HOLY $H!T - The FASTEST CPU on the Planet - AMD EPYC 9654

it will only last a month! for it burns up. damn AMD and Lenovo a bad combo!!!!

someone who uses this cpu will never run windows anyway

@Lunaire.- Nope just run a single server for Kubernetes.

if you need a task manager then you did something wrong with this beast :D

Yes, but Unix only use this Cores... . Fastes shit... .

@Jeffrey Allen if you had nvidia for your whole life you will be disappointed! The Software from AMD was and still is absolute garbage!

@fakecubed p

Yeah but at that point i guess you should consider a 360mm AIO with the size of the CPU because the 4090 is only that big because of the heatsink

The 4080 is also massive

The moment I read this lynus started talking about that

@ssokolow "[chuckle]" 🤓

[chuckle] Yeah. People generally don't think to make sure their programs handle "filename too long" gracefully. PTS was just trying to soldier on rather than aborting after the first failure.

They might as well just mount the 4090 externally at this point. It's HUGE.

Gonna get to the point where the GPU is going to be a desktop tower that you have a huge extra adapter for.

@cliff bird What you said about Nvidia scamming is baseless bullshit. The 4080 has 4 more gigs of vram than the 3080ti and its boost clock is nearly a gigahertz higher.

not just big but heavy as well. if the 50 series is going to be bigger very few will have cases big enough to fit it in. Ive got a full size case and the 4090 only just fits in. had to take some HHD enclosures out to get it in. mine came with a support rod to support the weight of it lol. it screws into the rear end closest to the front of the case and it sits on the bottom of the case to give it extra support. had probs trying to fit the support rod as i have an intake fan on the bottom of the case where the support rod is supposed to go. its the length that is the prob with them not how many slots they take up. No way will u be able to put 2 of them in a case and use SLI with them cause the motherboard would buckle under the weight of them. you also have to keep an eye on the 40 series cards as the power cables from the PSU to the GPU can melt and catch fire. Nvidia had to get new cables made so the newer cables should be fine but if u get 1 of the 1st gen cables that come with the GPU it could melt. Its a 16 pin cable but not the same type other GPU use. its 12 pins then another 4 smaller pins so an adaptor is supplied so it can be connected to a PSU and its the adaptor that has the fault. the adaptor has sockets to fit in 3 8 pin cables from the PSU. its where the adaptor fits into the GPU where it tends to melt and catch fire if its not put in correctly or becomes lose and its not a very tight fit so can come lose very easy. seems Nvidia r scamming a bit as well turns out the 4080 is a 3080TI rebadged and put in a bigger case

the 5 is going to be 4 full slots. the 4 series is only 3.5. so prepare for them to get even bigger

I use toothpaste instead of that "special" whatever. Never had a CPU burn on me.

man i can only imagine how surprisingly amzed look you would have

Some guys have all the luck.

@dylan chen yeah but how do you even create that "cache drive"? Wait a minute... I just realized that mean my entire initramfs can fit on cache? Maybe create a slightly different version called the initcachefs?

@Mark An older debian would probably run xfce with 768mb of ram reasonably well, you can probably fire up a few programs while using the desktop.

someone has actually done this already. I think with win95

@Mathew Gordon it would literally be instantaneous, only lag there would be is preloading those 200mb on the l3 cache and you know, having more free for other tasks

Storing bad apple in the cache

Windows cant eat all that cores you need linux to squeeze all that cores. Poor tim cook he must be salivating

@Thomas B You apparently forgot that AMD actually got a nice headstart in the GPU market by acquiring ATI. They already proved they were capable and weren't all that small. You make it sound like AMD cut their tiny little corner shop in half to make GPUs one day.

their limitation is being one company and they're smaller than intel and nvidia while competing against both, so half of AMD is making CPUs to compete against cpu behometh Intel and the other half is making GPUs competing against GPU behometh Nvidia

Ahh that makes sense

That's fucking bizarre

@Lux Aeterna performance won't make you all your cores run "all the time" at max frequency. Also, RHEL distros uses performance by default.

you sure its not "balanced"?

Gets annoying after awhile. That's why remote capability is so important. Put the noise elsewhere.

