RAM just got a LOT faster! Up to 10,000MHz!

[@DerekFletcher1]

This is honestly one of the nicer sets I've seen on the channel, guys. Seriously, I would prefer all talking head vids with this set up.

[@drblitz3092]

Nah. Looks to corporate office. Wheres the character .

[@---GOD---]

Lol what? This looks so boring... I don't want to get a lecture on finance in a board meeting...

[@Surms41]

@@drblitz3092 It's literally a window showing nature. That's the pretty part.

[@definitelyfunatparties]

@@Surms41 right.. we want watercooled rigs in the background, nature can stay outside, where it belongs.

[@HiddenHelixj]

Less is more

[@ScrakSFMs]

To be honest the scene is pretty nice! It feels like natural light more than a sad bunker background. No hate!

[@Nobody-vr5nl]

Ur not wrong. I worked in a sad concrete bunker. No windows in our small little tech shop. We had to roll up the garage door to get sunlight and air. It was depressing, lol.

[@ScrakSFMs]

@@Nobody-vr5nl lol hope the best!

[@theendurance]

nah bro, it has to be a dark rgb-filled "man cave". thats what the gamerz want, they dont want to look at grass even if its through a window

[@groenevinger3893]

@@theendurance maybe for some of you.. i like working on pcs for about 30 years now, but im a professional gardner tho.. so i like the grass and greens 😁

[@ScrakSFMs]

@@theendurance Damn that is sad. The saddest thing I have ever read from a Virginia.

[@reidsampey2045]

The sound in the video was fine. No discernible noise from the highway.

[@thatdrh]

My pockets are deep... but then again, theres a hole in the bottom...

[@AMDGuy88]

Well my arms are too short to reach my pockets and my wallet is glued shut!! HAHA

[@wossle73]

Games, are the hole in my Pocket😉 🤣WHY have Ultimate Speed, If you can't PLAY!👍

[@jeremylindemann5117]

Memory leak detected.

[@Duke49th]

My wallet is made of onion-leather. Every time I pull it out and open it, I start crying.

[@VicariousAdventurer]

They think we're DIMM

[@superpandabacon]

Damn. Looks like a great spot to film

[@kurobara9001]

Agreed. Put a plant in front of that electrical junction and Jay's got a spare set.

[@NocturnalMantis]

The more random the locations for the talking heads should be done. I hope the next one can be done in the closet or the bathroom or on the staircase any of these places would be fun.

[@OntarioGuy430]

I second the bathroom - it looks like a pretty nice size!

[@ChrisPza716]

I approve this message

[@Jeff-od3zk]

I approve the approval of this message

[@velvex77]

Bathroom approved

[@ash9803]

I feel like I'm being interviewed for an office job I don't want.

[@Z3r0_H0r1zon]

I work for my local university's chemistry department as their equipment tech. I can see benefits of cudimms in the computational chemistry. The atomic motion and molecular structure simulations that get coded could see a huge uplift in decreasing the time to compile and run them. I plan on doing a side by side test with cudimm vs regular udimm .

[@keeperofthegood]

I actually have to feel that "you" are the target audience with new tech, people that work with large data sets and large calculations. The tech of DDR4 really does not otherwise let users down. Minecraft isn't more creative, reading an email is still one word at a time and RGB is forever.

[@0Blueaura]

Cummdims

[@johnt.848]

I was going to say their use in crunching would, or should speed things up.

[@Angel2kinds]

are there ecc-cudimms? I can't imagine running anything serious without error correction. I work as a server architect, and the consistency of data is very important.

[@AHHGKS]

I would also hazard to guess that the biochemistry folks and/or molecular/drug design folks might see some benefit from this tech in the next few years, or whenever they decide to rebuild their lab's computer clusters. Though, that would probably be the higher end simulations which would require it. If I remember correctly, that would be the things which generally required being submitted to the Departmental servers to have the simulation data crunched. Probably in the 5-10 year range is when I could see the individual labs being able to afford to fully rebuild with the latest and greatest CPU/RAM technologies. Well, unless the individual professor's lab is flush with cash, which tends to be the organic/biochemistry folks who are working with medical related stuff.

