Question Zen 6 Speculation Thread

[MS_AT]

Why not?

Keep in mind that the cores in IOD will not share caches with CCDs. Every time you will put CCD to sleep you have to flush caches (you want to cut power to SRAM too as it always drains power). So when you wake the cores up you have to populate them again, etc. This costs both time and power.

In other words it becomes a balancing act as you have to decide when to handle over, you complicate scheduling and your user who bought 600$ CPU will get choppy experience when trying to run speedometer because you got the scheduling thresholds wrong. Only to save few watts that are drowned out by power innefficiences of the power supply. I mean sure, they might be able to run the workload good enough, but for good enough you could have bought n100

This is why I would dedicate LP cores only to predictable low interactive tasks, ideally background tasks. And ideally give the OS an API that would let software devs mark needed QoS level of the threads their program is using. It is also not to say they will be useless as they can handle background OS stuff giving big cores more time to serve the user.

This could work quite fine with this https://www.guru3d.com/story/windows-11-25h2-introduces-user-interactionaware-cpu-power-management/

But anyway this is more or less my reasoning why

 

[DrMrLordX]

I say their leak is garbage! My speculation is correct and "they can't prove it so they have no argument".

And you may well be correct. Or they may have misinterpreted data, as was the case with the Zen5 40%+ claim. They're not going to change their tunes in the face of counter-arguments.

It seems reasonable that AMD would prioritize a newer, expensive node for high-margin products. But, up until now, even their premium products have shared CCDs with consumer desktop (exceptions being Turin dense and Bergamo). Zen6 may change things a bit.

 

[ToTTenTranz]

If Medusa Halo has quad channel 256GB 10,000 MT/s RAM in addition to being attractively priced with PCIe 5.0 x16 slot, it will solve multiple problems at once:

1) Superb iGPU

2) Enough bandwidth for LLMs and AVX-512 workloads

3) No buying DIMMs and hoping they will hit the needed speed.

4) Less heat generation and lower noise and compact chassis

I will pay $2000 for it today if AMD wants a beta tester

And if things keep going like they are today, you'll find Medusa Halo in:

- One fugly tablet from Asus
- One laptop from HP that gets buried within hundreds of Intel+Nvidia SKUs
- A dozen oversized NUCs.



Yay.

 

[inquiss]

Not crippled by memory bandwidth. See previous post by 511.

And again, why pay more for TR if you can get the same MT performance much cheaper from NVL 52C on regular DT?

TR will only make sense if you need 64+ cores.

Because you want serious performance if you go TR. It's not more complicated with that. I'll say it again, the number of people who want serious multicore performance (and nova will not be serious because some of the cores are not big cores) and who are not willing to pay for it is tiny. I'm not invalidating your user case or your desire for it, but it's not a product with a big market. It's tiny. You are on a tech forum, and even then the number of people that want that here is small. You are not the average person and you are not a big market. DIY sales near that out. No one buys the 16 code as it is.

 

[511]

Because you want serious performance if you go TR. It's not more complicated with that. I'll say it again, the number of people who want serious multicore performance (and nova will not be serious because some of the cores are not big cores) and who are not willing to pay for it is tiny. I'm not invalidating your user case or your desire for it, but it's not a product with a big market. It's tiny. You are on a tech forum, and even then the number of people that want that here is small. You are not the average person and you are not a big market. DIY sales near that out. No one buys the 16 code as it is.

Bruh by Nova Lake there won't be a tangible difference between P and E cores in IPC the only thing will be clocks

 

[inquiss]

Bruh by Nova Lake there won't be a tangible difference between P and E cores in IPC the only thing will be clocks

Bruh, cool. So some cores will be better than others? Gotcha. So if you're serious about multicore performance it's one you would skip. For the tiny portion of hobbyists that want to tinker, fill your boots.

 

[eek2121]

Well, are we really expecting 24C Zen6 to beat 52C Intel Nova Lake-S in MT perf?

Yes.

AMD has SMT, and that will include the rumored LP cores.

 

[511]

Yes.

AMD has SMT, and that will include the rumored LP cores.

Nice you should check your calculations 🤣🤣

 

[maddie]

Nice you should check your calculations 🤣🤣

Did you really miss what he meant?

 

[Josh128]

If you need 52C Nova Lake-S level of MT perf, why would you get corresponding AMD TR if it’s more expensive?

You dont get the corresponding TR, you get one that stomps it. No one can convince me that if you NEED 52 cores MT perf, you dont NEED 128 or 192 cores MT perf all the more. Its just dumb. Nobody needs 52 cores in a desktop PC.

