AMD Core Stagnation - A Problem for Desktop Users?
HTML-код
- Опубликовано: 30 сен 2024
- Steve and Tim discuss why AMD's lack of cores in their productivity CPUs could be costing them, and is it a problem for desktop users?
Join us on Patreon: / hardwareunboxed
Join us on Floatplane: www.floatplane....
Buy relevant products from Amazon, Newegg and others below:
Radeon RX 7900 XTX - geni.us/OKTo
Radeon RX 7900 XT - geni.us/iMi32
GeForce RTX 4090 - geni.us/puJry
GeForce RTX 4080 - geni.us/wpg4zl
GeForce RTX 4070 Ti - geni.us/AVijBg
GeForce RTX 3050 - geni.us/fF9YeC
GeForce RTX 3060 - geni.us/MQT2VG
GeForce RTX 3060 Ti - geni.us/yqtTGn3
GeForce RTX 3070 - geni.us/Kfso1
GeForce RTX 3080 - geni.us/7xgj
GeForce RTX 3090 - geni.us/R8gg
Radeon RX 6500 XT - geni.us/dym2r
Radeon RX 6600 - geni.us/cCrY
Radeon RX 6600 XT - geni.us/aPMwG
Radeon RX 6700 XT - geni.us/3b7PJub
Radeon RX 6800 - geni.us/Ps1fpex
Radeon RX 6800 XT - geni.us/yxrJUJm
Radeon RX 6900 XT - geni.us/5baeGU
Is AMD falling behind on core count?
Disclaimer: Any pricing information shown or mentioned in this video was accurate at the time of video production, and may have since changed
Disclosure: As an Amazon Associate we earn from qualifying purchases. We may also earn a commission on some sales made through other store links
FOLLOW US IN THESE PLACES FOR UPDATES
Twitter - / hardwareunboxed
Facebook - / hardwareunboxed
Instagram - / hardwareunb. .
Music By: / lakeyinspiredg. .
As a programmer (not a gamer)... a wider register window and true AVX512 was an important incremental step forward. CCD latency has been a big step backwards. Fixing the latency, being able to run two GPUs at full speed and have a non flakey IMC that can handle 4 x 64Gb DIMMs would tick a LOT of boxes going forward.
As a programmer, do you think AVX512 will play a greater role in the future or will Intel's AVX 10.2 (and APX) be the new baseline for programmers to target? This could be an interesting long-term battle to watch. As AVX512 support is heavily fragmented, I wonder if AMD's AVX512 prowess could end out to be a pyrrhic victory eventually.
well thankx to me you know now trhat you should have being on X299 a long time ago, still vs mainstream intel and amd they kill it with that Chipset :D cuz they have full AVX 512 on them unlike mainstream intel and then theres quadchannel to consider, that platform even if its older screams for you unlike threadripper which is way to costly we are talking like 90% more or so for an 5 or 7000 series for NOT 90% more performance then lets say a 10980XE :)
my 10940X is running 256gigs and 2 gpus for davinci resolve (3090/30770) you can even put ANNOTHER GPU in there, cant do that on main stream, then imma live with the fact that they arent as fast, who cares, PCS arent always about speed you should realy consider X299 rather 10940X or 10980xe or maybe a 9960X they are too good for the price ;)
see ppl X299 isnt dead yet, maybe from that they arent made anymore or so but not dead for still using them
Add higher memspeeds next imc version (4000MHz memory clock for optimal ddr5 8000 operation) and they will fly in the 'gaming' tests at lower resolutions too. Since that's the bottleneck that didn't improve with Zen 5.
Edit: The latency socks, but if you are using them develop and run hpc models you can take that into account.
@@00wheelie00 well latency between ccds on x299 isnt an issue thankfully but ive got real 14 cores, not 2x7 core chip or in case of a 9960x which got 16 cores ive just have a 16 core die unlike a 5950x for example which is a 2x8core cpu ☺️👍 also bettwr in games, not the everythibg solution but realy nice having no nonsense glued cpus 👌☺️
AVX 512 is a retarded waste of die area
IMO, what consumers platforms need more of are PCIE lanes. 20-24 available to use have become quite restrictive, when considering how cheap PCIE storage is.
How much average user really needs more than 16TB on NVME? Now we have 8TB sticks and even low end systems can take at least 8TB of gen 3
@@PREDATEURLT the generation of PCIE don't matter. Because you have access to xx lanes of PCIEx does not mean you'll have access to 2*xx lanes of PCIEx -1
@@lamikal2515 NVME drives are so fast already that adding pcie lanes wouldn't even be noticeable for the average user. Hell i cant even tell the difference between a sata ssd and a nvme in real world applications considering its only a 10-20% load time improvement in games.
@@lamikal2515 I understand limitations of these things, what I mainly asking is is it limiting every day use of average user? Most people on planet use single GPU and single NVME.
@@jordanmackay6746 Well, Mobo manufacturers gives less and less SATA ports unless ultra high end, and CPU manufacturers don't give more PCIE lanes, so soon we'll have neither of those
As a 7800x3D user, i want a 12 core single ccd 3d chip. If they make that, it's an instant buy from me.
That's what my next upgrade will be. I don't care about 8core with 3-5% ipc increase.
7800x3D is already basically a supercharged console CPU. Game development will focus on 6-7 cores until a new generation of consoles. It would provide a good CPU for the future though with 12 cores, rather than hanging around in this 8 core console generation, buying the same things over again for increased 1080p performance. A 12 core just wouldn't make a difference in current gaming really for years, so it would be hard to market it when it has the same performance in gaming as the cheaper 8 core part.
