6:36 Dang, you cut the thermal paste application, i guess its appropriate, because 99% of the comments would have been about how wrong you were doing it.
9:55 Minor correction, you mention that the optane NVDIMMs use flash like any other SSD, but optane (3d xpoint) is its own thing, among other differences not really having an endurance limit.
Thanks to SkillShare for sponsoring today's video. Be one of the first 1000 people to visit skl.sh/craftcomputing06221 and get 1-month of Premium membership on the house... And no, I didn't forget the thermal paste. It was on the CPU.
It looks like the largest table I deal with on a regular basis has 1.15TB of data, 5.3TB of indexes , and 7,847,475,751 rows as of today. That's for 105 weeks of data as we remove everything older than that on a weekly basis. We're still a very small fish though as this runs on a single server with 4 Virtual Processors and 32GB of RAM.
I like the idea of buying old xeons and components in bulk and building a multi CPU server with obscene amounts of RAM for absolutely no reason at all. Like I really don't think I have a use for one besides a gaming server but I still want to build one.
I worked a few years ago (middle of 2019) with some prototypes of these memories. The idea was to study ways to optimize them and our BIOS to integrate with a new line of workstations from my company. The speed and latency results we got were insane. We had 6TB (12x 512GB) in a dual xeon machine, working as 50/50 (3TB RAM + 3TB AppDirect). The read/write speeds were near to 15GB/s on AppDirect mode.
Constructive criticism for the camera operator: when you were explaining the RAM slot colors, it would be better to have a better angle showing the slots, I could barely make out the blue if that's even the RAM slot. Also, the focusing could be improved to keep the subject Jeff's talking about in focus. Otherwise, framing of everything else is pretty good, and it's nice to have movement when doing something while talking. Good work camera operator!
Perhaps you should have been there so these comments could have been useful. Sometimes you miss things in the shoot, commenting like this isn't constructive, its irritating.
@@timellis7720 I consider it constructive since it’s something that can be considered for future videos, and pointing it out helps spread awareness. I wouldn’t know if they realized it or not. I’m not expecting this video to be reshot or anything, but I don’t see any harm in pointing it out. I’m not trying to be an asshole about it or demand this video be taken down entitely just for this small detail, nor hate the camera operator or anything, and I also want to give credit where credit is due, because like I said, the framing for everything else was great, there’s no doubt about that.
@@ReturnJJ no, everyone who has ever held a camera knows you need to keep important things in frame and keep it in focus. But they can still make mistakes, they don’t need to be reminded of them. You don’t go up to people and say “hey, here’s a helpful tip, those mistakes you made earlier, don’t do that”
Note that with Gen3, some Xeon Silver does support Optane, but only supports Optane 200 which is hellishly expensive. 4314 and 4316 both claim support while 4310 and 4309 do not.
I run a database recovery business and have to write hundreds of gigabytes of data to sql databases on a daily basis. This would probably be much more cheaper than the boatloads of ram and throwing SSD cache at the problem that I'm currently using.
I slowly upgrade to max memory on any given platform. My main rig only has 64GB right now, but my previous were maxed out at 128GB and 256GB. I'll wait for DDR5 to slide down in price before getting the next 64GB.
Lol at 15 min he totally called us out! That's exactly why I watch to reduce the risk of wasting my own money and pick up some ideas and practical how-to's along the way! Thanks man!
We have a couple of quad-socket VMware boxes with 3 TB memory each that we had to retire due to not being on VMware's HCL for ESXi 7. I set one of them up as a hardened Linux repository for Veeam. It's ridiculous. 56 cores. Current load average, 0.09.
i paused it to look up the beir and was impressed by the list of biers in the series/moniker. even though i know what lagunitas is now; id give the series a view.
Under 0.01 ms storage latency? My latest gear has latency measured in minutes. I wouldn't mind seeing your thoughts about using tape storage in the homelab. And one more thing, please take care of whatever is ailing your HPE server. IIRC it was also complaining the last time we saw it.
One of my favourite (and worst) hobbies is shopping our datacenter for boxes that are about to be recycled. If I lug it, I get it. xD Though, these days I do end up recycling most of it as you only need _so much_ hardware in a homelab.
