From our other sites
The story
The next-gen “PS6” console is expected to launch sometime between late 2027 and 2028, and new reports claim its CPU will move from the PS5’s AMD Zen 2 to the Zen 6 generation, boosting IPC (instructions processed per clock cycle) by up to 120%. The chip is said to be a semi-custom “Orion” design co-developed by AMD and Sony, built on TSMC’s 3nm process. On 5ch’s console-war board, users weighed in with skepticism — noting that the actual generation-to-generation IPC gains for Zen (about 19% for Zen 3, 13% for Zen 4, 16% for Zen 5) make it hard to see how four generations add up to 120% — alongside worries that the price will skyrocket.
New performance estimates have surfaced regarding the CPU in the next-generation “PS6” console, which is rumored to launch sometime between late 2027 and 2028. The PS6 is said to move from the AMD Zen 2 architecture used in the PS5 to the Zen 6 generation, which could bring a major boost to IPC (instructions executed per clock cycle). Current predictions suggest that CPU performance far exceeding the PS5’s could help maintain high frame rates even in games with complex physics simulations and large numbers of NPCs.
The PS6’s CPU is reportedly built around “Orion,” a semi-custom chip co-developed by AMD and Sony, combining TSMC’s 3nm process with the Zen 6 architecture.
Source: sumahodigest.com / Original article here
What people said
That was right when RDNA2 barely missed the cut, so Zen 3 (which launched around the same time as RDNA2) was probably a no-go too.
The Xbox Series X uses Zen 2 as well.
Doesn't matter how high-performance something is if nobody can afford it.
"Nobody can afford it" = wrong
"I can't afford it" = correct
Sorry to say, but the Switch 2's ARM chip is, broadly speaking, way more efficient per clock cycle than x86 CPUs.
If these specs are real, Steam games should run on it with room to spare. I really hope they go that route.
They probably don't want to bow down to Windows, so why not just ship it with Ubuntu instead 😁
Obviously — if people can't afford it, it won't sell. We're talking about a platform that lives or dies on sales volume here, you know?
Everyone would buy it even at a million yen. You're just the one who can't afford it.
I think a lot of people assume "surely it won't go above 100,000 yen," but even the much less powerful Steam Machine was 200,000 yen.
That's just a special bargain price for Japan. With the PS6, poor old Japan is probably getting cut loose.
The 2TB SSD and the 4nm chip make it expensive.
Since the Steam Machine is 200,000 yen, they'd probably want something like 170,000 yen.
The PS5's 6nm process is still relatively cheap.
Which means the PS6's 3nm chip will be even more expensive.
Maybe close to 300,000 yen just in production cost?
And that's a conservative estimate.
Say they sell it for 200,000 yen anyway —
to claw back a 100,000 yen loss through software royalties, you'd need to sell 300,000 yen worth of games just to net 90,000 yen.
In reality, they probably wouldn't even see that much in sales.
Storage is the weak point for pricing, so that's what they'll cut — 1TB at most, maybe even down to 500GB if it's bad.
If it crosses the $1,000 line, even the US market won't let that slide.
So they'll aim to keep it under $1,000, which means in Japan it'll probably come in under 200,000 yen.
Then just sell it for 300,000 yen.
We're talking about IPC here, right? Once you factor in the clock speed jump from shrinking 7nm down to 2nm, 120% doesn't even come close to the real gain.
IPC has stayed basically flat from the iPhone 12 to the iPhone 18, though.
A 120% IPC improvement is way too much of a stretch.
If they meant IPC improves 20% — reaching 120% of the original value — that might be realistic, though.
Agreed.
I think the correct reading isn't "a 120% improvement," but rather "a 20% improvement, bringing it to 120% [of the original]."
But even so, that's rough. The 8-core/16-thread setup is staying the same as the PS5, so compared to past PlayStation generations, the jump is way too small.
Going from Zen 2 to Zen 3 alone already gave up to a 19% boost.
https://www.4gamer.net/games/461/G046172/20201009109/
If you stack that all the way from Zen 2 to Zen 6, yeah, 120% starts to sound believable.
The PS6 reportedly packs two Zen 6 LP cores dedicated to handling the OS and background tasks, so unlike the PS5 — where 1-2 cores get tied up running the OS — there's no more interrupt lag eating into performance. That alone should be a big jump in real-world performance.
All 8 cores/16 threads become fully available for the game itself.
Zen 2 was an older architecture with IPC roughly on par with Skylake.
Zen 3 is where it caught up to Intel.
Everything since then has only seen margin-of-error-level gains.
If anything, AMD stopped chasing raw IPC and leaned hard into a server-optimized architecture instead — which is why their server market share keeps growing.
https://www.4gamer.net/games/612/G061273/20220925006/
Zen 4: 13% IPC improvement
https://ascii.jp/elem/000/004/210/4210323/
Zen 5: 16% IPC improvement
Various benchmarks went up proportionally too.
There was actually a console that was sold partly on being able to run Linux — it was called the PS3. Though Linux support ended up getting scrapped because it turned into a security hole.
Sony got sued on the grounds that if the PS2 and PS3 were going to use open-source Linux, then the PS3's own OS should also be made fully open-source and free. That's why they switched to a custom BSD and dropped Linux.
And that's exactly why it became impossible to use a PlayStation as a PC.
The Xbox's CPU and GPU were a newer generation and looked like they might win on raw performance, so Sony scrambled and cranked the PS5's clock speed way up to compensate. That's the whole reason it's being pushed so hard.
You just proved with your own data that the "120% improvement" claim is false.
Why are you dropping the word "up to"?
Even AMD's own marketing figure is already the "up to" maximum.
If the max per generation is only around a 10% improvement, four generations won't add up to a 120% improvement.
Measured in actual games, even a 50% improvement would be questionable.
For gaming especially, minor architectural tweaks don't move IPC much anymore — it's gotten to the point where you need brute-force moves like slapping on 3D cache to get real gains.
That's why the source I posted mixes average and max figures. There are cases where IPC jumps massively just from adding AVX-512 support, after all.
On console-war boards, Xbox is unconditionally ranked above everything else.
150,000 yen would already be a massive loss. No management team would sign off on eating a 1-2 trillion yen deficit like that.
Wake up from that dream already.
My guess is they'll scale back performance from the original plan and cut storage too. As long as they beef up the AI features, they can paper over the rest.
Background and Key Points of This Discussion
The “IPC” at the center of the PS6 CPU buzz is a measure separate from clock speed — it indicates how many instructions can be processed per clock cycle, and serves as a yardstick for real generational performance gains. Looking at AMD’s past Zen architectures, the published improvements were roughly 19% from Zen 2 to Zen 3, 13% for Zen 4, and 16% for Zen 5 — numbers that don’t easily stack up to 120% even when added together. That suggests the “up to 120%” figure is likely a maximum value for specific workloads, and actual in-game performance isn’t simply going to more than double. Alongside debate over how to interpret the number, the thread also showed real concern that the switch to TSMC’s 3nm process and larger storage capacities will drive up costs, pushing the console’s retail price well above the current PS5. IPC gains and retail pricing are technically separate issues, but in the thread they ended up tangled together.
※This article is excerpted and summarized from the 5ch (Hardware/Industry (Console Wars)) thread “[Good News] PS6’s Zen 6 Said to Boost IPC by Up to 120% Over PS5’s Zen 2“.
Leave a Reply