As a 7800x3d user I will upgraded based on factual data not theoretical discussions.
@@Mcnooblet Many games can take advantage of as many cores as you have... the whole "no game can use more than 8 cores" is really just ignorance...
@@dragam9534 Which games get any meaningful performance uplift with more than 8 cores then?
Higher core-count Zen 5 desktop parts are also constrained by memory-bandwidth in many scenarios. Adding more cores without also increasing memory bandwidth would exacerbate that and lead to diminishing returns. Hence why Threadripper historically uses four or eight memory channels vs. just two on X950x parts. I would not expect this to be any different for Zen 5.
It depends. First and foremost more cores means you can clock them lower.
Exactly! If you need more cores. Buy Threathripper!
@@seeibe Why trade single-thread performance for more cores? That’s dumb
AMD needs to make bigger CCD really. 16 cores on one CCD
Or, maybe just put 4 CCDs on a desktop chip. AMD would need some crazy PPA to put a 16 cores/CCD chip(they could kind of achieve this using C-Cores, and I think they might be planning something similar for Zen 6, but not to the level of 16 CCDs).
And lose price advantage aka forces to increase prices because of that… Not a good idea!
The main point of chiplets is to keep each chip small, cheap and reliable!
What AMD can do is to make cpu that has 3 or 4 8 core chips!
Actually AMD allready have that… It is called Threadripper!
Do you understand that cramming that many transistors on such a small place, increase temps drastically, also you'll have smaller bins due to failure ratio, which will increase prices? 4 Cores are fine for office applications, 6/8 are fine for gaming, over that its fine for multitasking or more complex situations. This of course will change with time as software and hardware evolves, but in truth, we are still leaps and bound over the old 4core i7 days.
Isn't that called Threadripper or Epyc?
@@vgernyc
It definitely is Threathripper! So there is an product allready in the market, if you need it!
PCIE lanes to the CPU, not cores: 16 for graphics, 16 for quad SSD card, 8-16 for high-speed network so we can use old gen3 server kit, a few more for playing with virtualisation, and some fun bits and pieces. Give us some things to play with without going over a multiplexed x4 link and USB.
I don't know if you need to go that far (40-48 lanes). 48 lanes would put you in workstation/HEDT territory. OTOH, quad SSDs is also an HEDT spec, so it's consistent. If I had 48 lanes to play with on desktop, I'd probably put 16 primary graphics, 8 secondary graphics (for console/desktop while a virtual gaming environment takes primary), 8 for SSDs, 8 for chipsets to more storage and USB, bringing the total to 40. Then the last 8 can go to whatever: expand secondary graphics to 16, more storage, or other PCIex4 slots.
@@bwcbiz 48 lanes is only HEDT because that's what we've been given recently. X79 was "enthusiast" without pushing into "you'll need a business case to support this purchase" territory occupied by pcie gen4 W xeons and threadripper. Quad-SSD carriers are < $30 new on ebay. No-one blinks at quad SATA ports. I'm struggling to find a use-case for a single x16 gen5 slot, but there are plenty for two x16 gen4 slots which provide the same bandwidth. Desktop virtualisation - I'd like to keep critical things and gaming/personal stuff separate. A cheap low-power card is fine for multi-screen non-gaming, if I can separate the two systems.
Suddenly so many gamers becoming micro processor designers and demand AMD to do this and that lol. There are more than enough cores.
I hard disagree with this take, and its because its way too narrowed in on gaming utility and not “what are you actually getting with this CPU vs Intel. Currently post Raptor Lake fix and ever since Alder Lake you have gotten similar game performance with a big mismatch is productivity.
Lets keep it simple, theres no reason where there needs to be TWO Ryzen 9’s. The 12 core 24 thread should be the Ryzen 7 and so on with that adjustment till the 6 core 12 thread is the Ryzen 3 and if they want, a 4 core Ryzen 1.
That would fix a lot and of course that comes with price drops. Its 2024 and we’re still paying $300 for a 6c/12t and $400 to $450 for a 8c/16t? Really? When people say Ryzen needs to increase core counts it isnt that the current amount if cores arent fine, its that the price for the current a mount of cores is too much and just not competitive to anybody who actually stops to think about it.
This video was straight up L.
They should revisit their Zen 2 reviews.
Exactly ,AMD keeps releasing 6c12th for 300$ which is ridiculous. At 450$, we should get a 12c-24th x3d on a single CCD. Hopefully they realise this by the time they release the 11800 x3d.
12c-24th x3d would be a big expensive waste. Target platform for gaming is currently PS5, and all consoles have 8 cores (1 dedicated to the OS). If there was no performance increases with a 12 core x3d, why would anyone buy it compared to a cheaper 8 core x3d that has the same gaming performance? CPU performance will be unified around consoles, not AMD suddenly increases core counts on their PC gaming line. Until new console generations, the target will remain 7 cores.
There is zero use for gaming.
I can understand wanting more cores on an R3 or R5, but R7 X3D is already peak performance with 1xCCD. It doesn't have the added latency or core parking issues from running two differing CCDs.
R7 X3D for the price of an R5? That would be the dream.
It would be 16 cores x CCD not 2x8.
Games barely uses more than 8 cores. First developers should adapt.
Watching all those current cores doing 3~8% utilization is really a problem. We need more cores!
They should retire 6-cores and make the base starting point 8-cores though. It's like Intel spending a decade on 4-cores.
Tim this is an L take.
Just like we say 8GB VRAM is not enough when paying 450-500 usd, similarly paying 300 usd is way too much for a 6 core part and 400 usd for only an 8 core.