 

[StefanR5R]

Because the only applications where that many cores would matter are used by professionals. Professionals won't bat an eye at the higher price of TR as it pays for itself many times over in the increased productivity.

The term "applications" encompasses programs and datasets; let's not forget the latter.

If they can get the same MT perf cheaper from 52C Nova Lake-S than from TR, why would they go with TR? If the performance is the same, they won’t get any increased productivity for going with TR.

They can't get the same performance. If, for instance, your finite element model has so many degrees of freedom that you have good use for fifty-two computing threads, then you also tend to have use for eight DDR5 DRAM channels. (And for memory error detection and correction too.)

— edit —

Zen 4 TR would have PCI-E lane advantage and memory size advantage the bandwidth advantage should be less when we compare Quad Channel DDR5-5200 vs Dual channel 8000 JEDEC and if Intel's IMC is to go by 10K shouldn't be difficult on NVL.

DDR5-10000? Are you only interested in when you get a computation result, or also interested in the result being the correct one?

For PCI-E lanes and memory size advantage it depends on the use case whether it’s of any importance. For a lot of use cases it’s not, and what you get with regular DT is good enough.

If the use case is not data intensive, then a four-core CPU is most often a perfect fit for the use case.

 

[maddie]

You dont get the corresponding TR, you get one that stomps it. No one can convince me that if you NEED 52 cores MT perf, you dont NEED 128 or 192 cores MT perf all the more. Its just dumb. Nobody needs 52 cores in a desktop PC.

Used by a capital challenged person bootstrapping a business, who will then upgrade to TR ASAP.

 

[511]

The term "applications" encompasses programs and datasets; let's not forget the latter.

They can't get the same performance. If, for instance, your finite element model has so many degrees of freedom that you have good use for fifty-two computing threads, then you also tend to have use for eight DDR5 DRAM channels. (And for memory error detection and correction too.)

Here is a typical Workstation at EPIC you will be able to replace it with 52C Nova Lake easily while costing less

https://dev.epicgames.com/documentation/en-us/unreal-engine/hardware-and-software-specifications-for-unreal-engine

 

[MS_AT]

Here is a typical Workstation at EPIC you will be able to replace it with 52C Nova Lake easily while costing less

https://dev.epicgames.com/documentation/en-us/unreal-engine/hardware-and-software-specifications-for-unreal-engine

View attachment 125085

I would not be so sure about easy, I switched in April from Tiger Lake to Raptor Lake HX at work, and I spent 3 days to debug why my compile times are worse than on Tiger Lake if I got 12 twelve extra e-cores...

But cinebench I guess will show matching numbers and it is all that matters (at least in some reviews)

 

[511]

I would not be so sure about easy, I switched in April from Tiger Lake to Raptor Lake HX at work, and I spent 3 days to debug why my compile times are worse than on Tiger Lake if I got 12 twelve extra e-cores...

What really? I jumped from a 8250u to a 185H and the performance difference for me is pretty insane for blender or when I use FFmpeg is it possible that windows was assigning things wrong way?

But cinebench I guess will show matching numbers and it is all that matters (at least in some reviews)

🤣🤣

 

[branch_suggestion]

Here is a typical Workstation at EPIC you will be able to replace it with 52C Nova Lake easily while costing less

https://dev.epicgames.com/documentation/en-us/unreal-engine/hardware-and-software-specifications-for-unreal-engine

View attachment 125085

Memory bandwidth is a thing that exists.

 

[StefanR5R]

I jumped from a 8250u to a 185H and the performance difference for me is pretty insane for blender or when I use FFmpeg

So, image rendering and video encoding are your focuses.
You mentioned a Zen 4-generation 64c/128t Threadripper. I can't find figures for the PRO version quickly, but here are ones for the non-PRO:
image rendering, video encoding
But Nova Lake-S's competitors in image rendering (if on CPUs) and video encoding (if on CPUs) will be Medusa desktop (that is, Zen 6-generation Ryzen) and Zen 5-generation Threadripper.

Furthermore, about all this talk about "52 core Nova Lake": a) Nova Lake-S may have the four LPE cores disabled, b) even if they were going to have them enabled, they do not provide throughput. It's better to think of the rumored top-end Nova Lake-S as a 48c/48t CPU. (Not that it matters much whether 48 or 52 hardware threads are crammed into the small desktop socket.)

 

[511]

So, image rendering and video encoding are your focuses.

along with code compilation and video games.

You mentioned a Zen 4-generation 64c/128t Threadripper. I can't find figures for the PRO version quickly, but here are ones for the non-PRO:
image rendering, video encoding
But Nova Lake-S's competitors in image rendering (if on CPUs) and video encoding (if on CPUs) will be Medusa desktop (that is, Zen 6-generation Ryzen) and Zen 5-generation Threadripper.

