Wednesday, September 3rd 2025

AMD "Zen 6" Processors to Use TSMC 2 nm Node for CCDs, 3 nm for IOD

AMD will transition its next-generation "Zen 6" desktop Ryzen and server EPYC processors to a split-node strategy, placing CPU core complex die (CCD) on TSMC N2P 2 nm node variant and building the I/O die (IOD) on the N3P 3 nm node. At TSMC, volume ramp for the 2 nm node is around the third quarter of 2026, opening the door for limited shipments as early as late Q3 or broader availability in Q4 2026. Under the split approach, the compute chiplet will house the Zen 6 cores and a substantially larger shared L3 pool, while the IOD will continue to host memory controllers, PCIe lanes, USB, and the integrated GPU. Earlier rumors suggest that each CCD could contain 12 Zen 6 cores with support for SMT, and a shared L3 cache that may increase to roughly 48 MB per CCD compared with prior generations. Typical consumer configurations would pair up to two CCDs with a single IOD, creating chips with up to 24 cores and 48 threads, while server variants could scale further depending on packaging.

Performance expectations center on achieving double-digit IPC gains, higher sustained clock speeds, and improved power efficiency, thanks to the advanced process nodes. The choice of N3P for IOD helps AMD cut some costs for non-core logic, while the N2P compute die aims to extract greater generational gains. However, as with any new TSMC node, it will carry a premium over the N3P, so using it only for the compute side is the most viable choice. Platform compatibility remains for one more generation, with Zen 6 expected to support the existing AM5 socket. Users can upgrade CPUs without a full platform replacement. If the ramp timing holds, Zen 6 will arrive into a crowded second half of 2026 desktop window. For datacenter applications, the EPYC "Venice" generation, based on Zen 6, will utilize sIOD (server I/O die), implementing PCIe Gen 6, which provides a generational doubling in bandwidth to GPUs, SSDs, and NICs. AMD is also claiming memory bandwidth of up to 1.6 TB/s for these CPUs, so we could see some of that trickling down into consumer configurations.
Source: Kepler_L2
Add your own comment

71 Comments on AMD "Zen 6" Processors to Use TSMC 2 nm Node for CCDs, 3 nm for IOD

#1
ymdhis
Does that mean the memory controller will stop sucking? Big if true. Will this open up the potential for getting new motherboard chipsets as well?

Here's hoping they'll also use the new interconnects from strix halo, which use a lot less power.
Posted on Reply
#2
dgianstefani
TPU Proofreader
ymdhisDoes that mean the memory controller will stop sucking? Big if true. Will this open up the potential for getting new motherboard chipsets as well?

Here's hoping they'll also use the new interconnects from strix halo, which use a lot less power.
MC is fine, the issue is it's on a separate chiplet on old style packaging, relying on IF to relay all data between IO/cores. This is an issue because IF hasn't really increased in speed since AM4, a Zen 3 part could easily run 4000 MT/2000 MHz memory in "1:1" with IF, so IF at 2000 MHz. This is essentially the same as AM5 with Zen 4/5, where IF is still at 2000 MHz, except now the memory is 6000 MT. If rumours are true, Z6 will be on next gen packaging with CCDs and IO die much, much closer together, with much lower latencies and bottlenecking. Similar issue to ARL with MC on separate die, and why the 200S boost works so well for gaming isn't the memory OC, but the die to die internal clock going up by about 1 GHz, alleviating the latency issue of having MC on separate die to cores.

If you compare the die shots/layout, chiplet/IO die haven't really changed in placement since Zen 2-5, so a closer to monolithic style arrangement with Z6 should be quite significant in implication, and about time.
Posted on Reply
#3
nowycham
Where is that info comming from ?
Posted on Reply
#4
Space Lynx
Astronaut
I always get confused on this stuff, so will I be able to use my same B650 Steel Legend motherboard with this chip? or will i need a new mobo? i probably will upgrade to this, late 2026 sounds like good timing as long as the main x3d version is under 550 bucks
Posted on Reply
#5
Geofrancis
Space LynxI always get confused on this stuff, so will I be able to use my same B650 Steel Legend motherboard with this chip? or will i need a new mobo? i probably will upgrade to this, late 2026 sounds like good timing as long as the main x3d version is under 550 bucks
if its still using DDR5 it will work on AM5.
Posted on Reply
#6
user556
dgianstefaniIf you compare the die shots/layout, chiplet/IO die haven't really changed in placement since Zen 1-5, ...
Zen 2 was first generation to use a separate I/O die.
Posted on Reply
#7
dgianstefani
TPU Proofreader
user556Zen 2 was first generation to use a separate I/O die.
True.
Posted on Reply
#8
Quicks
Meh, Zen 6 must come and go already.

