Zen 4 X3D Limited to 8-Core and 6-Core, No Meteor Lake in 2023: Frosty Year Expected for CPU Market

[btarunr]

A reliable source with CPU and platform leaks, ECSM_Official, made some new predictions about release timelines of upcoming desktop processors, and how 2023 could play out for Intel and AMD. 2022 is done, with no new desktop processor SKUs expected to launch from either brands. Intel is expected to flesh out its 13th Gen Core "Raptor Lake" desktop processor family in Q1 2023, with the addition of "locked" non-K SKUs spanning all four brand extensions (i3/i5/i7/i9). Besides these, Intel is expected to launch its new flagship, the Core i9-13900KS, with boost frequencies hitting the 6 GHz mark, in an attempt to ward off the threat from "Zen 4" with 3D Vertical Cache, a technology that springboarded "Zen 3" gaming performance to match that of "Alder Lake."

Both the i9-13900KS and AMD Ryzen 7000X3D processors are expected to launch toward the middle of H1-2023 (March-April). AMD is only expected to launch 6-core/12-thread and 8-core/16-thread SKUs with the 3DV cache technology. These would be single-CCD packages. There's no word on dual-CCD ones with 12-core or 16-core counts, so a Ryzen 9 7950X3D is not on the horizon. AMD is expected to debut its entry-level A620 motherboard chipset in Q2-2023. This chipset reportedly lacks CPU overclocking capability, is expected to lack PCIe Gen 5, and caps memory speed to DDR5-4800.



Intel is expected to refresh its 13th Gen Core processor lineup with new SKUs in Q3-2023. For now, all that's known about these "Raptor Lake Refresh" processors is that they come with 100-200 MHz speed bumps over existing 13th Gen Core SKUs of the time. This would also mean that the LGA1700 platform and "Raptor Lake" will be Intel's mainstay throughout 2023, and the 14th Gen "Meteor Lake" isn't launching until 2024.

While "Meteor Lake" will dominate Intel's mobile processor lineup, it will have a limited presence on the desktop side, due to its core-count of 6P+16E, despite IPC uplifts on both the P-cores and E-cores. These processors will, however, debut the next-generation Socket LGA1851 platform. 2024 will see Intel launch both the "Meteor Lake" and 15th Gen "Arrow Lake" processors. The "Arrow Lake" SoC will restore core-counts to the familiar 8P+16E, with IPC uplifts for at least the P-cores, over those on "Meteor Lake." In summary, 2023 will be a lukewarm year for new processor/platform launches, mainly due to the downturn in the PC industry. Intel and AMD will want to make less-risky bets.

View at TechPowerUp Main Site | Source

 

[Dirt Chip]

The market is on ice, but at least we have planty new worm CPU & GPU choice's to keep us through the long night.

 

[ratirt]

I guess Intel and AMD will start to put more efforts into new architectures of processors. Working on a totally new architecture that will fix some things in energy consumption and temps. I think it is for better. Maybe in 2024 we will get processors seriously revised and easier to tame with power and temps.

 

[usiname]

Lets the clown show with delay after delay of the 7nm and refresh after refresh of the 10nm start again

 

[Dirt Chip]

I guess Intel and AMD will start to put more efforts into new architectures of processors. Working on a totally new architecture that will fix some things in energy consumption and temps. I think it is for better. Maybe in 2024 we will get processors seriously revised and easier to tame with power and temps.

It is already a few years in progress.
New arch is a 5 years endeavor, you dont wait to lunch and the start to figure out new arch

Lets the clown show with delay after delay of the 7nm and refresh after refresh of the 10nm start again

And yet, those refresher pretty much dominant now in many cases.
Say a lot on the second player...

 

[usiname]

And yet, those refresher pretty much dominant now in many cases.
Say a lot on the second player...

Yes, with 10x the RED budget and still behind. Thats says a lot. Meanwhile do you remember where was AMD 6-7 years ago? And Intel?

 

[Hyderz]

2023 should be okay to build a nice mid range pc since most of the high end gpu are out this year

 

[john_]

Intel doesn't fear Zen 4 and AMD probably sees no reason to invest heavily in a platform that is still expensive and people do not buy it. I mean, even if they come out with a 16 core X3D model, how many are out there ready to invest in such a CPU? AMD will be better selling EPYC CPUs for now and hope that the upgrading wave on the AM4 platform will continue, securing them part of their CPU market share (AMD did suffer a huge dive in market share lately).

 

[ratirt]

It is already a few years in progress.
New arch is a 5 years endeavor, you dont wait to lunch and the start to figure out new arch

sure but from Intel's perspective the "new" launch was every year. Now Intel skips next years launch. 5 year but evey year you get a refresh of a refresh. I'm talking about a new thing and AMD and Intel skip next year so I guess they will revise everything. Every aspect of the CPUs and maybe we will get something very cool and fast next time.

 

[Dirt Chip]

Yes, with 10x the RED budget and still behind. Thats says a lot. Meanwhile do you remember where was AMD 6-7 years ago? And Intel?