That's the issue!
Did you guys forget your Zen 2 reviews? AMD provided value simply by providing higher core counts for every price category!!
EACH CCD SHOULD HAVE 10 CORES! WIL BE PERFECT! 9600X 8 CORES ... 9700X 10 CORES AND 9950X 20 CORES 40 TREADS!
The more cores, the more expensive!
Much better to have 3 or 4 8 core chips if you need more speed!
Aka buy Threathripper!
Core stagnation? just like intel that keep selling 4 core for mainstream user from 2008-2017?
And why did they do that? Because AMD was sleeping.
@@Dr.WhetFarts why? greed. which lead to intel downfall today.
@@stepbruv8780 And the same thing is happening to AMD. They keep Ryzen 5 as 6 cores and Ryzen 7 as 8 cores from 2017 to 2024.
Because AMD is even worse than Intel nowadays. At least Intel still has Core i3 and Pentiums for lower budget builders back then. AMD's lowest CPU for AM5 is Ryzen 5 8400F which is quite underperforming compared to Intel Core i5-14400F for similar price. Not to mentioned AM5 boards are more expensive and requires DDR5 which is also more expensive.
AMD core count is fine. The issue that people perceive is that the product lineup is tightly stitched together.
My history with AMD has been pretty good: K6-2/300 (1c/1t), Athlon 2650e (1c/1t), Phenom II X4 955 (4c/4t), FX-8370 (4m/8t), R5 3600 (6c/12t) and at this point I'm comfy. Was it weird going from an "eight core" part to six? Yes. It was also weird that was an empty sacrifice for 50% more threads. I'm at another crossroad for a same socket upgrade and while there's no push for me to upgrade, it will be one of those "nice" things much later on when I finally see a problem that my 3600 can't handle well. Could be a straight jump to 5000 series with the 5600. Could be a completely insane doubling of cores and threads with the 5900X or it could be another wild futureproof attempt of dipping into early X3D technology with the 5800X3D.
Whatever my choice, all I know is that I'm definitely NOT an AM5 customer and when it comes to the work that I do, I'm good. Core stagnation is not AMD's problem. CPUs are in a very good place. The GPUs need real attention. Core clock difference, memory stability, delivery on functional chips, a better quality hardware encoder, etc. GPUs should not be so fragile that delivery of a functioning modern product proves impossible. I'm still a bit upset about that.
Because for average desktop users we have gone past "good enough" and into batshit overblown territory? When we went from the Mhz wars to the core wars we knew this day would come and now we are here, where you have way more cores than Joe and Jane Normal are ever going to use. Hell I like to do some video and music editing as well as gaming and most of the time I have two or more cores sitting idle as these days apps take advantage of your GPU so what good would more cores do?
Ryzen 5 should be a Ryzen 3.
Ryzen 7 should be a Ryzen 5.
Make it happen!!!!!!!!!
What's the point of more than 32 threads when most applications are almost single thread...
The day old 'why need more?' anology.
In 1990 why need more than 1 core
In 2000 why need more than 2 cores
In 2006 why need more than 4 cores
Software side will always improve so getting more cores isn't the bad a thing.
@@animecutscenes3414 it is callled diminishing returns.
In desktop you dont need more than 32 threads. In a server with multiple VM instances yes.
The point is that they wont be in the future, more threads people have more pressure software devs have to utilise the power. If everyone had a processor with 100 threads and 50 cores, and you make a single threaded software your software is probably not very competitive. This is like saying whats the point of getting more than 2mb of memory in 1990 or something. When people got more memory then the software started utilising it more.
@@mattiarizzi Thats literally the same argument people have said 40 years now. "Nobody will ever need more than 640k ram" -bill gates in 1981. When more power becomes available software tends to adapt and use it.
@@tottorookokkoroo5318 it is clear that you never developed complex software.
Not every problem is scalable over multiple threads. And guess what? You will still find your 50 cores cpu waiting for a mutex to clear out
More cores are not essential BUT
They put pressure on lower core parts and result in cheaper parts in the long run.
It depends on how someone is going to use the CPU, and more cores can absolutely be essential, especially performance based cores. Just not for gaming since development targets will remain that of a PS5 for years to come at 6-7 core. That's just for gaming that is held back on core counts though, as productivity doesn't have those same limitations with console targets, and definitely benefits from higher core counts. Those cores still need to be good of course, as higher core counts with crap cores doesn't equal anything good.
If you really need more cores you can go with Threadripper. Adding >16 core Ryzens would just render the smaller Threadrippers pointless for compute-performance.
Why you aren't reviewing r5 7600x3d?
tha's a mc special they would have to import it from the u.s.
If AMD bumps core counts and offers:
6-core R3,
8-core R5,
12-core R7,
then 16-core R9,
the whole lineup would be so much cleaner and delineated plus breathing new life into R3s.
I know it's a long shot, but one can still hope 😅
intels e-cores and amd inter ccd latency which causes core parking is just annoying with scheduler issues. kinda wondering.. windows should be working closely with intel and amd and vice versa to make sure everything works fine.
I thought that with Zen 2, the 16 core part should have been the Ryzen 9 3900X, with the 12 core being the Ryzen 7 3800X. the 8 core the Ryzen 5 3700X and the 6 core the Ryzen 3 3600X. Later on, they could have introduced the non-X variants.
