AMD Zen 5 rumours

[Capitano]

I'm going to hijack this thread with a nod to https://forums.overclockers.co.uk/threads/zen-5-chat.18953487/

Made a decision to skip Zen 4 X3D, so naturally started looking for whats next.

Rumours say
not much increase in clock speed
different cache structure
substantial IPC increase
still 8 cores per CCD
and whatever in https://www.techpowerup.com/review/future-hardware-releases/#zen5

The big IPC increase is very exciting! wonder what sort of big.little core structure we will see though.. Will they have a 8950x for example that still has the 16 big cores, but also add on a bunch of little cores? Or will they reduce the big core count and add a lot of little cores which still gives over 32 threads, but with few big cores?

Interesting times lie ahead, can't wait to see all the leaks up coming!

 

[faceman123]

Using large little as an excuse to stagnate true performance is garbage. I want to run multiple virtual machines with the same core speed. Ryzen is already pretty energy-efficient. Heck, they ought to just go end game and figure out how to mainstream Threadripper. BIG NO to big little.

 

[Curlyriff]

Using large little as an excuse to stagnate true performance is garbage. I want to run multiple virtual machines with the same core speed. Ryzen is already pretty energy-efficient. Heck, they ought to just go end game and figure out how to mainstream Threadripper. BIG NO to big little.

Big-Little for generally desktop users make sense though. A lot of people use their PC for MS applications, discord, YT, Internet browsing and at those times a little core sipping power makes sense. The difference is I think it would be better to have around 4 little cores doing those little background tasks where windows assigns them to such based on the application and then you can still have 16-24 big cores for when gaming, video editing etc.

The route Intel have taken is in reverse to how it would make most sense in usage but done it to sell CPU with 'more cores' whilst actually being able to power the chip.

 

[Capitano]

Big-Little for generally desktop users make sense though. A lot of people use their PC for MS applications, discord, YT, Internet browsing and at those times a little core sipping power makes sense. The difference is I think it would be better to have around 4 little cores doing those little background tasks where windows assigns them to such based on the application and then you can still have 16-24 big cores for when gaming, video editing etc.

The route Intel have taken is in reverse to how it would make most sense in usage but done it to sell CPU with 'more cores' whilst actually being able to power the chip.

Yeah this is whats crucial, im talking about adding little efficient cores to the current chips with plenty of big cores so the total core count is way higher than what we have now. However if the top end AMD chip goes down to say 8 big cores and 12 little cores, thats a step down imo. Unless of course the little cores are as powerful as current big cores using much less power, which I highly doubt for the very next generation

 

[alec]

The big IPC increase is very exciting! wonder what sort of big.little core structure we will see though.. Will they have a 8950x for example that still has the 16 big cores, but also add on a bunch of little cores? Or will they reduce the big core count and add a lot of little cores which still gives over 32 threads, but with few big cores?

Interesting times lie ahead, can't wait to see all the leaks up coming!

So far the little core rumors are all in serverspace. Little cores are not coming to desktop, at least not in Zen 5 gen.

 

[Capitano]

So far the little core rumors are all in serverspace. Little cores are not coming to desktop, at least not in Zen 5 gen.

They will almost definitely come to the desktop, and there were early rumours saying it was going to be Zen5. I think Zen6 will be more likely though however

 

[Grim5]

Gigabyte puts our marketing statement saying AMDs next generation of AM5 CPUs is launching later this year

Gigabyte Says A New Generation of AMD Ryzen Desktop CPUs Will Be Released Later This Year

In a press release, Gigabyte has stated that a new generation of AMD Ryzen 7000 Desktop CPUs will be released later this year.

wccftech.com

 

[NinjaCool]

The 7000 desktop CPUs came out in Q3 2022 so it looks like we could very well see the next generation CPUs coming out by end of the year

 

[Grim5]

Looks like it might not be zen5 that Gigabyte referring too.

It looks like amd is going to launch a hybrid cpu first. The Phoenix APU has both P and E cores

AMD hybrid "Phoenix" APUs to feature Performance and Efficiency cores - VideoCardz.com

AMD Phoenix with Performance and Efficiency cores The company now has a name for its hybrid design cores in consumer APUs. According to the AMD Processor Programming Reference (PPR) guide for AMD Family 19h Model 70h CPUs (AMD Zen4/Phoenix), the hybrid architecture may support two types of...

videocardz.com

 

[gazref]

Are AMD slowing down production of the slow selling 7000 CPUs in anticipation on next gen release ? Wondering if we should expect some overdue better pricing or if they intend to throttle supply to keep pricing tiers more fixed

 

[Tetras]

Are AMD slowing down production of the slow selling 7000 CPUs in anticipation on next gen release ? Wondering if we should expect some overdue better pricing or if they intend to throttle supply to keep pricing tiers more fixed

Wouldn't have thought so, AM4 CPUs are still plentiful, so don't see why they'd cut supply for AM5.

