just bought WD Blue sata ssd. Are sandisk nands reliable?

Kdawg

[H]ard|Gawd
Joined
Aug 12, 2017
Messages
1,116
does sandisk actually make their own nands, or is toshiba still making them?


I had some bestbuy gift cards I got for a 20% discount, and was deciding between 860 evo and wd blue

but bestbuy jacked up their prices these last few days. I was going to get a 500gb 'renewed' 860 evo for $65, but they've hiked it to $75.
So I just pricematched walmart and got the 500gb wd blue for $64.
 
Sandisk is owned by WD now, and they previously manufactured their own NAND as part of a joint partnership with Toshiba. I'm not sure if all of Sandisk's drives include their own NAND or if some of them include NAND from others, though.

Regardless, the drive should be fine!
 
Kioxia (what Toshiba rebranded to) and WD (Sandisk) have a joint partnership and collaborate on the development of their NAND. It's not "Kioxia developing and making NAND for WD" as commonly thought. WD has its own process engineers at the Kioxia fab site and both WD/Kioxia work on the development. I wouldn't have any trouble trusting WD NAND devices.
 

WD SSDs should be reliable (but not proven yet) the Blue SSD line are technically Dramless ish now, the custom sandisk controller that is inside all WD ssds has 10MB or so (they haven't disclosed the size) of built in ram witch they believe is enough to manage page table and wear leveling so shouldn't die randomly or slow down when they are filled up past 40-60%

the WD Green SSD is True Dramless and should be avoided there are better Dram based SSDs available for less than $-10 to 10 more (WD green does not have the same meaning on the SSD line as the HDD line did witch was quite and low power, the Green uses more power then the WD Blue SSD due to the fact it lacks a Dram cache and has to work harder)

the 860 EVO will be faster if benchmarked, would you notice it in real world use not likely
 
I'm going to take apart this blue ssd to see if there is a separate dram chip.

This drive has the slowest 4k q1t1 speed ever. Write speed is less than half of 860 evo.
Not sure where that makes a difference.

This blue ssd has a 5 year warranty though

according to review in 2017, there's a micron chip and marvell controller...



WD-Blue-3D-NAND-500GB-SSD-Review-on-KitGuru-Open.jpg
 
I'm going to take apart this blue ssd to see if there is a separate dram chip.

This drive has the slowest 4k q1t1 speed ever. Write speed is less than half of 860 evo.
Not sure where that makes a difference.

This blue ssd has a 5 year warranty though

according to review in 2017, there's a micron chip and marvell controller...



View attachment 240490
that picture does how a dram chip to the bottom left (i assume it's one you pulled from the internet)

sure they changed to a new version of the WD Blue SSD Sata that was dramless in 2019 (but looks like WD Blue NVME SSD that is Hybrid dramless only)

i wouldn't be surprised if it was dramless after the stunt they pulled with the WD red SMR "nas ready" hdds
 
well, took mine apart and maybe voided my warranty.

But the design did not change.

500gb wd blue ssd still has a marvell controller and 256mb dram
with 4x 128gb nands.
manufactured in March 2020.


According to WD, this drive does not support encryption.

But Marvell's product page says it supports 256 AES.

So did WD disable encryption in the firmware? Or do I maybe have encryption support?

https://www.marvell.com/products/storage/ssd/88ss1074.html


blue.jpg
 
Last edited:
Couple points for you:
  • The WD Blue 3D shares hardware with the SanDisk Ultra 3D: Marvell 88SS1074 (w/DRAM) and BiCS3 (64L TLC). This controller is well-known, used in the MX300/1100 among others.
  • DRAM chip is 256M16 = 256M x 16b = 512MB which is the normal 1B:1KB ratio. This ratio matches the expectation of 32-bit (4-byte) addressing for every 4KB sector/cluster.
  • Four NAND packages at 128GiB each ("128G"), 64L TLC is 256Gb/die so these are QDP (quad die package). Four-channel controller so 4x4 interleaving which is the maximum.
  • The WD Blue 3D has a small, static SLC cache, likely ~6.25GiB at this capacity, but very fast direct-to-TLC speeds with full interleaving.
  • Very few consumer drives actually support SED (self-encrypting drive) functionality even if the controller optionally does, there's always on-NAND encryption which is separate. It doesn't matter since SED functions are not trustworthy.
  • Your warranty is not voided under US law.
  • The WD Blue 3D is up there with the best SATA drives, in my opinion. Bit slower than the 860 EVO and MX500 but it's in the same ballpark. I own a few of these.
  • If you want to learn more, check my signature link.
 
