Announcing the TechPowerUp SSD Specs Database

[xlen]

Welp this just killed the need for my database, especially if you will maintain it.

 

[Panyamin]

Amazing! Thanks!

 

[Wirko]

Hahaha i understand, and yes i also review SSDs but in my YouTube Channel in Brazil, i started a GOOGLE SHEETs database in 2019 then i met Wizzard in 2021 and started this project.

Oh no, it's not what I meant, sorry. I photoshopped (actually firefoxed) the list of reviews and added the red icon for TPU's reviews to stick out, that's it.

 

[Nuke Dukem]

I actually came upon the new DB by accident a few days ago when researching model specs and a search engine listed it as the top result. Good stuff

 

[konga]

This is excellent. Thank you Gabriel for all the hard work.

How will you handle drives that have had component swaps? Like when a manufacturer replaces the flash with a different type without changing the drive's model name?

 

[GabrielLP14]

Oh no, it's not what I meant, sorry. I photoshopped (actually firefoxed) the list of reviews and added the red icon for TPU's reviews to stick out, that's it.

ahhahah no prob, i noticed that, that would be cool to do but tbh i have no idea how to implement that yet

This is excellent. Thank you Gabriel for all the hard work.

How will you handle drives that have had component swaps? Like when a manufacturer replaces the flash with a different type without changing the drive's model name?

i'll do this, search SX8200 Pro

And yes so far i've found 12 variants from this drive alone hahahah yes you read correctly, not 4 not 5.... 12

 

[MagnuTron]

You guys rock...

 

[GabrielLP14]

You guys rock...

Thanks

 

[Pan]

Awesome!

Researching SSD's can be a pain in the butt! This will help a lot!

 

[Mussels]

Every suggestion i had to make was already implemented, looks good

i met Wizzard in 2021

I'm still trying to lure him down-under, but customs gets suspicious of people with suitcases of GPU's at the airport

 

[taiiat]

wonderful! i'm sure the Internet Community will appreciate the collated Data much how we appreciate the many other highly valuable Services which you provide. thanks, as always.

one thing i can think of though, hmm, perhaps the performance Numbers could express both officially claimed by the Manufacturer, as well as actual if there's sufficient Review data. like a sanity check for the Marketing.
could be something as simple as say....

Sequential Read: 7000 MB/s (Mfg. / official)
__________________6325 MB/s (tested)

and so on for the rest of the performance data. and any other data that has been tested, for that matter, because why not, right. claimed vs observed actual.

 

[b1k3rdude]

@GabrielLP14

The issue is the these IOPS are afaik based on 4K block size at a Queue Depoth of 32, the latter of which is complete bollox for the average consumer.

For example the Samsung 990 pro 1TB looks fantastic with IOPS of 1,200,000/1,550,000 r/w, but the actual IOPS at 4kQD1 is 22000/88000 (90/360MB/s) r/w - https://www.samsung.com/uk/business/memory-storage/nvme-ssd/990-pro-2tb-nvme-pcie-gen-4-mz-v9p2t0bw/

For day to day use the 990 pro isnt much faster than my old 960 pro @ 4kQD1 of 14000/5000 (on avg. I get 14600/63000 - 60/250MB/s). And the 960 dosent have the drop of in seqeuntial r/w after a few GB has been transfered, unlike the 990.

On a related not found this little site that is usefull for converting IOPS to MB/s - https://wintelguy.com/iops-mbs-gbday-calc.pl

 

[Wirko]

@GabrielLP14

"Are the IOPS figures in the DB based on the 4K block size? "

It is what manufacturers publish as "up to". Probably 4K sequential and with a deep queue because that's how every SSD can have a breathtakingly high IOPS figure.

 

[b1k3rdude]

It is what manufacturers publish as "up to". Probably 4K sequential and with a deep queue because that's how every SSD can have a breathtakingly high IOPS figure.

I updated my post above, its actually 4kQD32, which as I commented above is bollox for the avg consumer.

The 4kQD1 IOPS data is available, so @GabrielLP14 can this be added to the DB..?

 

[W1zzard]

and added the red icon for TPU's reviews to stick out, that's it.

That's actually a fantastic idea that I can use in a couple of other places

"w/ Heatsink" means with heatsink? maybe it's some native language English think, but couldn't it also mean without?

"w/" means "with", and "w/o" means "without" .. i think .. any native speakers?

 

[GabrielLP14]

wonderful! i'm sure the Internet Community will appreciate the collated Data much how we appreciate the many other highly valuable Services which you provide. thanks, as always.

one thing i can think of though, hmm, perhaps the performance Numbers could express both officially claimed by the Manufacturer, as well as actual if there's sufficient Review data. like a sanity check for the Marketing.
could be something as simple as say....

and so on for the rest of the performance data. and any other data that has been tested, for that matter, because why not, right. claimed vs observed actual.

its a good idea, i'll see what we can do @W1zzard

@GabrielLP14

The issue is the these IOPS are afaik based on 4K block size at a Queue Depoth of 32, the latter of which is complete bollox for the average consumer.

