Crucial MX300 2 TB

[Aquinus]

I already know that as do 99.9% of users visiting these forums (or atleast they should) but Drive manufactures still seem to persist with the lie and reviewers shouldn't still be falling for their BS and state the real formated capacity instead of the fairy tale that HDD/SSD manufacturers say they are then maybe they'd pick up their game and actually produce drives of the capacity they say they are

It's not really lying because the math actually does work out. The reason is because manufacturers base capacity off of a metric representation of size where 1000 bytes = 1 kilobyte where computers are basing it off powers of 2 which means that computers read 1024 bytes as 1 kilobyte (or more accurately 1kibibyte.) The fine print literally says this on hard drive packaging. Often, you're buying 2TB not 2TiB and that hard drive capacity is measured as a decimal representation not a binary one.

So lets say you have a 2TB drive, when you plop it into a computer. You'll see 1.819TB. That's because (2TB * (1000^4)) / 1024^4 = 1.819TiB. So the computer reads it as 1.819TiB (although, it's not usually telling you it's the binary representation,) but, it's still 2TB. Formatted or not, it's still a 2TB/1.819TiB drive. This has literally been the way things have been for decades.

 

[StrayKAT]

It's not really lying because the math actually does work out. The reason is because manufacturers base capacity off of a metric representation of size where 1000 bytes = 1 kilobyte where computers are basing it off powers of 2 which means that computers read 1024 bytes as 1 kilobyte (or more accurately 1kibibyte.) The fine print literally says this on hard drive packaging. Often, you're buying 2TB not 2TiB and that hard drive capacity is measured as a decimal representation not a binary one.

So lets say you have a 2TB drive, when you plop it into a computer. You'll see 1.819TB. That's because (2TB * (1000^4)) / 1024^4 = 1.819TiB. So the computer reads it as 1.819TiB (although, it's not usually telling you it's the binary representation,) but, it's still 2TB. Formatted or not, it's still a 2TB/1.819TiB drive. This has literally been the way things have been for decades.

Yeah! What he said

 

[bug]

Even if drives were advertised in TiB, you'd still not get the advertised capacity, because the file system reserves a variable amount anyway. So the discussion is really moot.

 

[newtekie1]

It's not really lying because the math actually does work out. The reason is because manufacturers base capacity off of a metric representation of size where 1000 bytes = 1 kilobyte where computers are basing it off powers of 2 which means that computers read 1024 bytes as 1 kilobyte (or more accurately 1kibibyte.) The fine print literally says this on hard drive packaging. Often, you're buying 2TB not 2TiB and that hard drive capacity is measured as a decimal representation not a binary one.

So lets say you have a 2TB drive, when you plop it into a computer. You'll see 1.819TB. That's because (2TB * (1000^4)) / 1024^4 = 1.819TiB. So the computer reads it as 1.819TiB (although, it's not usually telling you it's the binary representation,) but, it's still 2TB. Formatted or not, it's still a 2TB/1.819TiB drive. This has literally been the way things have been for decades.

Really, the storage manufacturers are the ones that are right, and not lying. TB is the International Standard of Units(SI) symbol for Terabyte. The SI says Tera means 1000^4. The software is actually lying to you by listing it as 1.81TB. The software should be displaying the size as 1.81TiB, which is the SI symbol for Tebibyte, which is defined as 1024^4.

I already know that as do 99.9% of users visiting these forums (or atleast they should) but Drive manufactures still seem to persist with the lie and reviewers shouldn't still be falling for their BS and state the real formated capacity instead of the fairy tale that HDD/SSD manufacturers say they are then maybe they'd pick up their game and actually produce drives of the capacity they say they are

Try some research. The HDD/SSD manufacturers are producing drives of the capacity they say they are, the software is lying to you.

 

[bug]

Really, the storage manufacturers are the ones that are right, and not lying. TB is the International Standard of Units(SI) symbol for Terabyte. The SI says Tera means 1000^4. The software is actually lying to you by listing it as 1.81TB. The software should be displaying the size as 1.81TiB, which is the SI symbol for Tebibyte, which is defined as 1024^4.