[@cr4zyj4ck]

It makes me giggle reading the comments, Jay's talking about extremely cutting edge memory tech and half the comments are about how nice his room is.

[@killerdoxen]

I watch very few in video ads, but your ifixit ones I don’t scroll past. Pretty funny stuff at times.

[@v1Broadcaster]

Always helps when it’s a product you already know you want for sure

[@KellicTiger]

I think CU + CAMM2 are were we are heading. The CAMM2 will allow the RAM to be closer to the CPU for shorter traces, and then you have the CU which also is going to help things.

[@scarecrow5848]

yeah, CU-CAMM2 on AMD would be so sick

[@NonLegitNation2]

and it shall be named CUMM

[@EthanAQueen]

Unless the prices go down exponentially vs CAMM modules it is going to be a non-starter. Dell charges over double for a CAMM module compared to the same capacity and speed DIMM... and they price their DIMMs way higher than buying from a 3rd party.

[@TX7410]

I really hope we see a push for more CAMM2. I would love it if they could move it to the backside of the motherboard as well, as that would free up a lot of space on the front, specially for mitx boards.

[@Demonoid1990]

I'm anxious to see CAMM2 on desktop systems. But Der8auer replied to one of my comments on CAMM2. Saying he's not expecting to see it adopted on LGA1851. Kinda left me a little confused considering several manufacturers at CES and Computex debuted several CAMM2 motherboards for Z790 & Z890. Either way I'm anxious to try out either one on Intels Arrow lake CPU, and compare to my home server with the 14700k. I had a lot of stability issues with it, not so much the CPU, but it's been constantly throwing errors for the memory. My personal system is currently a laptop with 9700k, which I'll be replacing with a new desktop system here relatively soon. Just gotta wait for reviews, and then wait some more for next Gen GPU's..

[@OntarioGuy430]

That is a really cool looking spot to film. I say keep it!

[@cerealport2726]

yeah, i have never really understood why youtubers spend so much on sets when this is just fine...

[@RetroNook]

Agree, a nice mix of tech and nature.

[@petermckenzie580]

Yeh the lighting and aesthetic is actually pretty good already.

[@kurtwinter4422]

Don't block the windows Jay

[@rtyler1869]

I agree. Keep this spot. Looks really good

[@welcometobobs]

Honestly this is a great place for videos where you do more talking / explaining! It’s a nice switch up

[@DannyCTalent]

I'll be honest, I would not mind if you had to shoot more videos in this setting for the future

[@dragonmaster1500]

I think this could be useful for heavy Data science tasks, especially in the GIS industry. We often deal with LiDAR or Orthophoto Datasets whose sizes are in the in the 10s of 100s of Gigabytes range, which means that most software just can't handle how big the dataset is. Now, you can tile it, basically cut it into smaller chunks, which makes it easier to process and then merge it back together at the end. But the thing about that, is it means that you're constantly loading data into the system memory, editing it, and then saving it to disk before grabbing the next piece, so anything that can increase the speed of loading, unloading and accessing the system memory can pay dividends, especially if you're dealing with a pipeline that can take more than a week to run.

[@LordRykard9376]

I run exponential random graph modeling on large social network datasets (hundreds of thousands of nodes and millions of edges) that requires tons of memory as many of the libraries are still based in R and not Cuda-oriented for parallel processing. This would be a nice step in that direction. Or maybe someday somebody could write igraph to support GPU processing!!!!!!!

[@dragonmaster1500]

@@LordRykard9376 Oh, I wish. I'd love to run more of my pipeline code on the GPU but barely any of the libraries I use, aside from the ones specifically dedicated to Machine learning like Pytorch and Tensorflow, actually take advantage of all the free processor cores.

[@LordRykard9376]

@dragonmaster1500 pytorch handles GIS data? I didnt know. Have you tried pyqgis on qgis? Or Geopandas? I might have to deal with Geospatial data in the near future. Might be good to know.