For example the Samsung 990 pro 1TB looks fantastic with IOPS of 1,200,000/1,550,000 r/w, but the actual IOPS at 4kQD1 is 22000/88000 (90/360MB/s) r/w - https://www.samsung.com/uk/business/memory-storage/nvme-ssd/990-pro-2tb-nvme-pcie-gen-4-mz-v9p2t0bw/

For day to day use the 990 pro isnt much faster than my old 960 pro @ 4kQD1 of 14000/5000 (on avg. I get 14600/63000 - 60/250MB/s). And the 960 dosent have the drop of in seqeuntial r/w after a few GB has been transfered, unlike the 990.

On a related not found this little site that is usefull for converting IOPS to MB/s - https://wintelguy.com/iops-mbs-gbday-calc.pl

Agreed completely, the issue is how the SSDs are tested, as i said previously if we change even the OS build numbers will change, which will cause issues. For example in TPU we use a AMD test system, but other websites uses an Intel. Each data in some scenarios do have a reasonable difference.
So its complicated to compile actual data. We could use our own, but bear in mind that the DB has 1200 SSDs, and TPU has tested what? 300-400 or 500 drives? See the issue? i'd love to add these numbers though the issue is that even the manufacturers don't even list.
Samsung, and Solidigm an Intel(in some cases) do advertise it though.

It is what manufacturers publish as "up to". Probably 4K sequential and with a deep queue because that's how every SSD can have a breathtakingly high IOPS figure.

Yes, the issues lies where i've replied above :/

 

[_JP_]

Congratulations on this awesome project @GabrielLP14 @W1zzard !! (for the lack of a better emoji)

Is there going to be a banner saying "Caveat Emptor Alert: Not the original configuration!" for the SSDs that become bait-and-switch?
Second question, is "Latest Known Firmware" a possible field for the entries?
Given that SSDs usually benefit from them and in the case of troublesome releases, to give users a way to understand if they are being impacted by an outdated FW.
Last question (until I remember more), link to the respective manufacturer's tool for managing it...possible?

 

[GabrielLP14]

- Samsung PM961 (Polaris, Samsung TLC)
- TeamGroup Cardea Z330 (RTS5763DL, SK Hynix 128L TLC according to VLO)
- TeamGroup MP33 Pro (RTS5765DL, SK Hynix 128L TLC)
- WD SN500 (WD Proprietary, SanDisk 64L TLC)

I will add more if I think of any

PM961 and SN500 added, will add today these others, thanks again
By the way, the PM961 review i found were using V3 MLC dies (48-Layers 256Gb)

 

[Wirko]

one thing i can think of though, hmm, perhaps the performance Numbers could express both officially claimed by the Manufacturer, as well as actual if there's sufficient Review data. like a sanity check for the Marketing.

That's ... peeking down the rabbit hole. Review results are a big bunch of numbers, some matter to some people, others matter to others. We often need graphs to present information clearly. Which transfer speed would you include in the database? The maximum of all measurements? That one is pretty much unimportant in use, most of the time.

 

[GabrielLP14]

That's ... peeking down the rabbit hole. Review results are a big bunch of numbers, some matter to some people, others matter to others. We often need graphs to present information clearly. Which transfer speed would you include in the database? The maximum of all measurements? That one is pretty much unimportant in use, most of the time.

unfortunately thats the case, if all reviews in the world had the same testing schemes and test systems we could do it more easily :/

- Samsung PM961 (Polaris, Samsung TLC)
- TeamGroup Cardea Z330 (RTS5763DL, SK Hynix 128L TLC according to VLO)
- TeamGroup MP33 Pro (RTS5765DL, SK Hynix 128L TLC)
- WD SN500 (WD Proprietary, SanDisk 64L TLC)

I will add more if I think of any

Can you provide me the link for these Teamgroups, if they are yours and you ran the VLO, pls send me the flash id.txt

 

[Selaya]

okay. suggestion time.

add the string as reported by/to the OS (viewable w/ something like CDI or HWInfo or w/e) somewhere prominent.
this is especially useful as to identifying different models sold as the same SKU (aka bait and switch) since sometimes the vendor actually doesnt bother to fix those.

also the firmware version(s) corresponding to the different models. another thing vendors tend to not fix (if that is possible indeed).
(see screenshot below)

 

[GabrielLP14]

okay. suggestion time.

add the string as reported by/to the OS (viewable w/ something like CDI or HWInfo or w/e) somewhere prominent.
this is especially useful as to identifying different models sold as the same SKU (aka bait and switch) since sometimes the vendor actually doesnt bother to fix those.

also the firmware version(s) corresponding to the different models. another thing vendors tend to not fix (if that is possible indeed).
(see screenshot below)

As for the part number / SKU we already have listed:

