@@Oliver-sn4be actually amd cards draw more power for the same performance as nvidia, sadly they are not improving as much, they don't have same money as nvidia to develop their gpus
@@kserkez1763 am looking forward to 5070 or the 5070 ti one gona see how the are if over price and bad ill go amd don't care anymore since prices are crazy this is gona be my last one until I win lottery or marry a rich Dubai woman 🐒 pay didn't even go up to make cards this price range 🥲 and we barely got cash for foods now days over here it like that and in germany I got left only about 200 or 400 😒 and work like a robots 🤖 ahhhh well ill go save up a bit going back ther 4 or more if I can survive the depression over ther once back amd and team Green ther gona be Auth ill get them then ❤ until then wish me luck robot 🤖 mode is one 😀
I mean, it does sound stupid in a consumer context, but in a data center context (what Falcon Shores is being targeted at) the equation looks a little bit different. So, for instance, if we were in a world with 300 watt 16 core CPUs that scored, for sake of example, 1,000 points in a benchmark for multithreading, and someone released an "efficient" CPU with 32 cores, used 450 watts, and scored 1,800 points on that benchmark, data centers would be all over it. Depending on your use case (compute versus cloud hyperscaling) you would actually be saving on the power use of the server mainboard itself (keep in mind, server chassis do use a lot of power in and of themselves), you save on the amount of cooling devices, space in the data center, number of racks, etc. In fact, in this specific example, I think you could have a 1:1 power ratio (300 watts versus 600) and a lot of data centers would still take the more power hungry chip for all the "non-CPU" density savings. The things we look for in terms of power efficiency in a residential context don't necessarily make sense when you're buying 10,000 devices for a specific enterprise use case. With that said, I'm pretty sure that Falcon Shores won't necessarily be the slam dunk they're hoping it'll be, and I'm pretty sure it's going to be too late to be relevant, or not deliver performance and density to justify its power consumption/price. I think in AI inference will go towards things like Groq, so general GPU compute is of a limited life span there, for training Nvidia has a stranglehold on the market, but Google's TPUs are interesting and we might see dedicated hardware for LLM training soon-ish, so I'm not convinced Intel has inroads there, and for scientific computing my understanding is that AMD's knocking it out of the park and has parts available already. I'm just not sure what Falcon Shores is supposed to do, to be honest.
It's a leak. Companies are way ahead in terms of technology. They can't just release a cpu without proper testing or they could be just gatekeeping it like Intel's 1st gen to 7th gen till competition ryzen was released.
Back when intel actually was making cpu and not silicon heating devices , they were 5 to 10 years ahead with there prototype and working prototype chips , then 2012/2014 came and they completely messed it all up .
@@Dokdokhu the term you're referring to is call "grasshopping generations," Nvideo already let it be know they're workin on three generation simultaneously.
That's how much power GN got out of their last Threadripper overclocking stream I think, but servers have such powerful cooling system that heat is not the problem, it's how much they pay to cool down these things and the humongous electricity bills.
Right????? It'll have to be setting performance records in a massive way!!!! Sure Nvidia AI accelerators suck the power like the world's gonna end, but they have the performance to back it up. Other than woefully bad integrated graphics, Intel doesn't have any track record to lean on when it comes to pushing the big sell. It just seems like they've run out of fresh ideas and the only way they can increase performance is just dump more power through it. At the rate of exponential growth in power consumption by Intel chips, soon you'll need one nuclear powerplant to run it, and a second one to cool it. 😂
I remember getting a cd burner on black Friday for 250 bucks from comp usa. It took me all year to save up. I miss having computer stores like that. Best buy isn't the same. I live outside the use but when I see family I drive a hour and a half to a microcenter.
@@charlesg5085 I remember a guy at work trying to be a hero, claiming that he didn't need to buy a burner because he overclocked his CDROM drive and it did it better than a real burner. I was like, in that case dude, I'll pay you $200 to do it to mine. Strangely, he never seemed to find the time, despite his promises. 😂😂
That's correct and the G series are always even number although they will never name it Ryzen 11000 with Zen 6 based on the new AM6 they will shorter the name from Ryzen 11000 to something like 1X or 11X or something else.
In fact AMD might change the naming scheme to match intel's again - they did it for notebook now coz AMD is trying to get a step in mobile (intel and Nvidia own the OEMs). Intel will change the naming scheme for desktop 15th gen aswell, so AMD would most likely follow.