An interesting thing about the retainer, the two holes on the corners that don’t have screws are different sizes and the posts on the sockets match them, so as long as you match the cpu triangle with the retainer triangle it will only be able to go on the correct orientation into the socket as long as you don’t force it At least that’s how my experience with these sockets are in the data center
I couldn't find anything explaining how the 1TB limit of Xeon Scalable v2 CPUs applied when using Optane; thank you for mentioning that the Optane PMem doesn't count towards this limit, as this as it was something I was really wanting to know! If you have encountered this information written down anywhere I'd love a reference to it.
FYI for anyone that's not removed one of these cpus from the cooler before, there is a "tim breaker" gap in between the CPU and the heatsink where you're supposed to stick in a screwdriver to break the tension of the thermal paste.
Thermal paste? Missed a step, unless you did it without showing the footage. I would have liked to see you application method on such a large heat spreader.
I hadn't looked at optane nvdimms, but just last week I was eyeballing what it would take to add some small (and surprisingly cheap) nvdimm-n dimms and battery to a Dell R740 for use as an extremely fast zfs zil, but found the documentation on the subject a little intimidating. Funny timing.
one real salient point here is about how the datacenter stuff of today will be homelab equipment soon and why it is such a good point is for the massive performance gains we will all see in the fat part of moore's law (which is extant) - we are talking about orders of magnitude improvements here and it is going to have real world impacts as this type of enterprise grade kit filters down - the prices are great and this is a non unobtanium jump due to the reasonable prices - yeah you were in the right place at the right time but deals like this are going to be commonplace before you now it as time and tech marches on - a couple of these are more affordable than a used car - you probably do want 2. pleae run some more benches to put into perspective how much work these types of machines can do. faster bandwidth and fiber proliferation is going to help uptake of this type of tech ---to fully take advantage you will want 100g - maybe that could be a followup that will fully uncork all the potential here - it is not much more per port vs 10gbe...
1 TB of ram..... hummmmmm you must play Cities Skylines too..... and @6:34 of course you remembered the thermal paste just off camera right? Oh re-reading that I thought "damn keyboard warriors" but it is a legitimate question for my ever curious mind... :) I hear the nerd, geek in your cadence and enjoy it so much a 40 years ago I was a nerd geek working with PDP-1170s I love your channel and wish I could afford just the ram that you installed lol... Thanks for the ride along and sharing your time to create, edit and UL these videos for me to geek out with you on... :)
with a dimm slot per of 6 max ram would be 6TB at 512GB per dimm As much as I could use optane at 512GB you would max out at 2TB at 1TB per side| personally if I was going to run optane in ram slots if technically feasible to do so it would1-3TB options when financially feasibility to do so.. you would likely max out a regular dimms 1-2TB sticks On a 6 dimm per side you would max our 4-8TB per side
I would love to see a method of using RAM as a read cache for databases within windows without moving the files from their slow medium. I know you can pull off a safe way using ZFS and linux for this read cache, but I've yet to see a method of doing this for windows. Current best guess is to create a RAMDISK/CACHE in Linux using ZFS and pass it to windows as an ISCSI device which really complicates things.
My biggest problem with ANY NAND-flash-based solution is that ALL of them has a finite write endurance limit to it, which I will undoubtably, wear out. HPC applications just produces SO. MUCH. DATA.
True, but for a home server/home lab use case, the endurance of these components is more than enough. I've got SAS SSDs from 2017 that measure write endurance in the >1 petabyte range. These components are made to live in big iron, inside real commercial data centers. I'm not doing high intensity compute or commercial databases or something; I won't be wearing these drives out for a good long time with my two blogs, a couple of minecraft servers, and my never-quite-done assortment of various status monitoring dashboards. Unless my usage pattern gets MUCH more intense than it is now, they should last for years. I suspect the same is true for these components in Jeff's servers.
@@sinisterpisces it depends on how much data is being passed through if his system is also hosting the raw video footage for his TH-cam videos. For my HPC/CAE/FEA/CFD applications, it's a never ending monsoon of data. It's not hard for me to generate about 110 TB of data in about 15 minutes if I am doing a FEA run using the direct, sparse solver, at which point, to average out the write endurance, it's just a matter/function of how often I perform that type of analysis. This is one of the reasons why I've decommissioned almost every single one of my SSDs because they're like brake pads on a car -- they're going to wear out. It isn't a question of IF they will wear out, but a question of WHEN they will wear out.
Damn I wish I could afford this or something like it. My whole server runs on an i5-3470t, 16gb ram, 2 x 1tb (raid1) and 2 x 3tb (raid1) in a microserver gen 8. I have HyperV for PiHole and Plex and then its just storage for everything else!
