WEBVTT

00:00:00.240 --> 00:00:06.759
if you buy Electronics on marketplaces like eBay or AliExpress you can easily

00:00:05.120 --> 00:00:11.160
end up with a product that doesn't live up to the performance or quality that

00:00:08.760 --> 00:00:16.480
you would expect but what if I told you that the same thing is happening right

00:00:13.480 --> 00:00:19.480
now on reputable websites like new EG

00:00:16.480 --> 00:00:24.000
and Microcenter and with trusted Brands

00:00:19.480 --> 00:00:25.560
like Kingston PNY and adata that's right

00:00:24.000 --> 00:00:31.400
all three of them have been accused of releasing new products then silently

00:00:28.640 --> 00:00:35.840
downgrading them after the fact it's the perfect crime right you get a wave of

00:00:33.719 --> 00:00:40.640
positive reviews from media and early Enthusiast customers right and for

00:00:38.239 --> 00:00:45.079
everyone who comes later well they have no way of getting the original product

00:00:42.559 --> 00:00:51.399
to compare so you can just handwave it away as your mileage may vary right

00:00:48.039 --> 00:00:54.559
wrong in March of this year I put out a

00:00:51.399 --> 00:00:58.480
bounty for a particular Drive the adata

00:00:54.559 --> 00:01:02.079
xpg sx8200 pro offering to pay our

00:00:58.480 --> 00:01:05.439
community members three times the MSRP

00:01:02.079 --> 00:01:07.600
if anyone could ship me a unique variant

00:01:05.439 --> 00:01:15.759
of this drive I mean how many could there be right two three eight wow

00:01:13.119 --> 00:01:21.280
that's a lot and we've got five of them to answer three questions once and for

00:01:18.720 --> 00:01:29.000
all are manufacturers bait in switching their customers why would they do that

00:01:24.040 --> 00:01:29.000
and does it really matter anyway

00:01:36.119 --> 00:01:44.799
on paper the xpg sx8200 pro is one of the top performing PCI Express gen 3 m.2

00:01:41.840 --> 00:01:50.399
drives it's rated at up to 3500 megabytes per second reads and up to

00:01:47.560 --> 00:01:55.079
3,000 megabytes per second rights and in third-party reviews it managed to meet

00:01:52.840 --> 00:02:00.200
and sometimes even exceed those lofty claims whether sequential speeds are a

00:01:57.719 --> 00:02:05.680
meaningful measure of SSD performance is a separate discussion but adata used it

00:02:03.439 --> 00:02:11.280
as the standard to measure this product and they delivered on it except for one

00:02:08.599 --> 00:02:16.400
small problem if you go to the store today and pull an sx8200 Pro Off the

00:02:13.920 --> 00:02:21.840
Shelf you might not actually be getting one of these it ended up being a Reddit

00:02:19.160 --> 00:02:28.319
user by the name of svart chimpan who blew the whistle on this they purchased

00:02:24.319 --> 00:02:30.360
three adata sx8200 Pro 2 TB drives at

00:02:28.319 --> 00:02:37.040
different points in time and and found that a brand new Drive purchased in 2021

00:02:33.480 --> 00:02:39.280
was performing about as well as a 94%

00:02:37.040 --> 00:02:44.519
full drive that had been in use for over a year big yikes as the kids say of

00:02:43.120 --> 00:02:49.200
course just because something was posted on Reddit doesn't make it accurate and

00:02:46.879 --> 00:02:54.760
in fact ad data pointed out that the user appeared to have connected one of

00:02:51.159 --> 00:02:56.560
the ssds directly to CPU controlled PCIe

00:02:54.760 --> 00:03:00.599
lanes while the other was running through the motherboard's PCH that

00:02:58.840 --> 00:03:05.720
actually could account for The observed performance difference but I'm not fully

00:03:03.239 --> 00:03:11.319
satisfied let's go through all of our different variants then and get them

00:03:07.879 --> 00:03:15.280
naked starting with what a data sent to

00:03:11.319 --> 00:03:17.239
the media all right there we go now

00:03:15.280 --> 00:03:21.879
there are three main Hardware components that affect SSD performance starting

00:03:19.400 --> 00:03:26.760
with the controller this acts as both the brain and Communications Hub of the