[@Danielle_1234]

I'm a data scientist. A bit of 102 information: All DS math done on data hits system ram, because SIMD is the fastest way to number crunch on a CPU today and SIMD hits ram. So any ram speed boost, even for smaller datasets is a speed increase. Also, it's not called tiling but streaming. Same idea, where if a dataset is larger than system ram parts are streamed in and out, kind of like page file in Windows. Today new libraries like Polars and DuckDB allow for easy streaming, which is really nice. It was only a few years ago when if a dataset was larger than ram on your local machine you'd put it up on a server with tons of ram and run it remotely. It's only very recently where we can do this kind of work locally with streaming, which is sooo nice. In these kinds of workloads you're bottlenecked by the speed of your hard drive. In the cloud you can't get really fast HDD speeds, but on your local machine you can splurge on your hard drive and it can reduce computation time from minutes down to seconds or hours down to minutes. For super number crunch heavy workloads it gets offloaded to the GPU but most workloads are more ram and hard drive speed limited before they become computation limited.

[@LordRykard9376]

@Danielle_1234 have you worked with GIS data? Specifically LiDAR? It's called raster/vector tiling (its specifically how GIS data is stored, processed, and visualized and software like ArcGIS handle it differently than sequential or columnar so DuckDB most likely wouldn't be the go-to for that type of data. Also it sounds like he's actually building these tiles so unless he or his work are building their own in-memory distributable systems, they will need to look to the big boys for processing or they have to build their own architecture if they get into terabyte territory. My work has racks of A100s we base virtual shared memory systems on almost exclusively through PyTorch. But it's all raster, no vector (way easier than what he's doing).

[@laddken]

Might look a little corpo but it’s clean so it looks good!

[@manitoba-op4jx]

needs some wood paneling

[@dongarippo7279]

I preferred it as it was before, this feels more professional, like he wants to sell us things, rather than the PC builder enthusiast he is. Well, as long as he does great stuff as always.. who cares, I guess :)

[@ChrisPkmn]

Put a tire on the wall and this'll be fine

[@haariger_wookie5646]

Love this set. Much better than all those sad basement setups that literally every KZbinr uses. Put a plant in front of that outlet and maybe some tech items (EVGA Shelf?) in the background and it’s a great set.

[@davidlt]

CUDIMMs are compatible with legacy systems. From SemiEngineering: PLL bypass mode: In this mode, the clock buffer just passes on the input clocks to output without any kind of signal buffering. The PLL bypass mode enabled CUDIMM devices behave like traditional UDIMMs without any buffering of the clocks. This is why it’s also referred to as legacy mode. Recommended CUDIMM operating speeds in PLL bypass mode are typically limited to 3000 MHz. To my understanding if CUDIMM is in "PLL bypass mode" (one of 3 modes) it will be running at default/base clock whatever that is. Probably one of JEDEC base clocks.

[@Drumaier]

yes I think he got that part wrong. In those cases the signal just wont get cleaned up but it will work just fine as a typical ddr5 (and without the max speed obviosuly, but is not like the system won't boot or anything)

[@abidlack1980]

I can concur the audio is fine and no noticable outside noise. Plus its a nice change from solid walls.

[@zeeunveil1324]

It’s been years since my last upgrade and Im still happy with my 3060 ti, 5800x3d, 4tb of nvme ssd up to 8k writing speed and 3200mhz rams. Don’t really have a reason to upgrade further, although having a 3090 ti would be nice but it just pointless when I have no use for it lol

[@keithasche]

I agree with you right now Jay. I feel that the innovation in the new laptop memory slot (LPCAMM2) is more relevant right now over CUDIMM.

[@Reckhs]

Thanks for the info. This was the first I heard of CUDIMM. I'm not on the bleeding edge myself, but it will be fun to see how it pushes things forward. In the meantime I'll enjoy making a new NAS and figure out self hosting a dedicated Satisfactory Game server. Fun to see the new space.

[@shannonrhoads7099]

Nappa: Vegeta, what does the scouter say about its megatransfer rate? Vegeta: (crush scouter) It's over nine THOUSAND!

[@adrastoso9727]

Never gets old!

[@flukshun]

Goku is what it looks like when Moore's Law keeps going

[@xshadygorillax5912]

Glad to see I'm not the only one that thinks of this when seeing 9000 anywhere

[@HupfderFloh]

It's ... 1006. Kick his arse, Nappa.

