I forgot to make a video about this $150,000 server

Linus Tech Tips
27 Mar 202418:54

Summary

TLDRThe video script discusses the discovery of a high-end server with Intel's Optane technology during a studio's spring cleaning. It delves into the features and benefits of Optane, including its use as persistent memory modules and its latency advantages over traditional NAND flash. The script also explores the reasons behind the discontinuation of Optane, citing factors such as delayed release, pricing, and competition. The video concludes with a demonstration of Optane's performance and a discussion on its current market availability and potential uses.

Takeaways

  • 💻 The studio stumbled upon a forgotten $15,000 server with two CPUs, eight SSDs, and 6.3TB of Optane persistent memory modules, initially intended for a detailed review.
  • 🤖 Intel's Optane business was discontinued in June 2022, leaving the advanced technology without further development despite its unique capabilities.
  • 🔍 Optane technology bridged the gap between conventional memory and storage devices, offering low latency and functioning as both SSD storage and RAM-like memory.
  • ⚡ The server utilized Optane in 'memory mode' to simulate an enormous amount of RAM, but with the drawback of data loss upon restart due to security measures.
  • 🚨 A test with SolidWorks flow simulation demonstrated the potential of Optane memory to handle large datasets, which could drastically reduce computational times for complex simulations.
  • 🤧 The script explains how Optane's 'app direct' and 'storage' modes work, allowing it to be used as high-speed storage or directly by applications for a hybrid approach.
  • ⚙️ Setting up Optane proved challenging due to poorly organized BIOS settings, highlighting an industry-wide issue with motherboard configurations.
  • 💰 Cost and compatibility issues with RAM made Optane a difficult sell despite its performance advantages, contributing to its market struggles.
  • ⭐ Optane's late arrival to the market and subsequent price drops in DDR4 RAM undermined its potential cost advantage during a global RAM shortage.
  • 🤴 AMD's competitive advancements with their EPYC processors further pressured Intel's Optane, emphasizing the importance of timely innovation and market adaptation.
  • 🚀 Despite Optane's discontinuation, leftover inventory offers value for specific applications, like DIY network appliances or as durable SSDs for intensive write environments.

Q & A

  • What is the main topic of the video script?

    -The main topic of the video script is about Intel Optane technology, its features, uses, challenges, and eventual discontinuation.

  • What are some of the key components of the $15,000 server mentioned in the script?

    -The $15,000 server includes two CPUs, eight SSDs, and 6.3 terabytes of RAM.

  • What is Intel Optane persistent memory, and how is it different from conventional RAM and SSDs?

    -Intel Optane persistent memory modules function somewhere between conventional RAM and storage devices. They offer lower latency than regular NAND flash SSDs and can be used in memory mode, where they appear as RAM to the operating system.

  • What are the two main modes of using Intel Optane persistent memory, and how do they differ?

    -The two main modes are memory mode and storage mode. Memory mode allows Optane modules to function as RAM, while storage mode presents Optane as an SSD with low latency compared to traditional NAND flash.

  • Why did Intel Optane face challenges in the market, according to the script?

    -Intel Optane faced challenges due to delays in its release, pricing compared to conventional RAM, compatibility issues with CPU models, and competition from other technologies like AMD's EPYC processors.

  • What is the significance of the 'large memory footprint version' of CPUs mentioned in the script?

    -The 'large memory footprint version' of CPUs is necessary for fully utilizing Intel Optane persistent memory. These CPUs support more RAM, allowing users to maximize the benefits of Optane.

  • Why did Micron, the manufacturer of Optane, decide to discontinue the product?

    -Micron discontinued Optane due to financial losses. Despite being a cool technology, it wasn't profitable, and Micron wanted to stop losing money.

  • What challenges did the creators of the video face in configuring the Optane server?

    -The creators faced challenges in configuring the Optane server due to poor organization of the BIOS settings, making it difficult to find the necessary configurations.

  • What is the recommended ratio of Optane to RAM for optimal performance, and how does it compare to the ratio in the server mentioned in the script?

    -The recommended ratio is 4:1 Optane to RAM for optimal performance. However, the server mentioned in the script had a ratio closer to 16:1, which is higher than the recommendation.