Behind what?
Preformance? Cost?
Dont forget that intel make their own cpu.
AMD outsource them so the complement on smaller node goes to tsmc, not AMD.

AMD make enormous change and uplifted its business tramendusly plus change the face of the market from the horrible bulldozer era.
Intel stagnated at that time but never was clear cut behind AMD.

Sounds you are cheering AMD while opposing Intel yet they are both the same giant tech companies that will do whatever thay need to make profit.

 

[Bwaze]

"AMD is only expected to launch 6-core/12-thread and 8-core/16-thread SKUs with the 3DV cache technology. These would be single-CCD packages. There's no word on dual-CCD ones with 12-core or 16-core counts, so a Ryzen 9 7950X3D is not on the horizon."

And the 3D counterparts will again have lower clocks and lower productivity, and only gain in very specific tasks,
- gaming fortunately being one of them.

But modern CPUs are very rarely bottlenecks, unless you specifically search for such situations - low resolution high FPS gaming with top end graphics cards. Once you introduce v-sync or other frame limiting tech, or use higher resolution or ray-tracing, effect of CPU speed largely dissapears.

 

[Bomby569]

i don't get it, launching new cpu's is "risky"? how so.

 

[Dirt Chip]

sure but from Intel's perspective the "new" launch was every year. Now Intel skips next years launch. 5 year but evey year you get a refresh of a refresh. I'm talking about a new thing and AMD and Intel skip next year so I guess they will revise everything. Every aspect of the CPUs and maybe we will get something very cool and fast next time.

There is nothing new thay can do for the next year. All the design is already finished.
What they can do is to choose not to OK the hell out of the CPU out of the box and leave it to the user.
News flash, it will not happen in the face of fierce compatition.

 

[Daven]

And the 3D counterparts will again have lower clocks and lower productivity, and only gain in very specific tasks,
- gaming fortunately being one of them.

The 5800X3D was downclocked due to the 140W limit of socket AM4. Since it was a 105W chip, that limit would have been exceeded by the addition of the extra cache and the higher turbo clock.

Now socket AM5 has a 230W limit. The 7600X and 7700X are lower wattage chips using only 105/180W versus the 170/230W of the 12/16 core parts. There is power headroom for cache to be added to the 6 and 8 core versions AND increase the clocks.

It was not some sort of physics that forced AMD to lower clocks after adding cache on the 5800X3D but the power limitations on the old socket AM4. AMD does not have that problem on AM5 unless they were adding cache to the higher power 12 and 16 core parts which go up to 230W. The leak says they are not adding cache to these parts.

I estimate higher clocks AND 3D cache on the 7600X3D and 7800X3D.

 

[ratirt]

There is nothing new thay can do for the next year. All the design is already finished.
What they can do is to choose not to OK the hell out of the CPU out of the box and leave it to the user.
News flash, it will not happen in the face of fierce compatition.

I disagree or You simply dont understand me. There is plenty to do even if the design is finished and maybe that is why there is nothing new next year and that is my point. It is for a reason. So instead releasing a refresh of a refresh, both companies would rather focus on some serious gains not just performance but also energy consumption and heat.
It does not matter how it will happen. What matters is, and i cheer for that, both companies draw conclusions about the power consumption and heat and will try to fix those things instead serving all of us power hungryfurnaces.

 

[Bwaze]

The 5800X3D was downclocked due to the 140W limit of socket AM4. Since it was a 105W chip, that limit would have been exceeded by the addition of the extra cache and the higher turbo clock.

Now socket AM5 has a 230W limit. The 7600X and 7700X are lower wattage chips using only 105/180W versus the 170/230W of the 12/16 core parts. There is power headroom for cache to be added to the 6 and 8 core versions AND increase the clocks.

It was not some sort of physics that forced AMD to lower clocks after adding cache on the 5800X3D but the power limitations on the old socket AM4. AMD does not have that problem on AM5 unless they were adding cache to the higher power 12 and 16 core parts which go up to 230W. The leak says they are not adding cache to these parts.

I estimate higher clocks AND 3D cache on the 7600X3D and 7800X3D.

5800X was rated the same as 5900X and 5950X, but it never reached the same power draw - and neither did the 5800X3D:

5800X3D still has lower power draw than 5900X and 5950X, even in most brutal tests. And those two don't even have any limitation on overclocking - so socket power delivery limitation surely can't be the only reason 5800X3D had a reduced boost frequencies and locked overclocking?

 

[Valantar]

I'm honestly entirely fine with this. While AMD is at a distinct disadvantage in terms of core counts and nT workloads, 6c X3D CPUs should be a killer gaming offering - they just need to come with appropriate price cuts across the range. Cut the 7600X to $200-230, sell the 7600X3D for $270-300, and do a similar move for the 7700X and the 7700X3D.