The monitor is called FO32U2P (or FO32U2 Pro) - there is no "Zero", it's the letter O. Monitor is great btw, already have it running for more than 200 hours (according to the monitor). For anyone buying it I would recommend adding a usb hub, if you want to use more than 2 devices with the KVM (which works great - got my DAC and insta cam connected via active hub)
@shutupnavikrant it doent work as hood as 1 8 core with x3d though as most games only use 1 core and if the none x3d ccd gets used its slower only by a little lol
We would be there already if we had invested in photonic chips since they were first thought of. They do like 20ghz+ without even making it to 1w energy because it uses light, not electricity, so we could have super small pcs with absolutely otherworldly clock speeds entirely passively cooled.
It's not quite that easy. Chips are already running at a frequency where the board designers have to make sure busses are the correct length, because even at that length the timing can be off. Literally the speed of light. So at terahertz, fuck me dead that'll take something special! Then again, those industrious Taiwanese chaps seem to pull something magical out of their ass every time someone claims we're basically at the limits of what is physically possible. So 🤷🤷🤷
@@POLARTTYRTM There's also one major problem. You can't get silicon to lase. It's been known for a very long time that the electron transition from the maximum energy valence band into the minimum energy conduction band it's a clean one. If the potential energy of the electrons and the electron holes that they're moving into, the transition back down is clean, as it's just an exchange of potential energy and results in the release of a photon. Aka the basis of lasers. With indirect band gaps, the transition back down to the valence band is indirect, as it takes a pitstop along the way. As a result you don't get that friendly little photon produced, and the electron not only gives up potential energy, but some of it's momentum to the crystal lattice - the silicon. Some how Intel have mastered the art of maximising that transfer of momentum to the silicon. Anyway, the point is that for us to produce photonic chips, we need our industrious Taiwanese friends to pull something seriously magic out of their asses in terms of some incredible delicately crafted impure silicon that magically has the properties required, another semiconductor altogether, or the ability to create the photonic elements at like microscopy lego pieces that they can then assemble on a silicon substrate. Or more likely, the know a fuck ton more about it than me and the solution won't look anything like what I've just ranted about. 😂😂😂 Either way, it's a good way down the road yet. Ironically, these power hungry Intel chips that we've been whinging about, combine with an amazing AI model trained specifically for assessing gagillions of materials in the hunt for the magical semiconductor needed to make it work. Ok I'll go and annoy someone else now. 🤣🤣🤣🤣🤣
I swear to god, if they change the 10k series to roman numerals because it lets them use the letter "X" even more than they already do I am going to write the most angrily worded postcard to Lisa Su.
@@pirojfmifhghek566 Xflation is real. At least on Intel's side, the X was reserved for their desktop HEDT lineup. Their 7th to 10th gen HEDT chips were actually great. HEDT motherboards are also a piece of art, they made "gaming" motherboards look like toys, even the most professional-looking ones like the Pro Art series.
8c will be for mobile and low end desktop. 16c will be for the mid-high end range of desktop and low end threadripper. 32c will be high end threadripper and epyc in general. Maybe Epyc will have some 16c as well, but only for the cheaper units. Likely binned 32c units that didnt' pan out properly. It's basically the only logical way for AMD to spread out that run of silicon, assuming there aren't any segmented products between it, once again, because of binned units.
If the CCD's are getting bigger in terms of core count. Will we not soon reach the point where we again just have a monolithic die for the CPU. I don't see standard desktop or mobile usage exceeding 16 cores anytime at all. Unless I am interpreting this incorrectly.
I do astrophotography and stacking days of short exposures takes a lot of compute. I'm looking forward to next Gen Ryzen. I do wonder how many cores the DDR memory subsystem can keep up with. That may be the bottleneck vs space and tdp under the ihs. Threadripper performance per dollar is horrible for home power user like me.
Tom from Moore's Law Is Dead leaked that AMD is bringing 3 chiplet variants: Classic, Dense and Dense Client, which is even denser than the standard Dense "C" core. This would seem to line up perfectly with that leak. 8 would be classic, 16 would be dense and 32 would be dense client.
Arrow Lake is meant to be getting a refresh with an 8p + 32e so I can definitely see AMD bringing a CPU with 8 Zen 6 cores + 16 Zen 6c which would actually be more threads than Intel's option so I think AMD will just see how competitive this Arrow Lake is and then release whatever they need to. But personally I think the sweet spot would be a single CCX with 16 high performance Zen 6 cores with stacked 3D v-cache, it would be amazing for gaming, very good for multi-threaded workloads, low latency, great power efficiency and cheaper to manufacture as it's a single chiplet.