[@tekjunkie28]

CU-DIMM looks to be pretty DIMM interesting

[@TheKazragore]

🤦 Goddamnit, lol

[@turambar304]

​@@TheKazragore😂

[@yuzukikittz3398]

​@@TheKazragore God DIMMnit

[@kevinerbs2778]

Its 30 to 40 years old tech coming back though.

[@xlr555usa]

Get Bit

[@jensdybmose]

I hope you will keep using the conference room for talkinghead videos, i love it having the highway and bushes and trees behind you, so cozy.

[@animedreamnetwork245]

i am liking the office setting for talking head videos. happy for the new move. The only con is that room has a bit of an echo.

[@Mr.NadzIsBetter]

Love the background, you deserve the new office man!

[@Reevesi]

I miss the raspberry at the end of the video. The "ting" doesn't feel right

[@iamaduckquack]

I hope their silliness doesn't get phased out. I like the humour.

[@krjal3038]

Maybe we could have the sound of Jay getting zapped by whatever he's unpacking on a given day

[@何浩平]

I know It's temporary but I really like the set. You're on the desk like a BOSS. The lively windows on the background is actually not bad.

[@Megaman_308]

Jay, do a video explaining the different Gears for memory overclocking and how they work, please.

[@nick066hu]

11:30 "Why is this 32 Gigs of DDR5 CU-DIMM are $400 and this 32 Gigs of non-CU-DIMM is $200 ? Why is the $200 gap?" Because on the CU-DIMM they added two extra chips: a PMIC (power management controller) and a clock driver IC. ... and these cost $0.17 each 😛

[@patricketheridge7853]

Jay, this looks like a good set to use in the future

[@mashpotato16]

i actually really like the way this video done with the background scenery - so clean and fresh - just feels natural

[@EidolonKaos]

The gray wall is a little sad but the window is great

[@rotnbazturd7569]

Rendering and audio/video stuff will benefit a lot from this. Another thing to think about is the CPU will be running harder to process the increase in data. At what point will the CPU bottleneck the memory ?

[@networkarchitect443]

Effectively that bottleneck would never happen with current CPU architectures unless the memory technology changed . RAM is relatively "slow" compared to the cache memory built into the CPU: both bandwidth and latency are orders of magnitude slower. The L3 cache in a modern CPU is capable of ~1000GB /sec compared to ~70GB/sec for DDR5 dual channel. Latency on the L3 cache is ~15ns while DDR4 is ~80ns. Look at the Ryzen X3D line of chips to see the impact of adding more cache on performance. Consumer CPUs are bandwith limited by having only 2 memory channels (4 sticks max). Workstation and server platforms like EPYC or Threadripper can have 8 or 12 memory channels which greatly increases bandwidth. Even on those platforms, memory bottlenecks are possible with large amounts of I/O.

[@Skelath]

I9 14900kf already throttling whenever the anti-virus does a full scan of my SSDs that have over-provisioning enabled to achieve the SSDs maximum rated speeds.

[@squidwardo7074]

@@networkarchitect443 Im guessing the actual traces on the motherboard would be the limiting factor long before the CPU is

[@kevinerbs2778]

​​@@networkarchitect443 the zen+ architechture could do 288 gbs of ram bandwidth on Epyc. If you go from that to to 12 channel you getting 3 times that bandwidth which does not count for increased speed too 288 x 3= close to 900 gbs add in the 1.25% increase and mhzand it would easily hit that 1,000 gbs bandwifth now. The bottole nwct is starting to be links like infinity fabric & pci-express. Along with the memory controler on NVMe drives which end up overheating from the high speeds.

[@TurboLoveTrain]

@@networkarchitect443 You're confusing L3 cache with CPU throughput. L3 is not part of the CPU as much as marketing teams try to convince people it is. There is absolutely a RAM bottleneck with x86 and literally why it runs so hot. They get around it by increasing the clock speeds not the bandwidth in L1 an L2 (which are the ones that really matter and are of ancient design on x86). Servers have more channels because they use different prefetch hardware. x86 has always had trash memory throughput on the actual CPU and people pretend that the added ram controllers on the dye ant extremely high clock cycles make up for the total throughput to the CPU--they were always a band aid just like the on board "AI "to predict fetch requests.