Try some research. The HDD/SSD manufacturers are producing drives of the capacity they say they are, the software is lying to you.

I'm reading this and imagining a customer that bought 2 pounds of apples only to find that he really got less than a kilo

 

[rippie]

please take a look at the graph of performance / $
i think you misrepresent the samsung 950 pro 256GB @ 19% (or any pci-e 4x m.2 ssd)

the performance is 2+ GB/s (4x over the MX300) and cheaper ($300?), so it should be more like 400%+, so all the pci-e 4x modules should be at the top performancewise

which numbers/formula are you using?

 

[W1zzard]

please take a look at the graph of performance / $
i think you misrepresent the samsung 950 pro 256GB @ 19% (or any pci-e 4x m.2 ssd)

the performance is 2+ GB/s (4x over the MX300) and cheaper ($300?), so it should be more like 400%+, so all the pci-e 4x modules should be at the top performancewise

which numbers/formula are you using?

Real life performance increases are not anywhere close to the maximum sequential throughput difference, look at the benchmarks.

 

[bug]

please take a look at the graph of performance / $
i think you misrepresent the samsung 950 pro 256GB @ 19% (or any pci-e 4x m.2 ssd)

the performance is 2+ GB/s (4x over the MX300) and cheaper ($300?), so it should be more like 400%+, so all the pci-e 4x modules should be at the top performancewise

which numbers/formula are you using?

Max sequential transfer speeds generally means squat irl.

 

[rippie]

I thought I asked a question, so I ask again what formula are you using to get to reallife benchmarks? seems tilted

 

[W1zzard]

I thought I asked a question, so I ask again what formula are you using to get to reallife benchmarks? seems tilted

no formula. run actual real application, with real data, measure time.

 

[rippie]

The chart you gave is a % number normalized(relative to) the crucial mx300. You saying there is no formula, is just obviously inaccurate.
'an actual real application' is an unquantified unit leaving me unable to verify, or form an opinion on.
That's why I ask questions, right, to understand what is going on...

So, now, instead of asking, let me try provide a light on the matter, why i think there is some oddity going on.
It is easily quantifyable by doing some _formula_ work with the data in your review.

let's first take a look at the cost
the samsung 950 pro 256 , cost about 140 euro.
the crucial mx300 2 TB costs 522 euro.
the crucial is a factor 522/140 = 3.7 more expensive

now let's take a closer look at the performance figure you provided
'samsung 950 pro 256' vs 'crucial MX300 2TB'
photoshop editing 49.6 vs 70.9 -> 42% advantage samsung
windows startup 17.3 vs 20.0 -> 15% advantage samsung
winrar 34.6 vs 37.2 -> 7.5% advantage samsung
office instal 163.4 vs 54.5 -> 67% disadvantage samsung

so let's take the _worst_ performing real application you provide, which is the office install. Btw this seems like interesting anomaly, since in all other applications the samsung outperforms the crucial at a fraction of the cost.

so if mx300 performance (54.5 seconds = 100%) costs 522 euro, then the samsung delivers 33% (54.5/163.4 = 33% , 67% behind) performance at 140 euro.
results in 100%/522 = 0.19% performance / euro, vs 33%/140 = 0.23% performance/ euro,

if you would normalize on the mx300
then mx300 would be 100% and the samsung would be 0.23/0.19 *100% = 123%
so 23% better performance / euro.

all other applications are relatively close together, so the factor 3.7 cost difference would lead to a bit different position.

please do check my math, but does it make sense that i do not understand 'a real application' performance / dollar results in 19% vs crucial MX300 2tb, since the evidence clearly points out otherwise?

 

[Glaring_Mistake]

Believe you've let some errors slip into this review.

For one I don't believe they're using 15nm TLC NAND, possibly 16 or 20nm as that's what Intel has said that the 3D MLC NAND uses (but they have also claimed that it contains about as many electrons as if built on a 50nm lithography).
Also they're not using the same NAND as the WD Blue 3D - WD Blue 3D uses Toshiba/WD's 3D NAND also known as BiCS NAND.
Like why would they purchase NAND from another manufacturer when they produce NAND themselves?