@_lk_s We had a similar issue back in the day with the AMD Abu Dhabi and earlier Opteron 12 core and 16 core processors (we had 250,000+ of them in our Northern Canada datacentre along with various RISC processors which was technically the fastest Supercomputer in the world at the time but we never reported it to the T500 list!) We got an enormous increase in processing speed by creating a super-optimized imaging, encryption, compression, advanced trigonometry plus vector array math routines, and expert-systems database and data-mining routines library entirely using locally-optimized CPU registers for storage and fine-tuned AMD SSE-1/2/3 instructions which we mapped to easy-to-use and network callable C++ functions. This custom library took the internal team TWO YEARS to code but it's still used today twenty years later in our older systems. All that assembler-code expertise did us VERY WELL when we finally went over to our internally-designed and manufacturered super-CPU's that took us over TEN YEARS to design and build for final tape-out which are CISC-based 128-bits wide running on GaAs (60 GHz) and RISC-based 128-bits wide opto-electronic-on-borosilicate (2 THz) substrates. Intel is NOT the only player in supercomputing and AMD is a great buy for those who want a specific type of parallism and multi-processing that Intel CANNOT give you! I definitely notice that SSE instructions mated to common imaging tasks BLOW AWAY the equivalent Intel performance metrics. Rendering a series of raytraced images and/or doing physics-based object/vehicle modelling and collision-deformation physics in parallel using SSE instructions on AMD Rome and Genoa EPYC cpus is an automatic 20-to-30% boost over Intel equivalents !!! Then add in Encryption, Compression and Decompression and long-form multi-language Unicode string searches ALSO make for an AMD strong-point! Intel though STILL is better at gaming on one to four threads. Past four-threads it's all AMD for massivelly parallel rendering work! I should note though, we (i.e. NCA-North Canadian Aerospace) STILL BLOWS AWAY ALL AMD, Intel, IBM, ARM, RISC-V processors on absolute 128-bit computing and outright computer fire-power using our in-house systems though! It's REALLY kinda hard to argue against massively parallel super-computer systems that range from One to 20 YottaFLOPS where a single one of our in-house designed and built fully-ITAR-free 2 THz combined CPU/GPU/DSP/Vector chips is over 50 PetaFLOPS in 128-bit trig, vector math and general imaging computational horsepower! Meh! V

​​@StarGateSG7o. Because linux cant match ibm computers on z/os on reliability.

​@Jeff Hendershota lot of software can use all the cores you can feed some genomics software can eat like 2TB ram and run for weeks nonstop.

​@Jeff Hendershot More cores means more IN-PARALLEL operations per second. For example: I am searching for a room temperature superconductor candidate and have narrowed it down to Yttrium-Copper-Aluminum-Calcium-Magnesium-Manganese-Vanadium-Zirconium-Gallium-Ferric-Silicon-Oxide BUT which actual formulas in what chemical formulas and proportions will make me a FLEXIBLE ceramics-based room-temperature superconductor? I need to test the various atomic/molecular bindings, their group thermal and conductive properties, their structural strength and longevity, and so many other properties such that I will assign ONE processing core for each candidate and iterate through all my data-mining criteria for EACH separate Superconductor Formula in parallel so I can check different formula properties all-at-once! The more cores I have, the more chemical formulas I can check per second! If I rack TWENTY dual-EPYC-processor motherboards that is a total of 96 cores running two application runtime threads per core for 192 threads per CPU, those twenty motherboards with 40 CPUs will have 3840 cores running 7680 runtime threads meaning I can run 7680 SEPARATE chemical formula searches AT THE SAME TIME so I find find the world's FIRST Room Temperature Superconductor! P.S. That above ceramic "formula" is one of the REAL CHEMICALS our parent Aerospace corporation is finding as a possible candidate for a real-world Room Temperature Superconductor! We found that Gallium is Key to superconductivity at -150C to +200 Celcius operation and are ALSO FINDING that the addition of Bismuth and Palladium are ALSO adding some "interesting EM properties" to our compounds! For those of you who need lots of cores, EPYC is no slouch when it comes to multi-threading to find the next Cancer Cure or the next ultra-strong and lightweight artificial Spider Silk cloth chemical formula! THAT is why we NEED MORE CORES!!!! It means that we can look-through and check more items per second in a simultaneous manner to find one or more candidates for a valuable and sellable end-product! V

@StarGateSG7 lol thnx for the info bro, . I remember it was a certain cpu that needed a different memory, I’ve searched and can’t find it, Mandela effect maybe?? It might be the one you mentioned but not sure if it was that far back. I’m guessing it was either a w variant, Xeon, extreme or even threadripper. I just remember Linus saying the model needed different ram and mobo

The problem is still the speed. Writing code that can utilize many cores for most applications is not possible. Rather have high GHz over high core count.

@slash2bot oh wow

We were drooling over the multi million 96 node 64 bit super computer that got installed in our uni when we were there. Threadrippers have been eating it up quite some time ago.