Ryzen10k sounds very interesting. You mentioned it yourself but ~ most what you're seeing here, is testing samples. Those may get into production, they may not. They are EPIC CCXs. Yes, the Ryzens use the same chips, but ... As I started ~ it's interesting, but I'd hold off concluding anything from this. We haven't even had the launch announcement yet, for Zen 5 (which is what I'm really interested in) so this is very early days, and leaks or not, speculation or not, a hell of a lot could change between now and when these go on sale. As just one example ~ all new CPUs coming out, will have both a trusted computing module (which I don't entirely trust) and a neural processing unit. That's a 'core' for AI. I don't WANT AI in my computer! Let me repeat ~ I am a Linux nerd, and while I don't go trawl through the source code for my OS or my programs, I could do it. That is central to the Linux + Open Sauce idea. [sigh] How is anybody going to go through the 'source code' for an AI engine? It's not human readable. You can't stick it in DEBUG and step through it one instruction at a time and figure out exactly how this piece of software, took that piece of data, and worked with it. At least half (more like 90%) of what you do to debug a c program, you just can't do with AI. So how do I know the CIA didn't place a back door in it? A key logger? An instruction to download and silently install some other piece of malware? Become part of a bot-net? What about Chairman Xi and his friends? What about Vlad the Invader? "That's what your trusted computing module fixes." No it doesn't. It provides hardware I can't examine, to do encryption and decryption I can't examine, and tell me something is legitimate, but which I can't check or confirm on my own Linux box. And yes ~ that is a downward / backward step from the check-sums we get today. AI is like coal, like nuclear ~ it has its uses. I'm not beating a drum that we shouldn't have it. What I'm saying, I don't need it in my computer. Same as I don't need a steam engine or a nuclear reactor. I don't need it in my operating system, or my word processor or anything else that sits under my desk.
As long as I can get 12 large cores for good price. Just the right middle, runs games well, Handbrake, compilation, not break the bank or draw too much.
The whole 2.1 cable is a bit overblown, you don’t need it! Plus it’s only the 2.1 80gbs certified that is short, you’re not going to be using that bandwidth for a while yet. So don’t hold back getting the FO32U2P, it’s an epic monitor, yes it’s expensive but having had mine a week. Money well spent 😊
why must everything be such a confusing mess these days normally when you have a display port 1.1 and 1.2 you know it's specs but 2.1 is divided in to many sub types that it's as stupid as the new HDMi standards.
Accelerators is a whole different ball game it's not for consumers it's for big data computing systems that are needed for all the AI and others things to come. These chips need to push lots and lots of data quickly and you need power to go fast.
Nvidia would be better off adding the fastest memory they can get their hands on for RTX 5000, if they keep using the slower memory like 4000 then 5k is just 4k with a different name
Please more than 8 cores on one CCD not because we need more than 16, but so we have more than 8 on one CCD and no cross CCD latency penalty for core heavy future games. So hopefully the 16 core CCDs go in consumer platforms.
I don’t want high performance cards or processor they are already working fast the heating issue is so much my pc goes above 75 percent playing games like valorant and cs2 There is no shuttering or screen freeze isssue but room gets too hot even in winters it’s like you need AC on for 24*7
It is always hold off and wait or this or that. Just pick a decent cpu a decent card and and ram and Just run the stuff its not complicated. Plus soon as you build the pc with in 6 months its not top of the line any more. SOOOOO what just build a mid grade system and run the dang thing
If current trends continue, there will never be a Zen 5 CPU and even Zen 4c could be at risk of being canceled. The reason is that x86 in any form cannot come close the performance per watt of desktop ARM processors from Apple, Qualcomm and now NVIDIA. Expect more to appear before Zen 5 could ship. Why would anyone pay more for a lower performance CPU that uses several times the power of an ARM processor?
Why do we need such high bandwidth interfaces for the display? Why not just use standard data transfer tech and put a video decoder for the lowest latency lossless format in the display, encoder in the GPU. That is good enough to use a display across networks without hiccups nor loss of quality. Definitely would be MUCH lower latency than a streaming gaming service even if run next door. Which still means questioning the point of having to use higher bandwidth connections for a display.
I'm so confused. How on earth did you count 12 CCDs on the Zen 4 chip when there are clearly only 4 there? Someone explain? I've clearly missed something!
That processor has 12 CCDs placed in 4 clusters of 3, the line between each is very difficult to see in the video, but it's probably easier for him to see it in the original image.
Many years ago, I told myself this when gtx 970 was released: Jeez! IF wattage for processing graphics keeps on increasing, at some point a nuclear power plant to play video games will be required! Am I becoming a prophet?