Yeah that is true

Furthermore, about all this talk about "52 core Nova Lake": a) Nova Lake-S may have the four LPE cores disabled, b) even if they were going to have them enabled, they do not provide throughput. It's better to think of the rumored top-end Nova Lake-S as a 48c/48t CPU. (Not that it matters much whether 48 or 52 hardware threads are crammed into the small desktop socket.)

i agree that the LP-E will not provide much throughput either for AMD Z6 or Intel NVL but they need to up the TDP to 300W and remove the horrible DLVR in desktop to actually reach full potential.

On a side note what is the beefiest AM5 Motherboard rn?

 

[igor_kavinski]

This could work quite fine with this https://www.guru3d.com/story/windows-11-25h2-introduces-user-interactionaware-cpu-power-management/

I'm assuming that they will just stop most of their stupid OS processes from churning away when no user interaction is done for a specified time period. That they are introducing this feature in 2025 is bonkers considering that it should be one of the most common sense features to have in an OS and should've existed already for decades.

 

[511]

I'm assuming that they will just stop most of their stupid OS processes from churning away when no user interaction is done for a specified time period. That they are introducing this feature in 2025 is bonkers considering that it should be one of the most common sense features to have in an OS and should've existed already for decades.

Telemetry would be on Microsoft had to know that you are not doing anything Nadella needs to know that you are not using your PC. 🙂

 

[Markfw]

He was talking about Intel, as was I in my response 🙂

along with code compilation and video games.

Yeah that is true

i agree that the LP-E will not provide much throughput either for AMD Z6 or Intel NVL but they need to up the TDP to 300W and remove the horrible DLVR in desktop to actually reach full potential.

On a side note what is the beefiest AM5 Motherboard rn?

@Det0x What is your choice ?
@511 Since he is our current highest overclocker, I would think his opinion would be very close, if not correct.

 

[igor_kavinski]

Nobody needs 52 cores in a desktop PC.

I wish I had a dollar for every time someone said we don't need this much of something.

Examples:

HDD and SSD storage
CPU GHz
CPU cores
RAM (We are already at 128GB with Strix Halo and some people, including me, want more!)
GPU VRAM

Once you have experienced something, you always want more of it and because the companies want to sell thousands or millions of units of something new and more exciting than what's already on the market. I guess Intel feels threatened by 48 Zen 6 threads so they were like, hey, let's go for 52 cores instead of just threads coz they have admitted defeat that they can't do HT better than AMD's SMT.

 

[gdansk]

Here is a typical Workstation at EPIC you will be able to replace it with 52C Nova Lake easily while costing less

https://dev.epicgames.com/documentation/en-us/unreal-engine/hardware-and-software-specifications-for-unreal-engine

View attachment 125085

Yeah because you're going to get 256GB of ECC on cheap NVL-S platforms? Go look up W880 prices. Any workload that wants 52 cores also wants ECC and almost assuredly lots of it.

If you're giving up ECC RDIMM then you can also afford to give up some cores too.

 

[511]

Yeah because you're going to get 256GB of ECC on cheap NVL-S platforms? Go look up W880 prices. Any workload that wants 52 cores also wants ECC and almost assuredly lots of it.

have you seen the cost of a 64 Zen 4 thread ripper it is $4400 even accounting for $1200 next gen motherboard and $1000 NVL-S CPU I would have $2200 to spare for the rest of the build.
7980WX Amazon

Here is the W880 MB

Pardon Our Interruption...

X14SAE | Motherboards | Products | Supermicro

@Det0x What is your choice ?
@511 Since he is our current highest overclocker, I would think his opinion would be very close, if not correct.

I am a bit confused here 😅.

 

[gdansk]

have you seen the cost of a 64 Zen 4 thread ripper it is $4400 even accounting for $1200 next gen motherboard and $1000 NVL-S CPU I would have $2200 to spare for the rest of the build.
7980WX Amazon

Here is the W880 MB

Pardon Our Interruption...

X14SAE | Motherboards | Products | Supermicro

And it'll be worse at everything a many-core CPU is supposed to be.
If Intel ships 16+32+4 NVL-S it'll be the most idiotic niche part unless it has phenomenal 1T performance and they make it the only good 1T part in their line up to make people pay for more than they want.

24 Zen 6 cores is enough to compete. It doesn't need to match it in CB nT, only beat it in 1T. Much like the 7700K didn't need to match the 1800X in nT for it to mop the floor with it in sales. Making a product for the one or two lunatics using a consumer CPU to transcode is an interesting business decision if real.