Too bad it's cancelled now, I had so much hope for 3DxPoint stuff. Imagine no need for RAM anymore, and having 3DxPoint NVRAM in DIMM slots for permanent/non-volatile storage.
I haven't watched this yet, but my first question would be are you doing some kind of RAM-DISK with with like Primo Cache? Not that i believe in 'overkill' lmao, but JESUS
6:36 Dang, you cut the thermal paste application, i guess its appropriate, because 99% of the comments would have been about how wrong you were doing it.
This guy gets it ^^^^
@@CraftComputing I thought the thermal paste troll was pretty funny.
Now he just gets a load of comments about how he didn't apply Thermal Paste instead...
@@BobHannent i noticed the sneaky troll cut in the video. hehe
Was the first thing I noticed. lol
Did i miss the TP getting put on the CPU lol
Nah he didnt put any on he yoloing it, it would seem.
Thermal Paste is for losers anyway
Yeah didn't see it either
he skipped it because everyone would have just told him how wrong he did it
Have think the same
9:55 Minor correction, you mention that the optane NVDIMMs use flash like any other SSD, but optane (3d xpoint) is its own thing, among other differences not really having an endurance limit.
It has endurance limit of course, around 300 DWPD for these 1st gen DCPMM.. CMIIW
@@alfred4683 what does DWPD stand for?
@@SeanCMonahan quick googling should tell you.. DWPD = Drive Writes Per Day
Hey Jeff, when changing the the CPU you forgot to mention putting new heat sink compound on.
cant tell if irony
I think it's an inside joke
I was looking for this comment!
@@skilz8098 me too! But you can tell he did because he wiggled the cpu side to side :P
@@skilz8098 me as well
6:22 I hope you did put on thermal paste in the video cut :D
Of course I did. But if I show *how* I install thermal paste, it leads to 300 comments about how I did it wrong.
@@CraftComputing unfortunately now you are going to get 300 comments about not putting paste on. :p
Video engagement is still engagement 🙃
@@CraftComputing All Hail The Mighty Algorithm!
@@CraftComputing I would have liked to see the right way of themal pasting.
Thanks to SkillShare for sponsoring today's video. Be one of the first 1000 people to visit skl.sh/craftcomputing06221 and get 1-month of Premium membership on the house...
And no, I didn't forget the thermal paste. It was on the CPU.
It looks like the largest table I deal with on a regular basis has 1.15TB of data, 5.3TB of indexes , and 7,847,475,751 rows as of today. That's for 105 weeks of data as we remove everything older than that on a weekly basis.
We're still a very small fish though as this runs on a single server with 4 Virtual Processors and 32GB of RAM.
Dare I ask what this table is?
What database software does your company use?
@@h.b.5577 We use Microsoft SQL Server mainly because the SQLXML Bulk Loader is really, really fast and works very well.
"What is my purpose"
"you pass the cpu"
"Oh my god"
512 GB Optane module for $500 is a steal.
I WANT MORE!!!!!
Yeah now that he made a video about it it will be sold out or 11dy million dollars..... Thanks Jeff
My favorite spud 😎
That's a stupid amount of memory!
I love it.
Now that I finally have my own house, I can fully dive into the wild world of servers and I can't wait!!!
Congrats on the house! Don’t go too overboard on it, it’s a deep rabbit hole haha
@@_mew i know but I'm super excited
@@_mew
I like the idea of buying old xeons and components in bulk and building a multi CPU server with obscene amounts of RAM for absolutely no reason at all. Like I really don't think I have a use for one besides a gaming server but I still want to build one.
No thermal paste?
He wiggled the CPU, so yes, thermal paste is here
I worked a few years ago (middle of 2019) with some prototypes of these memories. The idea was to study ways to optimize them and our BIOS to integrate with a new line of workstations from my company. The speed and latency results we got were insane. We had 6TB (12x 512GB) in a dual xeon machine, working as 50/50 (3TB RAM + 3TB AppDirect). The read/write speeds were near to 15GB/s on AppDirect mode.
Wonderful timing. My own optane just arrived - a 16 gig nvme stick of it, which I simply wanted to play around with on my core 2 quad system (lol)
Optane is magical, it's a shame they stopped trying to make consumer SSDs with it...