Intel Falcon Shores 1500W, it's a Server GPU, but apparently also a CPU, and it's to accelerate AI, not even graphics much probably; but again, Intel was never shy to use lots of Wattage, current Desktop CPUs are OC'ed to a fault and consume the highest of all CPUs. But that's thanks to Unnecessary Micro-Code: extra % Transistors/mm² and wiring, thus Micro-Code runs all the time, so it gets hot, making the GPU/CPU hotter and reducing freq. and performance, make can't more complex Processors, put more Core on CPUs, etc; Micro-Code is PIA and needs to go, go back to Direct Wiring: Processors can't be copied with a microscope anymore, and patents work at model level
32 core Ryzen9 would be a sellout and would be what RDNA4 and 5 gpus needs to not be a bottleneck. Whilst being a team AMD gamer is this being the products to save up for and start gaming in more openworld games without issue like GTA6 and Star Citizen.
Buy Intel if you have a cold house.
That used to be an AMD joke ..
Not it is intel and nevidia 😂
@@1989johno tables turn... Intel now drawing 250w+ while AMD beats them at performance and value at below 100w lol.
@@Oliver-sn4be actually amd cards draw more power for the same performance as nvidia, sadly they are not improving as much, they don't have same money as nvidia to develop their gpus
@@kserkez1763 am looking forward to 5070 or the 5070 ti one gona see how the are if over price and bad ill go amd don't care anymore since prices are crazy this is gona be my last one until I win lottery or marry a rich Dubai woman 🐒 pay didn't even go up to make cards this price range 🥲 and we barely got cash for foods now days over here it like that and in germany I got left only about 200 or 400 😒 and work like a robots 🤖 ahhhh well ill go save up a bit going back ther 4 or more if I can survive the depression over ther once back amd and team Green ther gona be Auth ill get them then ❤ until then wish me luck robot 🤖 mode is one 😀
"Efficency" and "1500W CPU / GPU combo" feels like mutually exclusive terms but okay Intel.
I mean, it does sound stupid in a consumer context, but in a data center context (what Falcon Shores is being targeted at) the equation looks a little bit different. So, for instance, if we were in a world with 300 watt 16 core CPUs that scored, for sake of example, 1,000 points in a benchmark for multithreading, and someone released an "efficient" CPU with 32 cores, used 450 watts, and scored 1,800 points on that benchmark, data centers would be all over it. Depending on your use case (compute versus cloud hyperscaling) you would actually be saving on the power use of the server mainboard itself (keep in mind, server chassis do use a lot of power in and of themselves), you save on the amount of cooling devices, space in the data center, number of racks, etc. In fact, in this specific example, I think you could have a 1:1 power ratio (300 watts versus 600) and a lot of data centers would still take the more power hungry chip for all the "non-CPU" density savings.
The things we look for in terms of power efficiency in a residential context don't necessarily make sense when you're buying 10,000 devices for a specific enterprise use case.
With that said, I'm pretty sure that Falcon Shores won't necessarily be the slam dunk they're hoping it'll be, and I'm pretty sure it's going to be too late to be relevant, or not deliver performance and density to justify its power consumption/price. I think in AI inference will go towards things like Groq, so general GPU compute is of a limited life span there, for training Nvidia has a stranglehold on the market, but Google's TPUs are interesting and we might see dedicated hardware for LLM training soon-ish, so I'm not convinced Intel has inroads there, and for scientific computing my understanding is that AMD's knocking it out of the park and has parts available already.
I'm just not sure what Falcon Shores is supposed to do, to be honest.
@@novantha1Even in data centers, 1500W is insane. Data centers usually have 20,000 GPUs => 30MW for Intel!
@@tringuyen7519Yeah lol, this impacts data centres even more than the average consumer
They're getting ready for Nvideo CPUs. 😅
@@tringuyen7519 H100 draws up to 700w and a DGX H100 is 8 of them so depending on how good Intel's GPU is it *might* be worth it for some use cases.
Ryzen 10k speculations before r 9000 release😂
BLA BLA BLA BLABLA BLA BLA BLABLA BLA BLA BLABLA BLA BLA BLABLA BLA BLA BLA
We are here for that 😂
It's a leak. Companies are way ahead in terms of technology. They can't just release a cpu without proper testing or they could be just gatekeeping it like Intel's 1st gen to 7th gen till competition ryzen was released.
Back when intel actually was making cpu and not silicon heating devices , they were 5 to 10 years ahead with there prototype and working prototype chips , then 2012/2014 came and they completely messed it all up .
@@Dokdokhu the term you're referring to is call "grasshopping generations," Nvideo already let it be know they're workin on three generation simultaneously.
Power is so important in big datacenters, no way in hell anyone buys the 1500W intel monster.
That's how much power GN got out of their last Threadripper overclocking stream I think, but servers have such powerful cooling system that heat is not the problem, it's how much they pay to cool down these things and the humongous electricity bills.