@Pixels Our (bio)fluidics researchers have told me that that is unfortunately not (yet) possible, so they are limited to CPUs. We have "graphics cards" (if you wanna call an A100 that) with up to 80GB of VRAM, so that is quite a high ceiling. I think it is something more technical. We have other GPU workloads (like neural net training) which again consume however many graphics cards they can get.

@Pixels thats not as easy. Sim on GPU got one big problem and thats limited Vram. and it will be problem for many years to come

i would run sims on it stored on RAM. 12TB is finally enough. When you have one sim scene that is 3TB without problem.

Ryzen 7700 65w coming January 10th.

God walk on 10 years in Minecraft dude

@Tom Smith I had an Atari 800, but it was the late 80's to early 90's when other kids I knew definitely had IBM compatible machines that ran DOS.

@Vivago No, there aren't, IBM mainframes are many of hundreds of individual CPUs working together but there is no single CPU that has that many cores. IBM's newest CPU, the Telum, is an 8 core CPU but it can scale to 32 chips per unit and multiple units can be interconnected. The older z15 had 12 cores and the z14 had up to 10, the thing to remember about a mainframe is the number of cores it says is never on one chip it is total for the entire system which is really a lot of server units linked together.

I see something are optical cpu that looks interesting 1000x faster then today's tech the problem we have with tech right now is its reaching limits with how small the semi conductor are at

I started with a 386 and an 80MB HDD lol

If you told me you could get 14 cores in a single package for 300 bucks back then, I would laugh at you.

@Corktail for x86? Sure but in general, no

Yeah bro I just recently started looking into pc parts again to upgrade from my 2016 build and was only seeking out intel 😭

Intel has a $15 billion R&D budget, Nvidia has a $5.27 Billion R&D budget, and AMD has a $2 Billion R&D budget.....AMD is still the underdog and we should all be hoping for their success as the best a consumer can hope for in a duopoly is a 50%/50% marketshare split for a balance of power and AMD is still far from that in x86 (especially in the most lucrative x86 segments mobility and enterprise)

@Corktail No there are more there's ARM as well

@Corktail "my brother in christ, there are only 2 cpu makers" Yeah. I know right? Nobody makes ARM chips and I'm 100% certain that there aren't specialized CPUs you can get from IBM or anything like that. Correction: There are only two mainstream desktop PC CPU brands in the consumer space.

But...will it run Crysis?

@Yuan Liu I wish 256 GB of RAM was available on consumer hardware already.

Bra 😂

@pv2b Entirely possible. Probably the computer equivalent of $800 kicks, but hey, it's whoever's money.

@argentaegis If he can afford 3 TB of RAM to put in his personal server, he must be doing something right with his personal finances.

"Okay, you know about computers. However, if you put 3TB of RAM in a personal server I question your knowledge of personal finance."

They are all super slow. It seems the newer the technology, the slower it is. It is funny how years ago, RD ram was faster than anything that is being produced today. Too bad people get hooked on useless information and play of words to make them think things are faster today. My P4 with 512megs of RD ram was still faster then anything that is on the market today. Trying to use a new computer today is like watching paint dry when you are trying to open notepad.

This guy rams

I have bookmarked this page for this response alone. Thank you for educating us.

@D-no OK. I think I saw it before it was edited and I understood it.

Yeah I didn’t understand any of what you just said

@BH4x0r So what? Use a decent machine with ARM and you'll see how little the instruction count matters in 99.9% of situations. I don't know what your point is, I literally just mentioned a CPU with an insane number of cores, I don't know what the hell you interpreted into that.

@Heath Mitchell a lot of uses will probably split those cores up for virtualisation. And the us3 cases that require that many cores would be very bespoke software applications, in much the same way as software that will take advantage of the upcoming genoa-x will be (genoa-x being the 3dvcache version of the chip, with way more l3 cache than even these have, but with less cores and lower clocks as the cache is more important than the speed or the core count for certain specific applications).

And AMD was the first with dual core.

AMD is putting the AMD in Amdahl's Law! /s =P

@SaHaRaSquad just looked it up. That's absolutely insane... They also just launched a new supercomputer a few weeks ago... Apparently it uses just 16 chips yet still has 13.5 Million (!) cores 😯😲 For others: Look up: "WSE-2 chips"

@Jimmy Jones or maybe, like me, I didn't feel like watching the whole thing but when I saw the comment it intrigued me. So for others like me, there's your time saved

At 12 inches it's huge, but the 3090 was actually 12.5 inches, so somehow even bigger.

It's comically large lol

@Derick D If you missed it then you must be blind.