[@TheBebe666]

This setup is nice and clean without any distraction behind.

[@alexandrea4316]

600 dollar Motherboard, 600 dollar CPU, 600 dollar DDR5 16GB kit. Woah Intel is really revolutionazing!

[@andrewt.5567]

GPU are already 600. Please dont give the PSU guys any ideas.

[@WindySilver]

It's such a mindfuck to see a different set after all these years. I like this one even though it has a bit of a corporate vibe, though! It's refreshing to see some nature - and since it's autumn where I live now and there's not so much green anymore, it's nice to see the greenery.

[@TheInternetHelpdeskPlays]

2:33 My first build used the brand new SDRam, the forerunner to DDR, my first sticks ran at 66Mhz. Eventually upgraded to 100Mhz

[@keeperofthegood]

Hey Jay :) I'm one of the "work form home" type people, and for my work system, I have an N100 Mini (I do have a personal rig for games). Locally run Teams and otherwise remote into my work station. While I can run Office365 local (sometimes do when chasing a work idea after hours) that is all I need in a computer. It works, works clean and fast and is mostly stable (yea Intel and Graphics in the same sentence is...). I really am not looking at the latest or greatest and feeling I need it. I know I've said it before here but it still applies; why drive a Lamborghini when you live in a school zone and the only driving you do can easily be done in a Ford Focus? As to that personal rig, still with early gen ZEN and DDR4 and not really found it lagging in a way that I would want to move up in generations. Could do with better graphics but then we always can do with better graphics 🤣

[@BlackHoleForge]

Hey Jay don't apologize for trying to look professional.

[@kobby2g8]

It's the perfect location to discuss such a topic. More of this!

[@BTCKYLE]

I love the grey and white rgb lighting

[@chillnspace777]

With ray tracing on the glass

[@robertm9906]

I haven't heard of CU-DIMMs until this video. I don't plan on upgrading since it's not that simple to use and there are a lot of factor to consider. Mainly I want to wait to hear the reviews of CU-DIMMs being tested and results of real world performance if it's even worth it the cost for that extra benefit. Btw, it feels odd watching Jay with greenery in the background. Maybe I haven't watch enough of Jay's videos especially the older ones. I can't remember seeing one video with plants in it. I could be wrong, but it's definitely rare like maybe trips to Micro Center. Those maybe the only videos I can think of.

[@rich48439]

Please include factorio in the u vs cu testing. Thanks!

[@glasseyemarduke3746]

Especially after the huge uptick in interest due to their new DLC that just launched

[@dddddbbb]

In case it helps Jayz - I am someone still on ddr4 as I didn't see enough of a jump in speed to justify the change to ddr5 yet as it seemed like the RAM speeds were far slower than I thought should be possible on that platform. With CUDiimm speeds now unlocking faster ram speeds it now might justify the change in platform to upgrade my mobo/ram/cpu. I look forward to seeing what the performance is like - particularly in games fps! Thank you for your content!

[@celcius_87]

Having no QVL seems wild

[@Jkend199]

Hell yea... in like 3 years when the prices come down to something resembling reasonable...

[@Fodiaoner]

I really like this angle!

[@EsotericArctos]

I can see this being useful for developer situations where they may be testing databases stand alone before deployment. The extra RAM speed and stability could be really beneficial for this.

[@SpaceOrphan]

This film location is incredible.

[@smithyMcjoe]

I actually bought an iFixit kit because of your adds. Honestly, one of the best purchases I've ever made!

[@Phil-q7h]

I ran DDR4 3200 next to DDR5 on 5600 on the same sort of set up. It made a grand total of……about 2fps. 80fps to 82fps, so realistically 2.5% faster. For a lot more cost. If you’ve got ddr4 board with a modern Intel processor, don’t bother changing.

[@ILoveTinfoilHats]

Look at your latency values. Games are very dependent on latency.

[@AKVvirus]

I didn't mind the look of this video because you were able to light it correctly and nailed the exposure so the windows did get blow out in the background. Can't wait for future videos in this new space.