 

[kute]

I look forward to the coming generation. I'm happy with the 7600x I got during that big sale in early december, so I'm skipping the X3D parts this time around. Unless the 7800x3d is really, really something.

Am I the only person satisfied with the direction Am5 took? Yes, the i9 is the fastest CPU, but it's uncoolable and power hungry. That's not quite the direction I want things to go. Realistically, all the AMD chips are just as fast, trading blows here and there, maybe an irrelevant amount behind at irrelevant resolutions.

The motherboard was 50 bucks more than it should have been, but the ram wasn't as expensive as I thought it'd be and my 5600 hynix easily oc'd to 6k/32

 

[Capitano]

the rumoured 22-30% IPC lift sounds very nice though

 

[humbug]

the rumoured 22-30% IPC lift sounds very nice though

That's:
2,530 to 2,700 in ST R23.
47,150 to 50,250 in MT R23. (assuming 16 cores)

At the same clock speed.

Who said this?

 

[Capitano]

That's:
2,530 to 2,700 in ST R23.
47,150 to 50,250 in MT R23. (assuming 16 cores)

At the same clock speed.

Who said this?

Hardware & Tech News - OC3D.net

OC3D is where you can find the latest PC Hardware and Gaming News & Reviews. Get updates on GPUs, Motherboards, CPUs, and more.

overclock3d.net

just early leaks of course, so take it with a grain of salt. But if true, its very impressive. The jump to 3nm should play a big part in it

 

[humbug]

Hardware & Tech News - OC3D.net

OC3D is where you can find the latest PC Hardware and Gaming News & Reviews. Get updates on GPUs, Motherboards, CPUs, and more.

overclock3d.net

just early leaks of course, so take it with a grain of salt. But if true, its very impressive. The jump to 3nm should play a big part in it

Red Tech Gaming.

The node size has nothing to do with it, IPC is all architecture.

Its early, you're right, ill believe it when i see it, but the rumours for Zen 4 turned out to be right.

Zen 5 is yet another new architecture but that IPC change is somewhat higher than previous architecture changes, IPC is the performance at the same clock speed and 22 to 30% is high, more than half of the 35% MT gain from Zen 3 to Zen 4 came from clock speed.

Its mindbogglingly fast, if they can pull it off...

 

[Grim5]

Hardware & Tech News - OC3D.net

OC3D is where you can find the latest PC Hardware and Gaming News & Reviews. Get updates on GPUs, Motherboards, CPUs, and more.

overclock3d.net

just early leaks of course, so take it with a grain of salt. But if true, its very impressive. The jump to 3nm should play a big part in it

If zen5 is 3nm it 100% ain't coming this year cause Apple owns 100% of all 3nm wafer supply for all of 2023

 

[humbug]

If zen5 is 3nm it 100% ain't coming this year cause Apple owns 100% of all 3nm wafer supply for all of 2023

Apple are always 1'st to the new node but they don't hog it to themselves forever and remember AMD are TSMC's No: 2 customer, that counts for a lot.

TSMC Might Cut 3nm Prices to Lure AMD, Nvidia

Industry sources say TSMC is considering lowering 3nm prices to stimulate interest from chip designers

www.tomshardware.com

 

[Capitano]

Red Tech Gaming.

The node size has nothing to do with it, IPC is all architecture.

Its early, you're right, ill believe it when i see it, but the rumours for Zen 4 turned out to be right.

Zen 5 is yet another new architecture but that IPC change is somewhat higher than previous architecture changes, IPC is the performance at the same clock speed and 22 to 30% is high, more than half of the 35% MT gain from Zen 3 to Zen 4 came from clock speed.

Its mindbogglingly fast, if they can pull it off...

IPC is architecture you're correct - its just the gains overall, not just IPC. Either way I think its safe to assume there will be an increase in IPC as well, but not to those levels.

If the majority of the gains come from an IPC increase vs the clock speed, that would be much preferred. There's only so much you can have high clock speeds before the power consumptions becomes too much

If zen5 is 3nm it 100% ain't coming this year cause Apple owns 100% of all 3nm wafer supply for all of 2023

We could still see the release in early 2024 say jan or feb, which isn't that much of an issue. Not to mention that despite Apple being their biggest customer, AMD is their 2nd biggest.

 

[humbug]

IPC is architecture you're correct - its just the gains overall, not just IPC. Either way I think its safe to assume there will be an increase in IPC as well, but not to those levels.

If the majority of the gains come from an IPC increase vs the clock speed, that would be much preferred. There's only so much you can have high clock speeds before the power consumptions becomes too much

Difficult to know.

You're right the article doesn't talk about IPC, and yet the slide does, but it also qualifies that with (1T) its as if it the slide was made by someone who doesn't understand what IPC is, qualifying threads is irrelevant when quoting IPC, you would only do that if you were including clock speeds which are usually different 1T to nT.
Some people use IPC and clock speeds interchangeably, which is idiotic, they have no understanding of it.... given the poor quality of todays tech journalists i'm going to assume they are idiots and agree its probably a combilation of clock speed and IPC.