC is in OP space, outside of user space, and is dedicated for the lifetime of the device. As such it has a separate garbage collection (GC) and wear-leveling zone from the native flash, which is different than how dynamic SLC operates. The wear on the drive is therefore the worse of two factors: the static SLC zone, and the native TLC zone. Writes will go to SLC first and if they do not end up being committed to TLC there is no effective wear. So it's possible to have more host than TLC NAND writes for a total WAF of <1.0.

 

If you use an appropriate program it should show SLC NAND writes as distinct from TLC NAND writes for this reason. The static SLC can handle far more program/erase cycles (factor of 10). So, for example, the drive endurance could be 30K P/E for SLC and 3K P/E for TLC, whichever comes first - however programs may be looking at the TLC writes which will be less than hosts in normal usage (since everything goes through SLC first, excepting sustained writes).

Thank you. What program can read slc wear? 

[Wooden Law - Black]

7 hours ago, RTXOutOfStockEdition said:

Thank you. What program can read slc wear? 

CrystalDiskInfo, the SSD have to show the parameter. 

[NewMaxx]

14 hours ago, RTXOutOfStockEdition said:

Thank you. What program can read slc wear? 

In addition to CDI as mentioned in the reply above this one, there's also Hard Disk Sentinel (click the S.M.A.R.T. tab for the drive). I have a 545s I can check myself but it's not currently hooked up, however in the past I remember checking it on HDS. In any case, TLC in SLC mode is usually rated for 30K or more PEC. It's worth mentioning that static SLC is more robust than dynamic SLC because the latter comes from all of the native flash (rotating based on wear) while the former is carved from the top/best blocks (in terms of retention).

 

The exact mechanisms of wear can vary but modern controllers can adjust how they do writes to balance the two zones (static SLC + native/dynamic). In fact they can also reduce the static SLC size over time also, if the space is better used for error correction (ECC) for example. Although I suspect I'm veering into technical territory here that you don't care about. Be mindful that if the drive lists SLC writes it will be parsed as 32MB, or in GB, or in sectors (512B) most commonly. Also, of course, in hex by default but easily converted to decimal.

[RTXOutOfStockEdition]

9 hours ago, Wooden Law - Black said:

CrystalDiskInfo, the SSD have to show the parameter. 

 

3 hours ago, NewMaxx said:

In addition to CDI as mentioned in the reply above this one, there's also Hard Disk Sentinel (click the S.M.A.R.T. tab for the drive). I have a 545s I can check myself but it's not currently hooked up, however in the past I remember checking it on HDS. In any case, TLC in SLC mode is usually rated for 30K or more PEC. It's worth mentioning that static SLC is more robust than dynamic SLC because the latter comes from all of the native flash (rotating based on wear) while the former is carved from the top/best blocks (in terms of retention).

 

The exact mechanisms of wear can vary but modern controllers can adjust how they do writes to balance the two zones (static SLC + native/dynamic). In fact they can also reduce the static SLC size over time also, if the space is better used for error correction (ECC) for example. Although I suspect I'm veering into technical territory here that you don't care about. Be mindful that if the drive lists SLC writes it will be parsed as 32MB, or in GB, or in sectors (512B) most commonly. Also, of course, in hex by default but easily converted to decimal.

Thank you both of you for the answers. I don't mind learning the more technical aspect of the technology. Even if I don't understand, someone else will probably benefit from reading your post.