00:03:24.280 --> 00:03:30.519
drive and better ones use smarter storage algorithms that have a profound

00:03:28.760 --> 00:03:35.640
impact on both performance performance and endurance next we've got the nand

00:03:33.640 --> 00:03:41.319
flash where all your data is actually stored fast high endurance nand means a

00:03:38.840 --> 00:03:46.040
fast high endurance drive and then finally we've got our dram cache this

00:03:44.080 --> 00:03:49.280
holds data temporarily so that it can be written in a way that okay this is

00:03:47.680 --> 00:03:54.720
becoming a bit of a pattern here makes the drive faster and longer lasting and

00:03:52.760 --> 00:04:02.439
this right here this is a good combination silicon motion SM 2262 EG

00:03:58.319 --> 00:04:05.200
controller four times 256 gigs of imft

00:04:02.439 --> 00:04:11.159
64 layer TLC n flash it's a total of 1 tbte and two packages of 512 megabytes

00:04:09.959 --> 00:04:17.720
of n ddr3l hold on a second given that speed

00:04:15.120 --> 00:04:22.919
and endurance are the only things really separating a premium SSD from a cheap

00:04:20.440 --> 00:04:27.880
one it stands to reason that changing even one of these components has the

00:04:25.160 --> 00:04:32.400
potential to turn this into an entirely different class of product so so what

00:04:30.680 --> 00:04:36.680
did ad data change I'm so glad you

00:04:34.800 --> 00:04:40.400
asked everything thanks to a tool developed

00:04:39.000 --> 00:04:45.520
specifically for silicon motion controllers we were able to generate

00:04:42.199 --> 00:04:48.479
this table with the full actual specs of

00:04:45.520 --> 00:04:52.720
each of our five drives and to summarize what's happening here compared to our

00:04:50.680 --> 00:04:58.320
original remember the one sent to the media not a single Drive had more than

00:04:55.919 --> 00:05:03.560
one of these three components I just talked about that match

00:05:00.680 --> 00:05:08.199
only this one purchased in July of 2019 even uses the same EG variant of the

00:05:06.000 --> 00:05:13.919
controller with the other three using the slower G variant none of our Samples

00:05:11.440 --> 00:05:19.199
used the same DDR3 as our original although I don't really expect that to

00:05:15.520 --> 00:05:22.360
be a problem and all five of them used

00:05:19.199 --> 00:05:25.160
not just different nand flash but

00:05:22.360 --> 00:05:30.199
different Nan Flash from completely different vendors that's right out of

00:05:27.919 --> 00:05:35.759
the top six Nan flash man manufacturers in the world we have five of them

00:05:32.720 --> 00:05:37.319
represented here which doesn't matter as

00:05:35.759 --> 00:05:43.680
long as they all meet the same bar for performance and endurance but the

00:05:39.880 --> 00:05:45.560
question is do they unfortunately

00:05:43.680 --> 00:05:50.520
without hundreds of samples we can't perform any kind of meaningful endurance

00:05:47.600 --> 00:05:55.800
testing but we can validate performance and let's just say some discrepancies

00:05:53.440 --> 00:06:00.240
were found right out of the gate the three drives using the slower G variant

00:05:58.080 --> 00:06:04.440
of the controller failed to to meet the up to sequential read and WR speeds

00:06:02.240 --> 00:06:09.800
advertised on the box while both of our en variant samples at least came very

00:06:07.080 --> 00:06:13.360
close in Crystal disk Mark but let's see if these differences hold up in PC Mark

00:06:11.840 --> 00:06:18.000
which is a benchmark that more closely simulates Real World Drive use now this

00:06:16.160 --> 00:06:22.560
is interesting much like in the synthetic Benchmark the faster clocked

00:06:20.080 --> 00:06:27.960
en variants managed higher Peak speeds for both reads and rights but their

00:06:25.400 --> 00:06:34.120
overall performance was actually measurably lower and when we reran the

00:06:31.440 --> 00:06:40.240
tests with all of the drives 3/4s filled with data the en variance took a more

