AMD Zen 6 CPU Specifications Leaks – Big Boost Unveiled

[erek]

"IPC and clock speed gains​

According to Red Gaming Tech, AMD’s Zen 6 CPUs should feature clock speeds of around 6.3-6.4 GHz. This is much higher than what AMD’s Zen 5 CPUs deliver. In terms of clock speed, AMD’s fastest Zen 5 CPU, the Ryzen 9 9950X, has a maximum boost clock of 5.7 GHz. This suggests that AMD’s Zen 6 clock speeds are at least 10% higher than AMD’s Zen 5 products. That’s a significant clock speed gain. Some earlier Zen 6 rumours claimed 7GHz speeds were possible, though this new leak suggests AMD will not hit that alleged performance target. Regardless, Zen 6 is expected to run much faster than Zen 5.

Additionally, AMD’s Zen 6 CPUs will reportedly feature 10-15% IPC (Instructions Per Cycle) gains. However, it is worth remembering that IPC gains vary significantly across workloads. After all, some software will benefit more from some architectural improvements than others. Red Gaming Tech says that some tasks will have lower IPC gains (around 8%), though this is common for new CPUs.

With more cores, significantly higher clock speeds, and a notable IPC boost, AMD’s Zen 6 architecture is set to impress. It is currently unknown when AMD’s Zen 6 CPU architecture will launch. Many currently expect AMD’s Zen 6 CPUs to arrive in early 2027, though they could still launch later this year."


View: https://youtu.be/G4CRO0fW69w?si=6ctaoX8WY5RWWFzH

Source: https://overclock3d.net/news/cpu_mainboard/amd-zen-6-cpu-specifications-leaks-big-boost-unveiled/

 

[DogsofJune]

Waiting on reviews before getting excited. Leaks always sound amazing.

 

[Gigantopithecus]

Waiting on reviews before getting excited. Leaks always sound amazing.

Can't wait for a $900 10800X3D!

 

[yinzerniner]

Can already see what clock speed improvements mean IRL, 9850x3d va 9800x3s. 7.7% clock speed is equivalent to ~2% gaming and 5% productivity. 15-20% IPC is OK, hopefully that comes with lesser power on a smaller TSMC node. 20% real life performance gains would be nice if they keep pricing increases in line.

 

[Ikasu]

Can't wait for a $900 10800X3D!

Hope that's not the case....But a possibility. Hate to say it though, I'll be picking one up even at that price ....My 3D apps can use every core I throw at it.

 

[Tsumi]

Definitely would take those clock speeds with a huge grain of salt, especially on the X3D models.

 

[zandor]

If there's a $900 part hopefully it'll be a Ryzen 9 X3D with 24 cores. That'd be easier to swallow.

This next round of procs is looking to be interesting. AMD is increasing core count & clocks, and doing a much needed overhaul on their interconnects between the CCDs and the I/O die. The interconnect improvement might be a bigger deal than anything else. Then Intel is going to 18A, planning "large last level cache" models to compete with X3D, and planning Core Ultra 9s with 2 compute dies and up to 52 cores (16P+32E+4LPE). I'd bet on higher clocks too. And of course both of them will be working on IPC.

edit: typo. X3D, not X2D

 

[Ikasu]

Makes me wonder if the 7ghz clock leaks were a stretch too far for them...or if it will still happen. Time will tell, although I won't really believe anyone except MLID on this matter, so maybe we'll get closer to 7ghz. Has always had reputable leaks there compared to the random fud spread by other youtubers.

 

[Sycraft]

This next round of procs is looking to be interesting. AMD is increasing core count & clocks, and doing a much needed overhaul on their interconnects between the CCDs and the I/O die.