  • Despite its discontinuation, is Intel Optane still available for purchase, and why?

    -Yes, Intel Optane is still available for purchase because Intel has over half a billion dollars of inventory left. They haven't aggressively promoted it, but it's still being sold, often at closeout deals.

Outlines

00:00

🔍 Rediscovering a Forgotten Server

During a spring cleaning session at their studio, the team stumbles upon a long-forgotten $15,000 server originally sent by Intel, equipped with two CPUs, eight SSDs, and an impressive 6.3 terabytes of Optane persistent memory. Initially, a deeper exploration of the server's features was planned but got delayed. The narrative shifts to discussing the now-defunct Intel Optane technology, its unique position between conventional memory and storage devices, and its low latency characteristics. The video script delves into the practical applications and limitations of Optane in professional settings, such as simulations that require significant amounts of memory.

05:03

🛠 Experimenting with Optane's Capabilities

The team embarks on an experiment to leverage the Optane memory by reducing the server's conventional RAM, aiming to understand the performance impact. However, they encounter limitations due to Intel's design, which requires Optane to be paired with standard RAM. They explore alternative modes to utilize Optane, such as 'App Direct' and 'Storage' modes, but face challenges due to the server's BIOS configurations. The segment highlights the difficulties in navigating motherboard BIOS settings and the industry-wide inconsistency in terms configuration and documentation.

10:05

📊 Analyzing Optane's Performance and Market Challenges

The script further examines the performance benefits of Optane through benchmark tests, comparing it against conventional SSDs and showcasing its superior latency. It then transitions to discussing the broader market challenges that led to Optane's discontinuation, including its delayed launch, the competitive RAM market, and the requirement for pairing with conventional RAM, which made it less economically viable. The narrative also touches on Intel's strategic missteps, such as the introduction of CPUs with large memory footprint versions that significantly increased costs, and how these factors, coupled with the rise of AMD's competitive offerings, contributed to Optane's decline.

15:06

🤔 Reflecting on Optane's Legacy and Lessons Learned

The final segment reflects on the remaining stock of Optane products and their current market positioning, suggesting that despite its technological innovations, Optane failed to capture a significant market share. The narrative critiques Intel's product strategy and pricing, suggesting that better management and market positioning could have altered Optane's fate. The script concludes with a sponsorship message, linking the theme of data privacy and control to the broader narrative of technological innovation and market dynamics.

Mindmap

Keywords

💡Optane Persistent Memory

Optane Persistent Memory refers to a series of high-capacity storage modules developed by Intel that combine the characteristics of RAM and SSD storage. In the video, these modules are highlighted for their unique ability to function in a 'memory mode', where they appear to the operating system as RAM, and in 'storage mode', where they act like super-fast SSDs with very low latency. The video discusses how these modules, equipped with 3D XPoint technology, have latency just 1/100th that of regular NAND flash memory, enabling significantly faster data access times.

💡3D XPoint

3D XPoint is a type of non-volatile memory technology developed jointly by Intel and Micron. It is the technology underlying Intel's Optane products. The video mentions 3D XPoint in the context of explaining the low-latency characteristics of Optane Persistent Memory modules, which allow for faster data retrieval compared to traditional NAND flash used in SSDs. This technology provides a bridge between dynamic RAM and traditional storage, offering a blend of speed, endurance, and density benefits.

💡Memory Mode

Memory Mode is a configuration option for Optane Persistent Memory that allows the modules to be recognized by the operating system as RAM. This mode is described in the video as allowing a computer to leverage Optane's high capacity for volatile memory use, albeit with the limitation that data is wiped upon restart due to security protocols. The video explains that Memory Mode can significantly increase the available 'RAM' without requiring special coding, enhancing the performance of applications that benefit from large amounts of memory.

💡App Direct Mode

App Direct Mode is another way to utilize Optane Persistent Memory, enabling applications to directly access the memory modules as a unique blend of RAM and storage. The video notes that this mode requires applications to be specifically coded to take advantage of Optane's features, which may have limited its appeal due to the high development costs versus the benefits. App Direct Mode offers potential for performance enhancements in applications that can efficiently leverage the low latency and persistence of Optane.

