The performance impact with a DRAM-less SSD

Currently reading
The performance impact with a DRAM-less SSD

DS620slim, DS415+
Operating system
  1. Linux
  2. macOS
  3. Windows
Mobile operating system
  1. Android
  2. iOS
Last edited:
Maybe interesting for some of you so I thought worth sharing.

I have two 3 ssd disk pools in my DS620slim.
The bigger 3.3 TB volume (slow moving files) is SHR, with 2 dirt cheap DRAM-less SSD 2TB and a 870 QVO 2TB.
The smaller volume consists of EVO and Sandisk SSD.

And although speed on this this bigger volume is not really important as the fast moving/frequently used data is on the other pool, I'd like to share a resource graph on the disks while on load, stressing the fact that there is no Dram cache.
I moved 50GB from the fast-SSD volume to the slow moving DRAM-less volume and see this:

Schermafbeelding 2023-04-28 om 11.31.18.png

So the first of the 2 DRAM-less disks (Drive 3) is 100% used during the full time of the move (15 minutes), and all other disks are virtually sleeping. A true bottleneck in performance.

Interesting to see the impact so clearly in the resource monitor, and also mainly on the first of the two DRAM disks.
Last edited:
Few months back I added 3rd party external ESATA enclosure to each of 3x NAS’s each stuffed with SSD. Using it as redundant (on 3xNAS) “long term storage”, the addition of which (and subsequent file removal from SHR pair) negated the need to purchase larger SHR drives, so it more than paid for itself.
I’m seeing what you indicate, but seeing that Synology suggests not to use SSD’s in their external ESATA enclosures… the interface might be limiting things, too. The NAS external ESATA is apparently slower than internal SATA. (I never bothered to try SSD in USBV3 to see if speeds were better in USBV3 vs ESATA. Being long term storage I can coexist with it being a bit slower).
Last edited:
That speed decrease Could be why you don’t see many expansion chassis out there. The single drive to ESATA chassis supporting 6Gb/s was $19, which places it solidly in the ‘Cost Effective’ column, with bandwidth far above what is presented to it.. For those with multiple NAS;s -- using them as redundant storage locations for long term data, on a budget.... This works
Today, swapped the 2x cheap no-DRAM SSD by 2x Transcend 2TD SSD with DRAM (these compare to MX500).
Again a 50GB folder with mixed contents. Copy time 2:20 minutes in stead of 15.
No clear bottleneck anymore. Also the load is not anymore mainly on the first of the SSD.
The samsung 870 QVO holds nicely.
What a difference does this DRAM make.

Scherm­afbeelding 2023-05-01 om 19.44.31.png

Also the other way around is ok (50Gb from the new 3.7TB to the "fast volume": A steady 120MB/s copy.
Scherm­afbeelding 2023-05-01 om 19.40.29.png
Interesting post. I'm surprised that the absence / presence of DRAM made such a difference. Is there a possibility that there was some other variable at play was the original 'Drive 3' close to its max TB write limit (SMART would confirm), or was the drive possibly un-TRIMMED and therefore experiencing slowdown due to Garbage Collection having to run during the test?

Is it possible to run smart & I/O tests on the former drive 3 now its out of the NAS? This may shed some further light on the issue...
DRAM and DRAM-less both have their place in the storage world. DRAM-less is often seen as the dead-end of SSD performance but DRAM has its own issues.

DRAM can add some latency and can have an odd interaction with the controller, multi-layer flash and small writes at low queue depths - which is where real performance is found, rather than daft performance tools that measure at incredible/impossible queue depths to produce the big numbers.

DRAM also introduces a hole where your data can be when the lights go out. This is bad news, especially for raid or indeed anywhere data is critical. For example, I would not use an SSD with DRAM in my NetGate router. If you buy a router from NetGate with a factory SSD you will find it has no DRAM either. Of course, you can buy SSDs with power-loss protection (usually enterprise grade) but it adds to the price and not all BIOSes and storage managers play well with it.

I did run the smart just before removing the A400 disks and they were 100% ok and on 98% of their estimated life time. Trim was not available on these disks as Kingston implements some mechanism to replace that according to specs.

Replacement was preventive, they are now over 3 years old (warranty period), an will be used in the off site DS415 as extra extra backup.
The samsung 850 and Sandisk ultra will reach 40000h soon, and are still going strong. Both are 2015.

I am not scared for DRAM data loss, the last power drop here was more than 12 years ago. We have excellent up time on electricity.

Create an account or login to comment

You must be a member in order to leave a comment

Create account

Create an account on our community. It's easy!

Log in

Already have an account? Log in here.

Welcome to! is an unofficial Synology forum for NAS owners and enthusiasts.

Registration is free, easy and fast!

Trending threads