[@Diafragmates]

I like the film location actually

[@mar504]

This technology will certainly help with deep space astrophotography processing. I collect thousands of raw photos which all go through a variety of treatments - measured and quality weighting, calibration, alignment, background normalization. Then the individual frames are integrated together to increase the signal to noise ratio, or sometimes drizzled to increase resolution (a process developed for the Hubble images). My last image took 10 hours of compute time to complete these steps, faster memory could certainly help get this done quicker.

[@glockmatics]

Corporate Jayz, how about those Mondays amiright?

[@HartenDylan]

Jay, this should definitely be the set for optimistic talking head videos! Lots of light, foliage, interesting angle. And then when you gotta rant about _____ company, you can do a darker moodier set to add that extra punch. Keep up the great work!

[@elitedimegaming5290]

The ram sticks are becoming sentient lol

[@AvatorGotti]

I’m liking the new vibe!! Can’t wait to see more content from the new lab!

[@Karn0010]

Once the prices come down it'll be nice. For just gaming, my guess is the price to performance will be dog shit for a while. If you do a lot of memory intensive stuff it'll be noticeable. It'll be interesting to see if Cu-DIMM and CAMM start a war within the memory ecosystem and which one wins out. I personally like CAMM from a motherboard space and clearance perspective. At least people are out there trying to get more out of memory.

[@kevinerbs2778]

CAMM DIMM's are just a different from factor, but if they can get its traces shorter by putting behind the board i would like that. You can eaily have both a CAMM CUDIMM.

[@ronaldmail00]

I love how he apologized for the sound in the room from the cars going by, right after all the explosions for his screwdriver set advertisement.

[@keniomo943]

Wow. It looks more vibrant

[@CeceliPS3]

I never bothered with RAM as the main thing that gives more FPS is the CPU. The only thing that could sell me this is if it gives either a huge bump in FPS or it makes damn near flawless when it comes to stutters.

[@schuck6566]

Some of us are old enough to remember rambus dram and intels deal with them. Look how well THAT worked out! 😂

[@HectorQuien]

Exactly!

[@PoisonNuke]

but Rambus was just too complicated and had to many ICs on the DIMMs. Therefore a lot of heat and integration was not easy. The CU-DIMMs here are way less complex then RD-DIMMs. And the interface is also less complex. And Rambus introduced super high latencies, wheras CU-DIMMs supposedly reduces the latency. So better chances for CU-DIMM to be successful. lets see the first real reviews and if the uplift in cost is really worth it.

[@greywizard2557]

Some of use are old enough to remember the first time round when RAM was soldered to the motherboard.

[@cloud1930]

Thanks for the memory update. Good explanation too.

[@xavstier]

Background so good it looks fake😳

[@BladeCrew]

I love the background Jay. When you make these talking head videos could you make them there. The lighting and background looks really good.

[@keithmyers2170]

The comment about latencies being better on CUDIMM is incorrect. Look at the announced CUDIMM kits, where primary CAS latencies are ~10 - 20ns worse compared to stock DDR5-6000 UDIMM's The higher Gear 2 clock speeds won't make up for the increase in primary latencies.

[@MrEricmopar]

Yeah, the reality is that DDR5 still can't outperform DDR4 3200/3600 because of the CAS Latencies. In fact Toms Hardware did some test suites of DDR4 vs DDR5 using current MOBOs with a choice of supporting either ram, and the results are shocking. Basically within the realm of errror, it's that close. Latencies still matter A LOT, with tons of games and programs and some run worse on the DDR5. Seriously... If people don't believe me, look up the Toms Hardware tests and others are out there too, from as recent as 2023.

[@kevinerbs2778]

​@@MrEricmoparthis has been true for ever new interation of ram going from ddr to ddr2 to ddr 3 to ddr4. Bandwidth always will out weighs that latency hit in the end. This old tech been done before, its just coming back.