💡Storage Mode

Storage Mode refers to a configuration of Optane Persistent Memory where the modules are presented to the operating system as SSDs, utilizing the high-speed memory slots for data storage with extremely low latency. The video discusses the challenges and potential benefits of Storage Mode, including the ability to set up volumes similar to RAID arrays, thereby improving data access speeds for storage-intensive applications.

💡Latency

Latency, in the context of this video, refers to the delay between a request for data and the delivery of that data. The script emphasizes the significantly lower latency of Optane Persistent Memory compared to traditional NAND flash memory, highlighting how this characteristic can improve the performance of applications by allowing for quicker data access and processing times.

💡3D Space Mesh Generation

3D Space Mesh Generation is mentioned in the video as a process used in applications like SolidWorks Flow Simulator. It involves creating a three-dimensional grid in which to simulate fluid dynamics. The script uses mesh generation to illustrate the vast amount of memory available with Optane, though it also notes that not all applications will require the extreme memory capacities provided, except in very complex simulations.

💡Data Wiping

Data Wiping, as described in the video, is a security feature of Optane Persistent Memory in Memory Mode where all data on the modules is erased upon system restart. This is due to the generation of new cryptographic keys as part of the module's security protocols. While this ensures data security, it also means that any persistent storage needs cannot be met by Memory Mode alone.

💡Mixed Mode

Mixed Mode is a configuration setting for Optane Persistent Memory that allows a portion of the Optane capacity to act as volatile memory (RAM) and the remainder as storage. This mode offers a 'best of both worlds' scenario, as mentioned in the video, enabling users to benefit from both the high-speed access of RAM and the persistent storage capabilities of SSDs without having to choose one over the other.

💡Latency Impact on Performance

The video discusses an experiment to measure the impact of latency on system performance by comparing Optane Persistent Memory to traditional RAM and SSDs. It explains that while Optane offers significantly lower latency, which is beneficial for certain applications, the overall system performance also depends on other factors such as sequential transfer speeds. This part of the video underscores the complex considerations involved in choosing between different types of memory and storage solutions based on specific performance needs.

Highlights

Spring cleaning at the studio led to the discovery of a $15,000 server with two CPUs, eight SSDs, and 6.3 terabytes of RAM.

The server was obtained from Intel three years ago and features Optane persistent memory modules.

Intel's Optane business was closed in June 2022, which affected the contacts who originally sent the server.

Optane is a family of products that bridge the gap between conventional memory and storage devices, offering low latency and high-speed performance.

The 3D XPoint chips inside the SSD have 100 times less latency than regular NAND flash.

The Intel Optane persistent memory module has 512GB of storage, appearing as a stick of DDR4 memory.

In memory mode, Optane modules are treated as a large amount of RAM by the operating system.

Optane memory has the downside of generating a new cryptographic key upon restart, wiping all data it contained.

No special coding is needed to use memory mode, and the CPU's memory controller automatically manages data transfer between RAM and Optane.

The server has 12 Optane persistent memory modules, totaling 6 terabytes, enough to store nearly 12 high-bit rate float plane exclusives in system memory.

Using Optane in memory mode can significantly speed up complex professional applications, potentially reducing processing time from a week to a day.

Optane's latency is several times higher than memory, which would impact performance in certain applications.

Intel's recommended ratio for optimal performance is 4 to 1 Optane to RAM; the server's ratio is more like 16 to 1.

Storage mode allows Optane to function like an SSD with super low latency compared to traditional NAND flash.

Setting up storage mode was challenging due to the poorly organized BIOS on the Intel motherboard.

Mixed mode configuration offers a blend of RAM and low-latency storage, managed by the BIOS.

Optane's price was a significant factor in its discontinuation, as it was more expensive than RAM and offered less bandwidth.

Optane's arrival was delayed, and by the time it was released, competitors like AMD had launched more advanced and affordable alternatives.

Despite its discontinuation, Optane is still available due to Intel's large inventory from 2022.

Optane's endurance is superior to NAND, making it ideal for applications requiring frequent data retention and overwriting.