Waiting for AM6.
Posted on Reply
#9
Hereticbar
QuicksMeh, Zen 6 must come and go already.

Waiting for AM6.
If the recent rumors suggesting that Zen 7 will be staying on AM5 socket turns out to be true, you will be waiting a long long time.
Posted on Reply
#10
Space Lynx
Astronaut
Geofrancisif its still using DDR5 it will work on AM5.
oh that is why i got confused, AM6 and Zen 6 just sound similar cause of the 6, heh nm I get it now
Posted on Reply
#11
FoulOnWhite
Sounds good but after so long on Intel i'm like a man in a barrel told to stand in the corner.
Posted on Reply
#12
Quicks
HereticbarIf the recent rumors suggesting that Zen 7 will be staying on AM5 socket turns out to be true, you will be waiting a long long time.
No worries I am not in a hurry. Seems AMD is also not in a hurry to make money then. 2029 / 2030 is perfectly fine. I am not a sucker for 10% -15% cpu gains anyways.
Posted on Reply
#13
Tek-Check
nowychamWhere is that info comming from ?
From Kepler, but it's almost the same to what Tom from MLID told us back in March.
We did not learn anything fundamentally new today that we did not already know in March from Tom.
Posted on Reply
#14
PrEzi
ymdhisDoes that mean the memory controller will stop sucking? Big if true. Will this open up the potential for getting new motherboard chipsets as well?

Here's hoping they'll also use the new interconnects from strix halo, which use a lot less power.
Why are you spreading FUD.
The MC has been fine for two-bank modules since years. With new CPU generation DDR5 6400 is quite normally reachable (with 1:1 ratio! Where intel sometimes needs 1:2 already). Yes AMD is more memory sensitive but it has been of a lesser impact with the 3D Cache versions lately.
Where they are still stuggling is a full 4 2-sided Module area, e.g. with my 4x32GB Modules I am reaching now 5600 with tuned timings and 1:1 ratio tops.... Some are reaching 6000 but it is a memory/mobo/cpu lottery atm.
This should improve with the new IO-Die. Of course 2-bank scenarios would also improve too, but they are not of a problem atm.
Posted on Reply
#15
AnotherReader
It's hard to believe that the IO die will move to TSMC's N3 process. N3 is fairly expensive and IO structures don't scale much with newer processes. Now, if they were aiming to use the higher logic density to make the IGP bigger or add a small NPU, that would make some sense, but again, even N5 would be sufficient for that.



The slide above shows the initially advertised improvements from N5 to N3, but in reality, SRAM density only increased by 5%. We don't know if analog structures improved by the stated figure, but given the miss for SRAM, it's likely that they didn't.
Posted on Reply
#16
BorisDG
Another perk of AMD. You can use same socket for few generations. Since I invested in high-end mobo, I'm ready to go for Zen 6 and may be 7?
Posted on Reply
#17
Darkholm
HereticbarIf the recent rumors suggesting that Zen 7 will be staying on AM5 socket turns out to be true, you will be waiting a long long time.
Yes but that does not mean that they will make it easy for 600-series or even maybe 800-series chipset motherboards. Let us presume that new MBOs will launch with Zen6. It is very possible to limit Zen6 to X870, X870E and new series and to dump 600 entirely and 800 lower end.

It still stings, that horsesh1t play from 2020 and 21 when ODMs and AMD tried so hard to explain how Zen3 is not compatible with 300-series chipset boards. But then, by pure magic, 19 months later, it was possible. Yay! BIOS chip has small capacity for new AGESA codes. Yea right.

Until the launch day, I will take this info with a truck of salt.
Posted on Reply
#18
Makaveli
dgianstefaniMC is fine, the issue is it's on a separate chiplet on old style packaging, relying on IF to relay all data between IO/cores. This is an issue because IF hasn't really increased in speed since AM4, a Zen 3 part could easily run 4000 MT/2000 MHz memory in "1:1" with IF, so IF at 2000 MHz. This is essentially the same as AM5 with Zen 4/5, where IF is still at 2000 MHz, except now the memory is 6000 MT. If rumours are true, Z6 will be on next gen packaging with CCDs and IO die much, much closer together, with much lower latencies and bottlenecking. Similar issue to ARL with MC on separate die, and why the 200S boost works so well for gaming isn't the memory OC, but the die to die internal clock going up by about 1 GHz, alleviating the latency issue of having MC on separate die to cores.