Oh how I wanted this to come to HEDT with the ability to boot directly off the optane dimms.
Constructive criticism for the camera operator: when you were explaining the RAM slot colors, it would be better to have a better angle showing the slots, I could barely make out the blue if that's even the RAM slot.
Also, the focusing could be improved to keep the subject Jeff's talking about in focus.
Otherwise, framing of everything else is pretty good, and it's nice to have movement when doing something while talking. Good work camera operator!
Perhaps you should have been there so these comments could have been useful. Sometimes you miss things in the shoot, commenting like this isn't constructive, its irritating.
I consider anything better than Kyle's camera operator (his other hand) to be good enough on these solo channels :)
@@timellis7720 I consider it constructive since it’s something that can be considered for future videos, and pointing it out helps spread awareness. I wouldn’t know if they realized it or not. I’m not expecting this video to be reshot or anything, but I don’t see any harm in pointing it out.
I’m not trying to be an asshole about it or demand this video be taken down entitely just for this small detail, nor hate the camera operator or anything, and I also want to give credit where credit is due, because like I said, the framing for everything else was great, there’s no doubt about that.
@@kenzieduckmoo True that.
@@ReturnJJ no, everyone who has ever held a camera knows you need to keep important things in frame and keep it in focus. But they can still make mistakes, they don’t need to be reminded of them. You don’t go up to people and say “hey, here’s a helpful tip, those mistakes you made earlier, don’t do that”
Slowly watching.... No thermal paste?? 🤯
I used Schrödinger's Thermal Paste. It's best not to observe its application.
That server has more RAM the my PC has in SSD storage! How many chrome tabs can it do?
2, but it crashes.
I'm guessing actually 1000 (due to software limitations). My PC currently has 280 open.
@@laloajuria4678 Rough.
@@Big_Flubba I would guess about 954 but if it werent chrome but a browser like Opera GX or something similar it would probably be more than 1000.
@@Kaotix19 Plausible!
Note that with Gen3, some Xeon Silver does support Optane, but only supports Optane 200 which is hellishly expensive. 4314 and 4316 both claim support while 4310 and 4309 do not.
Can’t believe he killed Geordi LaForge and has his visor on display.
*slaps ram* these babies can hold so many Call of Duty installations
I run a database recovery business and have to write hundreds of gigabytes of data to sql databases on a daily basis. This would probably be much more cheaper than the boatloads of ram and throwing SSD cache at the problem that I'm currently using.
I slowly upgrade to max memory on any given platform. My main rig only has 64GB right now, but my previous were maxed out at 128GB and 256GB. I'll wait for DDR5 to slide down in price before getting the next 64GB.
Hmmm, did you have to remove it again to add thermal paste?
Lol at 15 min he totally called us out! That's exactly why I watch to reduce the risk of wasting my own money and pick up some ideas and practical how-to's along the way! Thanks man!
We have a couple of quad-socket VMware boxes with 3 TB memory each that we had to retire due to not being on VMware's HCL for ESXi 7. I set one of them up as a hardened Linux repository for Veeam. It's ridiculous. 56 cores. Current load average, 0.09.
It would make an awesome ProxMox server!
*me having just bought 128gb thinking I am going to have a baller system* "A terabyte of RAM" ME: AWWWW MAN
So, the CPU thermal paste - has it ever been applied?
That 500optane chips are going to need hundreds of videos to be paid!!! Cant wait the playlist
This one’s been on my bucket list to try out. Thanks for a great episode!
17:40 Are you going to try to run a minecraft server with this? If so I would love to see that.
i just keep on watching when the drink goes down
Awesome Video! What's with the Creeper though ? :-)
i paused it to look up the beir
and was impressed by the list of biers in the series/moniker.
even though i know what lagunitas is now; id give the series a view.
Great Video. Thanks for putting in the time!
Was there thermal paste on the new CPU?
Under 0.01 ms storage latency? My latest gear has latency measured in minutes. I wouldn't mind seeing your thoughts about using tape storage in the homelab.
And one more thing, please take care of whatever is ailing your HPE server. IIRC it was also complaining the last time we saw it.
I do wonder how well these would work on TrueNAS with it using the zfs ram caching and such
Is it me...or did you not put new thremal compound between the CPU and the heatsink?
He cut it to avoid thermal paste comments
One of my favourite (and worst) hobbies is shopping our datacenter for boxes that are about to be recycled. If I lug it, I get it. xD Though, these days I do end up recycling most of it as you only need _so much_ hardware in a homelab.