Although Optane showed potential in latency reduction, its speed was limited compared to modern Gen 5 drives.

Transcripts

00:00

we just kicked off Spring cleaning at

00:01

our studio which wouldn't normally be a

00:03

video except we came across this a

00:08

$15,000 server with two CPUs eight ssds

00:12

and a whopping

00:14

6.3 terabytes of ram when we got this

00:18

from Intel 3 years ago we did a quick

00:21

unboxing promising a deeper dive into

00:23

some of its more exotic features and

00:25

then promptly forgot about it well that

00:28

Ram I showed you guys is is n Ram at all

00:30

but optain persistent memory modules you

00:34

might know that Intel's optain business

00:36

was sadly shuttered in June of 2022

00:38

which means the contacts who originally

00:41

sent us this have sadly moved on we on

00:44

the other hand haven't so all these

00:47

years later I am finally going to do

00:50

what I said I was going to do and finish

00:52

building the rer piece I mean sorry I

00:54

mean digging into what made optain so

00:58

special and cool and

01:00

what ultimately killed it just like I'm

01:03

going to kill this segue to our sponsor

01:05

ug green they're getting into the Nas

01:07

Game and they claimed their new Nas sync

01:09

dxp 4800 plus is the most powerful Nas

01:12

under $500 it's got four Bays for

01:15

storage a 12th Gen 5 Core Intel

01:18

processor 8 gigs of ddr5 memory that's

01:21

expandable and more check it out at the

01:23

link down

01:28

below

01:33

before we go any further some of you

01:35

probably need a refresher isn't Intel

01:37

optain and SSD what's it doing in a ram

01:40

slot well optane was more of a family

01:43

than an individual product ranging from

01:45

inexpensive hard drive accelerators for

01:47

pre-builts where the manufacturer was

01:49

too cheap to install a real SSD to

01:52

incredibly low latency not to mention

01:55

expensive storage for the fastest

01:57

servers in existence what all of these

01:59

uses had in common though was they

02:02

looked like conventional memory and

02:04

storage devices but functioned somewhere

02:07

in between and that's exactly what we're

02:10

looking at here this appears to be a

02:12

normal SSD but the 3D Cross Point chips

02:15

inside it have just 100th the latency of

02:19

regular Nan flash as for this guy well

02:22

it looks like a stick of ddr4 memory but

02:25

it's actually an Intel optain persistent

02:27

memory series 100 module with a whopping

02:30

512 gigs of storage on it and this

02:33

server has 12 of them for a total of 6

02:37

tabes that's enough to fit nearly 12

02:40

High bit rate float plane exclusives

02:42

solely in system memory wait a second

02:45

lus memory I thought you just said it

02:47

was storage I did but the simplest way

02:50

to take advantage of it is in memory

02:52

mode where your optain modules just show

02:55

up to your operating system as a whole

02:57

whack load of RAM now the bad thing

03:00

about this mode is that it forces optane

03:03

to have one of the big downsides of RAM

03:06

for security reasons every time an

03:09

optain module in memory mode gets

03:11

restarted it generates a new

03:13

cryptographic key effectively wiping all

03:16

of the data that it contained the good

03:18

news is that optane can handle many more

03:21

wipes than regular ssds so this is

03:24

unlikely to harm it and what's also cool

03:27

is that no special coding is needed to

03:29

immediately use memory mode and the

03:31

cpu's memory controller will even

03:33

automatically move colder data off of

03:36

the memory onto the optain and vice

03:39

versa let me show you how this would

03:41

work in a demanding professional

03:44

application right now you're looking at

03:46

solid works flow simulator but we aren't

03:49

simulating anything at the moment

03:50

because first we need to generate a mesh

03:53

which is essentially a 3D Space full of

03:56

data sampling points where some fluid

03:59

let's say air is going to interact with

04:02

some object say this heat sink it's not

04:05

using that much RAM though what 45 gigs

04:08

48 gigs that's a lot of ram okay it's a

04:11

lot of ram but it's 2% of what we have

04:14

the issue is that it just takes hours

04:16

for the CPU to process a large enough

04:18

mesh like this for it to fill up all of

04:21

that Ram let alone the optain that's in