[@SeventhCircle77]

True but it seems it's more so jdec of 6400 meaning it would be at 1.1v

[@mojojojo6292]

​@@MrEricmoparThat test is terribly outdated. The ddr5 tested is very slow with poor timings. The game suite tested is also very outdated

[@krisc1684]

​@@MrEricmopar I'm not saying I can guarantee that it was a bad test, but I wouldn't take any testing done by Tom's Hardware at face value, not this day and age

[@harkkyn]

Truthfully I'd keep the scene and this space for this style of talking video's. I dig the vibe it's putting out. 👍

[@nbrowser]

ah HA! JayzTwoCents is a Corporate shill....CONFIRMED! hehe! I kid I kid dude, the background actually looks classy. Looks like the move was a good idea.

[@alexandrea4316]

He's got a good Intel sponsor recently that's for sure.

[@sylvie_v2939]

CU-DIMMs are becomming only one of two options in Laptops -- soldered memory or CU-DIMM memory (which vendors may have used interim standards for*). For trace-length, and possibly using all 4 DIMMs - CU-DIMMs will become necessary in these applications even before 6400mhz+ speeds are considered.

[@adamdais2824]

Liking the new rooms

[@jk_46]

Jayztwocents is still such a gem of a channel! Keeping it real and interesting!

[@Kapono5150]

I would build a second PC to try out faster Ram speeds but I just don’t trust Intel after how they reacted to Raptor lake

[@truefoa]

Wow. Things like this make think about gaming in 10 years time when new (now) tech is old and standard.

[@Just_Pele]

All I want is a new HEDT platform so I can build a new pro workstation, because regular consumer grade stuff from Intel and AMD is flaky AF and just doesn't cut it.

[@bradhaines3142]

those are dead because they eat into the server hardware. they want you to either live with high end consumer or low end server parts

[@bocahdongo7769]

​@@bradhaines3142AMD releases new Zen 4 non-Pro TR not too long ago If you forget those stuff exist

[@kevinerbs2778]

​@@bradhaines3142what are you talking about? Threadripper has TRX50 and WRX90. None of you buy it because its cost is in between expensive serves & desktops in cost. HEDT is right there, matter of fact zen 5 threadripper 9000 will be faster than current desktop. Since it has higher bandwidth ram.

[@thephantomchannel5368]

@@bradhaines3142 Desktop from both AMD and Intel are much closer to mobile/ laptop than they are to server/ workstation. There is room for desktop to evolve to something better than what could be considered "mobile-extreme" which is what consumer desktops have become. The price for consumer desktops is probably the most disappointing aspect about the platforms for how gimped they are with I/O compared to workstation/ server. Desktop needs an overhaul and should be considered HEDT by default and the pricing for current desktop needs to drop by at least $100-200 dollars for being basically a glorified console. As another comment points out, Threadripper does exist but it is priced well beyond what the average consumer would spend unless they actually had a purpose for using it for its advantages. Even a stripped down non-Pro Threadripper with half the capability and half the price would still end up being twice as good as the best desktop and priced the same. If either Intel or AMD could come out with a desktop platform with a quad channel memory controller and at least 32-48 PCI-e gen5 lanes direct from the CPU for the same price as the "high-end" consumer desktops that we have now ($500-600) I would guess there would be quite a large demand. Sure it might eat into the lowest tier Threadripper sales but those platforms advance far faster than consumer so it won't be long before the performance delta is even larger than the hypothetical HEDT system that I just came up with. Given how the new Zen5 server parts are performing, the new Threadrippers when they eventually come out will be insane and just make consumer desktop look like even more of a joke and a ripoff by comparison. As it is, Intel Arrow Lake will cost a minimum of a thousand dollars to build out a mid level system assuming the CPU is $500, the mobo is $500, and the new CU-DIMMs are $250-500.

[@neolordie]

Threadripper

[@rudymaldonado667]

Great location to film these kind of videos. Very mature and refined lol

[@jacobthomas90]

*me rocking ddr4 2666 cl 16* Ah yes, I can use this info... haha, Keep up the good work Jay and Co!

[@abidlack1980]

Jay, youre known for a lot of things, but thats why youre fun to watch.

[@macvirii]

The 5th grade in me thinking how CU in Brazilian Portuguese means "a****hole" xD Thanks for the info, I was researching about it last week

[@tedll75]

I just got to say that the IFIXIT commercial is the best commercial ever!