I can be a recycling center; my homelab needs more hardware :-)
I don't know DCPMM in memory mode doesn't have CPU limitation, that's good news for in-memory database like SAP Hana, Oracle DB, NoSQL, etc
loved the creeper... actually, it would be interesting to se how it would handle as a mc server... with that amount of memory and speed?
An interesting thing about the retainer, the two holes on the corners that don’t have screws are different sizes and the posts on the sockets match them, so as long as you match the cpu triangle with the retainer triangle it will only be able to go on the correct orientation into the socket as long as you don’t force it
At least that’s how my experience with these sockets are in the data center
I couldn't find anything explaining how the 1TB limit of Xeon Scalable v2 CPUs applied when using Optane; thank you for mentioning that the Optane PMem doesn't count towards this limit, as this as it was something I was really wanting to know! If you have encountered this information written down anywhere I'd love a reference to it.
You forgot thermal paste or is it a magical connection between CPU and cooler
At 6:55 I am screaming "Thermal Paste!"
Bonkers indeed - but you're having fun and so am I watching your antics. Don't expect me to be following your lead though.
Just wodering if the CPU is running a bit Warm? I didn't see a thermal paste applied to the new CPU. Or maybe it was lost in the edit..?
here i am upgrading a cpu on a 5 yr old b350. we are the same!
Am I the only one that saw him put the CPU back in but no thermal base
I'm just waiting for the day someone makes a 1 Petabyte of RAM computer
FYI for anyone that's not removed one of these cpus from the cooler before, there is a "tim breaker" gap in between the CPU and the heatsink where you're supposed to stick in a screwdriver to break the tension of the thermal paste.
Also FYI, its possible to forgo the plastic clips entirely during installation and somehow manage to not wreck the LGA
Was it Me or Did He not Apply Thermal Paste on the New CPU ???
Great video as always but I need to ask because you are sending mixed signals.
William Shatner or Mark Hamill?
No need for Thermal Paste. Sweet
I feel like I would be really happy with about 16 or 32GB of ram for my laptop as right now I am still stuck with 8.
17:25 IO/PS makes no sense, considering P is for Per. :D
Much like normal CPUs. The larger ones almost always have side of the socket that has a gold bar. Line that side up with the triangle on the chip
You didn't put thermal paste on the new cpu? or did you cut that out.
Thermal paste? Missed a step, unless you did it without showing the footage. I would have liked to see you application method on such a large heat spreader.
The man has me worried...
I was wondering if he missed that step until he shimmy the CPU while its on the cooler block. lol.
I just want this for my Plex database.
Ooo. I like Blue. what is blue in the rack?
6:40 I feel like you forgot to put there the thermal paste...
Man that intro really got me haha
Hi Jeff, I have a test for you. Just try to create a single chia plot. Let it work harder ;)
I hadn't looked at optane nvdimms, but just last week I was eyeballing what it would take to add some small (and surprisingly cheap) nvdimm-n dimms and battery to a Dell R740 for use as an extremely fast zfs zil, but found the documentation on the subject a little intimidating.
Funny timing.
Love you Visor mate!
Just today I updated a server to 96 gb of ram... and noticed boot drive only has 93 gb of space. I salivate at prospect of 1 Terabyte of RAM!
did I miss the thermal paste?
@ 6:40 ... no thermal paste?
Optane mem is almost dirt cheap now. crazy how prices fall a year later. I think its because Intel may not support it anymore or something?
What can you do with it? ZFS give it all to me, Muggle I demand it!!!
one real salient point here is about how the datacenter stuff of today will be homelab equipment soon and why it is such a good point is for the massive performance gains we will all see in the fat part of moore's law (which is extant) - we are talking about orders of magnitude improvements here and it is going to have real world impacts as this type of enterprise grade kit filters down - the prices are great and this is a non unobtanium jump due to the reasonable prices - yeah you were in the right place at the right time but deals like this are going to be commonplace before you now it as time and tech marches on - a couple of these are more affordable than a used car - you probably do want 2. pleae run some more benches to put into perspective how much work these types of machines can do. faster bandwidth and fiber proliferation is going to help uptake of this type of tech ---to fully take advantage you will want 100g - maybe that could be a followup that will fully uncork all the potential here - it is not much more per port vs 10gbe...
