Friday, December 2nd 2022

AMD Readies 16-core, 12-core, and 8-core Ryzen 7000X3D "Zen 4" Processors

AMD is firing full cylinders to release a new line of Ryzen 7000-series "Zen 4" Socket AM5 desktop processors featuring 3D Vertical Cache, at the earliest. Faced with a significant drop in demand due to the slump in the PC industry, and renewed competition from Intel in the form of its 13th Gen Core "Raptor Lake" processors, the company is looking to launch the Ryzen 7000X3D desktop processors within January 2023, with product unveiling expected at AMD's 2023 International CES event. The 3D Vertical Cache technology had a profound impact on the gaming performance of the older "Zen 3" architecture, bringing it up to levels competitive with those of the 12th Gen Core "Alder Lake" processors, and while gaming performance of the Ryzen 7000 "Zen 4" processors launched till take match or beat "Alder Lake," they fall behind those of the 13th Gen "Raptor Lake," which is exactly what AMD hopes to remedy with the Ryzen 7000X3D series.

In a report, Korean tech publication Quasar Zone states that AMD is planning to release 16-core/32-thread, 12-core/24-thread, and 8-core/16-thread SKUs in the Ryzen 7000X3D series. These would use one or two "Zen 4" chiplets with stacked 3D Vertical Cache memory. A large amount of cache memory operating at the same speed as the on-die L3 cache, is made contiguous with it and stacked on top of the region of the CCD (chiplet) that has the L3 cache, while the region with the CPU cores has structural silicon that conveys heat to the surface. On "Zen 3," the 32 MB on-die cache is appended with 64 MB of stacked cache memory operating at the same speed, giving the processor 96 MB of L3 cache that's uniformly accessible by all CPU cores on the CCD. This large cache memory positively impacts gaming performance on the Ryzen 7 5800X3D in comparison to the 5800X; and a similar uplift is expected for the 7000X3D series over their regular 7000-series counterparts.
The naming of these 7000X3D series SKUs is uncertain. It's possible that the 16-core part is called the 7950X3D, and the 12-core part 7900X3D; but the 8-core part may either be called the 7700X3D or 7800X3D. Quasar Zone also posted some theoretical performance projections for the 7950X3D based on the kind of performance uplifts 3DV cache yielded for "Zen 3" in the 5800X3D. According to these, the theoretical 7950X3D would easily match or beat the gaming performance of the Core i9-13900K, which begins to explain why Intel is scampering to launch the faster Core i9-13900KS with a boost frequency of 6.00 GHz or higher. The report also confirms that there won't be a 6-core/12-thread 7600X3D as previously thought.
Source: harukaze5719 (Twitter)
Add your own comment

153 Comments on AMD Readies 16-core, 12-core, and 8-core Ryzen 7000X3D "Zen 4" Processors

#101
dwarfcow
VarioFirst, you are comparing some fairly old architectures. Second, game developers understand that the trend is moving towards an 8 core mainstream, and will design upcoming games accordingly.
The PlayStation 2 could be argued to have had 8 cores (it was 8 separate processing parts, and the PlayStation 3 and 4 both had 8 cores, so it's definitely not a "new" trend.
PunkenjoyNo, Dev target a specific performance level.

Still today, A 6 core CPU that have more single thread performance than a 8 core will have same or better gaming performance.

Core aren't slot that Dev fills to make their game. Think about it like a boss with a set of employee. Let say he have 6 incredible employee and another boss have 8 average employee. Both boss run a ghost writer shop and need to produce a 100k word book. (Aka A Frame for gaming)

There are two task, writing the words and at the end editing it. Editing would be done only by 1 staff to make it consistent. (It's the main thread in gaming that control and sync all the other)

Boss A 6 employee can write 1300 words per hours each and review at 5200 words pour hours
Boss B 8 employee can write 1000 words per hours and review at 4000 words per hours

so each boss need to start scheduling the work. He first assign all employee on the first hours into writing words

Boss A would be at 7800 words and Boss B would be at 8000 words.
Then one of the employee would start to review the text for editions,

For boss A, the 5 remaining employee would be done writing 11.8 hours later but they would have to wait for the reviewer. He would be done 19.23 hours later. (for a total time of 20.23 hours)

For boss B, the 7 remaining employee would be done writing 11.5 hours later but the editor would only be done 25 hours later (for a total of 26 hours.)

