Announcing the TechPowerUp SSD Specs Database

[Wirko]

Thank you.
I'll see what i can do, it's really difficult to get detailed information on these older planar NANDs
One of the issue is some sites claim it's Samsung 30nm and others claim 32nm lol

There's a Samsung product selection guide from 2H2011, found on the archive.org, that has a bit of info too. The part number matches what I posted before - but it's for the 9.5 mm thick version, the thinner 7 mm version has a different part number. Note that the SSD drive called PM810 exists too, so it could be the very same thing as the 470.

Wayback Machine

The drive:

The NAND package:

And hey, Samsung had "3bit MLC" chips in 2011, that must be the thing that much later became known as TLC!

There's something interesting about the S3C29MAX01-Y340 controller too. As it appears, it was sold separately as a general purpose microcontroller, I found it here, with no mention of SSD applications.

 

[GabrielLP14]

And hey, Samsung had "3bit MLC" chips in 2011, that must be the thing that much later became known as TLC!

3-bit MLC is TLC, they just try to trick people, because MLC is "Multi" Level Cell, so in theory if it has more than 2 it can be called MLC, if in the future they release a PLC die (5 bits ber cell) they'll definitely call it 5-bit MLC

There's something interesting about the S3C29MAX01-Y340 controller too. As it appears, it was sold separately as a general purpose microcontroller, I found it here, with no mention of SSD applications.

That's NEAT

There's a Samsung product selection guide from 2H2011, found on the archive.org, that has a bit of info too. The part number matches what I posted before - but it's for the 9.5 mm thick version, the thinner 7 mm version has a different part number. Note that the SSD drive called PM810 exists too, so it could be the very same thing as the 470.

I'll try to add these as well

hey @Wirko Check it out.

Sci-Hub | A 159mm2 32nm 32Gb MLC NAND-flash memory with 200MB/s asynchronous DDR interface. 2010 IEEE International Solid-State Circuits Conference - (ISSCC) | 10.1109/ISSCC.2010.5433912

The following picture is a Samsung MLC 32Gb 32nm Die from Samsung, the ones used in the Samsung 470 series

 

[Wirko]

3-bit MLC is TLC, they just try to trick people, because MLC is "Multi" Level Cell, so in theory if it has more than 2 it can be called MLC, if in the future they release a PLC die (5 bits ber cell) they'll definitely call it 5-bit MLC

hey @Wirko Check it out.

Sci-Hub | A 159mm2 32nm 32Gb MLC NAND-flash memory with 200MB/s asynchronous DDR interface. 2010 IEEE International Solid-State Circuits Conference - (ISSCC) | 10.1109/ISSCC.2010.5433912

The following picture is a Samsung MLC 32Gb 32nm Die from Samsung, the ones used in the Samsung 470 series
View attachment 272658

That's a lot of data in there.
I see just a small error in the database, the block size is 1024 KB or 128 pages, not 1024 pages.

Those old chips, they had fully binary organisation, there was 2^n of everything. It's quite different in new ones.

 

[GabrielLP14]

That's a lot of data in there.
I see just a small error in the database, the block size is 1024 KB or 128 pages, not 1024 pages.

Those old chips, they had fully binary organisation, there was 2^n of everything. It's quite different in new ones.

oh great catch thanks
Corrected

It actually much harder to find data on these older dies then newer ones lol

 

[Mussels]

Oh no, my WD SN730 isnt listed


*Hysterical crying*

 

[GabrielLP14]

Oh no, my WD SN730 isnt listed


*Hysterical crying*

I'll try adding this line-up, to be honest i didn't even remember them hahaha

Oh no, my WD SN730 isnt listed


*Hysterical crying*

i have added it, unfortunately i couldn't find much

Western Digital SN730 1 TB Specs

1 TB, TLC, PCIe 3.0 x4, M.2 2280

www.techpowerup.com

 

[Mussels]

I'll try adding this line-up, to be honest i didn't even remember them hahaha


i have added it, unfortunately i couldn't find much

Western Digital SN730 1 TB Specs

1 TB, TLC, PCIe 3.0 x4, M.2 2280

www.techpowerup.com

They're an OEM model - Mine came from a batch of failed AN1500 RAID cards, I kept getting sent 1TB models (2x512) instead of 4TB (2x2TB) and in the end took the SSD's out and ditched the overheating cards, after they let me keep them despite refunding me
Ended up with 6 of the bloody things, good thing they're somewhat decent


What we need now is an SSD chart like the GPU one, showing loose performance comparisons...