That is the part that really interests me. I'd like an X3D processor since they are the gaming kings. Realistically I could get a 9800X3D and be happy... BUT I do use my system for things that make use of more cores. Nuendo and Vegas in particular. No it isn't the hugest deal, I'd probably be fine with an 8-core system, but my 13900k can crank through shit faster than a 9800X3D can. Ok well a 9950X3D would be at least as fast if not faster still... but it has issues because of the interconnects. None of them deal breakers, but from my research it has more issues with low latency realtime audio than Intel CPUs.

So I stay where I am for now. But if they get their interconnects up to snuff where there isn't such an issue with inter-CCD communication, it could well be that a Zen 6 is the best CPU both for gaming AND audio/video shit.

 

[philb2]

So what explains AMD's apparent engineering prowess. Seems better than Intel. That April Fool's joke about AMD buying Intel. Maybe there is a kernel of truth there. Who knows. Certainly not me!

 

[erek]

Who knows. Certainly not me!

Hmm. I think you know more than you're letting on

 

[Tsumi]

That is the part that really interests me. I'd like an X3D processor since they are the gaming kings. Realistically I could get a 9800X3D and be happy... BUT I do use my system for things that make use of more cores. Nuendo and Vegas in particular. No it isn't the hugest deal, I'd probably be fine with an 8-core system, but my 13900k can crank through shit faster than a 9800X3D can. Ok well a 9950X3D would be at least as fast if not faster still... but it has issues because of the interconnects. None of them deal breakers, but from my research it has more issues with low latency realtime audio than Intel CPUs.

So I stay where I am for now. But if they get their interconnects up to snuff where there isn't such an issue with inter-CCD communication, it could well be that a Zen 6 is the best CPU both for gaming AND audio/video shit.

A much quicker interconnect would be a huge deal for sure. Not only would it help with stuttering, but it would also mean that the entire X3D cache could theoretically be used to feed one CCD. Maybe some programming magic will be needed to prioritize what data goes on which CCD, but if it prevents fewer fetches out to RAM that could be huge.

 

[erek]

"AMD Details Upcoming Zen 6 PQOS Extensions: Advanced Bandwidth and Privilege Controls​

by btarunr Thursday, 13:15 Discuss (12 Comments)
Imagine you're a web hosting vendor leasing out specific number of CPU cores of a large core-count processor. You'd want to specify QoS limits on the shared L3 cache performance for those cores, so they don't hamper performance of other tenants. AMD this week released a technical document detailing the Platform Quality of Service (PQOS) ISA extensions for its next-generation Zen 6 microarchitecture. These ISA enhancements provide sysadmins and cloud providers with greater control over processor and memory subsystem performance. The latest document outlines three primary additions to the Zen 6 PQOS feature set, Global Bandwidth Enforcement (GLBE), Global Slow Bandwidth Enforcement (GLSBE), and Privilege-Level Zero Association (PLZA). These features are designed to scale performance management across complex multicore environments by allowing software to regulate bandwidth and execution privileges more effectively across expansive groups of logical processors. The development shows that AMD is steering toward a more closely collaborative hardware QoS solution for its multicore processors.

A highlight of the Zen 6 PQOS updates is the implementation of Global Bandwidth Enforcement (GLBE), which allows system software to specify L3 external bandwidth limits for groups of cores that span across multiple traditional QoS Domains. By grouping these into a unified "GLBE Control Domain," AMD enables a competitively shared bandwidth ceiling for specific Classes of Service (CoS). This upgrades older architectures that only provided L3 external bandwidth enforcement on a strictly per-domain granularity. Next up, AMD introduced Global Slow Bandwidth Enforcement (GLSBE), a parallel feature that applies the exact same multi-domain bandwidth limiting principles to system memory explicitly designated as "Slow Memory." Both GLBE and GLSBE provide granular controls via specific model-specific registers.