The fact that the boss B had more "Slot" to put the work into didn't mean they had the job done earlier.

Game work exactly that way. They are way more multithreaded than in the past, but there is still a main thread that is criticial. This is why gaming performance don't scale linearly. (Unlike rendering that have an almost linear scaling since all the job can be done by themselves).

What you need to think is there is by example for each frame x amount of CPU operation per seconds to do in the main thread, and there x amount of cpu operation to be done that can be multithreaded.

A faster core is able to do more operations per second. A really faster core can even do those one after another before the slower 2 core can complete them.

It still the case today, and it will be still the case in the future. Because game have to run a main core as fast as possible then spread the defined amount of load to other core. If faster core finish earlier, they can grab more work before a slower cpu with more core complete it's first job.



Again, in the current gen. 7600x 6 core beat all 8+ core from the previous generation. This is also why the E-core doesn't really help.

One of the gain right now of having a more core CPU is they are binned to run at higher clock. I would like to see what a 7600x running at the same frequency as the 7950x would do. I am pretty sure they would be very close.
I like this metaphor, and let's keep it going. Boss A relies heavily on his star employees, who have to grind it out over and over, day after day, and eventually get burned out and light the building in fire, while boss B's employees are content and just keep chugging a long. Over time Boss B gets better and better at breaking up the work load to be more balanced, better utilizing his resources, Boss A recovering from smoke inhalation and his employer's autopsy of why the business failed pushes him out of the job.
Posted on Reply
#102
efikkan
Article…and a similar uplift is expected for the 7000X3D series over their regular 7000-series counterparts.
If I'm not mistaken, this article has taken the performance gain of 5800X->5800X3D in selected benchmarks, and multiplied Zen 4's performance with this gain to estimate the performance level with Zen 4 + 3D V-Cache. This is wrong on so many levels that it's completely useless. Firstly Zen 4 has different performance characteristics and a much better front-end, which means the gains of additional L3 is likely to be less significant. Secondly, and most importantly, gaming performance is only indirectly related to CPU performance and there are diminishing returns in gaming performance with faster CPUs. So for these reasons, Zen 4 with 3D V-Cache is likely to get less gains in gaming than Zen 3.
Why_MeA 7900X3D / 7700X3D makes sense. A 7950X3D not so much imo.

Keep in mind that graph shows average FPS, which is a very bad metric across games. If one CPU boost extra FPS in a game at ~400 FPS vs. another performing well in the rest, then the weighting will be skewed towards the high FPS games. So such graphs will amplify cases where a handful games performs extra well.
pjl321I could also easily see AMD pulling in Threadripper 6000 (Zen 4 based) as they don't need to worry about not having enough chiplets for Epyc like they did during Zen 3, things aren't flying off the shelves right now (hence the massive price cuts to Ryzen) so they have more than enough production for all variants for all markets.
Threadripper have been produced in incredibly low numbers lately due to yield requirements. You need exceptionally good quality chips to run them at these clock speeds.
pjl321I also don't believe AMD can afford to take their foot off the gas right now, they need all the performance crowns they can get to help fight Nvidia mindshare for their Radeon division. So if Zen 4 3D wins in gaming, Zen 4 Epyc wins in servers and Zen 4 Threadripper wins in workstation then the 7900XTX will subconsciously get a much better reception because 'AMD' is seen as a dominant winner.
That assumes extra L3 cache will make them "win" in gaming, and it's not certain that 3D V-Cache will enable Zen 4 to thoroughly beat Raptor Lake. Don't forget that in the weeks leading to Zen 4, many were estimating (regular) Zen 4 to far exceed Alder Lake and Raptor Lake in gaming performance, and that didn't happen.