AMD should have also relocated the GPU to the I/O die with the APU variants of Zen 2, at least with regards to the desktop parts. And with Zen 3 they should have introduced an 8 core APU with 64MB of Infinity cache on the I/O die, that could become L3 CPU cache when a discrete GPU is installed. Of course, they could have also offered CPUs with just 2 Navi cores and no Infinity Cache on the I/O die, potentially with 64MB of V-Cache on the CPU, starting with Zen3.
Certainly, once Intel introduced e-cores, AMD should have corrected this naming convention.
10/23-24 Cores for the X600-x950 Chips would we great. Im running most my stuff virtualized in my own server - yes even Gaming. And being able to have more breathing room with CPU cores would be sooo good.
Also: Fix RAM issues, holy moly. I dont want to go with just 64GB of RAM on my server, nor go full 10x on price for modern hardware with actually viable single core performance (Gaming). Just let me have more RAM and PCIe, please!
im waiting for a 7950x mobo/ram combo to be available or see what arrowlake combo microcenter will have.
I disagree so much. Some of those who start professional application will have limited budget at the beginning. And that's where Intel offers better value.
And I believe Ryzen 5 is the best-selling AMD CPU instead of Ryzen 7 and 9. So, that should be where AMD focus to offer the best value of performance (both gaming and productivity) if AMD want to improve the market share. The biggest market share AMD has managed to claim was 37-39% during 2021. But now in 2024, AMD CPU market shared is being pushed back to 33%.
Honestly I think 16 cores is enough, seeing as the 9950x trades blows with the 14900k. The biggest issue is the infinity fabric and ccd latency. From what I've heard, there's essentially 2 ports per ccd for infinity fabric but only 1 is used on Ryzen vs the 2 in epyc. Also using some sort of like silicon interposer or 2.5d connections vs regular trace connection could probably help. I think latency is the biggest issue, along with ram bandwidth hence why the x3D sees such an uplift. Also some more pcie lanes would be nice, like having 1 x16 and 2x4 off the cpu.
This is why i am waiting for ryzen 10,000. Leaks say it will have 32 cores. You won't need a new cpu for a long time. They talk about fixing infinity fabric design. It's all about latency. Currently, it affects memory speed. It all adds up to amd needing to fix a problem they have had since ryzen 1000 and killed the flagship card for this next generation. Ryzen 10,000 design, if it works correctly, could be the revolutionary design we have been waiting for. Ryzen brought the cores, but few games use that many cores other than games like city skylines 2.
I do think we're suffering a bit of performance stagnation because AMD is not seeing serious competition from Intel, which is a repeat of what happened to Intel for years when AMD was not presenting meaningful competition. Besides this, we're also at a point where few games are leveraging more than 6 cores and almost none are leveraging more than 8, so increasing core-to-CCD count beyond this just isn't going to yield meaningful performance benefit. Hobbyists and workstation users who actually have a use for more than this are the people the dual CCD parts are for. Beyond this, like dedicated workstations are the platform Threadripper is designed for.
I've seen some other people mention wanting more PCI-E GEN 5 lanes, particularly for storage. Ignoring that this IS available on higher end X670E motherboards, there are diminishing returns for the typical gaming-centric end user. Particularly, PCI-E Gen 5 disks are only faster in sequential reads and writes and games don't rely on that and since the last time I checked, nobody has a disk that results in faster asynchronous speeds than even Gen 4 lanes can handle anyways. But even if we really want all our drives to run at Gen 5 speeds, the typical user is only going to leverage those speeds for no more than 2 disks at a time, and typically only when copying from one to the other. So a PCI-E Gen 5x8 link feeding a bus of 4+ disks would make the most sense for an end user case.
Wow people are idiots, want more cores for what ?. game dev hardly use greater then 4-8 cores and even 8 is a stretch. And if you want obscene amount of cores of whatever reason get threadripper gamers like complaining about problems they don’t have 😑
Shuffling around will not work. AMD s 12 core CPU in many tasks and games is worse than 8 core cpu.
I played Monster Hunter Rise with a 4 core cpu days ago. 60 fps. I don't think that core count will matter much in the future.
This is easy to prove: Is the average core count of sold CPU higher now than 3 years ago? Is 6 core and 8 core selling less and 12 core and 16 core selling more every year?
If this is happening then we know that people need more cores.
As a gamer, 5600G + RX 6600 does the job, even with 3440x1440. How long did Intel keep 6 core CPUs behind the HEDT paywall? Till 2700X outed. Forced. Nuf said.
The true reason is Greed. Why offer more cores than we can sell same 8 core all over again. AMD is adopting intels strategy 4 cores thats it. Also more cores will affect their threadripper business.
Cpu stagnation, gpus worth more than diamonds, motherboards expensive for no appearant reason, at least ram is still relatively cheap...
probably don't want to cannibalize threadripper. but threadripper's biggest selling point is probably io
I really think a boost in per ccd core count would be a smart move at this point. 10 cores per ccd and tightening the ccd to ccd latency would be huge. Amd also really needs to work on better memory controllers and whatever else is needed to get higher memory speed and bandwidth.