Right????? It'll have to be setting performance records in a massive way!!!! Sure Nvidia AI accelerators suck the power like the world's gonna end, but they have the performance to back it up. Other than woefully bad integrated graphics, Intel doesn't have any track record to lean on when it comes to pushing the big sell. It just seems like they've run out of fresh ideas and the only way they can increase performance is just dump more power through it.
At the rate of exponential growth in power consumption by Intel chips, soon you'll need one nuclear powerplant to run it, and a second one to cool it. 😂
If they have the performance to back it data centers will definitely buy..
ill wait for the Ryzen 50,000,000 lineup before upgrading
I'm still running the ancient threadripper 1950x and it's chugging along just fine so far
@@sneakysquirrel1990 8700k here work just fine.
until going to the 5900x was running an I7 4770k
@@darkomenz794 im using the i7 9700K. wont upgrade for another 3 years
I’m sittin here feeling like a proud daddy with my 7800X3D. Not watching any more news and new releases until it melts. 😁
should've just called them the Ryzen >9000 series and codenamed the architecture 'Vegeta'
🤣😂🤣
*Zen 5 = Nirvana*
*Zen 6 = Morpheus*
*Zen 7 = Prometheus*
But what did the scout say?
Intel making silicon heating devices since 2014 .
You have a great smooth voice, dude. No harsh or hard sounds or consonants.
The best radio voice ever, right? Not sure what microphone he uses, but damn does it sound clear, clean and soft!
5:28 -> Paulo GOMES, not GAMES.
I used to live by Micro Center and Comp USA while growing up in the 90s.
I still go there though Comp USA went a long time ago.
I remember getting a cd burner on black Friday for 250 bucks from comp usa. It took me all year to save up. I miss having computer stores like that. Best buy isn't the same. I live outside the use but when I see family I drive a hour and a half to a microcenter.
Micro Center reminds me of Fry's Electronics which never came back after the pandemic.😢
@@charlesg5085 I'm glad to still get to experience it when you can and i agree. Remember Circuit City too?
@@melaninfuture I've never been to Fry's but no i am curious!
@@charlesg5085 I remember a guy at work trying to be a hero, claiming that he didn't need to buy a burner because he overclocked his CDROM drive and it did it better than a real burner. I was like, in that case dude, I'll pay you $200 to do it to mine. Strangely, he never seemed to find the time, despite his promises. 😂😂
Actual Ryzen 10k might be just their G series / APU parts like 4k, 6k and 8k, making Zen 6 11k
That's correct and the G series are always even number although they will never name it Ryzen 11000 with Zen 6 based on the new AM6 they will shorter the name from Ryzen 11000 to something like 1X or 11X or something else.
In fact AMD might change the naming scheme to match intel's again - they did it for notebook now coz AMD is trying to get a step in mobile (intel and Nvidia own the OEMs). Intel will change the naming scheme for desktop 15th gen aswell, so AMD would most likely follow.
it's over 9000!!
The monitor is called FO32U2P (or FO32U2 Pro) - there is no "Zero", it's the letter O. Monitor is great btw, already have it running for more than 200 hours (according to the monitor). For anyone buying it I would recommend adding a usb hub, if you want to use more than 2 devices with the KVM (which works great - got my DAC and insta cam connected via active hub)
A Ryzen 16 core x3d hell yes on 1 CCD ...... take my cash now
Upcoming next gen 16 core X3D is on both CCD's.
@@winnb4968a single CCD with 3d cache is always way to go for gamers though
@shutupnavikrant it doent work as hood as 1 8 core with x3d though as most games only use 1 core and if the none x3d ccd gets used its slower only by a little lol
@@winnb4968 i may look at that if the reviews are good 👍
@@Tainted-Soul no you're wrong, this ain't 2009, modern games do utilize a lot more cores and threads
I can't wait for an actual evolution in our transistors and chips. I want terahertz chips
We would be there already if we had invested in photonic chips since they were first thought of. They do like 20ghz+ without even making it to 1w energy because it uses light, not electricity, so we could have super small pcs with absolutely otherworldly clock speeds entirely passively cooled.
@@POLARTTYRTMI know I know, but where is the fun in that? It's all business bud.
@@aeternus80 extremely powerful computers that could render an 8k 120fps video or play games at 8k 240fps barely hitting 1w of energy.