3D V-Cache works in certain server markets because some workloads are very cache sensitive, which is why this technology was developed in the first place. I haven't seen many workstation workloads which shows the same characteristics, so I doubt a 3D V-Cache Threadripper would make sense.
Posted on Reply
#103
wheresmycar
I've been hearing this console+8-core revolution for several years now. Since the quad 7700K kaby lake (5/6 years ago) i've been observing how the desktop 6 v 8 core "gaming" skirmish played out with each newly launched family Gen of CPUs. There is definitely a broader difference in performance jumping from a quad to 6-cores but moving to 8 doesn't necessarily produce any "significant" benefit hence hardly a compelling reason to pay +$100-$150 launch premiums. From what i've seen, 8 core or higher parts tend run on a marginally faster boost clock (usually 100-300mhz) where small but negligible performance gains are evident. In other words, on a 6 core pedestal single core supremacy is still king although i'm not discounting 8-core parts considering there are a few extremely CPU-battering titles which can definitely benefit (maybe not so much with climbing FPS but more-so with smoother gameplay).
Posted on Reply
#104
Ownedtbh
why not release the 3D Cache from the begin with?
Posted on Reply
#105
Count von Schwalbe
Ownedtbhwhy not release the 3D Cache from the begin with?
The engineering for the basic die had to be completed first, before looking at stacking them.

Also, it helps clear up ASEGA bugs for when new CPUs launch.

And they probably wanted to drop them as an "answer" to raptor lake.
Posted on Reply
#106
efikkan
wheresmycarI've been hearing this console+8-core revolution for several years now. Since the quad 7700K kaby lake (5/6 years ago) i've been observing how the desktop 6 v 8 core "gaming" skirmish played out with each newly launched family Gen of CPUs. There is definitely a broader difference in performance jumping from a quad to 6-cores but moving to 8 doesn't necessarily produce any "significant" benefit hence hardly a compelling reason to pay +$100-$150 launch premiums. From what i've seen, 8 core or higher parts tend run on a marginally faster boost clock (usually 100-300mhz) where small but negligible performance gains are evident. In other words, on a 6 core pedestal single core supremacy is still king although i'm not discounting 8-core parts considering there are a few extremely CPU-battering titles which can definitely benefit (maybe not so much with climbing FPS but more-so with smoother gameplay).
For the vast majority of games, 6 (fast) cores is going to be plenty for a good while. Some seemed to think that we should have 16 cores to feed a GPU by now, but the overall trend in game development is actually that the CPU is going to less important (relatively speaking), providing it meets a certain performance level. This is because the fact that complexity is gradually being moved to the GPU (shader) code. So we are not likely to see games running at high load on 16 cores, unless talking about special use cases like encoding multiple streams, etc.
Ownedtbhwhy not release the 3D Cache from the begin with?
Because of availability of the cache chiplets. These wafers are extremely costly, and takes up capacity which otherwise could be used to produce more CPUs. The cost is offset by using the best cache chiplets on Epycs, so the overall volume of such chips is going to be limited.
Posted on Reply
#107
Vario
wheresmycarI've been hearing this console+8-core revolution for several years now. Since the quad 7700K kaby lake (5/6 years ago) i've been observing how the desktop 6 v 8 core "gaming" skirmish played out with each newly launched family Gen of CPUs. There is definitely a broader difference in performance jumping from a quad to 6-cores but moving to 8 doesn't necessarily produce any "significant" benefit hence hardly a compelling reason to pay +$100-$150 launch premiums. From what i've seen, 8 core or higher parts tend run on a marginally faster boost clock (usually 100-300mhz) where small but negligible performance gains are evident. In other words, on a 6 core pedestal single core supremacy is still king although i'm not discounting 8-core parts considering there are a few extremely CPU-battering titles which can definitely benefit (maybe not so much with climbing FPS but more-so with smoother gameplay).
Disregard consoles, the difference is 8 core parts on desktop PC being mainstream in 2022. PC game developers will make use of them.
Posted on Reply
#108
Tek-Check
Could someone, please, explain those Korean extrapolated graphs?
Posted on Reply
#109
pressing on
Tek-CheckCould someone, please, explain those Korean extrapolated graphs?
The Korean site references wccftech as the source for its information, and that source does not include any graphs. The graphs would seem to be pure guesswork and nothing else. The suggested unveiling date for these CPUs is CES early next year so it won't be long before some more definitive information is available.
Posted on Reply
#110
Minus Infinity
V-cache models brought forward basically a whole quarter, amazing how good competition is for the consumer. Alas prices will rise even further over Raptor Lake. The funny thing is Zen 4 is already plenty good enough for gaming, it doesn't need help there IMO I need v-cache to show solid gains in a lptmof non-gaming software especially stuff like sims run in Matlab, Mathematica, fluid sims like Ansys, OpenFOAM, Comsol etc.
Posted on Reply
#111
Garrus
1d10tWhat a fine example.