After 16 cores on a mainstream desktop platform I struggle to see the use case for the majority of users. If you need more cores than that then you are in the minority and thus not worth it for amd or Intel to manufacture for. I do however wish amd would continue to make thread ripper chips or make epyc chips more readily available for the average consumer. But as a lot of other commenter's are saying, we need more pcie lanes. Most b550, b650 boards lack the availability of pcie gen 4x4 or higher m.2 slots. Which is a problem considering gpus are so thick now that they cover most pcie slots. I have an msi b550 pro board and it has 2x m.2 slots but one is gen 3 and the other is gen 4 but my rtx 4080 covers 3 pcie slots and the remaining pcie x16 slot is gen 3 and only x1. So I can't use an m.2 expansion card and get full speeds or even better speeds than sata ssd. Also because I have both m.2 slots used it disables my sata connections. This is a serious issue. With most people switching to m.2 because the speed and cost are better than sata ssd, we need more pcie lanes available at higher pcie gens. But really if the whole motherboard used pcie gen 5 then the slots could be x1 and still be sufficient for even the fastest nvme drives available. Which pcie gen 5 nvme drives are very much overkill unless you are doing some serious transfers or reading and writing to large asset libraries like unreal engine 5. I would much prefer to see nvme drives get larger capacities while being cheaper. $800 for 8tb is insane. But the longevity of nvme over hdd is so much superior it's a not even a consideration to use hdd anymore. Now I suppose the issue of storage could be fixed by creating a Nas or similar large storage server which you connect to via ethernet, but you still need pcie slots and lanes to use an expansion card for 10gb Lan, which again can be an issue with massive gpus.
I see at the 0:47 second mark you mounted the aio on the CPU socket and then put the CPU on the block😂 that should be more than enough cooling for that chip 😂
Why waste money on more cores when the competition uses more cores and cant keep up? At this point you guys are just nagging.
Next time try the same tests first placing cpu and then the AIO not the other way around 😂
core count is fine but the price isn't. 8 core should be 300 at most now
Zen 6 will double the cores to 32 Cores and maybee 16 cores are the new 8 cores.
iam still happy with my 5800x and i dont care about stagnation
The only issue with AMD not increasing core count is the lackluster improvement over 7000 series if they had increased core count at the same tier of pricing that would've been great
Reintroduced R3 6 core
R5 becomes 8 core
R7 dont really like 12 cores over 2 CCDs but yeah
R9 stays 16 cores but gets a slight decrease
If you really think about it, they've havent increased core count in the mainstream range since Zen launched
i called it after the zen 5 flop they've stagnated the market like nvidias stagnating vram in the ocnsumer market for over a decade now
It's crazy to think that the $300 CPU that I bought in 2011 (I7 2600k) is only 2 cores less than the new $300 AMD CPU that exists today (9600x). IMO 9600x only makes sense if it's $200, just like 7600 today.
Yeah but the Instruction Per Cycle Per Core greatly improved over time(tho it's a bit stagnating for now), if you compare one core of the i7 2600K to one core of a Ryzen 5 7600 you'll see that one 7600 core is A LOT faster than your i7 2600K core(if we are using the same frequency), a 7600X and a 9600X are way faster than this i7 even if you disable two of the cores...
I guess the current cinebench cores have done t their job🤣
As a 60 FPS gamer...no big deal.
threadripper is giving you 96 cores
We want more cores AMD!!!
i miss when moar corez where an ironic joke and not a sad reality
We went from 'We want more cores Intel!' to 'We want more cores AMD!'. Ironic.
No we don`t… or actually you can buy more cores allready! Buy Threathripper!
@@haukionkannelA thread ripper cost 3000$, a brand new expensive motherboard and has worse performance than ryzen in gaming 💀
@@nayan.punekar Exatly!
That is why we don`t need more cores! We need faster cores!
But when all games use 16 cores… Then we need cpu that has more! So far we can see that 6 core 8 core, 12 core and 16 core cpus are about same speed in games! So… if you Only game, 6 core is sweet spot, 8 core is plenty and anything more is wanted for the gaming, untill games start to use more cores… it can be a long wait…
I want a 24core 48 thread
AMD stuck on 16 cores. The new Intel.
Dunno there is always threadripper and epyc
@@Zetraxes yeah and you had X series CPUs too >.>
96 core threadripper 7000 has entered the chat
watch when they do a Ryzen 9 9900x 32 core 64 thread.
@@Zetraxes yea if you have $3000 to spend on measly 24 cores as an entry point. threadripper is unobtainium.
Nah intel has been mostly 4 cores while AMD offers affordable 6 cores cpu which is great for lower end Virtualization, multitasking and running other multiple instances
Intel 13th gen i5 provides minimum 25% better multicore performance compared to latest Zen 5 Ryzen 5 part.
300 USD/€ for 6 cores CPUs isn't what I (and a lot of people actually) call affordable.
I have a 13700k for my office pc and I like the 2 different type of cores. I can assign the "P" cores to a large task even cpu mining and also assign the "E" cores to the web browser, it's practically seamless with no lag or stutter. I can even play older games like Batman Arkham Asylum with no lag and stutter on the "E" cores while the "P" cores are busy at something else. It's like having 2 pc's in 1. Assigning cores on my 5800x3d gaming pc doesn't work anywhere close to as well, it lags, stutter etc.
69 cores plz
thanks
AMD moves the 9700 8 core 16 thread to Ryzen 5k to directly compete with i5 14600k, make 12 core 24 thread 9900 a Ryzen 7k to match i7 14700k. Problem solved with dumb branding.
Rather than more cores they should focus on improving igpu. Why haven't they made quicksync competitor I dont know
igpu already eats up a ton of the IOD on which everyone is saying more PCIE lanes, faster memory controller.. Of course now they are a gen behind by recycling the same IOD for Zen 5.
AMD did improve iGPUs just for gaming, and it's probably why they can sell 6c/12t CPUs with iGPUs for 300 USD/€ while the 7500F is sub 200 USD/€.
AMDs big problem is that dual CCD is bad for gamers but does good in productivity - Intel does both just fine with their approach but watt usage is a problem, should be fixed with TSMC 3N and 18A eventually tho. I can't wait to see what Intel can deliver using TSMC 3N personally.