I'll accept donations of one of every SSD ever, and work my way through if needed

 

[GabrielLP14]

They're an OEM model - Mine came from a batch of failed AN1500 RAID cards, I kept getting sent 1TB models (2x512) instead of 4TB (2x2TB) and in the end took the SSD's out and ditched the overheating cards, after they let me keep them despite refunding me
Ended up with 6 of the bloody things, good thing they're somewhat decent


What we need now is an SSD chart like the GPU one, showing loose performance comparisons...


I'll accept donations of one of every SSD ever, and work my way through if needed

Oh yeah, i noticed, i read some reviews of these AIC AN1500 and saw these SN730 there. Fortunately, they are indeed good drives.
It might be tough to do it since SSDs are more susceptible to have performance impact if you change the testing platform like Intel x AMD, and even if we use our own testing data which W1zzard has over the years, it would still be missing literally hundreds of drives.

 

[Wirko]

Oh no, my WD SN730 isnt listed


*Hysterical crying*

The database won't be complete until my 80 GB Intel X25-M G2 planar MLC drive makes an appearance. Maybe I can send the VLO ... if the utility runs on Windows XP.

 

[GabrielLP14]

The database won't be complete until my 80 GB Intel X25-M G2 planar MLC drive makes an appearance. Maybe I can send the VLO ... if the utility runs on Windows XP.

yes it runs ahahaha, i've never heard of this drive before hahaha
The issue is, if the controller is proprietary you won't be able to run it.
Only controllers like, Phison, Silicon Motion, Innogrit, Maxiotech etc

 

[_JP_]

i have added it, unfortunately i couldn't find much

Would this this doc help?

 

[GabrielLP14]

Would this this doc help?

Holy crap, it does help a lot indeed thanks man ahahaha.

Would this this doc help?

 

[_JP_]

The thing with OEM drives is that they usually have really detailed documentation by my experience, due to the number of specialized applications other manufacturers might require, but those are usually hard to find, or behind a partner portal.
With the exception of the latter, because NDAs likely prohibit, it takes a bit more time to find the right stuff.
I'm glad I could help!

 

[GabrielLP14]

The thing with OEM drives is that they usually have really detailed documentation by my experience, due to the number of specialized applications other manufacturers might require, but those are usually hard to find, or behind a partner portal.
With the exception of the latter, because NDAs likely prohibit, it takes a bit more time to find the right stuff.
I'm glad I could help!

Makes sense indeed
If you find any other feel free to mention me

 

[Mussels]

Would this this doc help?

The details!


I love how WD hide certain stats on some drives, try finding the TBW on their kamikazee drives aka the WD greens

"1 million hours" and a few asterisks

SSD-Z name was already tried by someone in the past, but we need a TPU-Z program that can read these details like GPU-Z does and upload specs to be verified

It could be like userbench, without being shit!

 

[_JP_]

Makes sense indeed
If you find any other feel free to mention me

Will do!
But you can also reach-out to me if you're having a hard time finding more info on a drive, just mention the model and I'll see what I can do.

I love how WD hide certain stats on some drives, try finding the TBW on their kamikazee drives aka the WD greens

"1 million hours" and a few asterisks

You'd think that they actually test that thoroughly?

 

[GabrielLP14]

Will do!
But you can also reach-out to me if you're having a hard time finding more info on a drive, just mention the model and I'll see what I can do.

You'd think that they actually test that thoroughly?

sure, right now i'm trying to find a doc sheet on the Micron MLC 32Gb 34nm Die so i can add that ancient Intel SSD our friend suggested

 

[_JP_]

Hm, that info requires registration at Micron, but I'd say it's doable. I'll see what I can do.

 

[GabrielLP14]

Hm, that info requires registration at Micron, but I'd say it's doable. I'll see what I can do.

no need just try to find on google scholar, its free and no registration required, there you can find these docs but i cant find it, for example, check it out

 

[Wirko]

Thanks a lot, fellow archeologists.

I see that the same Intel's controler was used in several other SSDs, including enterprise models, so there's a high probability that the same NAND dies went into those too.

The controller is Intel PC29AS21BA0, at least in the 160 GB model, and it's a 10-channel controller. A funny model number I must say, since "29" was reserved for EEPROM chips and all their ilk, including NAND, sincer ever at Intel. Just like 27 was EPROM, 82 were/are peripherals, and we all know 80.