00:06:36.919 --> 00:06:43.880
noticeable performance hit between 7 1/2

00:06:40.240 --> 00:06:46.639
to 9% that's compared to less than 1% on

00:06:43.880 --> 00:06:51.680
the g- drives so in the case of the controller swap you're likely to end up

00:06:49.000 --> 00:06:57.000
with better performance in the real world but that doesn't explain aata's

00:06:54.800 --> 00:07:02.360
loot box approach to Nan flash procurement now this one is more likely

00:06:59.639 --> 00:07:06.599
down to pricing and Market availability the SX 82 Pro launched right around the

00:07:04.560 --> 00:07:11.360
time that 96 layer flash was making its way onto the market there's an R&D cost

00:07:09.280 --> 00:07:15.800
associated with these transitions but adding more layers generally lowers the

00:07:13.680 --> 00:07:20.879
cost per Gigabyte incentivizing this kind of development except that nand

00:07:18.800 --> 00:07:26.280
pricing is subject to supply and demand pressure like any commodity so if Apple

00:07:24.000 --> 00:07:30.520
let's say booked a huge order for 96 layer nand tightening the supply that

00:07:28.319 --> 00:07:35.720
could drive up the price forcing smaller players like ad data to buy older 64

00:07:33.240 --> 00:07:40.319
layer Tech or whatever's available all of which can be fine just like you'll

00:07:38.400 --> 00:07:45.039
look fine in a stealth hoodie from LTT Store.com the thing is sourcing

00:07:42.479 --> 00:07:49.199
components from two or more suppliers isn't a new strategy and sometimes it's

00:07:47.000 --> 00:07:53.840
a necessary one Samsung uses both Snapdragon and exos S so's in their

00:07:51.520 --> 00:07:57.879
Flagship phones and apple was sourcing modems from both Intel and Qualcomm

00:07:56.039 --> 00:08:02.879
simultaneously at one point while they made the transition this this practice

00:08:00.080 --> 00:08:06.039
can alleviate supply chain issues and it can force your vendors to stay cost

00:08:04.599 --> 00:08:10.280
competitive if they know you can just buy it from someone else AMD famously

00:08:08.840 --> 00:08:15.280
pulled a bait and switch with one of their products the ryzen 5600 a couple

00:08:12.840 --> 00:08:20.919
years ago but in that case what they did was quietly moved it from 14 NM to 12 NM

00:08:18.840 --> 00:08:26.440
giving their customers a slightly better version of what is fundamentally the

00:08:22.879 --> 00:08:29.000
same product at a great price no harm no

00:08:26.440 --> 00:08:33.880
foul it's just that the two rules to follow are that proper validation is

00:08:31.479 --> 00:08:39.240
needed to ensure that both parts are indistinguishable to the end user and

00:08:36.560 --> 00:08:44.200
failing that the changes must be disclosed a data doesn't appear to have

00:08:41.760 --> 00:08:49.560
done either of those things in their statement to Tom's Hardware they said we

00:08:46.600 --> 00:08:53.480
do guarantee that every sx8200 Pro will perform to spec within the regular

00:08:51.519 --> 00:09:00.240
variations since the drives have been tested stringently but while that may be

00:08:56.680 --> 00:09:03.160
technically correct anything between 0

00:09:00.240 --> 00:09:09.279
and 3500 megabytes per second is technically up to 3500 megabytes per

00:09:06.440 --> 00:09:15.240
second it also has the potential to be extremely misleading there was a similar

00:09:12.160 --> 00:09:17.480
Scandal back in 2014 when Kingston and

00:09:15.240 --> 00:09:22.160
PNY were called out for this same bait and switch tactic actually as we were

00:09:20.320 --> 00:09:27.800
finishing writing this video news surfaced that PNY had allegedly quietly

00:09:25.279 --> 00:09:34.839
reduced the accelerate CS 3030's Endurance by nearly 80% % on their 500

00:09:30.600 --> 00:09:37.200
gig 1 tbyte and 2 tbte models yet there

00:09:34.839 --> 00:09:42.120
was no revision number or announcement and it's still labeled on shelves as the

00:09:39.519 --> 00:09:46.040
same drive that means that customers are left without the proper tools to make an

00:09:44.120 --> 00:09:51.279
informed decision now these vendors might say who cares about a 10%

