AMD's 3nm-based Zen 5 processors appear in leaked roadmap

midian182

Posts: 7,081   +62
Staff member
Rumor mill: We’ve heard plenty of rumors about Zen 4, but what about its successor? Despite being a couple of years away, at least, a leaked partial roadmap has revealed some possible details about AMD’s future CPUs, including their Ryzen 8000 branding and use of the 3nm manufacturing process.

The leak, which comes from Chinese social media site Weibo, claims that the follow-up to AMD’s Zen 4-based Ryzen 7000 (Raphael) mainstream desktop CPUs will be the Ryzen 8000 series, codenamed Granite Ridge. The same generation notebook APUs, codenamed Strix Point, will also use the Ryzen 8000 name.

It’s believed that Zen 5 will be based on the 3nm manufacturing process, presumably from TSMC (N3). As with Zen 4, Zen 5 is said to use the AM5 socket. The change means concerns over bending one of the CPU’s pins may be a thing of the past as AMD is reportedly switching from a pin grid array (PGA) design to the same land grid array (LGA) design favored by Intel for the majority of its chips.

The Strix Point CPUs are also said to be taking a leaf from Intel’s book. Much like Chipzilla’s upcoming Alder Lake processors, AMD’s APUs are said to use a hybrid big.LITTLE architecture that mixes Zen 4 and Zen 5 cores.

Unlike the 3nm-based Zen5, Zen 4 is expected to use the 5nm process, but that could change in Strix Point. The little cores in the APUs have the codename Zen4D and could be an enhanced version of the architecture, one based on the 3nm process. The chips are also rumored to use a new L4 cache system that will work as system-level cache.

It’s unclear whether AMD will adopt a hybrid design for its Granite Ridge desktop CPUs as well, though it seems unlikely. Zen 4 isn’t expected to arrive until next year, while Zen 5 is slated for 2023, so it’ll be a while before we know for sure.

Hopefully, the chip shortage the world is currently enduring will be long-forgotten by the time Zen 5 rolls around. AMD recently admitted that the current situation had forced it to prioritize the manufacture of its high-end processors over budget offerings.

Permalink to story.

 

yRaz

Posts: 3,816   +3,960
No pins, eh? There was always something satisfying about putting a cpu in the socket and feeling it drop when the pins lined up. LGA is better, though in respect of not bending pins...

I have heard that LGA sockets can't pump as much juice in a CPU as pins but I don't know how true that is. It's also not relevant to me as I don't really overlook anymore. I play around with it but after a few benchmarks I usually just leave it stock.
 

RedBlu

Posts: 44   +56
The change means concerns over bending one of the CPU’s pins may be a thing of the past

Has that ever been a real issue? If you're in a position that you need to use enough force to bend a pin then the layout probably won't matter.
 

Irata

Posts: 1,675   +2,812
Sounds interesting but not sure why I‘d get excited over something that‘ll be released in a couple of years.

Also, I‘m not sure if following Arm‘s approach with big and little cores is a good idea for desktop.

Then again, perhaps this is something different, mixing cores with a different focus / strengths (besides power consumption) instead. That could be interesting.
 

avioza

Posts: 242   +217
Has that ever been a real issue? If you're in a position that you need to use enough force to bend a pin then the layout probably won't matter.

I have received procs with bent pins before from resellers.

Hot tip: Mechanical pencil tip slipped over the pin gives good control for straightening. :D
 

yRaz

Posts: 3,816   +3,960
I have received procs with bent pins before from resellers.

Hot tip: Mechanical pencil tip slipped over the pin gives good control for straightening. :D
I've found that even a 0.5mm pencil can't straighten them. I've had much more success with sliding a credit card between the rows of pins to line them back up
 

wiyosaya

Posts: 6,393   +4,725
Out of all the AMD CPUs I've used over the years I never had any issues with pins. You line up the marked corner on the CPU with the socket and she drops right in. But, I suppose if it's one less thing to worry about then that's a good thing.
LGA's have simply moved the pins to the socket. As I see it, it does not really get rid of the issue.
 

Avro Arrow

Posts: 1,727   +2,046
TechSpot Elite
I have to assume that AMD is switching from PGA to LGA to allow for more pins. The pins on their PGA CPUs are considerably thicker and sturdier than the pins found in LGA sockets. I accidentally bent a corner pin once on one of my Phenom II CPUs but it wasn't a big issue to just bend it straight again with a butter knife.
 

Markoni35

Posts: 1,234   +507
It essentially moves the issue to a less accident prone area.

Or you could say, it moves the expenses from CPU to motherboard. So that CPU can become cheaper, and thus more attractive, hoping that user won't notice the whole system still costs the same.
 

yRaz

Posts: 3,816   +3,960
Or you could say, it moves the expenses from CPU to motherboard. So that CPU can become cheaper, and thus more attractive, hoping that user won't notice the whole system still costs the same.
Normally I wouldn't reply to garbage like this, but look into the manufacture of the shape of the "negative" pins in an LGA socket
 

Markoni35

Posts: 1,234   +507
Normally I wouldn't reply to garbage like this, but look into the manufacture of the shape of the "negative" pins in an LGA socket

Yeah, but you forget expenses related to bent pins. Since the CPU has no pins, they cannot bend. Now the mobo manufacturers have the problem, since their pins can bend. They have to deal with improper installation of CPU which can bend the pins and check whether it was their or customer's mistake. The CPU manufacturer wash their hands. Not their problem.
 

captaincranky

Posts: 17,211   +5,965
As long as we have smart watches that will tun "Grysis", before the sun burns out, it will have not died in vain.

This thread however, will die long before that happens, of acute banality. :facepalm::poop:

In the meantime, if you should need me, I'll be over at Quora, upvoting all the anti Trump threads, and turning all the requested answers to stupid computer questions in as spam
 

Theinsanegamer

Posts: 2,661   +4,141
Has that ever been a real issue? If you're in a position that you need to use enough force to bend a pin then the layout probably won't matter.
Yes, AMD CPUs have a nasty habit of getting stuck to heatsinks because their PGA socket has no retention bracket, and when they get pulled out, pins tend to bend.
No pins, eh? There was always something satisfying about putting a cpu in the socket and feeling it drop when the pins lined up. LGA is better, though in respect of not bending pins...

I have heard that LGA sockets can't pump as much juice in a CPU as pins but I don't know how true that is. It's also not relevant to me as I don't really overlook anymore. I play around with it but after a few benchmarks I usually just leave it stock.
I cant imagine power is an issue looking at what intel's 11900k pulls through a LGA socket.
Yeah, but you forget expenses related to bent pins. Since the CPU has no pins, they cannot bend. Now the mobo manufacturers have the problem, since their pins can bend. They have to deal with improper installation of CPU which can bend the pins and check whether it was their or customer's mistake. The CPU manufacturer wash their hands. Not their problem.
The most common cause of bent pins on a CPU is them getting pulled out of their socket on an angle during heatsink removal. LGA sockets have retention brackets and thus do not suffer from this issue. When is the last time you heard of an intel CPU pulling free from a motherboard?

They can also be bent during transportation in their box, not an issue with LGA CPUs. You can put protective covers on motherboards (as all mobo makers do) but doing so to a PGA CPU is much harder to do properly.
 

dnous

Posts: 39   +40
Has that ever been a real issue? If you're in a position that you need to use enough force to bend a pin then the layout probably won't matter.
Having worked at a retailer, having bent socket pins was a way bigger issue than bent CPU pins. A lot of consumers mess up placing (Intel) CPUs. I just hope AMD goes for a similar installation method as with the Threadripper CPUs.
 

hk2000

Posts: 165   +86
Yeah, when they stop using pins, I'll start considering AMD. Until then, not a chance. Then again, it's kinda too late isn't it? Since Intel went away from pins, I've built hundreds of systems using their CPUs as a matter of course, with no consideration whatsoever for AMD processors!
 
Last edited:

mattferg

Posts: 180   +173
No pins, eh? There was always something satisfying about putting a cpu in the socket and feeling it drop when the pins lined up. LGA is better, though in respect of not bending pins...

I have heard that LGA sockets can't pump as much juice in a CPU as pins but I don't know how true that is. It's also not relevant to me as I don't really overlook anymore. I play around with it but after a few benchmarks I usually just leave it stock.

The answer is really simple. Bent CPU pins are AMD's RMA problem, bent motherboard pins are the AIB's problem. Both are pretty damn fragile.
 

mattferg

Posts: 180   +173
Yes, AMD CPUs have a nasty habit of getting stuck to heatsinks because their PGA socket has no retention bracket, and when they get pulled out, pins tend to bend.

I cant imagine power is an issue looking at what intel's 11900k pulls through a LGA socket.
The most common cause of bent pins on a CPU is them getting pulled out of their socket on an angle during heatsink removal. LGA sockets have retention brackets and thus do not suffer from this issue. When is the last time you heard of an intel CPU pulling free from a motherboard?

They can also be bent during transportation in their box, not an issue with LGA CPUs. You can put protective covers on motherboards (as all mobo makers do) but doing so to a PGA CPU is much harder to do properly.

You know you can just add a retention bracket to a PGA socket right?