Edit: not entirely true, the 320 (consumer) and 710 (enterprise) SSDs had the same controller but 25 nm NAND.

 

[GabrielLP14]

The database won't be complete until my 80 GB Intel X25-M G2 planar MLC drive makes an appearance. Maybe I can send the VLO ... if the utility runs on Windows XP.

Hey buddy great news for you.
Look at what i found... hahah

Since i now have more in depth data from these dies, i can surely add the Intel X25-M G2 with more reliable data.

Thanks a lot, fellow archeologists.

I see that the same Intel's controler was used in several other SSDs, including enterprise models, so there's a high probability that the same NAND dies went into those too.

The controller is Intel PC29AS21BA0, at least in the 160 GB model, and it's a 10-channel controller. A funny model number I must say, since "29" was reserved for EEPROM chips and all their ilk, including NAND, sincer ever at Intel. Just like 27 was EPROM, 82 were/are peripherals, and we all know 80.

Edit: not entirely true, the 320 (consumer) and 710 (enterprise) SSDs had the same controller but 25 nm NAND.

i'll try to find more data on the controller now and then move to add them to the database.

 

[Wirko]

Another bit of archeological info: how Intel increased the write endurance for their enterprise SSDs.

He's probably referring to the same 25nm type of die that's also found in the Intel 320 (consumer) and 710 (enterprise) SSDs, also analysed and compared here by Tom's Hardware. HET means High Endurance Technology, which means whatever Intel tells us or doesn't tell us. They had to reveal at least some info because back then it was necessary to convince enterprise users that MLC can be trusted.

As it appears, the 710 has better binned dies, writes at lower speeds (so it can charge the cells with more precision), and also writes at lower voltages. On top of that, it has an incredible amount of overprovisioning: 320 GB raw, 200 GB available. All that was supposed to result in 33x more rewrite cycles than in the consumer-oriented 320.

I believe SSD manufacturers still have to make similar compromises today: higher write speed and lower endurance, or the other way around.

 

[GabrielLP14]

Another bit of archeological info: how Intel increased the write endurance for their enterprise SSDs.
View attachment 273292
He's probably referring to the same 25nm type of die that's also found in the Intel 320 (consumer) and 710 (enterprise) SSDs, also analysed and compared here by Tom's Hardware. HET means High Endurance Technology, which means whatever Intel tells us or doesn't tell us. They had to reveal at least some info because back then it was necessary to convince enterprise users that MLC can be trusted.

As it appears, the 710 has better binned dies, writes at lower speeds (so it can charge the cells with more precision), and also writes at lower voltages. On top of that, it has an incredible amount of overprovisioning: 320 GB raw, 200 GB available. All that was supposed to result in 33x more rewrite cycles than in the consumer-oriented 320.

I believe SSD manufacturers still have to make similar compromises today: higher write speed and lower endurance, or the other way around.

Quite interesting i'll be sure adding that as well in the near future

I believe SSD manufacturers still have to make similar compromises today: higher write speed and lower endurance, or the other way around.

Not always though, Samsung, SK Hynix, Kioxia have capped their die throughput in some SSDs to have higher endurance.
For example, get the Samsung 990 Pro, the 2TB have 32 dies of Samsung V7 512Gb 176-Layers, these dies in a day to day scenario can beat 150-160MB/s of write throughput. So 32 dies x 150 MB/s = ~4.800 MB/s
But as you can see in the graph below, the Drive writes at an average of 1.4 GB/s
They did that to not only increase endurance (probably) but to generate less heat output and a higher efficiency (MBps / watt)

 

[Mussels]

3-12 months of data retention is still scary AF, I power up my laptop like twice a year

 

[_JP_]

3-12 months of guarantee that there will not be rotting bits (You'd be surprised how many people didn't know this was a thing for USB pens since those showed-up, where the guarantee was in hours to weeks, at best).
It doesn't mean that after that time your data simply becomes unrecoverable, there's still ECC.
Really comes down to how the PCB is done, but regardless of that, it's worse for higher layered NAND.
Too bad SLC/MLC SSDs for lower capacities aren't a thing anymore, besides halo products.
The Transcend 370S was the last one I could find "widely available", being MLC, back in 2018. And I say it this way because I could find them at a few stores, but those had little stock and had them costing a small fortune for 512GB.