Some of devastating loss were 4c8t i3-10105F beating 8c16t Ryzen 7 2700x, 4c8t i3-12100 beating 8c16t 9900k or in a manner of speaking 6c12t 5600x beating 10c20t i9-10900K. I don't get the idea of 8 core being a bare minimum for playing console port game.
You also showed the 8 core 5800X3D 30 percent faster versus the 5800X because of cache. 8 cores will be the best. 8 core 7800X3D will slaughter every other CPU in this game. Yeah a 6 core one would work also, but it won't exist, so...

Also your chart proves that 8 cores help. 3700X versus 3600X for example, it is faster.
Posted on Reply
#112
Legacy-ZA
In my opinion, they should have just released these versions from the get go.
Posted on Reply
#113
Mussels
Freshwater Moderator
LionheartIt doesn't suck at everything else, it's just normal/mediocre
Hardly even that, the performance deficit vs a 5800x (I own both) is tiny


Some people compare an 8 core AMD CPU to a 16 core AMD or 20 core intel and think it's terrible because the multi threaded results are lower... but it's an 8 core CPU.
GarrusAlso your chart proves that 8 cores help. 3700X versus 3600X for example, it is faster.
Every step up on the AMD side has a roughly 200Mhz advantage, when clocked the same they perform the same - that's not the cores giving the boost


One big thing is that my 5800x3D is faster than my 5800x in games by a considerable amount (20% more or so on average, and the 1% lows and 0.1% low are just... gone) - while using a LOT less power.

An 8 core x3D 7000 series CPU should absolutely demolish these gaming results, since they've got better max/peak FPS values already, the cache should bring up the minimums and if it follows the 5800x3D's trend it'll use less power doing so, running a lot colder and negating a lot of issues with the AM5 chips
Posted on Reply
#114
JustBenching
The KingWhen it comes to gaming you can't beat the price performance of the 5800X3D period. 7600X does out perform 13600K on AVG.
Really? Most cpus beat the 5800 3d in terms to price performance, what are you talking about? Especially in their until very recent 450€ pricing, it was a horrible price to performance CPU. I can name 10 recent cpus that beat it in that regard
Posted on Reply
#115
The King
fevgatosReally? Most cpus beat the 5800 3d in terms to price performance, what are you talking about? Especially in their until very recent 450€ pricing, it was a horrible price to performance CPU. I can name 10 recent cpus that beat it in that regard
Sorry, should clarify that if you already on AM4 platform you can't beat the 5800X3D price performance in gaming. ;)