It's not quite that easy. Chips are already running at a frequency where the board designers have to make sure busses are the correct length, because even at that length the timing can be off. Literally the speed of light. So at terahertz, fuck me dead that'll take something special! Then again, those industrious Taiwanese chaps seem to pull something magical out of their ass every time someone claims we're basically at the limits of what is physically possible. So 🤷🤷🤷
@@POLARTTYRTM There's also one major problem. You can't get silicon to lase. It's been known for a very long time that the electron transition from the maximum energy valence band into the minimum energy conduction band it's a clean one. If the potential energy of the electrons and the electron holes that they're moving into, the transition back down is clean, as it's just an exchange of potential energy and results in the release of a photon. Aka the basis of lasers. With indirect band gaps, the transition back down to the valence band is indirect, as it takes a pitstop along the way. As a result you don't get that friendly little photon produced, and the electron not only gives up potential energy, but some of it's momentum to the crystal lattice - the silicon. Some how Intel have mastered the art of maximising that transfer of momentum to the silicon.
Anyway, the point is that for us to produce photonic chips, we need our industrious Taiwanese friends to pull something seriously magic out of their asses in terms of some incredible delicately crafted impure silicon that magically has the properties required, another semiconductor altogether, or the ability to create the photonic elements at like microscopy lego pieces that they can then assemble on a silicon substrate. Or more likely, the know a fuck ton more about it than me and the solution won't look anything like what I've just ranted about. 😂😂😂 Either way, it's a good way down the road yet. Ironically, these power hungry Intel chips that we've been whinging about, combine with an amazing AI model trained specifically for assessing gagillions of materials in the hunt for the magical semiconductor needed to make it work.
Ok I'll go and annoy someone else now. 🤣🤣🤣🤣🤣
I remember the days when you could just pull and plug in memory on your video cards and memory cache on the motherboard.
I'm happy with my I9 though :) (I didn't have to put on my heater all winter)
5:50 "This is Brazil", the foreigners can't tell us how much VRAM we have to use😂😂😂 just kidding
After 10,000 they should somehow change the naming scheme, it's just too many numbers and too long to say. There has to be a better way to name stuff.
I swear to god, if they change the 10k series to roman numerals because it lets them use the letter "X" even more than they already do I am going to write the most angrily worded postcard to Lisa Su.
@@pirojfmifhghek566 Xflation is real. At least on Intel's side, the X was reserved for their desktop HEDT lineup. Their 7th to 10th gen HEDT chips were actually great. HEDT motherboards are also a piece of art, they made "gaming" motherboards look like toys, even the most professional-looking ones like the Pro Art series.
1:45 - Yeah, I noticed that when I bought the DP 2.1 cable. It's too damn short!
The only use of 'efficiency cores' is to inflate Cinebench numbers. That's it.
Tell Micro Center to open stores in the Pacific Northwest and I'll be one of their best customers.
Paying PGE in Oregon is not on my list at 1500 watts way to much.
Intel wants to innovate. They want to built a CPU heater
8c will be for mobile and low end desktop. 16c will be for the mid-high end range of desktop and low end threadripper. 32c will be high end threadripper and epyc in general. Maybe Epyc will have some 16c as well, but only for the cheaper units. Likely binned 32c units that didnt' pan out properly.
It's basically the only logical way for AMD to spread out that run of silicon, assuming there aren't any segmented products between it, once again, because of binned units.
AMD bumping up their Core Count? That's a banger for me 🤩
It's over 9000!!
I'd be thrilled if I only got 8 cores with a massive bump in IPC.
Very curious. Will be intriguing to see what becomes of this.
The monitor isn't what makes me excited, it's the VR application that excites me.
Its over 9000!!!!!!
If the CCD's are getting bigger in terms of core count. Will we not soon reach the point where we again just have a monolithic die for the CPU.
I don't see standard desktop or mobile usage exceeding 16 cores anytime at all. Unless I am interpreting this incorrectly.
Unfortunately there's no micro centre's outside of the US !!
There aren’t even enough in the USA.
Unfortunately. But even if we had it, we would still have sky-high taxes that would drive prices to the moon even in a chain like micro center.
Que legal ver o Paulo Gomes aparecendo aqui no Gamer Meld
I do astrophotography and stacking days of short exposures takes a lot of compute. I'm looking forward to next Gen Ryzen. I do wonder how many cores the DDR memory subsystem can keep up with. That may be the bottleneck vs space and tdp under the ihs.
Threadripper performance per dollar is horrible for home power user like me.
I'm waiting for Ryzen 1mill, rumors are it will beat Superman and Goku at the same time.
Isn't that just Super Brainiac?
Nah that will be the Ryzen Over 9000
Bro , I just noticed the Buc-cees hat. Are you a real Texan? San Antonio here…
2:56 Why we don't have this in France 😭
I’m only interested in the Ryzen 15000 or above.
The others just don’t cut it for my work flow.
I bet those Zen 6 32c either are cores with reduced functionality or straight out ARM.
Tom from Moore's Law Is Dead leaked that AMD is bringing 3 chiplet variants: Classic, Dense and Dense Client, which is even denser than the standard Dense "C" core. This would seem to line up perfectly with that leak. 8 would be classic, 16 would be dense and 32 would be dense client.