Rounding out the Zen 6 microarchitecture updates is Privilege-Level Zero Association (PLZA), a unique performance and security monitoring feature that grants hardware the ability to automatically associate highly privileged execution with a specific CoS or Resource Monitoring Identifier (RMID). Historically, AMD PQOS mechanisms tied these identifiers to each logical processor on a standard per-thread basis. PLZA, however, empowers the system to actively override this thread-level association whenever a logical processor executes at Privilege Level Zero (CPL=0). This ensures that system software, such as the host OS kernel or a hypervisor running with AMD SVM enabled can utilize tailored resource limits regardless of the thread's standard user-level configuration."

 

[DukenukemX]

What really matters are Geekbench results and if AMD has adopted a similar Binary Optimization Tool like Intel has. I'm kidding of course.

 

[Marees]

So what explains AMD's apparent engineering prowess. Seems better than Intel. That April Fool's joke about AMD buying Intel. Maybe there is a kernel of truth there. Who knows. Certainly not me!

TSMC, is my best guess

also AMD usually designs for Data Centres first & later scales down that same design for client (unlike Intel which does multiple specialized designs)

I would hold on the zen 6 claims — as usually all AMD designs are server first. client benefits are usually a side effect of concentrating on server

 

[philb2]

Hmm. I think you know more than you're letting on

No I know nothing. Apologies to Sgt Schultz.
View: https://www.youtube.com/watch?v=34ag4nkSh7Q

My background is in software. I read this forum mostly to learn about new hardware developments.

 

[Lakados]

What really matters are Geekbench results and if AMD has adopted a similar Binary Optimization Tool like Intel has. I'm kidding of course.

I will happily shit on Intel for the iBOT implementation as it currently stands, but it is a big deal, if they can get support for it more universal it could be the biggest thing since Hyper Threading as it could feasibly replace it in a much more efficient manner.

I would love to see iBOT eventually running as a hardware VM in-front of the CPU with it natively and automatically doing its thing for every command dispatched its way regardless of developer implementation.

Of course developers should still maintain best practices and blah blah blah, but with CPU architectures rapidly evolving and OS’s being rather slow to adopt the optimizations for that hardware, (looking at you Microsoft….) and virtually no developer incentive to update older software to make use of those features, it would be absolutely HUGE if Intel could pull that off.

 

[WhoBeDaPlaya]

Won't hold my breath after Zen 5% (and the XBox Game Bar crap), especially with the ongoing Eyes Wide Shut AI wankfest
Don't bother mentioning X3D, since the physical implementation for Zen4 was gimped and thus the gains look more impressive

 

[Tengis]

I use 3d modeling software, work with 3d scans, gaming, Photoshop, Illustrator, I do some 3d rendering of the stuff I model... and my Ryzen 7 5800x chugs along great. Ive even dabbled with local ai models that tax my CPU pretty heavily. When I read reviews I feel like I have zero reason to spend any money to upgrade. Back in the day, upgrading your CPU and/or GPU every generation was borderline mind blowing and now I sit stagnant for years.

I feel like these constant upgrades people do are basically Gucci purses for dudes that dont have anything else to spend money on

 

[TangledThornz]

Difficult to get excited for new CPUs when we can't find or afford GPUs and RAM to go with it all

 

[Randall Stephens]

Leaks always sound amazing.

Depends

 

[Gigantopithecus]

Depends

Nicely done!

 

[DukenukemX]

I will happily shit on Intel for the iBOT implementation as it currently stands, but it is a big deal, if they can get support for it more universal it could be the biggest thing since Hyper Threading as it could feasibly replace it in a much more efficient manner.

I would love to see iBOT eventually running as a hardware VM in-front of the CPU with it natively and automatically doing its thing for every command dispatched its way regardless of developer implementation.

Of course developers should still maintain best practices and blah blah blah, but with CPU architectures rapidly evolving and OS’s being rather slow to adopt the optimizations for that hardware, (looking at you Microsoft….) and virtually no developer incentive to update older software to make use of those features, it would be absolutely HUGE if Intel could pull that off.

Until Intel's BOT works universally, I just don't believe it's nothing more than a method to boost Geekbench scores. The people behind Geekbench don't like it and will flag results as invalid. BOT doesn't even work in Geekbench 6.7 as it only works with Geekbench 6.3.