Forgot about the forum police in the comments section, my bad. :p
Posted on Reply
#116
Mussels
Freshwater Moderator
fevgatosReally? Most cpus beat the 5800 3d in terms to price performance, what are you talking about? Especially in their until very recent 450€ pricing, it was a horrible price to performance CPU. I can name 10 recent cpus that beat it in that regard
I mean, the 5800x3D dropped massively in price and has stayed down
Maybe it hasnt where you live yet, but it's down ~35% from launch and a lot cheaper than a barely competing 12900K or 13900K that has worse 0.1's and 1% FPS results
(And no, don't go comparing the CPU alone. the x3D can be cooled by a damn wraith stealth on any budget board and keep the good performance, nothing else can)
Posted on Reply
#117
JustBenching
The KingSorry, should clarify that if you already on AM4 platform you can't beat the 5800X3D price performance in gaming. ;)

Forgot about the forum police in the comments section, my bad. :p
Even then, of course you can. The 5600 literally wipes the floor with the 3d in terms of performance per price
MusselsI mean, the 5800x3D dropped massively in price and has stayed down
Maybe it hasnt where you live yet, but it's down ~35% from launch and a lot cheaper than a barely competing 12900K or 13900K that has worse 0.1's and 1% FPS results
(And no, don't go comparing the CPU alone. the x3D can be cooled by a damn wraith stealth on any budget board and keep the good performance, nothing else can)
I said EVEN with the reduced price, even if you already have an am4 board, its not a value cpu. Its actually pretty average / mediocre in terms of performance per price, even in gaming. The 5600 at 130 euros literally wipes the 3d easily. Ive no idea how you can argue otherwise

Regardless of that, you think the 3d beats the 13900k? Lolwut

I know this is with launch prices, but other cpus have fallen in price as well, even more than the 3d has. It sits at the bottom of the performnace per dollar graph

tpucdn.com/review/amd-ryzen-7-5800x3d/images/performance-per-dollar.png
Posted on Reply
#118
Slimmie205
Space Lynxmy 13600kf and 6800xt is already maxxing out everything i throw i at at 165hz 165 fps 1440p. Unless I start playing more demanding games, I may never need to upgrade again.

Anything above these mid-range options is strictly for high refresh 4k gamers imo.


that being said, I might be stupid and get a 7900 xtx :D
Im on the same boat. Have a 12900ks and a 6950xt. Have a 144hz 1440p monitor. No FPS issues yet... I can wait a couple of years I guess
Posted on Reply
#119
Xaled
7950X3D does make sense. It is a prestige thing. You don't want your best CPU to fall behind neither opponents nor your own cheaper, lower-tired CPUs.
5800X3D was just a proof of concept, now it is time for complete implementation
Posted on Reply
#120
Punkenjoy
We are no longer in the last decade. Devs do not target a number of cores. They have X amount of operation witch X amount can be send to other thread. They can scale it out to more thread but the more you scale out the multithreaded part, the more the single threaded part become the problem.

Before modern low level API, Game were hard to multithread and they had to decide what to run on other cores, during thoses days it was true that a game could benefits from having x cores. Some game had a core for physics, for sims, sound, etc. Big large thread. That was not very effective and performance were very poor if you didn't had enough thread to run them.

But today, it's no longer the case, The modern API allow the devs to send a lot of operation to others thread. By example all draws calls (command sent to the GPU to perform an action) can now be multithreaded. There are a lot of them each frames. We wouldn't require a 10k cores CPU for that just because there is that much thread generated. They can be run one after another. A faster core will be done with one thread way quicker and can start another one quicker too leading to more performance than a slower CPU with more cores.

Even more today since all those thread have data in common that be cached, if they are not too far, it can lead to very nice performance gain since the CPU will not waste time waiting for the data from main memory.

That bring back to the point of 3D-Vcache. This is useful when a software reuse frequently the same data and that data can fit or close to fit in the cache. Game really behave like this since it's generating frame one after another. And you realise that if you want high FPS (60+) and you have to wait for memory latency, your working set need to be quite small. 250-500 MB depending on memory and other stuff.

So having the possibility to cache a significant portion of it allow you to either do more access to the memory or run thing faster.

Many well threaded application that are not game that are run by consumers are either full memory access (like video encoding witch is streaming from memory) or 3D rendering (were the apps have been designed to work with dataset that fit well into L2/L3 without 3D-vcache.)