04:24

there and this is a really silly way of

04:26

doing this so what you're saying is flow

04:29

doesn't need that much RAM obtaine is

04:31

pointless no no no Flo Sim does need

04:34

heaps and heaps of ram just a model like

04:37

this doesn't necessarily if we were

04:40

trying to run something that's super

04:41

complex like crashing a car into a

04:44

barrier having terabytes of memory could

04:45

be the difference between a finishing in

04:47

a day and a finishing in a week of

04:49

course though this isn't memory its

04:51

latency is several times higher which

04:53

would obviously impact performance the

04:55

question is though how much to find out

04:59

we came up with a rather terrible idea

05:02

why don't we just remove a bunch of the

05:05

actual Ram so that the system has no

05:07

choice but to use the obtaine and then

05:10

we can do a before and

05:16

after it's not going to do anything is

05:19

it

05:20

nope oh jeez there goes the alarm wait

05:25

maybe there's no way no sadly Intel

05:31

thought of that and requires each stick

05:33

of obtain pm to be paired with a stick

05:36

of ram in order for the system to

05:37

function at all and they have a

05:39

recommended ratio of 4 to1 optained to

05:42

RAM for Optimal Performance ours however

05:46

it's more like a 16 to1 ratio and the

05:49

server came straight from Intel

05:52

so what do we do with the rest of it as

05:55

it turns out many customers simply used

05:58

higher ratios but there are some other

06:01

ways to take advantage of the persistent

06:03

part of our persistent memory app direct

06:06

mode and storage mode which Alex is

06:09

showing us now yes app direct mode

06:12

exposes the optain directly to the

06:14

application which can use it like

06:16

storage Ram or somewhere in between

06:20

unfortunately we aren't able to show

06:22

this off because doing so would require

06:23

us to code our own program that

06:26

explicitly takes advantage of it and

06:28

this is a task that almost no company's

06:30

worth billions of dollars found to be

06:32

worthwhile so doing it for a YouTube

06:33

video would have a dubious Roi to say

06:36

the least what we can show you is

06:38

storage mode have you ever wondered what

06:41

would happen if you could plug your SSD

06:43

into a high-speed memory slot well

06:46

basically this storage mode presents the

06:49

obtained tier operating system much like

06:51

an SSD just with super low latency

06:54

compared to traditional Mand flash

06:56

setting this bit up though was what

06:58

caused this project to be stalled for

07:00

months because we just couldn't figure

07:02

out how to do it not because Alex is

07:05

dumb or because I'm dumb but rather

07:08

because the BIOS on this Intel

07:09

motherboard is so poorly organized now

07:11

okay okay sorry this is going to be a

07:13

bit of a tangent but motherboard

07:15

manufactures can you please figure your

07:17

shiz out I mean we have this

07:19

industrywide problem where every single

07:21

motherboard maker calls the exact same

07:24

setting something ever so slightly

07:26

different and then puts it in a

07:27

different place meaning even if you are

07:30

quite knowledgeable and not dumb and you

07:33

know exactly what you're looking for

07:34

there's a solid chance you won't be able

07:36

to find it we were easily able to figure

07:40

out how to use app direct mode on Dell

07:43

super micro Lenovo and HP server boards

07:46

since they had decent documentation but

07:48

no not the Intel one we even called up

07:50

Patrick from serve the home who

07:52

literally wrote the guide on how to do

07:54

this and he couldn't figure it out

07:56

either okay look at this Lis I'm going

07:59

to go back to the main

08:00

menu we're looking for the memory

08:03

controller spot where do you think that

08:07

is well I would go to Advanced and then

08:11

I would probably go to memory

08:14

configuration so this is why we needed

08:15

Emily to help us find it yes it is under

08:18

PCI configuration um option ROM

08:21

controller and that's where the memory

08:24

controller is of course I can't believe

08:27

we didn't think of that Anyway by using

08:31

storage mode we're able to select how

08:33

much of the optain we want to be

08:35

presented as an SSD and we can create

08:37

volumes with it much like with a raid