I don't think AMD will do something like BOT, but it goes to show how much influence these synthetic benchmarks have over the mind share of consumers. I'd wish the tech community would dump this awful tool as it doesn't represent real world performance, but 99% of crap YouTube tech reviewers use this and Cinebench. Don't be surprised that Zen 6's Geekbench results are barely any better than Zen5.

 

[Lakados]

Until Intel's BOT works universally, I just don't believe it's nothing more than a method to boost Geekbench scores. The people behind Geekbench don't like it and will flag results as invalid. BOT doesn't even work in Geekbench 6.7 as it only works with Geekbench 6.3.

I don't think AMD will do something like BOT, but it goes to show how much influence these synthetic benchmarks have over the mind share of consumers. I'd wish the tech community would dump this awful tool as it doesn't represent real world performance, but 99% of crap YouTube tech reviewers use this and Cinebench. Don't be surprised that Zen 6's Geekbench results are barely any better than Zen5.

AMD does (sort of), back in 2024 they started contributing to the LLVM Bolt project which is the open source version.
Intel has been working with LLVM Bolt longer, and it is platform agnostic.

iBOT is just the Intel proprietary version of Bolt with a scaled down scope so they could focus it and get it out faster, same idea though.

https://www.phoronix.com/news/Intel-Clear-BOLT-Like-Optimize
https://github.com/llvm/llvm-project/tree/main/bolt

 

[LukeTbk]

I don't think AMD will do something like BOT, but it goes to show how much influence these synthetic benchmarks have over the mind share of consumers.

Geekbench is almost 100% application based workload, reading PDF, sqlites databases tasks and browsing the web like geekbench is not what people tend to mean by synthetic benchmark. There is a bit of synthetic benchmark in it that test memory latency but not a big part of it. Geekbench is quite a good representation of common days to days task, not sure what the better single number one would come up with.

I'd wish the tech community would dump this awful tool as it doesn't represent real world performance,

How so ?

Until Intel's BOT works universally, I just don't believe it's nothing more than a method to boost Geekbench scores.

do not forget boosting game scores...

after Zen 5

zen4 to 5 did not had a significant TSMC node update (N4P to N4X or something like that), zen6 will a ton of "free" upgrade just by jumping from TSMC 4 to TSMC 2 (and TSMC 3N on the io), plus new TSMC packaging for the latency in between them, even if AMD does nothing more than achieve to tape out on those newer node it should be a good generation leap.

 

[Sycraft]

do not forget boosting game scores...

It is funny how some of the very same people who will complain about game developers not optimizing their stuff will turn around and be mad when Intel introduces something to optimize code. At this point it isn't super exciting as it only works on a select number of games, but the idea is one I am totally for. If there's a way to have a took that can swap in more efficient instructions when the code itself isn't complied with them, I'm all for it.

That was the original idea behind LLVM. While they like to pretend that "Nah bro, it's just a complier and toolchain alternative to GCC, that's all it ever was!" it used to mean "Low Level Virtual Machine" and the idea was looking at recompilation of code on a system to increase speed. It didn't end up working out that way, needless to say, but that's what the original research was about. While it didn't work out, that doesn't mean the idea is one that isn't worth continuing on. The idea of being able to look at the binary of something and then reoptimize it for the system it is on is a compelling one. It may never actually work at scale, but it is a concept worth trying.

 

[Lakados]

It is funny how some of the very same people who will complain about game developers not optimizing their stuff will turn around and be mad when Intel introduces something to optimize code. At this point it isn't super exciting as it only works on a select number of games, but the idea is one I am totally for. If there's a way to have a took that can swap in more efficient instructions when the code itself isn't complied with them, I'm all for it.