For those looking, it's 9:16

​@The Silver4 dies not 4 cpus. A cpu is a whole package

@Nirmal Singh wait are you seriously not getting this joke

@Nirmal Singh If I was an engineer for them I'd be buying myself an Epyc and making a "why does this even exist" rig for the lulz, like the 10k systems people ask Linus to make. But why? Because I too would like 512gb of RAM. I stilll have my laptops' 512mb RAM stick I wanted to sell lol

@The Silver so? its called interconnect, its the future, from google to IBM everyone is using amd Epyc cpus, if that goes over your head, just focus on application performance, everything will be chiplets in future...

@Nirmal Singh You don't need to be an engineer to see it's clearly 4 CPUs into one. Also, stop this bullshit "muh if you don't like AMD you must be an Intel user!!"

That thing is a straight up brick. It can, unironically, be used as a murder weapon.

@The Echelon Well most people won't want a "shit brick house" as their case, but GPUs aren't going to get any smaller. The HAF 700 is definitely overkill, and intentionally so since I planned on a dual GPU system. However, mini-itx is pretty much a dead form factor for modern GPUs and it's likely ATX and EATX will be the only form factors that allow future GPUs, with even the standard ATX becoming cramped and possibly causing thermal limiting in air cooled setups.

@ノルtacon my previous system was a dual GPU system, but they were workstation GPUs so you're on the right track. My original intention was to run dual 4090s, but right now the single 4090 is handling my workload without it being really necessary to get another. The sick thing is that any more than two 4090s in my system would require me to hire an electrician to add circuits to handle the load. I'm hoping when Nvidia comes out with the workstation version of the 4090 (presumably called the A7000) that it's more energy efficient and not a crazy markup over the rtx 4090.

Pioteko You should move down here to the states. The 4090s have been fairly well restocked in microcenters here in the US, they even had founders editions in stock yesterday, but I get your point. I definitely refused to pay the scalpers prices. I got mine by just checking the microcenter website a few times a day and when I finally saw a liquid suprim x in stock I ran to my car and drove there as fast as I could. Yesterday was the first time I have seen them have a founders edition in stock though.

@Imad R Dont worry, you willp rob get 4090 performance in 3 gpus generations, so the RTX 7060 might achieve similar 4090 performance.

@Leonhart damn

It's the way they "cheated" the transistor size limitations to get more performance. Just stick more of them inside, even if they aren't smaller, easy. You just end up with twice the size and twice the heat and power consumption....

Well, er yeah, obviously that would be great. Count me in. Then you just know an AIB partner in 2036 is going to push 800watts through that puppy to power the 80,000 shader units and the GDDR11XX. Until we hit some kind of ceiling or have some kind of sensible standard this shit is just going to get crazier and crazier. Unless there's some kind of architecture or engine that renders all that power obsolete. I started to get hyped for the Euclidean thing until I realised it was all voxel... But something. Just don't knock the people that made it all possible. These guys are legends. They made all this real. Like Dave Haynie and the guys at Commodore are my heroes.The whole 'boomer' shit is just insolent, infant bullshit to piss people off and get attention. Rise above that. You're better than that.

It's called bad engineering. Boomer tech. I can wait for thin small nice gpu with 75 watt to play 4K 240fps. No cable needed.

@Pedro Silva micro itx, you're probably right!

Yeah man, that thing is insane. I bet theres ITX cases with less volume than that

@flameshana9 not true. Imagine you had 30 people, which represents cores. You tell the 30 people to do a task, such as cleaning a house. They can split up and all clean different portions of the house and do it 30 times faster. This is a task that can utilize multiple people. Now I tell them to write a 30 page story about a cat. They cannot write them all at the same time, they don’t all know what the story should be, and need someone to write page 1 first, then based off of what was written on page 1, write page 2, etc. this is a task that cannot utilize more cores, so it doesn’t matter how many you have. Lots of tasks a pc does are dependent on the previous results and cannot be multithreaded. Having 128 cores does not make it faster if the cores are slower than a fast 8 core. For a lot of tasks the 8 core would faster than the 128 core.

@Asmodin "running well" is so subjective. If it can run cp2077 over 60hz at 1440p ultra settings I would say that was running pretty well. If it's noticably slower in Photoshop then the next guy will call it trash. If you run old servers as workstations then a i7 9700 equivalent with 96 frickin cores is just peachy, I'd imagine.

@alaskan hybrid it all just depends if the specific workload scales well to running in parallel or not. Or you want to tell me finishing calculations in less time isn't faster? Yes, single threaded workloads obviously don't run well on this, but that's not what it's designed to do in the first place

@flameshana9 doing things fast with many cores is powerful doing things fast with one core is quick.