As for the firmware revision, unfortunately some reviews don't even list it, for example, TPU does, but like tom's Hardware which is one i consider one of the best 3 top SSDs reviews in the planet doesn't

 

[Selaya]

yeah, hence the prominent part of it (i did in fact miss it) - i'd consider that to be more useful information than the advertised name tbh but yea

as for the firmware, just list those which are known ig? and ask people to submit the missing information - sure it'll take time but thats what building dbs is all about innit

 

[GabrielLP14]

yeah, hence the prominent part of it (i did in fact miss it) - i'd consider that to be more useful information than the advertised name tbh but yea

as for the firmware, just list those which are known ig? and ask people to submit the missing information - sure it'll take time but thats what building dbs is all about innit

it's not a bad idea at all
i'll just finish some adjustments and i can implement that in the very close near future, since there are literally over 1200 SSDs

 

[MachineLearning]

unfortunately thats the case, if all reviews in the world had the same testing schemes and test systems we could do it more easily :/


Can you provide me the link for these Teamgroups, if they are yours and you ran the VLO, pls send me the flash id.txt

MP33 Pro review, with a different controller (a much better one actually).

This is the VLO readout of mine:

v0.15a
OS: 10.0 build 19044
Drive : 2(NVME)
Scsi : 2
Driver : W10
Model : TEAM TM8FPD002T
Fw : VC0S036G
HMB : 32768 - 65536 KB (Enabled, 64 M)
Size : 1953514 MB [2048.4 GB]
LBA Size: 512
Fw Str : [REALTEK_RL6577 _p_tH3V6V] []
Bank00: 0xad,0x89,0x28,0x53,0x0,0xb0,0x0,0x0 - Hynix 3dv6-128L TLC 16k 1024Gb/CE 1024Gb/die
Bank01: 0xad,0x89,0x28,0x53,0x0,0xb0,0x0,0x0 - Hynix 3dv6-128L TLC 16k 1024Gb/CE 1024Gb/die
Bank02: 0xad,0x89,0x28,0x53,0x0,0xb0,0x0,0x0 - Hynix 3dv6-128L TLC 16k 1024Gb/CE 1024Gb/die
Bank03: 0xad,0x89,0x28,0x53,0x0,0xb0,0x0,0x0 - Hynix 3dv6-128L TLC 16k 1024Gb/CE 1024Gb/die
Bank04: 0xad,0x89,0x28,0x53,0x0,0xb0,0x0,0x0 - Hynix 3dv6-128L TLC 16k 1024Gb/CE 1024Gb/die
Bank05: 0xad,0x89,0x28,0x53,0x0,0xb0,0x0,0x0 - Hynix 3dv6-128L TLC 16k 1024Gb/CE 1024Gb/die
Bank06: 0xad,0x89,0x28,0x53,0x0,0xb0,0x0,0x0 - Hynix 3dv6-128L TLC 16k 1024Gb/CE 1024Gb/die
Bank07: 0xad,0x89,0x28,0x53,0x0,0xb0,0x0,0x0 - Hynix 3dv6-128L TLC 16k 1024Gb/CE 1024Gb/die
Bank08: 0xad,0x89,0x28,0x53,0x0,0xb0,0x0,0x0 - Hynix 3dv6-128L TLC 16k 1024Gb/CE 1024Gb/die
Bank09: 0xad,0x89,0x28,0x53,0x0,0xb0,0x0,0x0 - Hynix 3dv6-128L TLC 16k 1024Gb/CE 1024Gb/die
Bank10: 0xad,0x89,0x28,0x53,0x0,0xb0,0x0,0x0 - Hynix 3dv6-128L TLC 16k 1024Gb/CE 1024Gb/die
Bank11: 0xad,0x89,0x28,0x53,0x0,0xb0,0x0,0x0 - Hynix 3dv6-128L TLC 16k 1024Gb/CE 1024Gb/die
Bank12: 0xad,0x89,0x28,0x53,0x0,0xb0,0x0,0x0 - Hynix 3dv6-128L TLC 16k 1024Gb/CE 1024Gb/die
Bank13: 0xad,0x89,0x28,0x53,0x0,0xb0,0x0,0x0 - Hynix 3dv6-128L TLC 16k 1024Gb/CE 1024Gb/die
Bank14: 0xad,0x89,0x28,0x53,0x0,0xb0,0x0,0x0 - Hynix 3dv6-128L TLC 16k 1024Gb/CE 1024Gb/die
Bank15: 0xad,0x89,0x28,0x53,0x0,0xb0,0x0,0x0 - Hynix 3dv6-128L TLC 16k 1024Gb/CE 1024Gb/die

I visually checked the controller to verify that it's using the RTS5765DL.

The Cardea Z330 gives an identical NAND output, except the controller name is different. I also visually checked the controller here as well, RTS5763DL.

Team Group have relabeled the NAND, so it's entirely possible it's prone to change.