08:40

array with the maximum performance

08:42

occurring when you interleave across

08:44

multiple optained dims that are

08:46

operating on the same CPU which

08:48

essentially creates a raid zero array on

08:51

opan now as you might have guessed this

08:54

does come with the risk of data loss so

08:56

if you don't care as much about

08:58

performance you might want to mirror the

09:00

data across both CPUs or I don't know

09:03

maybe just buy an

09:04

SSD but all of that aside the coolest

09:08

configuration is mixed mode this allows

09:10

you to set a target for how much of your

09:13

optain will be storage and how much will

09:15

be Ram creating a Best of Both Worlds

09:18

situation where the BIOS handles all of

09:20

the complex bits and you get both cheap

09:23

RAM and super low latency storage all

09:26

right what's your memory percentage goal

09:28

what do you mean goal it's doesn't

09:29

really do exact it just kind of gets

09:32

close to the percentage that you want I

09:34

think we all know the answer to that

09:36

nice so we

09:38

reboot

09:41

need sorry seemed

09:44

mad okay all right no you don't seem mad

09:47

you

09:48

seem rightfully like we need to discuss

09:51

this come on come on create name space

09:53

here we go there we go how about now

09:58

there we go

10:00

I love how surprised it is it's making a

10:02

YouTube thumbnail face speaking of which

10:05

we're working on a really cool video now

10:06

that YouTube has AB testing for

10:08

thumbnails Maria is creating a like

10:12

cringe YouTuber face version of every

10:14

thumbnail for a little bit and then just

10:16

like a nice relaxed one and then we're

10:18

going to make a video about which ones

10:20

people actually prefer whatever they say

10:22

they prefer it could change our

10:24

thumbnail practices there we go all

10:26

right okay what we're looking at right

10:28

now is our G Drive which is a

10:30

conventional nbme SSD and our o Drive

10:34

which is our optain drive and this

10:37

measure is seconds per transfer which is

10:40

essentially the latency of each

10:42

operation on the drive now what we're

10:44

going to do is we're going to run

10:45

Crystal disc mark on both of these

10:47

drives at the same

10:49

time and

10:51

theoretically yes yep the regular nvme

10:56

is way more which is bad o is opan opan

11:04

which makes sense because it barely even

11:06

appears that's how low the latency of

11:09

this is which doesn't matter for

11:10

everything and in fact if we have a look

11:12

at our Crystal dis Mark

11:14

results it's not that fast in terms of

11:18

sequential transfer speeds it's just

11:20

that it's really really responsive wow

11:24

we've done it we go okay we can actually

11:26

see what the crap we're looking at now

11:29

here we go this is what we expect to see

11:33

all three tests are running right now

11:36

there we go yes and our SATA is much

11:40

higher in terms of latency not to

11:42

mention that it can be very inconsistent

11:43

see that initial Spike when we first hit

11:45

it that's a yikes and while nvme is a

11:48

lot better it doesn't have that super

11:52

super flat line when all of them are

11:54

running the same test that we see with

11:56

optane where the latency is negligible

11:59

which which can be a huge benefit for

12:01

certain applications and not terribly

12:04

important for others which is cool in

12:08

some cases really cool so then why does

12:11

optain no longer

12:12

exist the answer is many many reasons

12:16

the first and biggest problem is that

12:19

optin arrived 2 years late it was

12:22

supposed to come with Skylake in 2017

12:25

with speeds that matched the ddr4 2666

12:28

that would be paired with those

12:29

processors but obtaine didn't arrive

12:32

until Cascade Lake which used 2933 megat

12:36

transfer per second dims that meant that

12:38

by choosing optain you would be giving

12:41

up memory bandwidth to every application

12:44

that doesn't need the extra capacity

12:47

that is not ideal also in 2017 there was

12:52

a global drram shortage as phones like

12:54

the Samsung Galaxy S6 Google pixel and

12:57

HTC1 M9 started using ddr4 while Samsung

13:02

and Micron were struggling to ramp up

13:04

production 16 gigs of RAM hit prices of

13:07

over

13:08

$200 that would have been a perfect

13:11

opportunity for an almost Ram product at