@alaskan hybrid power = speed, bruh

@Bob Lister lol I wasn’t even born at that point

@mcl48YT for gaming that’s mostly true

@mcl48YT if a game or video editor can fully use more than that why not?

@W Z Crazy how far we've come in such a short time

@mcl48YT If you talk strictly about gaming, then yeah sure. 6 cores are optimal. Even so, new games nowadays now utilized more than six cores.

bro is built for efficiency

@-- Wendell is great but he is not accessible for more entry level tech nerds, his content is for intermediate to expert level nerds. The way he talks, the way he talks about the topics, it’s very clear to those who know our field but to some pc gamer who doesn’t realize their interests can become a career with more digging around, it’s very heavy. LTT has a style of presentation that’s energetic and concise and beginner friendly, people like Wendell and Steve from gamers nexus have a style of far more passion but expectation you know what they’re talking about to a degree. I love them, but I know plenty people who watch LTT but couldn’t grasp l1t or GN. It’s why Linus has so many subscribers compared to other tech channels

@TranquilCam definitely not a loser, it's just hard to make exciting content when it comes to servers. The average person simply cares more about FPS in games than pi numbers crunching.

@Sugg me too, which is why I'm subscribed to Level1 and STH, however, compare the numbers between them and LTT. There's not enough of us to justify making more server content in LTT.

@TranquilCam I think the topics are very interesting, they just need to be delivered in an interesting manner. and ltt media has managed to make videos on computing compelling to watch, so I'm sure they can do the same to data center infrastructure

Makes sense. Thanks for the answers. I remember old computers used to take a while to check the memory when booting up. I also wouldn't guess that it would also check the hard drive. Servers usually have a lot of space and memory so I can see that taking a while to check 5tem on boot.

@Mud Kip Cause the motherboard needs to check absolutely everything. The ram and the hard disks usually take the longest to check.

@Milan Putnik I worked for SoftLayer before IBM bought them for their cloud. We had super micro servers and they actually didn’t take that long to boot up in comparison to AWS servers. When I worked for a AWS manufacturing plant doing diagnostics, those were by far the worst. Lenovo had some shit boot times too. I actually prefer Supermicro boot times over Lenovo and AWS boot times anyday lol. Unless you have one of them 4us with 2/3TB of RAM then forget it.

​@Mud Kip in datacenter we use a special type of RAM memory with the integrated error correction code or ECC. With the introduction of DDR5 the ECC comes to the wide consumer market, the ECC technology is part of the default design of every DDR5 module. Now back to the question you asked - compared to the classic desktop setups based on the CPUs you've mentioned, within the ecosystems using the server grade buffered RAM with ECC we have a lot more stuff going on during boot time (as one might have guessed) - with the primary time consuming one being the so called "memory training". Now, that alone is a whole new topic with a ton of settings in BIOS and it prolly deserves its own chapter on Linus TT or Steve's GamersNexus... Anyway, memory training is a one-time event that's performed by the system on its first boot or subsequently after any significant or otherwise unique change within the particular system, during which the system sets up, tweaks and tests the RAM memory and all its advanced features like the ECC so that it works at its best. That takes time. Of course, in this video Linus showcases the latest CPU by AMD and also explains how there's a bug present in the microcode that causes the long boot times nonetheless and which AMD announced a fix for in the upcoming days hopefully.

Don't know much about servers but why does it take so long to boot ? Do all SuperMicro servers take a long time to boot ? A little off topic but I fell out of gaming/PC building for a while but my old AMD Athlon X2 system booted much faster than my current Ryzen 7 system (due to UEFI I would guess ?) You would think with faster hardware it wouldn't take so long to boot.

The use case I'm really excited about (whenever I may be able to afford anything like this and especially have adequate power supply) is Blender. I could render things _sooo_ fast on this! And my current 120GB of RAM (yes, 120, not 128; I added 96 to my previous 24) is already pretty utterly fantastic but can still just _slightly_ be limited if I make things large and detailed enough (not enough that I feel much urgency to upgrade to my current motherboard's max supported 192 to be fair; I think I'm more limited by CPU speed at this point, finally). One of these systems could support such an absurd amount of RAM for even larger renders! I bet even rendering an entire Hong-Kong size island with super dense, highly-detailed foliage and NO LODs would hardly have any trouble at all (other than, maybe single-thread performance should geometry nodes remain the same; that's my current bottleneck rendering slightly smaller things today which can take some minutes to set up a render).