If you compare the die shots/layout, chiplet/IO die haven't really changed in placement since Zen 2-5, so a closer to monolithic style arrangement with Z6 should be quite significant in implication, and about time.
excellent post i'm looking forward to 1:1 with DDR5 8000 on Zen 6 :)
DarkholmYes but that does not mean that they will make it easy for 600-series or even maybe 800-series chipset motherboards. Let us presume that new MBOs will launch with Zen6. It is very possible to limit Zen6 to X870, X870E and new series and to dump 600 entirely and 800 lower end.

It still stings, that horsesh1t play from 2020 and 21 when ODMs and AMD tried so hard to explain how Zen3 is not compatible with 300-series chipset boards. But then, by pure magic, 19 months later, it was possible. Yay! BIOS chip has small capacity for new AGESA codes. Yea right.

Until the launch day, I will take this info with a truck of salt.
This is a possibly we shall see 670E and 870E are the same chipset with just USB 4.0 added. So would be a kind of a dick move lol.
Posted on Reply
#19
mkppo
If that IMC is anything like a desktop version of the Turin's current IMC, it's going to be lethal.

Makes me want to downgrade from my 4585PX and just hold off for Zen 6. I'm dying trying to stabilize 8400MT/s with this one, but 10,000MT/s should be a piece of cake for that IMC.

Bring it on.
Posted on Reply
#20
Geofrancis
Its my understanding that all AM5 boards have a much larger bios chip than what AM4 shipped with, so I doubt we will see the limited bios support that happened with AM4 where support for older chips was removed to support the later generations so I would expect every AM5 board (other than oems like dell/HP/lenovo) to support every AM5 chip after a bios update.
Posted on Reply
#21
Tek-Check
dgianstefaniMC is fine, the issue is it's on a separate chiplet on old style packaging, relying on IF to relay all data between IO/cores. This is an issue because IF hasn't really increased in speed since AM4, a Zen 3 part could easily run 4000 MT/2000 MHz memory in "1:1" with IF, so IF at 2000 MHz. This is essentially the same as AM5 with Zen 4/5, where IF is still at 2000 MHz, except now the memory is 6000 MT.
It's practically not an issue for better performance because IF data bandwidth substantially increased from Zen3 to Zen4/5, from GMI2 to GMI3.
dgianstefaniIf rumours are true, Z6 will be on next gen packaging with CCDs and IO die much, much closer together, with much lower latencies and bottlenecking. Similar issue to ARL with MC on separate die, and why the 200S boost works so well for gaming isn't the memory OC, but the die to die internal clock going up by about 1 GHz, alleviating the latency issue of having MC on separate die to cores.
Memory OC is literally the main reason why 6-7% uplift on average is possible with the Boost feature. When the Boost feature is enabled on the same speed memory kit, the "uplift" is on average ~1%, as tested by Tom's Hardware below. Some specific games had a bit more, but hey, it is what it is.

www.tomshardware.com/pc-components/cpus/we-tested-intels-unreleased-200s-boost-feature-7-percent-higher-gaming-performance-thanks-to-memory-overclocking-now-covered-by-the-warranty

Phoronix found 6% uplift in a comprehensive re-test in Linux 400 benchmarks back in April, before he tested the Boost feature a few days later. The uplift came mostly from new BIOS optimizations and polishing drivers.
www.phoronix.com/review/intel-arrow-lake-ubuntu-2504/10

When the Boost feature was introduced, he re-retested it again in ~70 workloads including games, and found, in his words, "the performance benefits of the 200S Boost mode were very muted"
www.phoronix.com/review/intel-200s-boost-linux/6