WD SSDs should be reliable (but not proven yet) the Blue SSD line are technically Dramless ish now, the custom sandisk controller that is inside all WD ssds has 10MB or so (they haven't disclosed the size) of built in ram witch they believe is enough to manage page table and wear leveling so shouldn't die randomly or slow down when they are filled up past 40-60%

the WD Green SSD is True Dramless and should be avoided there are better Dram based SSDs available for less than $-10 to 10 more (WD green does not have the same meaning on the SSD line as the HDD line did witch was quite and low power, the Green uses more power then the WD Blue SSD due to the fact it lacks a Dram cache and has to work harder)

WD Blue 3D still has DRAM. The WD Blue SN500/SN550 NVMe drive is, however, DRAM-less. It should have ~32MB of SRAM, only a portion of which is used for mapping/metadata. It's enough on that drive because it has a powerful controller (scaled-down from the SN750's tri-core design), good BiCS4 (96L) flash, a fully static SLC cache (particularly good at 1TB), and high direct-to-TLC speeds (no folding state), in addition to the fact NVMe reduces latency and is more efficient in comparison to AHCI. It actually does very well when fuller, I have one that I tested up to 95% actually, again this is due to the overall design (check my SN550 posts at Reddit for more details).

The WD Green SSD is garbage. The DRAM-less SSD Plus (SanDisk) is variable, however the newest largest SKUs actually switched to the Marvell 88SS1074 (as on the Blue/SanDisk Ultra 3D) but with less DRAM. DRAM-less drives take less power technically since they don't have a DRAM chip, however under load they will eventually take the same or more power since it takes them longer to finish background management. This is one reason the SN550 is okay being DRAM-less, since the controller + NVMe protocol allows it to finish tasks faster and more efficiently. DRAM-less SATA drives should be considered at less than their marketed capacity because they tank in performance when fuller.
 
Couple points for you:
  • The WD Blue 3D shares hardware with the SanDisk Ultra 3D: Marvell 88SS1074 (w/DRAM) and BiCS3 (64L TLC). This controller is well-known, used in the MX300/1100 among others.
  • DRAM chip is 256M16 = 256M x 16b = 512MB which is the normal 1B:1KB ratio. This ratio matches the expectation of 32-bit (4-byte) addressing for every 4KB sector/cluster.
  • Four NAND packages at 128GiB each ("128G"), 64L TLC is 256Gb/die so these are QDP (quad die package). Four-channel controller so 4x4 interleaving which is the maximum.
  • The WD Blue 3D has a small, static SLC cache, likely ~6.25GiB at this capacity, but very fast direct-to-TLC speeds with full interleaving.
  • Very few consumer drives actually support SED (self-encrypting drive) functionality even if the controller optionally does, there's always on-NAND encryption which is separate. It doesn't matter since SED functions are not trustworthy.
  • Your warranty is not voided under US law.
  • The WD Blue 3D is up there with the best SATA drives, in my opinion. Bit slower than the 860 EVO and MX500 but it's in the same ballpark. I own a few of these.
  • If you want to learn more, check my signature link.



is there a tool I can use that tests slc cache size?


Lastly, the 4k Q1T1 speed of this drive are some of the worst of all SSDs ever.

I tested ~30 MB/s read, 60 MB/s write, which is inline with some reviews out there.
most other drives are 33% faster read, and 100% faster write with 4k Q1T1.

In what real world scenario would I be penalized by such low speed?


I've seen one person with wd blue ssd somehow achieve 100MB/s write on 4k q1t1, which is a little better, but still less than 120MB/s 4k q1t1 that other drives get.


I'm wondering if I should return this because of such low random 4k q1t1
 
4k 1qd reads are typically around 40mb/s (what your getting is to be expected (only Samsung are the king in 4K 1QD)

what does crystaldiskmark show (actual screen shot)

Toshiba XG3 NVME 1TB.png

even on a NVME 4K 1QD is not any better then a SATA SSD)
 
i currently have the drive removed
and i didn't save a screenshot of crystal 7.0.0, but the last line was a little less than ~30/60 like i said before.

To get an apples to apples comparison, I looked at this review....
https://www.kitguru.net/components/ssd-drives/simon-crisp/wd-blue-3d-nand-500gb-ssd-review/6/

AS SSD benchmark shows 40/120 at 4k

ASSSD-2.jpg


4k write is double what mine is, even though it's the same hardware, same Marvell controller. Same everything, according to their teardown pics.


kitguru tested on a 7700k.

my laptop is a 6700hq. Would that alone account for the big difference in 4k throughput?


my 970 evo+ nvme isn't much faster than that 7700k+WDBlue500g above....

fefefefef.png
 
Last edited:
So, if you know anything over the past couple of years, then you know on the "high end" there have been some ghastly mistakes. And one of those was from SanDIsk which caused an entire line of drives to stop working after about 4 years. This probably affected many, but made the news due to HPe's revelation about the problem (noting this is the 2nd of these in the past 5 years for HPe).