A big reason gamers are not going above 8 cores is the CCD latencies and each CCD having their own last level cache.
CCDs not sharing LLC is especially bad with 3D cache. Which is why AMD hasn’t yet shipped models with two CCDs each with 3D cache. That setup would still have very slow latency between CCDs. Better to core park every game thread to one CCD. But that limits games to 8 cores. No scaling beyond that.
Zen 6 apparently has 16 big core and 32 small code CCDs for professional parts, but consumer CCDs still have 8 cores. So this problem is not going away. 16 core CCD with 3D cache would have been super interesting for gaming.
All the talk about we have enough cores already, games dont use many cores is outdated or soon to be. Modern open world games together with background tasks and everything else going on already reaching the limit. Just take ryzen 5800x3d,7800x3d and open bunch of windows chrome with 20+tabs, outlook, teams, plex server running, plex media player running, discord with someones stream, and then add game to it running UE5 that has big open world and you see your 8core 16thread cpu running up to 80% with some cores oversaturated. We really need a 12core single ccd ryzen cpus. Or put more cache on both ccd. Also have 14900k too, and its same story 8 power cores not enough for games in similar scenarios mentioned above, and so ecores starve P cores of power and Pcores cant boost as high anymore too. Game trying to use all 24cpu cores and when limited manually with process lasso just makes game laggy.
We want Zen 2 days back!!
300 usd 6 core CPUs too little gaming advantage over Intel and a huge gap in productivity.
I mean for fs, the AMD iGPU isn't even all that useful in Resolve and Premiere unlike Intel's iGPU.
Adding C/E cores would in most cases likely hurt gaming due to latency. On the laptop chips there is quite a lot of latency between the clusters which hurts gaming performance.
On-package memory will help to make multicore workloads more competitive versus dedicated GPUs.
As an old Linux-head who tinkers with things, just for fun, a 9950X is (will be) just about perfect. Ok, reducing the latency and such would make it even better, but even now, it compiles like a champion, it ploughs through most of the workloads I throw at this old box.
Like what?
Like I just went and tried to get a Win11 24H2 but that's only available if you suck up to Microsoft. But you can get it by UUDump. The easiest way looked like doing it in Linux, but that involves installing a bunch of tools I don't know, which means some PPAs I don't know...
Let's do it from inside the Win 11 install I have, inside a virtual machine, and then export it, then install it. That's taken quite a while, but I now have two fully activated Win11 installs, a 23 H 2 and a 24 H 2.
Do I get paid to do that?
Hell no ~ it's like a little girl playing with a doll house. I shift Ken around and change his shorts because it amuses me. The curtains should match the drapes. But it worked this old 6700 pretty hard for a couple of hours. If I had that 9950X now, it'd plough through that in about 10 minutes.Getting back to the original question, I think 16 Zen 5 cores is enough, for any reasonable hobbyist / home user / techno-geek. If you want to go up again, order a ThreadRipper, but that goes from being a $1,100 AUD part to a $5,ooo Aust $ part. That's not reasonable. That's not a home geek hobby class chip, despite people trying to present it that way. I can justify an eleven hundred dollar CPU (cough) sort of. There's no way on gods green earth I can justify a five grand CPU.
A lot of college students like myself do production workloads, but can’t afford an i9 or Ryzen 9. That’s where a CPU like a 13600k comes. Insane value when I beat a 7800x3d by up to 30% for $150 less. When it comes to gaming, I mostly would get the same fps as I use a 4070ti and play at 1440p. AMD needs to step it up on the lower end/mid range.
Not true for gaming, I switched from 13600kf to 7800x3d with a 4070 ti and gained a significant amount of performance plus way more stable frame graph. I agree with workloads tho
@zeogamingmc This really depends on the title, settings & resolution. If you're always completely GPU limited a 13600K or a 7800X3D will matter very little, if at all, when paired with a 4070Ti.
@@zeogamingmc I’m assuming the games you play utilized 8 cores, because from online sources, there’s like only a 2% difference with a 4070ti. The difference increases to 10-15% with a 4090 tho.
That's what Threadripper is for
its too expensive and exotic for small scale use, really.
@@sgredschExactly!
@@sgredsch cheaper than a decent used car
If tou need more cores buy threadripper???? Am i dumb?
Threadrippers are insanely expensive.
When u guys do a podcast can ya'll invite more people? It feels like podcast vs q/a feels same, except in one you edit out parts to speed up and the other is continuous...
Each chiplet/CCD needs to be expanded beyond 8 cores. Each CCD is split into 4 CCX's each containing 2 cores, right? And the 6-core chiplets just have 1 CCX disabled? CCX's need to be expanded to 4 cores each. Each chiplet could then supply 16 cores, and the Ryzen 9's would be bumped up to 32 cores.
The problem with thatbis that each chip would become more expensive and there would be more errors!
AMD can increase the amounth of small chips… now we have one or two chips. In future, if needed, they can put 3 or 4 small chips in one cpu!
Much better price and realibility wise!
as long as consoles keep having 8 core cpus, there will be no noticeable advantage to higher core count. the only reason those most expensive ones are a tiny bit faster is that they're better binned and a bit higher clocked.
but those 6-core parts really have to go
I think you guys are missing a big point here. The software adapts to the hardware not the othe way around. If 8 cores or more becomes standard then the software will adapt to it and will utilise them. The 9000 series should have had more cores for it to be a real generational upgrade.
Software that can utilize a lot of cores has been doing so for a long time. Software that can't will never do so. The simple matter of fact is that games are not a great workload for multiple cores. And even getting a game to properly utilize those cores is a ton of work for the devs.
My favourite example of multithreading stupidity is The Last of Us Part 1. The game maxes out so many CPUs while doing basically nothing. You can be standing in a house with no other characters, no action happening and your CPU will still be running at 80+ % utilization on modern 6+ core CPUs.
Well that’s just not true. We still get games nowadays like Hogwarts Legacy where it utilizes less than 4 cores, and Microsoft Flight Simulator which utilizes less than 6 cores
@@yancgc5098 Yep and the games would probably run better if they utilised more cores. So thats a competitive advantage they are not taking an advantage of, while many other games do and they tend to run way better. Both of those games have been criticized for poor performance.
intel are the new amd amd are the new intel nvidas the new IBM
7950X and 9950X are faster than 14900K in most productivity workloads while also being much more efficient.
Biggest joke I’ve heard. 14900k is 20% faster in production workloads. You probably watch videos from reviewers that don’t undervolt or have proper cooling to support the all core frequency. Watch any video and you see the 14900 score between 39000-43000 on r23. 7950x can’t even pass 36k. Meanwhile 9950x is around 39k.
@@ProVishGaming R23 is a productivity workload? LOL!! Watch GN review of 9950X. Both AMD CPUs are leading in Adobe workloads (except Photoshop edit: Premiere only, Photoshop IIRC still AMD in front) and Blender for example.
@@jsbfe9395 photoshop literally uses a couple threads. That’s not a multcore workload. Heck a 9950x barely beats a 7700x which has half its cores. Please tag me when you use proper benchmarks that use all cores.
@@ProVishGaming So let me get this straight: AMD 9950X AND 7950X are beating Intel's 14900K in the following tests (Gamers Nexus 9950X review ruclips.net/video/iyA9DRTJtyE/видео.html) : Blender 4.1 & 4.1.1, Adobe Photoshop, 7-Zip compression, Chromium Compilation, 7-Zip Decompression, SpecWS etc but you're going to disregard all this benchmarks and mention only R23 results where you are admiting a 43K over 39K advantage for 14900K which is a 10% lead, nowhere near enough to the 20% that you claimed). That's really pathetic TBH my friend. Really sad.
@@ProVishGaming So let me get this straight: AMD 9950X AND 7950X are beating Intel's 14900K in the following tests (Gamers Nexus 9950X review ruclips.net/video/iyA9DRTJtyE/видео.html) : Blender 4.1 & 4.1.1, Adobe Photoshop, 7-Zip compression, Chromium Compilation, 7-Zip Decompression, SpecWS etc but you're going to disregard all this benchmarks and mention only R23 results?
they realy need to fix ccd latency
intel gen 12,13,14 i5 have consistently outperformed ryzen 7s in multi-thread workloads. it is time to rename 8 cores to ryzen 5, 6 cores to ryzen 3, and 12 cores to ryzen 7, and price accordingly.
You dont need more cores lol. People forget that even the i9-14900k is what? 8p cores and 16e cores. In most workloads, you'll get better results on 16 more robust cores vs spread out on 16 e-cores. If anyone needs to be increasing core count, its Intel. The e-core technology is cool stuff but the gimmick has worn off and we could really go back to 12 or 16 p-cores and 6-8 e-core technology.
I mean I see the 13600K embarrassing the 9600X and 9700X.
@@Deeptesh97my 13600k oced to 5.6ghz beats a 9700x pboed by 20%. A 13600k is $200 on Amazon. Why should anyone pay $350 for a 9700x??? In gaming they trade blows because of my OC. Like makes no sense. Before I get those “my cpu is going to die” comments, my voltage is 1.27V. It’s a hard limit, meaning it will never go above it. Makes no sense to buy a 9700x.
They couldn't physically fit those P-cores in the same die space with that amount of E-cores & remain in the same powerenvelope on top of the physical space constraint...
The only real issue with the E-cores is a software issue, because their instructionset isn't uniform, so you need a scheduler to make the correct decisions when assigning a task to a certain core.
From what we know from leeks and rumors .zen 6 will double the core count to 16 per Ccd .
I assume the line up will be this
R5 10500 12c
R5 10600 14c
R7 10700 16c
R9 10900 24c
R9 10950 32c
That'd be interesting. Much more so than the 9000 series.
Lies, compiling.
you need your medication?
@@Deathscythe91 Compiling linux from scratch can do that to a bot, more to be pitied than hated. 🙃
@@Deathscythe91oh it’s this idiot again. This is the guy that spend $175 more on a CPU to get the same fps as my Intel 😂😂😂😂. Bro could have gotten a 4080 but Intel bad AMD good 😂😂😂. This is the problem with most AMD shillers.
The 14700K was the best choice for me for good enough gaming performance and great multi-core performance with DDR4-compatibility to reduce costs (as I already owned a good kit).
The tables have turned. Now Intel is providing more multi-threading performance while AMD focusses on gaming.
not true. multicore perf is largely same between 16 p-cores amd and 8p+16e-cores intel, cause garbage e-cores are about equal to a single thread on amd.
9950x is peak desktop multicore, 7800x3d is peak gaming, 7995wx is peak workstation. intel currently wins at nothing but consumption, cost and instability, which is bad for competition and consumers
@@elu5ive not true at all lol! 14900k beats a 9950x in multi core workloads by 10%. Most trash RUclipsrs always show low scores for the i9 because they are unable to cool it. A 14900k all core frequency is 5.7ghz, and at those frequencies, it scores 43k on cinebench r23. At the lower end, my 13600k provides 15% more performance in production workloads compared to a 9700x. 9700x is $350 on Amazon, while a 13600k is $200. Makes absolute no sense to go with AMD. Overpriced and underwhelming.