Let's not romanticize Intel Boost. Elephants don't fly.
QuicksNo worries I am not in a hurry. Seems AMD is also not in a hurry to make money then. 2029 / 2030 is perfectly fine. I am not a sucker for 10% -15% cpu gains anyways.
No worries. AMD is not in a hurry to make money from you. They will make money from tens of millions of other buyers.
AnotherReaderIt's hard to believe that the IO die will move to TSMC's N3 process. N3 is fairly expensive and IO structures don't scale much with newer processes. Now, if they were aiming to use the higher logic density to make the IGP bigger or add a small NPU, that would make some sense, but again, even N5 would be sufficient for that.
Don't forget that they use IOD for two generations at least. One IOD served Zen2/3, another one Zen4/5, and the new one will most likely serve Zen6/7, so they need to have ~4 years of IOD longevity. It looks like N3 is universal for all client IODs, including mobile. Another reason is power efficiency. Yet another reason is introduction of new Infinity Fabric that is projected to use fan-out, dense design.
DarkholmYes but that does not mean that they will make it easy for 600-series or even maybe 800-series chipset motherboards. Let us presume that new MBOs will launch with Zen6. It is very possible to limit Zen6 to X870, X870E and new series and to dump 600 entirely and 800 lower end.
That's not going to happen. There would be a huge backlash against AMD. They would lose the momentum and community trust. I would immediately switch back to Intel if that happens. The worst case scenario is that 900 series boards will simply have all new features, which is to be expected, such as faster memory, a few extra PCIe lanes, integrated USB4 or similar.
DarkholmIt still stings, that horsesh1t play from 2020 and 21 when ODMs and AMD tried so hard to explain how Zen3 is not compatible with 300-series chipset boards. But then, by pure magic, 19 months later, it was possible. Yay! BIOS chip has small capacity for new AGESA codes. Yea right.
Exactly. They must have learnt that lesson. Current BIOS chips are fine.
Posted on Reply
#22
_roman_
Is there a rumour tag in the topic?

Very small at the end
Source: Kepler_L2
Speculation: Somewhere and somehow there will be sometimes in the future another AM5 processor. It may work with current AM5 mainboards. Therefore you will not need to buy another motherboard. If the better process node will increase performance we will see. ....

Are there any facts in the topic?
Space LynxI always get confused on this stuff, so will I be able to use my same B650 Steel Legend motherboard with this chip? or will i need a new mobo? i probably will upgrade to this, late 2026 sounds like good timing as long as the main x3d version is under 550 bucks
There are no facts.

When a future processor has the same socket & you have firmware which supports your processor it will work.

The firmware is the uefi, sometimes named bios, update of your mainboard which also includes the agesa version for your processor and other stuff.

It's best to look out for the AGESA version.

two spreadsheets to keep around as a amd motherboard user

docs.google.com/spreadsheets/d/12zg6yT_H7H-W1voyw1ZoIrj0GSE7WI4Ug-uLlv-Asa8/edit?gid=858530609#gid=858530609

docs.google.com/spreadsheets/d/1NQHkDEcgDPm34Mns3C93K6SJoBnua-x9O-y_6hv8sPs/edit?gid=446857487#gid=446857487
Posted on Reply
#23
GenericUsername2001
AnotherReaderIt's hard to believe that the IO die will move to TSMC's N3 process. N3 is fairly expensive and IO structures don't scale much with newer processes. Now, if they were aiming to use the higher logic density to make the IGP bigger or add a small NPU, that would make some sense, but again, even N5 would be sufficient for that.



The slide above shows the initially advertised improvements from N5 to N3, but in reality, SRAM density only increased by 5%. We don't know if analog structures improved by the stated figure, but given the miss for SRAM, it's likely that they didn't.
There are rumors that the i/o & iGPU die will also be packing 4+4c Zen 6 cores, along with an 8 CU iGPU, and said chip by itself will be used as a laptop & low end desktop chip. Again, rumors, but at least then N3 would make sense if the same chip is replacing Krackan & Hawk Point chips in the market. Again, take that with a shaker of salt.
Posted on Reply
#24
_roman_
PrEziSome are reaching 6000 but it is a memory/mobo/cpu lottery atm.
My previous asus mainboard which I used for over 2 years was crap.

I changed to MSI mainboard with same CPU, DRAM, NVME, gnu linux, w11 pro. Just by testing another cpu tower cooler, another thermal paste and another mainboard I get much better DRAM clocks. Some mainboards are just crapware. I tested the same AGESA version on both the ASUS and MSI mainboard. MSI is far better when you are able to put in the numbers by hand. Stock values are worse for both mainboards.
Posted on Reply
#25
N/A
AnotherReaderIt's hard to believe that the IO die will move to TSMC's N3 process. N3 is fairly expensive and IO structures don't scale much with newer processes. Now, if they were aiming to use the higher logic density to make the IGP bigger or add a small NPU, that would make some sense, but again, even N5 would be sufficient for that.
If you look closely at the Lunar Lake N3 and Arrow lake N6 IO tile, you'll see that the memory controller is scaling down on the N3.
Posted on Reply
Add your own comment
Nov 19th, 2026 03:13 PST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts