• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)

Page 110 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
intel is dumping ht going forward. amd's design choice according to adrock thuston is that Zen will maintain 16 cores at the top end for now for zen 5 but zen 6 will include dense cores of the same type, these will have less cache but also have ht. Cores will increase with this zen 6 processor according to adrock a few hours ago.

Intel may top out at 24 or 32 cores and 24 or 32 threads. AMD may top out at 24 cores 48 threads and likely use less power and have less bugs than intel's novel approach.
Do you have any link to the source of all this information?

If source is only the forum user adrock_thuston, how do we know whether it's accurate? It's contradicting a lot of other source online. E.g. I've seen no source online saying Zen 5 will be 24C/48T for regular desktop. Instead all sources seem to say it maxes out at 16C.
 
Do you have any link to the source of all this information?

If source is only the forum user adrock_thuston, how do we know whether it's accurate? It's contradicting a lot of other source online. E.g. I've seen no source online saying Zen 5 will be 24C/48T for regular desktop. Instead all sources seem to say it maxes out at 16C.

It is just a conjecture, but based on the fact that AMD will have 8c, high performance chiplet and a 16c high density chiplet. The chiplets are interchangeable, so AMD can add one of each, potentially, for a specialized SKU.

Far easier to achieve than Intel having to tape out a new, monolithic 8+32 CPU chiplet - which you are assuming Intel will...
 
Here are some recent sources mentioning 8P + 32E for desktop Arrow Lake:



Are you for real right now?

Hardware times is a joke of a rag. They're worse than wccftech.

Digital trends mentions they have nothing but...

We don’t have any raw specifications for Arrow Lake processors as of yet, but there have been some tantalizing rumors that we can consider.

And videocardz bases their crap on whatever hasan's rag aka wccftech publishes.
 
It is just a conjecture, but based on the fact that AMD will have 8c, high performance chiplet and a 16c high density chiplet. The chiplets are interchangeable, so AMD can add one of each, potentially, for a specialized SKU.

Far easier to achieve than Intel having to tape out a new, monolithic 8+32 CPU chiplet - which you are assuming Intel will...
Rock never mentioned a 8+16 setup for zen either. You are the only one who keeps mentioning this.
 
It is just a conjecture, but based on the fact that AMD will have 8c, high performance chiplet and a 16c high density chiplet. The chiplets are interchangeable, so AMD can add one of each, potentially, for a specialized SKU.
Ok, so you're guessing there will be a Zen 5 desktop CPU with 1x8C Zen5 chiplet + 1x16C Zen5C chiplet, within same 170W TDP power budget as Zen4? Is there any source for that information?
Far easier to achieve than Intel having to tape out a new, monolithic 8+32 CPU chiplet - which you are assuming Intel will...
From what source do we know Arrow Lake desktop CPUs will be monolitic and not chiplets?
 
This is my opinion (so take it for what it's worth) but Intel's problem isn't packaging. Both MTL/ARL will have the best packaging tech available for client products. They're not really behind in packaging - at all.

The interposers will do the job, but the cost of MTL may prevent it from being able to compete on the low end of the market.

And there may also be a capacity issue for Intel, in how many Intel can package:

 
who wants to tell him the 8+32 he thinks is a fact was a guessed number spread out over 2 years ago when komachi initially found the raptorlake leaks on bapco? with it being a 8+16 part u nlike the 8+8 part of alder it was assumed each new generation would get a +8 e core addition. mtl-s was allegedly a 8+24 part. kind of like some thought nova would be an 8+48 setup.

while I wouldn't put it past intel to make such a stupid move even they're not that completely stupid to sell a fireball waiting to explode and put your flat on fire.
 
I was just quoting what JoeC said, i.e.:

"AMD will have 8c, high performance chiplet and a 16c high density chiplet"

But thanks for confirming it's inaccurate. I assume you're then implicitly saying it'll only be 16C for Zen 5 on desktop after all.

As in CPU tile is one hueg 8+32 ring with cores.
Either way why even care about stillborn products.

I never said that. Sources online just say 8P + 32E CPU cores for Arrow Lake desktop CPUs. Some mention that it'll use chiplets. But regardless they say 8P + 32E CPU cores. And if accurate, I have a hard time seeing how a 16C Zen5 CPU will be able to compete with that in MT workloads.
 
I never said that. Sources online just say 8P + 32E CPU cores for Arrow Lake desktop CPUs. Some mention that it'll use chiplets. But regardless they say 8P + 32E CPU cores. And if accurate, I have a hard time seeing how a 16C Zen5 CPU will be able to compete with that in MT workloads.
The sources you claim are sources are not sources but are referring to a rumor that was made up in 2021 when the original raptorlake es benches were found online on a databse showing an e core count increase. it was then presumed intel would add e cores each generation. that isn't the case.

that rumor then got spread by mlid and the other channel, the greasy fat boy.
 
Last edited:
I was just quoting what JoeC said, i.e.:

"AMD will have 8c, high performance chiplet and a 16c high density chiplet"

But thanks for confirming it's inaccurate. I assume you're then implicitly saying it'll only be 16C for Zen 5 on desktop after all.



I never said that. Sources online just say 8P + 32E CPU cores for Arrow Lake desktop CPUs. Some mention that it'll use chiplets. But regardless they say 8P + 32E CPU cores. And if accurate, I have a hard time seeing how a 16C Zen5 CPU will be able to compete with that in MT workloads.

8+32 seems a little farfeteched. A lot of people don't care much for e cores, and 16 seems plenty. Amdhal's Law. Of course if you are "Megatasking" to bring back an old phrase it could be useful.
 
8+32 seems a little farfeteched. A lot of people don't care much for e cores, and 16 seems plenty. Amdhal's Law. Of course if you are "Megatasking" to bring back an old phrase it could be useful.
would keep my house toasty in winter. beats splitting wood in late summer and waiting for it to dry out over the next couple of months.
 
8+32 seems a little farfeteched. A lot of people don't care much for e cores, and 16 seems plenty. Amdhal's Law. Of course if you are "Megatasking" to bring back an old phrase it could be useful.
There'll most likely be 8P +16/24E variants of Arrow Lake too. Not everyone needs to buy the top end.

But just wondering performance-wise, what you guys think 16C Zen5 CPU would be comparable to in MT workloads? Arrow Lake 8P + 12E, 8P + 16E, 8P + 24E, or 8P + 32E?
 
The interposers will do the job, but the cost of MTL may prevent it from being able to compete on the low end of the market.

Why would it be expensive? Packaging in raw materials is incredibly cheap. The TSMC silicon on MTL/ARL isn’t expensive either.
 
Back
Top