That was the original idea behind LLVM. While they like to pretend that "Nah bro, it's just a complier and toolchain alternative to GCC, that's all it ever was!" it used to mean "Low Level Virtual Machine" and the idea was looking at recompilation of code on a system to increase speed. It didn't end up working out that way, needless to say, but that's what the original research was about. While it didn't work out, that doesn't mean the idea is one that isn't worth continuing on. The idea of being able to look at the binary of something and then reoptimize it for the system it is on is a compelling one. It may never actually work at scale, but it is a concept worth trying.

At this point there are so many instruction sets, with so many architectures, in so many possible configurations that properly optimizing code for a CPU is just not feasible on an industrial level, unless you are building for a specific hardware architecture.

Game developers can certainly do their best but focusing on GPU and texture optimization covers more bases with more immediately visible results.

Developers need help optimizing for CPUs, and that need is why Meta and Intel originally started the LLVM Bolt project. They needed help optimizing code for the Data Center.

 

[DukenukemX]

How so ?

It doesn't represent real world applications. Specifically when it comes to Apple M-series chips vs AMD and Intel. Most synthetic tests give Apple a huge win, but with real world applications it's either a slight win or a slight loss. Geekbench just so happens to be the worst offender.

do not forget boosting game scores...
View attachment 795577

Most of those games are old with a few exceptions. Shadow of the Tomb Raider is 9 years old.

It is funny how some of the very same people who will complain about game developers not optimizing their stuff will turn around and be mad when Intel introduces something to optimize code. At this point it isn't super exciting as it only works on a select number of games, but the idea is one I am totally for. If there's a way to have a took that can swap in more efficient instructions when the code itself isn't complied with them, I'm all for it.

I believe Intel's BOT tool is just replacing the binary with one hand optimized. I'd like to be proven wrong because it would be cool if a tool just auto optimized code for applications, but I don't think this is the case. Also, the better solution would be to just release the source code and let others try and optimize the code.

AMD has been working hard in optimizing Linux for their CPU's, while Intel fired theirs. AMD even created their own Clear Linux, and calls it Rocky Linux. In the end, AMD will still be faster than Intel, even with their BOT tool.

That was the original idea behind LLVM. While they like to pretend that "Nah bro, it's just a complier and toolchain alternative to GCC, that's all it ever was!" it used to mean "Low Level Virtual Machine" and the idea was looking at recompilation of code on a system to increase speed. It didn't end up working out that way, needless to say, but that's what the original research was about. While it didn't work out, that doesn't mean the idea is one that isn't worth continuing on. The idea of being able to look at the binary of something and then reoptimize it for the system it is on is a compelling one. It may never actually work at scale, but it is a concept worth trying.

I imagine Intel would have made the attempt to hand optimize the applications that BOT is now doing, but likely no company would let them look at the source code. The best solution would be to let other companies like AMD and Intel to look at the code to see if something can be done. I believe AMD did this with Starfield.

 

[LukeTbk]

Most of those games are old with a few exceptions. Shadow of the Tomb Raider is 9 years old.

yes the older the better for this I would imagine, boosting game that were not really limited by those cpu anyways.

It doesn't represent real world applications

Does it not represent well the application it claim to represent, like how fast a browser or opening PDF are on those different cpu ? for which real life scenario tested in geekbench like compressing files, pdf rendering, webbrowsing, clang compiling, searching or tagging picture in a large photo library, HDR processing of your pictures, object detection/blurring background does geekbench cpu ranking score in those category differ significantly of when an user does it day to day ?

I believe Intel's BOT tool is just replacing the binary with one hand optimized.

That what Intel are saying, not sure it is really by hands too, but it is exe by exe after a profiler looking at the isa calls, find a way to re-order/change some of them that take advantage of more recent calls and cache size.

Also, the better solution would be to just release the source code and let others try and optimize the code.