I'm more "worried" (not that current CPUs aren't stupidly fast already, so no real worries, but still...) that Meteor Lake is seeing yet another delay. I wonder what the catch is. Low fab yields? Low packaging yields? Something else?

socket power delivery limitation surely can't be the only reason 5800X3D had a reduced boost frequencies and locked overclocking?

The 5800X3D had reduced clocks partially for thermal protection due to the extra insulating layer of (blank) silicon on top of the CPU cores, and partially because the cache die couldn't tolerate voltages above 1.35V at all. This is also the reason for it being locked for OCing, as OCing necessitates voltage control, and higher clocked Ryzens often boost above 1.45V, even if it's for short periods of time.

The question is whether the cache die for Ryzen 7000X3D will have its own voltage plane, decoupled from core/on-die cache voltages, which would be a possible way around this - if such a separation is possible.

 

[Bwaze]

I'm more "worried" (not that current CPUs aren't stupidly fast already, so no real worries, but still...) that Meteor Lake is seeing yet another delay. I wonder what the catch is. Low fab yields? Low packaging yields? Something else?

Perhaps Intel's analysts predict a really harsh economic downturn, so they are now focusing on trimming down the company and reducing costs, and will focus on delivering a new CPU generation when the conditions are better?

 

[Valantar]

Perhaps Intel's analysts predict a really harsh economic downturn, so they are now focusing on trimming down the company and reducing costs, and will focus on delivering a new CPU generation when the conditions are better?

That's definitely a possibility, especially as AMD clearly aren't very focused on competing in consumer markets right now for whatever reason.

 

[Vayra86]

Lets the clown show with delay after delay of the 7nm and refresh after refresh of the 10nm start again

Exactly, what else is new, gosh a new node, gosh delays...

and indeed... we're practically still exploring 10~7nm revisions, despite what the marketing dictates. Except now with EUV, which allows them to really execute on the number... too bad marketing is a few steps ahead of reality still.

 

[Daven]

5800X was rated the same as 5900X and 5950X, but it never reached the same power draw - and neither did the 5800X3D:

5800X3D still has lower power draw than 5900X and 5950X, even in most brutal tests. And those two don't even have any limitation on overclocking - so socket power delivery limitation surely can't be the only reason 5800X3D had a reduced boost frequencies and locked overclocking?

Companies tend to be conservative when it comes to product limits. Even though a third party shows higher power draw over the 140W limit, doesn’t mean AMD is using the same metric for internal testing that decides product specs. Stability, longevity and other factors play a role.

Either way, the AM5 six and eight core parts are the ‘lower’ power parts on the platform and have more headroom. The 5800X was already at the ‘higher’ power level for AM4 before adding cache. These are the levels and limits as defined by AMD and not third parties.

 

[Bwaze]

Companies tend to be conservative when it comes to product limits. Even though a third party shows higher power draw over the 140W limit, doesn’t mean AMD is using the same metric for internal testing that decides product specs. Stability, longevity and other factors play a role.

That chart shows whole system power draw, not just CPU. The fact is, 5800X and also 5800X3D had a lower power draw than 5900X and 5950X - and those two processors could also be overclocked on top of that. Socket power draw limitation was therefore certainly not the factor in lowering the boost frequencies of 5800X3D. Or locking the overclocking. Maybe the combination of both would come close to power draw of overclocked 5950X, but Valantar above posted a much more probable reasons.

 

[Pumper]

How disappointing. Who do they thing will be buying the 7600X3D for over $400?

 

[Dirt Chip]

I disagree or You simply dont understand me. There is plenty to do even if the design is finished and maybe that is why there is nothing new next year and that is my point. It is for a reason. So instead releasing a refresh of a refresh, both companies would rather focus on some serious gains not just performance but also energy consumption and heat.
It does not matter how it will happen. What matters is, and i cheer for that, both companies draw conclusions about the power consumption and heat and will try to fix those things instead serving all of us power hungryfurnaces.

You will still get the same refresh but with slight power modification at best. If "serious gains" is only power related than we are agree.
But the node will stay the same as originally pland, the new arc benefit will not change so IPC will not change, core number also will stay the same as pland. Those are the thing, imo, that you need to change in order to get "serious gains"

 

[Daven]

That chart shows whole system power draw, not just CPU. The fact is, 5800X and also 5800X3D had a lower power draw than 5900X and 5950X - and those two processors could also be overclocked on top of that. Socket power draw limitation was therefore certainly not the factor in lowering the boost frequencies of 5800X3D. Or locking the overclocking. Maybe the combination of both would come close to power draw of overclocked 5950X, but Valantar above posted a much more probable reasons.

Again private companies may have quality control criteria that cannot be independently verified by third parties. TPU testing values have nothing to do with internal AMD spec decisions. AMD already admitted that they downclocked due to platform limitations. This is already an admission of a big negative. There is no reason for a cover up due to this admission.

And again, the AM5 has more power headroom and therefore the 3d cache parts could be clocked higher than the non 3d cache parts. I’m conjecturing based on available information as I am not an AMD employee.