Those are the most commonly benched case but there are plenty of scientific or others types of apps that would benefits from having a larger L3. The 7950x3d might be niche for gamers(luxury item) but it could really benefits some specialized workload.
Posted on Reply
#121
Slizzo
MusselsBecause these 3D chiplets need lower voltages, they cant run at the same clockspeeds and wattages as the regular chips - they'd get slaughtered by the basic websites that only care about the clock speed and the cinebench scores

going beyond a single CCX is a big risk because it hurts what these chips are best at - low latency cached data
MusselsAn 8 core x3D 7000 series CPU should absolutely demolish these gaming results, since they've got better max/peak FPS values already, the cache should bring up the minimums and if it follows the 5800x3D's trend it'll use less power doing so, running a lot colder and negating a lot of issues with the AM5 chips
This has been touched on already; but Robert Hallock stated that the limitations that were present for Zen3 stacked V-Cache are not present on Zen4 stacked V-Cache (i.e. being limited by voltage and whatnot).

So the much lower clocks and whatnot experienced by the 5800X3D will not be an issue with the Ryzen 7000 series. Also, with Zen4 being designed from the outset to work with 3D V-Cache should help in its ability to make better use of that cache vs. Zen3 where it was retrofitted as a proof of concept.


I am currently waiting on these processors to drop to upgrade to (hopefully) a Zen4 X3D processor from my current 10940X. If they offer a 7900X3D, I'll probably jump on that providing the performance is there in gaming. One of the issues with some games is that they don't like high core count CPUs for some reason.
Posted on Reply
#122
Minus Infinity
Xaled7950X3D does make sense. It is a prestige thing. You don't want your best CPU to fall behind neither opponents nor your own cheaper, lower-tired CPUs.
5800X3D was just a proof of concept, now it is time for complete implementation
But most people are thinking gaming when they think v-cache, who buys the 7950X for gaming. AMD needs to really do a good job of highlighting v-cache benefits outside of gaming. They should have slides prepared showing it's benefits in other types of software. We only have results for server applications mostly from Milan-X to see how it can benefit non-gaming.
Posted on Reply
#123
GhostRyder
I cant wait to see the performance of these chips! I will be curious if the 3D stack helps alleviate some of the chiplet problems we have seen on the 12 and 16 core variants as I may go for the 12 core instead of the 8 depending on the performance.
Posted on Reply
#124
beedoo
Minus InfinityBut most people are thinking gaming when they think v-cache, who buys the 7950X for gaming. ...
I will. Thanks :)
Posted on Reply
#125
wheresmycar
Minus InfinityBut most people are thinking gaming when they think v-cache, who buys the 7950X for gaming. AMD needs to really do a good job of highlighting v-cache benefits outside of gaming. They should have slides prepared showing it's benefits in other types of software. We only have results for server applications mostly from Milan-X to see how it can benefit non-gaming.
streaming, encoding, transcoding, rendering, etc etc + gaming, no?

I pally up with 3 mates whilst gaming... two of them (partners in crime) are weddings/events videographers + trained drone pilots (or something like that).... so a ton of video editing/rendering on the go.... both buy into these types of higher core count CPUs. They've already bought into AM5 with a 7950X for strictly work purposes.... and both of them are holding off on their gaming+work rigs for something like a 7900XD/7950XD/etc. Another friend who's not looking to upgrade is currently on a 10900K for gaming, streaming and transcoding... I wouldn't be surprised if he jumps on the same bandwagon at some point.

Plus, i regularly see people on PCPP forums requesting build lists with similar demands. In short, there's definitely a market for it. Not for me though, i keep my gaming rig separate from work systems (not that my workload demands/benefits with 16 cores or whatnot)

X3D is welcome on all Zen 4 tiers but its a darn right shame they're gonna skip on a 7600X3D... thats baffling!
Posted on Reply
Add your own comment
Nov 21st, 2024 12:56 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts