Three problems:
1) You ignored parts in the article you linked that you didn't like (or simply didn't read it to completion in a rush to prove your nonsense points, just like your claim that DRAM-less drives should be avoided due to corruption). For example:
Today, however, systems have capabilities such as data (or cache) pre-fetching that enable HMBs to work well and achieve low latency. Again, it’s a bit more of a hassle to get the same performance as an SSD with DRAM, but it’s certainly not impossible.
If you had read my posts correctly, you would have understood that my comments were regarding how DRAM cache on SSDs helps
when data is being written, not read.
Now be so kind please and explain how the hell would pre-fetching ever help with written data. And how.
Your linked article says nothing about using the DRAM as a cache FOR THE DATA ITSELF. Only that it is used as a cache when writing data and that said cache includes the the mapping table and metadata, exactly as I've pointed out numerous times. It is explained in that very article what the cache is used for:
I never said that DRAM is used for caching user data.
I only said it helps when data is being written. I never said that mapping table is not stored in DRAM.
I'm gonna extend my previous statement and say that DRAM helps with burst writes and sustained writes once (p)SLC gets exhausted, also random writes benefits are not negligible.
It is possible to see that DRAM-equipped drives have dramatically lower amount of speed fluctuations during sustained write test.
Phison explicitly states DRAM purpose for storing metadata and mapping table as another purpose for DRAM:
On an SSD, DRAM is used for several purposes: 1) as a cache when writing data to the drive, 2) as a storage location for indexed metadata and a digital map of where data is being stored across the SSD’s NAND flash memory cells, and 3) as a go-between that handles communication protocols between the SSD and the device’s processor.
There's cache
EDIT: Where the hell did part of my response go?!
Okay, once more ...
There's caching data when writing stated as purpose 1) excplicitly. I don't know what Phison is using DRAM for in purpose 1), but it is something different to purpose 2), what you are always pointing to.
It's up to SSD maker to use that drive's DRAM for any purpose they deem necessary. Of course, DRAM is
used for wear-leveling, mapping table, garbage collection, etc, but that does NOT mean it can't be used for anything else.
What are you even talking about, did you even read that review?
I actually did. That drive is doing really great for DRAM-less drive, but only thanks to huge (p)SLC cache. Look how that writing speed fluctuates.
Also, take a look at DRAM-less drive test when such huge cache was not present and compare it to DRAM-equipped drives at that time.
Crucial's BX500 is designed to bring blazing SSD throughput to your PC at a low price, but it has a few caveats.
www.tomshardware.com
Of course, today's drives are much more advanced than SATA drives back then. NVMe controller is much better at handling this situation now, no discussion about it.
However, this does not mean that the negative impacts of lacking DRAM are still not present in some way even today.
Writing speeds are much more stable, though they may be solid (high) on average as much as with SSDs with DRAM.
The Kingston KC3000 is built using the Phison E18 controller and Micron's best 176-layer TLC NAND flash. In our performance testing, the drive can beat the Samsung 980 Pro and is the fastest SSD we ever tested. It shares that performance throne with the WD Black SN850.
www.techpowerup.com
Samsung 990 Pro is the company's flagship M.2 NVMe SSD. Compared to the 980 Pro, it comes with an improved controller and more modern flash. Our review confirms: this is the fastest PCI-Express 4.0 drive you can buy right now, beating the competition from WD, Phison and Solidigm.
www.techpowerup.com
The Samsung 980 is the company's first attempt at releasing a DRAM-less SSD for the consumer market. While such designs usually are slow and have terrible random write rates, Samsung has found a way around that. Our Samsung 980 review confirms that this is one of the fastest SSDs you can buy.
www.techpowerup.com
Your logic makes zero sense, even if we assume DRAM is being used as a cache for the data itself (it isn't), how in the world does that help in a scenario where cache isn't allowed to recover? You'd fill up the 200 MB of free DRAM (Samsung 990 Pro 1 TB only has 1 GB of DRAM and most of that is going to be used for the mapping table and metadata) in tiny fraction of a second and then there would be zero benefit for sustain workloads and likely performance drops as the DRAM has to clear data out to make room for new metadata and mapping table changes. This is why the DRAM isn't used as a cache, it would be an extremely poor use of a highly value resrouce. In a sustained write scenario where cache isn't allowed to recover the write speed always reverts to the native speed of the flash once cache runs out. That bench is more a test of the controller and NAND, not of the differnces between DRAM and DRMA-less.
My logic says that any
non-volatile volatile memory is used basically for caching purposes. And there's a reason why SSD manufacturers still implement DRAM cache on premium SSD drives.
990 Pro 1TB has 1 GB RAM because there's a so-called ratio: 1 MB of RAM for every 1 GB of NAND storage.
Again, how does controller handle DRAM usage and in how many various ways it utilizes the DRAM is defined by firmware.
Going with your logic, I don't understand why SSD makers still make drives with DRAM. What's the purpose if DRAM-less are on the bar with DRAM-equipped?
DRAM is another piece of silicon on the PCB, thus increases costs. It would be wise but unethical move to silentnly remove DRAM with newer model and sell it at same price.
My opinion stays the same - all PC components should be as independent as possible from other PC components.
Because the less things for OS to manage, the better, especially for P'o'S Win11 so called OS.
I think you kind of mismatched my arguments with what iameatingjam was debating with you before.
Now, I'm over with this discussion in this thread, let's not f*ck it up any more. Dude should finally come to a decision on buying a drive.