It is not really the source code (Intel does not know the game source code and does not rewrite a new one to compile it with a more recent compiler), it is a different assembly binary code, old game often goes in limbo and become a bit of headache for game studio to do anything significant to them (remember the little crisis about having to make a 64 bits build to support physx), if they wanted their game engine to use the latest AVX and rebuilt their games not sure how much they would need Intel help. And part of the cache optimisation will be specific to those Intel gpu and starting to make multiple game binary for each using a different Intel compiler tag is too much.

AMD has been working hard in optimizing Linux for their CPU's, while Intel fired theirs.

Not sure what it has to do with any of this, yes division Intel sold years ago are stopping to get drivers update from them...

 

[DukenukemX]

yes the older the better for this I would imagine, boosting game that were not really limited by those cpu anyways.

Yea but the older the game then the higher the frame rate already is on newer hardware. Going back to Shadow of the Tomb Raider, it's already nearly 300 fps. This is also a game that's frequently benchmarked on Apple M-series chips, and does usually win on Apple hardware. It's also possible these games were chosen because they don't have Denuvo. I know Borderlands 3 had it removed in 2021, but not sure about the other games. I doubt Denuvo would be happy having binaries altered by Intel's BOT.

Does it not represent well the application it claim to represent, like how fast a browser or opening PDF are on those different cpu ? for which real life scenario tested in geekbench like compressing files, pdf rendering, webbrowsing, clang compiling, searching or tagging picture in a large photo library, HDR processing of your pictures, object detection/blurring background does geekbench cpu ranking score in those category differ significantly of when an user does it day to day ?

Again, no it doesn't. If you look at how much Apple's M-series chips win with Geekbench vs real world applications, it's giving a false positive. This also includes 3D Mark tests as Apple's been winning big in those. Much like Qualcomm's Snapdragon X chips, it does poorly in real world applications.

It is not really the source code (Intel does not know the game source code and does not rewrite a new one to compile it with a more recent compiler), it is a different assembly binary code, old game often goes in limbo and become a bit of headache for game studio to do anything significant to them (remember the little crisis about having to make a 64 bits build to support physx), if they wanted their game engine to use the latest AVX and rebuilt their games not sure how much they would need Intel help. And part of the cache optimisation will be specific to those Intel gpu and starting to make multiple game binary for each using a different Intel compiler tag is too much.

Then game studios should release source code like John Carmack has done and let others tweak it. It's clear there's a lot of performance left of the table.

Not sure what it has to do with any of this, yes division Intel sold years ago are stopping to get drivers update from them...

Embrace source code. Deny rebuilding binaries. You don't see performance issues on Linux because all of the source code is available, and as a result we get distros like CachyOS. x86 has a massive optimization issue because modern x86 have become very good at vector/SIMD. You know like AVX-512 and Intel's AMX. What Intel's BOT does is remake applications in favor of SIMD. Geekbench obviously doesn't vectorize parts of their tests because lazy or it doesn't favor Apple. Which is only made worse since most x86 applications use generic binaries made over a decade ago.

 

[LukeTbk]

Again, no it doesn't. If you look at how much Apple's M-series chips win with Geekbench vs real world applications,

which geekbench score category does not match the real world application ? That an easy claim to disprove, are the pdf opening and webbrowsing actually not matching their score ?>

Geekbench obviously doesn't vectorize parts of their tests because lazy or it doesn't favor Apple.

Again, geekbench for the most part use popualr real world library to do things as close as possible as regular people will do, they do not try to go fast or slow and not the one doing the relevant part of the code running, those library not using those vectorisation can for things like security, old device support and so on.

 

[Mackintire]

Zen 5 was developed by the same engineering team that designed Zen 2 and 2+. Compared to Zen , Zen 3 and 4 which was designed by a different team.


The team working on Zen 5 made some very controversial optimization and tooling choices when designing Zen 5. As such Zen 5 was a bit underwhelming in some scenarios. The good news is that the Zen 5 baggage is not being inherited into Zen 6's design.