It seems that the fiber optic era in display data transmission is not far away.
Arrow Lake is meant to be getting a refresh with an 8p + 32e so I can definitely see AMD bringing a CPU with 8 Zen 6 cores + 16 Zen 6c which would actually be more threads than Intel's option so I think AMD will just see how competitive this Arrow Lake is and then release whatever they need to.
But personally I think the sweet spot would be a single CCX with 16 high performance Zen 6 cores with stacked 3D v-cache, it would be amazing for gaming, very good for multi-threaded workloads, low latency, great power efficiency and cheaper to manufacture as it's a single chiplet.
What's a microcenter?
Intel is actually targeting a very specific market, those who can afford a cpu or heating but not both
Congratulations on video quality upgrade
Fourth time: DO NOT TALK DIRECTLY INTO THAT MICROPHONE. Banging in the ears.
Ryzen10k sounds very interesting. You mentioned it yourself but ~ most what you're seeing here, is testing samples. Those may get into production, they may not. They are EPIC CCXs. Yes, the Ryzens use the same chips, but ...
As I started ~ it's interesting, but I'd hold off concluding anything from this.
We haven't even had the launch announcement yet, for Zen 5 (which is what I'm really interested in) so this is very early days, and leaks or not, speculation or not, a hell of a lot could change between now and when these go on sale.
As just one example ~ all new CPUs coming out, will have both a trusted computing module (which I don't entirely trust) and a neural processing unit. That's a 'core' for AI. I don't WANT AI in my computer!
Let me repeat ~ I am a Linux nerd, and while I don't go trawl through the source code for my OS or my programs, I could do it. That is central to the Linux + Open Sauce idea.
[sigh] How is anybody going to go through the 'source code' for an AI engine? It's not human readable. You can't stick it in DEBUG and step through it one instruction at a time and figure out exactly how this piece of software, took that piece of data, and worked with it. At least half (more like 90%) of what you do to debug a c program, you just can't do with AI.
So how do I know the CIA didn't place a back door in it? A key logger? An instruction to download and silently install some other piece of malware? Become part of a bot-net? What about Chairman Xi and his friends? What about Vlad the Invader?
"That's what your trusted computing module fixes."
No it doesn't. It provides hardware I can't examine, to do encryption and decryption I can't examine, and tell me something is legitimate, but which I can't check or confirm on my own Linux box. And yes ~ that is a downward / backward step from the check-sums we get today.
AI is like coal, like nuclear ~ it has its uses. I'm not beating a drum that we shouldn't have it. What I'm saying, I don't need it in my computer. Same as I don't need a steam engine or a nuclear reactor. I don't need it in my operating system, or my word processor or anything else that sits under my desk.
Cant wait to get a first look at ryzen 15000
As long as I can get 12 large cores for good price. Just the right middle, runs games well, Handbrake, compilation, not break the bank or draw too much.
So we wait for Zen 6 as it is already in development but we have the main information already while Zen 5 is not even out yet?
@8:20 That awkward moment you forget how to count past six, and stutter...
With such a level of core density per CCD, imagine what sort of APU's could become a possibility...
4K back again... It better be here to stay this time Gamer Meld. 😡
I thought Ryzen 10000 would be using Zen5 architecture.. isnt Zen4 already used in Ryzen 7000?
The future's so bright, I got to ware water cooled shades.
I got Ryzen 5 7600X today xd should i buy 7900 GRE Nitro+ or 4070 they same price in my country
a zen cpu using the zen5C chiplet,overlayed with an x3d chiplet would be my choice there. 16 cores of x3d on a single ccd, oh sign me up.
The whole 2.1 cable is a bit overblown, you don’t need it! Plus it’s only the 2.1 80gbs certified that is short, you’re not going to be using that bandwidth for a while yet. So don’t hold back getting the FO32U2P, it’s an epic monitor, yes it’s expensive but having had mine a week. Money well spent 😊
I think the actual core count will depend on the size of those 16c CCDs. If its the same size it will probably be a Ryzen part. 🤔
AORUS FO32U2 1500$ cad, not too bad.. and its 1000 nits.. HDR gona look great on this.. have to buy it.
why must everything be such a confusing mess these days normally when you have a display port 1.1 and 1.2 you know it's specs but 2.1 is divided in to many sub types that it's as stupid as the new HDMi standards.
1500watts? What it's running on?
Accelerators is a whole different ball game it's not for consumers it's for big data computing systems that are needed for all the AI and others things to come.
These chips need to push lots and lots of data quickly and you need power to go fast.