[@deptofcarstereorepair]

i'm so glad memory manufacturers found a new gimmick to jack up prices again!

[@intuitivme]

Fine background for a talking head video. Jayz enjoying his new office! Congrats!

[@davidhines7592]

cudimm is going to make ddr5 cheap like ddr4 so im all for it. i can move to ddr5 while everyone is on faster cudimm and working through the bugs

[@CeceliPS3]

I'm still on DDR4 and I've got a 4090. I saw no significant benefit of changing everything on my build just to accommodate DDR5. If you're thinking of making an upgrade and you're going to that when you know new tech is already released, you're just shooting yourself on the foot.

[@EbbeChristensen]

Can we change the name to CUMM

[@IIHydraII]

​@@CeceliPS3FPS %iles (general smoothness of games). Xmp ddr5 > tuned b die.

[@victorkalinin553]

@@IIHydraII ddr4 has less CL, so is difficulty to say that ddr5 is 100% better at gaming

[@ziggs123]

​@CeceliPS3 why though. A 4090 to get "bottlenecked" by RAM is such a weird decision imo. AM5 will be there for a long time

[@maconijnr]

No matter how many years i have been watching, the ifixit ad still brings me a chuckle :)

[@skallamadura3277]

NGL.......... I like this 'scene' lol.

[@TheAlethian]

the best ad that was actually watchable. good job

[@KiwiHelpgeek]

Wasn't Dell's CAMM supposed to be the "next big thing" when it came to RAM? Is that already a dead duck technology?

[@bradhaines3142]

i dont think its dead but I haven't heard anything past dell laptops using it, might have to give it a couple years

[@yumri4]

Outside of what reads to me like workstation tasks I don not have a use case that quicker system RAM would help. I am hoping the cost of CU-DIMMs goes down as they might be trying to force it onto the market due to how few avg consumer processes will actually benefit from it. Spreading out the cost over millions will help the cost go down instead of a few hyper specific processes which will lead to very few people buying it.

[@kaseyboles30]

Will there be CU CAM2 modules?

[@CatalystDestiny]

While I say you should include regular UDIMM's in arrow lake, yeah you really need to have BOTH kinds of ram tested with arrow lake to really give it the real picture of how arrow lake will do, especially with X3D so soon in order for people to make proper choices of which platform they want to go with based on the complete full data set of all the capabilities of both sides offerings.

[@user-le9gu6lp3x]

What about Gearing and Latency?

[@thangskhs3685]

Nature has this calming effect

[@patrickprafke4894]

And in a side note. We are starting to plateau in PC hardware. We are starting to run out of gains in certain areas.

[@familhagaudir8561]

Recent CPU releases sure have been a let down, as if we are near a limit. Even AMD's supposed energy efficiency gains on its "performance-stagnating" latest chips come with a massive asterix of "only for certain workloads" (not gaming). Intel's new releases also seem to boil down to "now you won't cook your room running our CPUs"... well good job catching up to AMD there Intel.

[@crystal_reverie]

Because of the RC load of circuit trace, the edges of signal change is like a slope, and when frequency go up to a certain point those slope will stack on each other and no longer to know where is "1" or "0". So Jay is not completely correct since CU-DIMM only has buffer on clock to help the quality but not on the data.

[@onthefritzbiking]

For the old timers over here… doesn’t the background give some 1st studio vibes from like 7 ‘ish years ago?

[@Sptn051]

Please include both UDIMM and CUDIMM results in your Arrow Lake video! I'd love to see the differences between the two as far as performance is concerned.

[@GalvinW]

Every time I hear CU I think See You

[@johnt.848]

I think of Copper.

[@Oldmanyellsatcomputers]

Cool the tech is advancing, but limited use when VRAM is what drives me more. I would be interested in the effect on compiling large projects like a bitbake image or something.

[@NVMDSTEvil]

11:02 incorrect, board bios update is all that is needed to use cudimm on any board. It just runs in bypass mode.

[@PhoenixChimera]

Can’t help feeling there should have been one more explosion at the end of the iFixit, also thanks for covering this subject as it’s really important! (More explosions?)