As a result Zen 6 will punch above the avg generational uplift values

 

[LukeTbk]

The good news is that the Zen 5 baggage is not being inherited into Zen 6's design.

which one (the main zen4->5 difference like getting twice as wide seem to be kept....) ? as for AMD being big enough for that separate/different teams, that seem an exageration to me.

zen6 punch above its value will be a lot because it is rare for TSMC node and packaging/chiplet interconnect to change so much between 2 generation like that.

 

[philb2]

which one ? as for AMD being big enough for that separate/different teams, that seem an exageration to me.

zen6 punch above its value will be a lot because it is rare for TSMC node and packaging/chiplet interconnect to change so much between 2 generation like that.

^^ THIS is why I participate in these threads, not because I really know anything. Post #16 above.

 

[ChadD]

Not sure what it has to do with any of this, yes division Intel sold years ago are stopping to get drivers update from them...

Intel has let go almost everyone they had working on Linux support. But not everyone. They ended clear yes. But they still do acritecurial work. I mean not having Intel server CPUs work with server OSes isn't an option.

The gutting of those projects and people though is relevant. Intel consumer cpu optimizations are effected.

On the magic bit of Intel code for their new chips. Open source it or I don't trust you. My guess is it isn't doing anything people assume it's doing. It's probably just telling the thread director to ignore e cores when x or y software is detected... And or power limits on the p cores. I doubt highly it is on the fly optimizing shit. Just compiling simple software with link time optimization doubles compile time... But Intel with a few hundred lines of code is recombining for simd in real time? I call bs.

 

[LukeTbk]

But Intel with a few hundred lines of code is recombining for simd in real time? I call bs.

it detect it and use pre made code too, it is exe by exe specific (with a very limited amount of them done) from my understanding so the rack machine could have took all nights optimising it if they need too.

The giant jump of performance (over 30%) in specific stuff that would benefit from SIMD (HDR image processing do sound like something that would do the same math heavy operation to millions of different pixel as parrallel as you can go) and nill in otherse make it look "legit", instruction count reduction has been observed by reviewer I think.

Not that we should trust that it does not take security shortcut, lot of attack vector historically has been with multithreadding-performance optimisation (spectre, meltdown and what not), being slow on purpose so the attacker cannot deduce from the time an operation took to know anything about it is not uncommon, as long as it stay just exe per exe and on/off, not an issue, but would it become systematic that could become one.

Intel has let go almost everyone they had working on Linux support. But not everyone.

almost everyone ?
https://commandlinux.com/statistics...te Contributors,volume of second-place Google.

we are talking about the biggest and by far Linux corporate contributor in the world historically

 

[ChadD]

almost everyone ?
https://commandlinux.com/statistics/linux-kernel-contributors-lines-of-code-statistics/#:~:text=Linux Kernel Corporate Contributors,volume of second-place Google.

we are talking about the biggest and by far Linux corporate contributor in the world historically

Historically yes. Will be interesting to see the numbers for 2026 when we get there. I know they still have plenty of Linux people around. It is also true that they have let a ton go as well. Obviously not supporting Linux isn't an option. Maybe... I mean it is possible they mostly trimmed fat. Time will tell.

Those numbers are mostly pre cuts.

 

[Mackintire]

which one (the main zen4->5 difference like getting twice as wide seem to be kept....) ? as for AMD being big enough for that separate/different teams, that seem an exageration to me.

zen 6 will punch above its value will be a lot because it is rare for TSMC node and packaging/chiplet interconnect to change so much between 2 generation like that.

The expectation AMD set was +12% gen over gen

Zen 5 was closer to +8% with the singular exception of AVX 512 performance which was 215%

Zen 6 's improvements are going to land in the 15-18% gen over gen improvement range, while its single thread performance uplift will land at roughly 8%

Intel isn't sleeping at the wheel either, they will likely land in the 12-15% uplift range with Diamond Rapids. The magic question is how will Intel's Arrow Lake successor integrate those planned changes, I haven't a clue.