@@ProVishGaming You got lucky for now and your 14900K hasn't had its degradation/oxidation issues YET.
There's literally no reason for someone on Intel right now to buy the latest Intel, when they refuse to fix the problem. I'm on a 10980XE that's overclocked.
What option do I have except the 9950x? nothing. People don't want to take massive risks on unstable parts. Same as why I'll be on my 3090 till either I hear that the upcoming 5090s have stopped burning randomly through the connector power spikes, or the 60 series rather. I don't want to risk buying dumb products that won't last, and if I do buy them pray it doesn't happen "to me as well".
I will only buy (upgrade) again when AMD starts giving us CCD with 10 or 12Core. They just don't do it because PROFIT
Why AMD use multible chips with low core count?
That keeps the price of single chip low, the chip is reliable, there is less errors in the chip!
AMD definitely should not make huge 12 core chip!
They should make cpu that has for example 3 or 4 small chips instead on one and two we have now!
The main point allmost nobody needs 4*8 cores aka 32 cores!
For those people,AMD allready has Threstripper thatbis much better for multiplayer cores! It has wider memory bandwide, so it can actually feed those zillions of cores. Even 16 core 9950 is more memory starved than core starved, so Treatripper is the better choise if you need more cores in anyway!
@@haukionkannel If AMD gives a 16cores CCD to the treatripper, it could give it to us too. I don't want a CPU with two CCDs. You don't need it because someone told you that you didn't need it, it seems like talk from the time of Intel 4Cores and intel they said that 4Cores was enough....
@@starlightHT so you want to pay Threathripper price for that one big CCD? The bigger the CCD, the more there will be falty ones! That is why AMD makes multible small chips for consumer products. There will be less faulty ones, so they can keep the price low! If you want big chips buy Threathripper. If you want to have cheap. Buy consumer versions that has small and cheap chips.
I want games and programs that use 16 cores or more! We just don`t even have games that use more that 6 cores even we have bigger cpus. So the problem at this moment are the programs, not the amounth of cores in CPUs!
And as i did say. If we some day need 24 cores tuplat games. AMD can and should release cpu that has 3 8 cores chips and keep price low instead of making expensive monolith chip! That is what did make AMD Zen cpus so good and so popular. Cheap smal chips and use multible of them, is someone need more cores!
@@haukionkannel I want something exciting and worth my money, is that too much to ask? How much do you think those two pieces of sand that the 9950x has cost? That costs around 50/100 euros and AMD charges 700 euros. Are you telling me that increasing the CCD size a little is too much to ask?
@@starlightHT
They are small, so they cost very little!
If they increase CCD size expect 16 cores to cost more than douple what 7950x cost now!
Most likely it would triple the cost. The very reason AMD did go to chiplets was to reduce the price! And the smaller one chiplet is, the cheaper it is.
All in all the defectivs increase exponentially when you go bigger chips. There are some internet Pages that explain how much harder it is to get working chip when the size is X vs when the size is Y. But in general the correlation is strong!
No, not e-cores. E-cores, while interesting, aren't as useful as they need to be. I have heavy reservations about Intel removing hyperthreading and relying on e-cores for the workloads.
As for AMD, a product shift would be nice. But also more cores.
For instance, launch prices for an upcoming product stack I'd like to see:
6-core Ryzen 3 for $199
8-core Ryzen 5 for $299
12-core Ryzen 7 for $399
16-core Ryzen 9 for $699
That's a shifted stack, eliminating the 900x lineup and shifting the rest of the prices down.
And then in a generation or two, say Zen 8, or w/e, work toward increasing core counts again, primarily in the middle and high-end:
6-core Ryzen 3 for $199
8-core Ryzen 5 for $299
16-core Ryzen 7 for $399
32-core Ryzen 9 for $699
Binning is straight forward. The core piece is the 8-core CCD. The Ryzen 3 is a low yield with 2 cores disabled. The Ryzen 7 is two CCDs. The Ryzen 9 is 4 CCDs.
4 CCDs does seem like a lot, and the chance for latency high, so maybe a core-increase would just logically bump the entire product stack, so there'd be at most 2 CCDs, like we currently have.
You are wild for saying ryzen 3 and 5s launch price should be 200-300$ were intel i3 and i5 launched for 110 and 200$
@@animecutscenes3414 I was literally copying the last known prize of the Ryzen 3 launch price. Not my suggested price at all.
@@ElladanKenet" launch prices for upcoming product stake i would like to see"
I assumed you were suggesting for those absurd launch prices
@@animecutscenes3414 Yes. Note LAUNCH price. Those are the same launch prices of Zen 5. Which, however much consumers would like, AMD is likely to either meet or exceed.
So if the previous Ryzen 3 was launched at $199, then the next one to launch would also be at $199, or higher.
It's not a great price. It's simply the price the old stuff was at. Of course I'd like something more reasonably priced.
Getting rid of HT will be so much better for security though. It's quite a dated hardware design that's had its use & can now be laid to rest because we have other options available now.
4 cores is enough, 6 cores is the sweet spot, 8core is plenty for now and later, 12 and 16 is for pros.
The sad thing is that AMD pioneered the advanced packaging with Fury and Vega only Intel to be the first making CPUs with similar packaging.
Why are gamers the biggest cryers on yourltube? They need sunlight. @JayzBeerz-tearz