My point is that problems can happen to anyone, even with supposedly high end gear.

WD probably was not my first choice of "quality victor" for all storage. But, I do hope that they treat all their acquisitions with respect, learn from them, and not just look for how to rebrand using only their cheapest and possibly worst quality fabs.

Here's a capture (to file) CDM for my HP 512G Z-drive (1st ed.), SAMSUNG MZHPV512, underneath it all.

-----------------------------------------------------------------------
CrystalDiskMark 6.0.2 x64 (UWP) (C) 2007-2018 hiyohiyo
Crystal Dew World : https://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
* KB = 1000 bytes, KiB = 1024 bytes

Sequential Read (Q= 32,T= 1) : 2228.920 MB/s
Sequential Write (Q= 32,T= 1) : 1568.892 MB/s
Random Read 4KiB (Q= 8,T= 8) : 727.359 MB/s [ 177577.9 IOPS]
Random Write 4KiB (Q= 8,T= 8) : 424.905 MB/s [ 103736.6 IOPS]
Random Read 4KiB (Q= 32,T= 1) : 182.353 MB/s [ 44519.8 IOPS]
Random Write 4KiB (Q= 32,T= 1) : 157.491 MB/s [ 38450.0 IOPS]
Random Read 4KiB (Q= 1,T= 1) : 38.613 MB/s [ 9427.0 IOPS]
Random Write 4KiB (Q= 1,T= 1) : 108.403 MB/s [ 26465.6 IOPS]

Test : 1024 MiB [F: 0.0% (0.2/476.8 GiB)] (x5) [Interval=5 sec]
Date : 2019/08/06 22:21:27
OS : Windows 10 Professional [10.0 Build 17134] (x64)

Actually found a newer run... so adding this on. Same drive:

------------------------------------------------------------------------------
CrystalDiskMark 7.0.0 x64 (UWP) (C) 2007-2019 hiyohiyo
Crystal Dew World: https://crystalmark.info/
------------------------------------------------------------------------------
* MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
* KB = 1000 bytes, KiB = 1024 bytes

[Read]
Sequential 1MiB (Q= 8, T= 1): 2260.472 MB/s [ 2155.8 IOPS] < 3707.16 us>
Sequential 1MiB (Q= 1, T= 1): 1662.255 MB/s [ 1585.2 IOPS] < 630.20 us>
Random 4KiB (Q= 32, T=16): 731.024 MB/s [ 178472.7 IOPS] < 2865.78 us>
Random 4KiB (Q= 1, T= 1): 34.536 MB/s [ 8431.6 IOPS] < 118.06 us>

[Write]
Sequential 1MiB (Q= 8, T= 1): 1580.170 MB/s [ 1507.0 IOPS] < 5296.20 us>
Sequential 1MiB (Q= 1, T= 1): 1573.440 MB/s [ 1500.5 IOPS] < 665.29 us>
Random 4KiB (Q= 32, T=16): 425.229 MB/s [ 103815.7 IOPS] < 4925.45 us>
Random 4KiB (Q= 1, T= 1): 72.318 MB/s [ 17655.8 IOPS] < 56.20 us>

Profile: Default
Test: 1 GiB (x5) [Interval: 5 sec] <DefaultAffinity=DISABLED>
Date: 2020/02/23 16:06:06
OS: Windows 10 Professional [10.0 Build 18363] (x64)
 
there is nothing wrong with the SSD, its to be expected and different motherboards can to result in different 4K results (userbench is not completely reliable, but all the other ones are fine)

if you need high 4K read writes then get a current model Samsung (but really you only see random 4K read/write at boot up witch to a SSD is not really that hard compared to a HDD)

kitguru review of the WD has very high 4K write (higher than my NVME SSD) and its older version of Diskmark
 
Last edited:
is there a tool I can use that tests slc cache size?

I'm wondering if I should return this because of such low random 4k q1t1

Let me ask you one question upfront: does your drive say WD Blue 3D or WD Blue? Since the old version was 2D with planar TLC which is slower at 4K. It's possible you got the old version by mistake. I'm an expert on these drives and that sounds like what happened. The old drive used the same controller and therefore could have matching firmware technically.

As for SLC, it's easy to test if you have multiple SSDs just by copying a large file or something. There's a fast zone (SLC) and a slow zone (TLC). Although with the WD Blue 3D the base TLC is very fast not least due to it only having static SLC, so it's not a huge deal.
 
Let me ask you one question upfront: does your drive say WD Blue 3D or WD Blue? Since the old version was 2D with planar TLC which is slower at 4K. It's possible you got the old version by mistake. I'm an expert on these drives and that sounds like what happened. The old drive used the same controller and therefore could have matching firmware technically.