13:14

a lower price except that by 2019 that

13:17

same 16 gigs of RAM was down to more

13:20

like a hundo also remember how we

13:22

mentioned that each stick of obtained PM

13:25

requires a stick of ram to pair with it

13:27

in order for it to work

13:29

well yeah so companies that were looking

13:32

into optain they weren't just comparing

13:34

the price of opan versus Ram they had to

13:36

look at the cost of an optain module and

13:40

a stick of ram compared to just getting

13:42

two sticks of ram that uh oh by the way

13:45

are going to run faster Tom's Hardware

13:47

has an article from 2019 about the

13:50

pricing quoting a single stick of this

13:53

512 gig optain at a cool

13:57

$675 meaning that this server that we

14:00

casually lost when it was new had

14:04

$81,000 worth of optane in it that

14:07

sounds like a lot because it is but in

14:11

2019 a single stick of 128 gigs of ECC

14:15

memory was roughly

14:16

$4500 meaning optain was only about

14:19

onethird of the price so wait hold on a

14:21

second that was in 2019 that still

14:24

sounds pretty good I mean the demo we

14:26

just did looked pretty good optained for

14:29

the win right well I do love your

14:32

optimism remember guys we spoiled the

14:34

ending earlier and everyone dies but it

14:37

doesn't have to be a total loss there's

14:39

maybe a valuable lesson here for the

14:41

industry to maybe learn this time intel

14:46

if you're listening and everyone you

14:48

need to stop locking your products

14:51

because I didn't buy enough of your

14:55

products the Cascade Lake CPUs that did

14:58

eventually sh with obtaine had two

15:00

variants a standard version and a large

15:03

memory footprint version in the case of

15:06

the Zeon plattinum 8260 that we've got

15:08

in this server The Standard Version

15:11

supported only 1 tab of ram a number

15:14

that's pretty easy to hit with regular

15:16

old Dam meaning that if you wanted to

15:18

really take advantage of optane you

15:20

needed the zeeon Platinum 8260 L well I

15:25

mean that's no problem right I mean how

15:26

much more could it

15:30

$3,400 oh yeah and since you're buying

15:33

two of them it's an extra

15:36

$6800 before you buy a single flipping

15:39

optane module can see why they used L as

15:43

the suffix for those adding insult to

15:46

injury you know what also happened

15:48

between 2017 and 2019 while intel was

15:51

struggling to get optane out of the door

15:53

AMD launched epic Rome and then epic

15:56

Milan featuring more cores more memory

15:59

channels support for four ttes of ram

16:03

without a silly large memory footprint

16:05

tax oh and by the way they were also

16:08

cheaper this added up to not a whole lot

16:11

of people buying optane and Micron who

16:14

owned the Fab where all the optane was

16:15

being built finally said you know what

16:17

enough is enough it was and still is a

16:20

very cool technology but in 2021 Micron

16:23

figured you know it be a lot cooler how

16:26

about not losing 400 million ion dollar

16:29

a year so one year later Intel killed

16:33

optain which um raises one interesting

16:36

question why can you still buy it well

16:39

about that even though there hasn't been

16:41

a Fab making the stuff for years now in

16:44

20122 Intel revealed that they still had

16:47

over half a billion dollars of inventory

16:50

left over and they haven't exactly been

16:53

pushing it that hard I mean what is this

16:55

pricing Intel do you want this on your

16:58

books forever

16:59

there are some awesome closeout deals

17:01

out there these 16 gig accelerators are

17:04

amazing there are go-to drives for a DIY

17:08

Network Appliance or a NZ because

17:10

they're literally five bucks and opan

17:13

features substantially more right

17:15

endurance than nand which is great if

17:17

you want to constantly retain and

17:20

overwrite comprehensive system logs and

17:22

this one on new egg is still among if

17:24

not the best drive on the market even if

17:27

it is starting to show its age in some

17:29

ways for example it'll have outstanding

17:32

latency like we talked about before but

17:34

it runs at a quarter the maximum speed

17:37

of a modern Gen 5 Drive which is a bit

17:40

of a yikes so in

17:43

conclusion Intel do you want this thing

17:45

back we uh honestly really don't know