Would downloading the weather data set and trying to predict the up-coming weather use up the available memory?
1 TB of ram..... hummmmmm you must play Cities Skylines too.....
and @6:34 of course you remembered the thermal paste just off camera right?
Oh re-reading that I thought "damn keyboard warriors" but it is a legitimate question for my ever curious mind... :)
I hear the nerd, geek in your cadence and enjoy it so much a 40 years ago I was a nerd geek working with PDP-1170s I love your channel and wish I could afford just the ram that you installed lol...
Thanks for the ride along and sharing your time to create, edit and UL these videos for me to geek out with you on... :)
with a dimm slot per of 6 max ram would be 6TB at 512GB per dimm
As much as I could use optane at 512GB you would max out at 2TB at 1TB per side|
personally if I was going to run optane in ram slots if technically feasible to do so it would1-3TB options when financially feasibility to do so.. you would likely max out a regular dimms 1-2TB sticks
On a 6 dimm per side you would max our 4-8TB per side
07:00 no thermal compound?
did you 3d print that visor/stand? stl?
wait no thermal paste on the new one?
I would love to see a method of using RAM as a read cache for databases within windows without moving the files from their slow medium.
I know you can pull off a safe way using ZFS and linux for this read cache, but I've yet to see a method of doing this for windows.
Current best guess is to create a RAMDISK/CACHE in Linux using ZFS and pass it to windows as an ISCSI device which really complicates things.
0:30
needs more egg with the camera
It appears that you forgot to put on fresh thermal paste, unless you did that off camera...
What’s the HDStor box about the board with the SATA and power cables coming out of it? Is it just somewhere to put 2.5” drives?
You left out the part of putting thermal paste on the CPU :)
My biggest problem with ANY NAND-flash-based solution is that ALL of them has a finite write endurance limit to it, which I will undoubtably, wear out.
HPC applications just produces SO. MUCH. DATA.
True, but for a home server/home lab use case, the endurance of these components is more than enough. I've got SAS SSDs from 2017 that measure write endurance in the >1 petabyte range.
These components are made to live in big iron, inside real commercial data centers.
I'm not doing high intensity compute or commercial databases or something; I won't be wearing these drives out for a good long time with my two blogs, a couple of minecraft servers, and my never-quite-done assortment of various status monitoring dashboards. Unless my usage pattern gets MUCH more intense than it is now, they should last for years.
I suspect the same is true for these components in Jeff's servers.
@@sinisterpisces
it depends on how much data is being passed through if his system is also hosting the raw video footage for his TH-cam videos.
For my HPC/CAE/FEA/CFD applications, it's a never ending monsoon of data.
It's not hard for me to generate about 110 TB of data in about 15 minutes if I am doing a FEA run using the direct, sparse solver, at which point, to average out the write endurance, it's just a matter/function of how often I perform that type of analysis.
This is one of the reasons why I've decommissioned almost every single one of my SSDs because they're like brake pads on a car -- they're going to wear out.
It isn't a question of IF they will wear out, but a question of WHEN they will wear out.
Can you run grok?
Hey Jeff.
What rails do you use for your server case? 🤔
Why is it called ram instead of rom remembering memory instead of random access memory ?
Damn I wish I could afford this or something like it. My whole server runs on an i5-3470t, 16gb ram, 2 x 1tb (raid1) and 2 x 3tb (raid1) in a microserver gen 8.
I have HyperV for PiHole and Plex and then its just storage for everything else!
Too bad it's cancelled now, I had so much hope for 3DxPoint stuff. Imagine no need for RAM anymore, and having 3DxPoint NVRAM in DIMM slots for permanent/non-volatile storage.
You didn’t use any thermal paste with the new CPU!
My lab has a Dell R930 with 4TB of RAM and 4 CPU in it
Server or workstation? What cpu?
I haven't watched this yet, but my first question would be are you doing some kind of RAM-DISK with with like Primo Cache? Not that i believe in 'overkill' lmao, but JESUS
no need to with optane, it's already a ram/disk.
Thermal paste on new CPU?!
but can you mine chia on it?
6:49 what about thermalpaste?
6:25 thermal compound?
How does it affect the dram memory speed (fsb?)
Thermal paste ???
Doofenshmirtz cosplay when
🎵🎶 Craft Computing Evil Incorporaaaaated! 🎶🎵
Nice, I want a terabyte of ram and a storinator :(