VOGONS


Reply 20 of 26, by pentiumspeed

User metadata
Rank l33t
Rank
l33t

data center SSDs to consider: SATA 100GB intel S3700, 100GB Micron P400m and few others. First two are rated around 1PB written endurance. That's about 1000TB in writes.
Ditto to intel 710 100GB, micron P300 100GB but less endurance. All also have capacitors for power loss protection.

If you want much more capacity, P400m 400GB, Micron M500DC 480GB or 800GB, etc.

Cheers,

Great Northern aka Canada.

Reply 21 of 26, by darry

User metadata
Rank l33t++
Rank
l33t++
The Serpent Rider wrote on 2022-05-04, 21:38:
darry wrote:

Additionally, I don't see why the SD controller's integrated wear levelling functionality (AFAIU, all SD controllers have some and I would expect it has gotten better over the years) would not be effective in random (or more accurately in this case mixed random and sequential) write scenarios .

They are effective to some degree, yes. It's just that specific SD won't be any better than others, which are not rated for continuous linear write or other marketing shenanigans.
But that industrial grade SD from Kingston is basically just SLC flash (or TLC converted to SLC) and should be very robust for Windows 9x.

At least some of the high endurance ones are pMLC (TLC use as MLC) See https://ripitapart.com/2020/07/16/reverse-eng … microsdxc-card/ . Not sure I really need SLC level endurance .

darry wrote on 2022-02-10, 04:24:

EDIT: Based on the packaging of my Sandisk 128GB Max Endurance, it should be good for 60000 hours of recording at 1080p (which the packaging equates with 26Mbps), that means about 5355 Program/Erase cycles or 669 TBW . That being said, even if the 26Mbps is a misprint, as I doubt security cameras often record 1080p at much more than 5Mbps (h.264) or 2.5 Mbps (h.265), we would still be talking about over 500 Program/Erase cycles or over 60 TBW . That isn't bad at all in my book for an SD card .

Reply 22 of 26, by darry

User metadata
Rank l33t++
Rank
l33t++
pentiumspeed wrote on 2022-05-04, 22:24:
data center SSDs to consider: SATA 100GB intel S3700, 100GB Micron P400m and few others. First two are rated around 1PB wr […]
Show full quote

data center SSDs to consider: SATA 100GB intel S3700, 100GB Micron P400m and few others. First two are rated around 1PB written endurance. That's about 1000TB in writes.
Ditto to intel 710 100GB, micron P300 100GB but less endurance. All also have capacitors for power loss protection.

If you want much more capacity, P400m 400GB, Micron M500DC 480GB or 800GB, etc.

Cheers,

In a primarily DOS centered Pentium 2 laptop, I don't believe I need quite that much endurance, then again, that Sandisk microSD is no slouch on endurance if my math is correct (see my previous post).

Reply 23 of 26, by chiveicrook

User metadata
Rank Newbie
Rank
Newbie

I still think that cheap "smartphone" sdcards or any other cards with "A1" class are best suited as hdd replacements. Not only do they have guaranteed random r/w performance but also they are more likely to have wear leveling algorithms optimized for such use case.
Not sure how lazy (or more likely how limited by budgeting) are engineers in these companies but as an embedded software dev I would definitely tune such controllers to their intended use cases.

Reply 24 of 26, by The Serpent Rider

User metadata
Rank l33t
Rank
l33t

I still think that cheap "smartphone" sdcards or any other cards with "A1" class are best suited as hdd replacements.

Yes, buying a bunch of reasonably priced generic SD cards is also an option. Backup data and replace a card if it dies.

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 26 of 26, by darry

User metadata
Rank l33t++
Rank
l33t++
darry wrote on 2022-02-17, 06:24:
I tried and apparently succeeded in setting drive capacity of a 250GB Samsung 860 EVO to 128GB using [1] and unplugging/plugging […]
Show full quote

I tried and apparently succeeded in setting drive capacity of a 250GB Samsung 860 EVO to 128GB using [1] and unplugging/plugging the drive in a USB enclosure (apparently not recommended to use one of those) . Drive capacity went from [2] to [3] . I was also able to reverse the operation by setting sectors back to 488397168 and then redo it .

Unfortunately, I can't test this in my 4010CDT, as all the IDE to SATA adapters physically won't fit into the drive bay. I could work around this by using a 44-pin male to female IDE cable, but that would cost me about 30 CAN$ or more for a quick delivery or 20ish CAN$ if ordering with an expected arrival date some time in May . Neither of those options is very enticing for the moment .

I was, however, able to test a Max Endurance Sandisk 128GB card . First, here is a test in my Windows 10 PC's integrated SD card reader [4] .
And here are some test results in my Toshiba 4010CDT when running through an FC1307 based MicroSDXC to IDE adapter in CrystalDiskMark [5] and ATTO [6] .

[1]

sudo hdparm --yes-i-know-what-i-am-doing  -Np268435456  /dev/sdb

root@bob-VirtualBox:~# hdparm -g /dev/sdb

[2]

/dev/sdb:
geometry = 30401/255/63, sectors = 488397168, start = 0

[3]

root@bob-VirtualBox:~# hdparm -g /dev/sdb

/dev/sdb:
geometry = 16709/255/63, sectors = 268435456, start = 0

[4]
max_endurance.png

[5]
crystaldiskmark_4010cdt.png

[6]
atto_4010cdt.png

EDIT: Fixed a typo

The LBA48 capable BIOS (including the latest beta one) that manages the SIS963L based IDE controller on my Asus P4S800-MX does not like the manipulated amount of sectors on my test Samsung 860 EVO . The BIOS believes the drive is about 8GB in size, either through LBA or CHS, rather than 127GB (or its initial actual 250GB) . As a comparison, an unmodified 500GB 860 EVO is detected at full nominal capacity in LBA mode .

Interesting .