As for SLC, it's easy to test if you have multiple SSDs just by copying a large file or something.
yes he has the 2D version
the 2D nand version WDBNCE5000PNC (slower then 3D nand drives)
the WDS500G1B0A WDS500G1B0B seem to tbe the 3D nand versions (think it actually says 3D nand on the box and disk)
 
yes he has the 2D version
the 2D nand version WDBNCE5000PNC (slower then 3D nand drives)
the WDS500G1B0A WDS500G1B0B seem to tbe the 3D nand versions (think it actually says 3D nand on the box and disk)

Yeah, I caught that too. Seems to be the case and explanation.
 
hold up

where are you getting this is 2D?

It says 3d nand on the case and box.

I would be very surprised if WD was getting rid of their old 2D stock with false advertising. This drive is only sold at Bestbuy and Walmart.
and the mfgr date was last month!

IMG_20200427_010024.jpg

IMG_20200425_143403.jpg



I just ordered a 860 EVO arriving tomorrow, so I'll test that too. And if I get shit 4k speeds with samsung, welp, I guess it's my shit laptop motherboard

Asus called this a gaming laptop, but drives suffer 25-30% speed penalty.
 
Last edited:
"WDBNCE5000PNC" was also used for the 2D/planar Blue, hence the confusion. Although your pictures from before match the 3D internally. It was worth double-checking.

Might very well be the laptop. I know my older Ivy Bridge laptop runs SATA drives at sub-par levels for 4K, for example. I have a Lexar NS200 in it (basically a Crucial MX500) and it doesn't hit what you find on review sites, etc.
 
haha,

looks like new wd blue is just a shit drive.

Tested new 860 evo and got at least 50% faster on 4k q1t1.

mehh... now that I've compared Anvil benches, I can see where the WD is slower, even in sequential reads.

WDC WDBNCE5000PNC_500GB_1GB-20200428-1718.png

Samsung SSD 860 EVO 500GB_500GB_1GB-20200428-1830.png




I wonder if I would get the same shitty 4k results if I got another WD Blue from bestbuy.
I have a feeling every blue drive in this batch has such shit results.

860a.png
 
Last edited:
For some reason, makes no difference if meltdown/spectre are fixed or not....


wdbluecdm2.png


Definitely going to exchange this ssd for another sample.


meanwhile, 860evo faster than ever, but oddly also not affected by meltdown/spectre patches...

860evocdm3.png



Interesting results with a Ryzen laptop.

squeezed out a little more 4k performance from ryzen with the WD BLUE
but ryzen cripples ssd in other metrics.

Untitled1.png
 
Last edited:
the 850 onwards samsung ssds the 4k results generally beat most other ssd manufacturers, really you won't notice unless your doing benchmarks

if you're comparing it to a HDD there is no comparison even a Dramless SSDs or even QLC based SSDs tend to smoke HDDs (unless you Write a lot to a QLC SSD then they can drop to 80mb/s on writes until they have been left alone for 10 minutes to flush the SLC cache out)
 
if you're comparing it to a HDD there is no comparison even a Dramless SSDs or even QLC based SSDs tend to smoke HDDs (unless you Write a lot to a QLC SSD then they can drop to 80mb/s on writes until they have been left alone for 10 minutes to flush the SLC cache out)

Yes, indeed, there are cases for some SSDs where performance is way way way down. And even less than spinny disk for bigger writes. Just a thing.

Another disadvantage that isn't discussed much is idle shelf life. You can pick up that ancient HD from 5 years back and get to your data. Things get very dicey after two years on a shelf for SSD.
 
Yes, indeed, there are cases for some SSDs where performance is way way way down. And even less than spinny disk for bigger writes. Just a thing.

Another disadvantage that isn't discussed much is idle shelf life. You can pick up that ancient HD from 5 years back and get to your data. Things get very dicey after two years on a shelf for SSD.


ssd data just disappears if you don't power up the drive for a long time?

I've had usb flash drives with data on them that hasn't been accessed for years, still accessible. is NAND different?
 
NAND will lose data through leakage when stored. JEDEC's official tested requirement is 52 weeks (1 year). As the voltage changes the value gets closer to boundaries and thus harder to read, requiring more and more levels of ECC (LDPC soft sensing). It's possible to correct some bits through parity/RAID (spare area of each page) but eventually it corrupts.

Older USB drives, and I actually have one at 128MB (that's not a typo), used SLC and MLC back in the day which is of course less prone to this. Also, flash drives tend to run hotter and you store them cooler than JEDEC's rating (40C/30C respectively) which could be multiple years.
 
i powered up a old samsung SSD it seemed to still be working fine, Temperature matters with the way they can lose data over time but that's what the ECC is there to handle voltage drift

you have to leave one alone for 2 and 3 years to see what happens (i could guess QLC could have shorter off life as it has multi voltage states for part of the NAND so probably wouldn't take much to make a area unreadable)
 
Back
Top