About the Marvell controller being a new variant with support for TLC NAND is it really new or is it just that you don't consider it that old yet?
Because Marvell 88SS1074 BSW2 has been used in a number of drives previously, like in the SanDisk X400 which does use 15nm TLC NAND.

400 TBW endurance. 400 * 1024 = 409,600 GB

5 years = 5*365 = 1825 days

400 * 1024 / 1825 = 224 GB per day

Know you may have simplified things a bit but reaching the TBW may not mean that you've used up all of its rated P/E cycles let alone that it will die after reaching the TBW.
Though if you have worn it out that much it may be a good idea to begin to consider replacing it.

 

[W1zzard]

office install

I have looked into the reason for the low office scores, and the underlying reason is a single step in the office installation that does something with the installer archives. it's 100% repeatable and seems to be due to some way microsoft accesses the file data in their installer.

since it is a valid usage scenario, and office is probably one of the most installed apps in the world i see no reason to exclude it just because it gives undesired scores

performance / dollar results

you are aware that these are normalized by capacity?

WD Blue 3D uses Toshiba/WD's 3D NAND also known as BiCS NAND.

Of course, not sure why I missed that. Conclusion updated. Do you have a source re 16/20 nm ?

About the Marvell controller being a new variant with support for TLC NAND is it really new or is it just that you don't consider it that old yet?

What I meant was that 88SS1074 is newer than 88SS9189

 

[rippie]

you are aware that these are normalized by capacity?

so that is then the thing. in the review it is explicitly mentioned capacity was excluded.
by that you mean you didn't exclude looking at capacity, but you included it the formulas normalization with capacity too

okay, end of confusion but you might consider rephrasing that one

 

[W1zzard]

so that is then the thing. in the review it is explicitly mentioned capacity was excluded.
by that you mean you didn't exclude looking at capacity, but you included it the formulas normalization with capacity too

okay, end of confusion but you might consider rephrasing that one

"Please note that this score was normalized to exclude the capacity of the tested drive."

How would you rephrase it?

 

[rippie]

results are normalized to the crucial mx300 2tb and normalized to take the drive capacity into account

 

[Glaring_Mistake]

Of course, not sure why I missed that. Conclusion updated. Do you have a source re 16/20 nm ?

It's not necessarily mentioned all that often so it can be somewhat tricky to find but here is one source for 16nm:

https://www.intel.com/content/www/u...-series/dc-s3520-480gb-2-5inch-6gbps-3d1.html

And one for 20nm:

Like I said before this is technically just for 3D MLC NAND that they've actually said use 16-20nm lithographies but I think it holds true for their 3D TLC NAND as well.
However if their claim of it holding as many electrons as a 50nm lithography is true then it doesn't really matter what lithography they say they're using.

 

[Glaring_Mistake]

Of course, not sure why I missed that. Conclusion updated. Do you have a source re 16/20 nm ?

Might want to update the first and second page as well seeing as you mention it being 15nm TLC NAND there too.

And I know this may be because you have not tested WD Blue 3D yourselves but according to reviews it outperforms the MX300 and especially under heavy loads.
Think that may be down to both the NAND and that SanDisk (which WD acquired) is pretty good at writing firmware for Marvell controllers.

 

[newtekie1]

Might want to update the first and second page as well seeing as you mention it being 15nm TLC NAND there too.

And I know this may be because you have not tested WD Blue 3D yourselves but according to reviews it outperforms the MX300 and especially under heavy loads.
Think that may be down to both the NAND and that SanDisk (which WD acquired) is pretty good at writing firmware for Marvell controllers.

The thing about measuring the lithography size is they can basically say it is anything they want.

The nm name has become nothing more than a milestone name, not an actual measurement of anything in the silicon chip. So if Crucial says 15nm TLC NAND, then it is 15nm TLC NAND. Just like AMD says Polaris and Vega are 14nm FinFET, or anyone else that is using GlobalFoundries' 14nm FinFET, when they are really just using FinFET on the old 20nm lithography but are just calling it 14nm. This is also why chips made with the Global Foundries' 14nm FinFET consume more power than identical chips made with 16nm FinFET from TSMC. The TSMC process is actually using a smaller lithography than GlobalFoundreis, despite the names.