Nvidia would be better off adding the fastest memory they can get their hands on for RTX 5000, if they keep using the slower memory like 4000 then 5k is just 4k with a different name
Please more than 8 cores on one CCD not because we need more than 16, but so we have more than 8 on one CCD and no cross CCD latency penalty for core heavy future games. So hopefully the 16 core CCDs go in consumer platforms.
1500w? Are they nuts?
I don’t want high performance cards or processor they are already working fast the heating issue is so much my pc goes above 75 percent playing games like valorant and cs2
There is no shuttering or screen freeze isssue but room gets too hot even in winters it’s like you need AC on for 24*7
3 feet is plenty for me, my tower sits right next to the screen
It is always hold off and wait or this or that. Just pick a decent cpu a decent card and and ram and Just run the stuff its not complicated. Plus soon as you build the pc with in 6 months its not top of the line any more. SOOOOO what just build a mid grade system and run the dang thing
If current trends continue, there will never be a Zen 5 CPU and even Zen 4c could be at risk of being canceled. The reason is that x86 in any form cannot come close the performance per watt of desktop ARM processors from Apple, Qualcomm and now NVIDIA. Expect more to appear before Zen 5 could ship. Why would anyone pay more for a lower performance CPU that uses several times the power of an ARM processor?
Why do we need such high bandwidth interfaces for the display? Why not just use standard data transfer tech and put a video decoder for the lowest latency lossless format in the display, encoder in the GPU. That is good enough to use a display across networks without hiccups nor loss of quality. Definitely would be MUCH lower latency than a streaming gaming service even if run next door. Which still means questioning the point of having to use higher bandwidth connections for a display.
cant wait for the 10800x3d
I'm so confused. How on earth did you count 12 CCDs on the Zen 4 chip when there are clearly only 4 there? Someone explain? I've clearly missed something!
That processor has 12 CCDs placed in 4 clusters of 3, the line between each is very difficult to see in the video, but it's probably easier for him to see it in the original image.
@@Ts6451 ahhhhh i seeee. i barely have any reception where i am atm so i was watching on 144p or something. i was never gonna see that lol
I just want to build a future proof pc to host an LLM and image generation ai software.
We need microcentre in the UK now.
Many years ago, I told myself this when gtx 970 was released: Jeez! IF wattage for processing graphics keeps on increasing, at some point a nuclear power plant to play video games will be required!
Am I becoming a prophet?
Monitors unboxed, is aside channel of hardware unboxed
I live in Idaho. No where near microcenter. We have a bestbuy but they never have anything in stock
Intel Falcon Shores 1500W, it's a Server GPU, but apparently also a CPU, and it's to accelerate AI, not even graphics much probably; but again, Intel was never shy to use lots of Wattage, current Desktop CPUs are OC'ed to a fault and consume the highest of all CPUs.
But that's thanks to Unnecessary Micro-Code: extra % Transistors/mm² and wiring, thus Micro-Code runs all the time, so it gets hot, making the GPU/CPU hotter and reducing freq. and performance, make can't more complex Processors, put more Core on CPUs, etc; Micro-Code is PIA and needs to go, go back to Direct Wiring: Processors can't be copied with a microscope anymore, and patents work at model level
Well, I live in Phoenix area.... Oh all right, I'll get a plane ticket and then run to my nearest Microcenter.... Hint...Still waiting Microcenter!!!
the closes micro center near me is about 4 hours away. i cant "just go" bruh
its over 9000!!!!!!!!
over 9000!
hear this sht all day buy this buy that waiting
Ryzen Zen 6 cant have 32 cores with dual channel DDR5 memory because more cores requires more data feeding unless it moves to AM6 and DDR6.
Hope its on AM5
Don’t buy wait for everyone to buy the faulty ones 1st 😂
AMD not using HBM memory on consumer cards after using us as guinea pigs is frustrating.
Nah, intel should switch to being a heater manufacturer😂
Amd 10000 will use DDR6?
When dark mode?
It’s over 9000
32 core Ryzen9 would be a sellout and would be what RDNA4 and 5 gpus needs to not be a bottleneck. Whilst being a team AMD gamer is this being the products to save up for and start gaming in more openworld games without issue like GTA6 and Star Citizen.
I wish micro center would open around me!
That's it. The clickbaiting is on another level with this one.
Imma stick to my gamers Nexus weekly recap
Wouldn’t it be Ryzen 11000? Being as desktop has gone 5000 > 7000
Yeah 1500W is for data centres no way is it for consumers , cooling 1.5K watts would require a cooler the size of 3.2 square meters ( copper)
Wish we could see...32 cores...64 threads....and something akin to 780M....on one chip...
I can see the US converting to 240V because of PCs LOL.