00:09:48.560 --> 00:09:55.079
performance difference or it's not our fault it's down to the inconsistent

00:09:52.959 --> 00:10:00.959
Supply and it's unmanageable to make a new skew every time there's a change but

00:09:57.959 --> 00:10:02.760
my counter to that is a people pay a lot

00:10:00.959 --> 00:10:09.560
of money for 10% more performance in some cases and B if you can't be ar to

00:10:06.440 --> 00:10:11.800
test and label and get the specs right

00:10:09.560 --> 00:10:17.560
for your own products maybe you need some oversight I mean this is why

00:10:14.040 --> 00:10:20.600
organizations like Vasa exist prior to

00:10:17.560 --> 00:10:23.800
vesa's display HDR certification program

00:10:20.600 --> 00:10:25.920
HDR ready was casually thrown around

00:10:23.800 --> 00:10:31.200
without any regard for the level of experience that product would actually

00:10:28.079 --> 00:10:33.600
deliver now the current system ain't

00:10:31.200 --> 00:10:38.440
perfect either I me display HDR 400 shouldn't even exist but at least that

00:10:36.639 --> 00:10:43.399
label means that you can trust that product to meet a certain performance

00:10:40.240 --> 00:10:45.760
standard camera storage labels like u3

00:10:43.399 --> 00:10:50.079
and v90 on SD cards serve a similar purpose they give consumers a quick and

00:10:48.320 --> 00:10:54.120
easy way to ensure that their media will handle the speeds their camera requires

00:10:52.240 --> 00:10:58.480
now this hasn't really been an issue with SSD boot drives up until now

00:10:56.480 --> 00:11:03.440
because the main competition for them has been from hard drives but as

00:11:01.000 --> 00:11:06.639
genuinely faster ssds become a requirement for gamers who want to take

00:11:05.200 --> 00:11:11.720
advantage of features like direct storage it would be really nice to see

00:11:09.480 --> 00:11:16.760
someone hold these guys accountable finally for playing so fast and loose

00:11:14.440 --> 00:11:19.880
with their product specs oh wait I guess that's

00:11:17.560 --> 00:11:25.800
us SSD manufacturers stop it we see you all of

00:11:23.560 --> 00:11:29.399
us you see that view counter down there everyone knows about this now and if you

00:11:27.519 --> 00:11:33.920
guys don't cut it out we're going to be stuck telling everyone to just stick to

00:11:31.720 --> 00:11:39.760
vendors like Samsung who are vertically integrated and don't randomly change

00:11:36.600 --> 00:11:42.639
suppliers for literally every core

00:11:39.760 --> 00:11:47.200
component of their product just like we tell everyone to

00:11:45.000 --> 00:11:52.760
stick to buying technology and computer hardware at Microcenter Microcenter has

00:11:49.839 --> 00:11:56.320
25 locations across the United States and you can check out microcenter's

00:11:54.320 --> 00:12:00.360
custom PC Builder to spec out the best PC for your budget they'll help you

00:11:58.560 --> 00:12:03.920
ensure all your parts are compatible find stock available at your nearest

00:12:01.760 --> 00:12:07.680
Microcenter location and once you add it to your cart you can arrange same day in

00:12:05.800 --> 00:12:11.480
store pickup for a fee you can check the box marked same day Pro assembly and

00:12:09.680 --> 00:12:15.000
Microcenter expert technicians will assemble your PC for you if you want

00:12:13.519 --> 00:12:19.279
help deciding what parts to put in your new custom gaming PC join the online

00:12:17.279 --> 00:12:23.199
Microcenter Community it's a great place to discuss Tech with other enthusiasts

00:12:21.320 --> 00:12:26.880
so don't wait check out the link in the video description to learn more about

00:12:24.560 --> 00:12:31.600
Microcenter and get a free pair of wireless bluetooth headphones valid inst

00:12:28.720 --> 00:12:37.399
store only no purchase necessary if you guys enjoyed this video maybe check out

00:12:34.040 --> 00:12:38.800
the uh lonus was right video it's

00:12:37.399 --> 00:12:44.440
doesn't it's not what it sounds like it's the other Linus and we're talking

00:12:40.440 --> 00:12:44.440
about ECC memory offer