Zen 3 is rumored to be flaunting monumental IPC gains in early testing

mongeese

TS Maniac
Staff member

As AMD confirmed at the HPC-AI Advisory Council UK conference, their next-generation server architecture, Milan, is being tested by their biggest customers. Milan and Ryzen 4000 will share the same Zen 3 architecture and TSMC’s 7nm+ EUV process, meaning that most we hear about Milan applies to 2020’s Ryzen, too.

According to Red Gaming Tech and Bits n' Chips, a multitude of AMD’s customers have tested Zen 3 and all say the same thing. The new architecture’s redesigned floating-point units have increased IPC in scientific tasks by something close to 50%, while integer work has seen a 10-12% boost. The combination of the two has resulted in a ~17% or greater IPC increase in mixed workloads.

Zen 3 will deliver performance gains "right in line with what you would expect from an entirely new architecture."

A few leaked AMD slides, confirmed and elaborated upon by industry sources, explain how these IPC improvements have come about. Unlike the jump from Zen+ to Zen 2, which merged two quad-core CCXs (core complex) into one octa-core CCD (core complex die), Zen 3 is reported to have unified octa-core CCDs. This does two things, theoretically: radically reduces the average inter-core communication latency and increases the amount of L3 cache each core has available to it.

Having a unified octa-core CCD is very powerful. In our 4 GHz comparison, we found best-matched threads in the 9900K to have 46.5 ns of communication latency and the worst to have 52.6 ns. The 3700X, with two CCXs, had a best result of 30.2 ns and a worst result of 84.6 ns. That means that in applications with minimal inter-thread communication like Cinebench, Zen 2 can beat Intel, but in communication heavy applications like games, Intel has the advantage.

The rumored Zen 3 octa-core CCD design flips that on its head and could take away Intel’s latency advantage in games. Testing has shown that Zen 3 apparently has "significantly enhanced" latency across the chip. I doubt many Milan server CPUs have been tested with games, but when consumer Zen 3 does arrive, the IPC increase in games could be more than what we're hearing about currently.

Reworking the CCD designs requires the cache to be reworked, too. In Zen 2 each CCD has 32 MB of L3 cache, but it’s split into two discrete 16 MB packages, one for each CCX. With a unified CCD design, the full 32 MB becomes available to all eight cores at high speeds. AMD’s slides confirm this, but also suggest that there may be more than 32 MB per CCD. According to Bits n’ Chips, there may be an increase of "at least" 50%, but we’ll have to see if this applies to server versions of the chip only.

Industry sources have also hinted that the L3 cache could get a 40% increase in bandwidth. AMD finally matched Intel on L3 bandwidth with Zen 2, so pushing past them could put Intel on the backfoot there, too.

Apart from IPC, there might not be many other changes with Zen 3. Core counts are expected to stay the same, for a few good reasons; the market isn’t ready, AM4 can’t support more, the software is still trying to catch up, etc. Clock speeds on the server chips have apparently increased by 100 MHz to 200 MHz, but AMD has already pushed their clocks aggressively with Zen 2, so we wouldn’t expect much there.

TSMC is upgrading AMD to the 7nm+ EUV node, and node changes offer its advantages. While the boost clocks on the new chips might not improve much, the base clocks might. With the introduction of EUV (extreme ultraviolet lithography), yields should give AMD further breathing room on pricing. Of course, this is all theoretical as AMD is months away from deciding on prices.

Which perhaps brings us to the most important point: Zen 3 is technically complete, but we’re still anywhere from 6 to 12 months away from launch. Even if everything we suspect now is correct, it could all change by the time Ryzen 4000 series CPUs are released.

Permalink to story.

 

Vulcanproject

TS Evangelist
The unified cache and 'native' 8 core design should be enough by itself to improve Ryzen's performance in games to at least parity. Right now if you run 1440p or higher with good RAM the difference is mostly academic between a 3950x and a 9900k anyway. 10 percent at 1080p, less the more resolution you pile on. That's the trend as it is.

AMD still need more though, they need to be able to say we can match or beat Intel in every area for the consumer. Gaming is the last bastion of Intel supremacy and I think it will finally fall to Zen 3. Although we probably won't be seeing it for another 9 months.

Needless to say game developers are already wrestling with Zen 2 and Navi in new consoles. That's another bonus AMD have in their pocket. When most devs become familiar with the Zen architecture it will surely only help, even just a little, as those games drop on PC.
 

Adi6293

TS Guru
Well my cash is waiting, I will most likely jump for the 4700X that should be at least 30% gain over my 2700X
 

Lionvibez

TS Evangelist
This looks excellent.

Glad I went with a 3800x for my build now.

And may move up to a 12 code model for Zen 3.

However if a 8 core zen 3 cpu offers 15-20 more IPC and a nice overall performance improvement that would still be a nice upgrade.
 
  • Like
Reactions: m3tavision

veLa

TS Evangelist
Zen3 is Am4 compatible where did you read it was not?

AMD won't be changing sockets until DDR5 and Zen 4 most likely.
So I haven't seen anything concrete, and AMD has stated AM4 has a lifespan of up to 2020, but with major architectural changes, you can't be 100% sure about that. With that being said, AMD has been great with socket compatibility so far.
 

Lionvibez

TS Evangelist
Ryzen 4000 or Zen 3 will see a x670 chipset released with it but will still be AM4 compatible.

PCIe 5. DDR5 will require a new chipset and socket but we won't see those until probably 2022 or late 2021
 

Adi6293

TS Guru
So I haven't seen anything concrete, and AMD has stated AM4 has a lifespan of up to 2020, but with major architectural changes, you can't be 100% sure about that. With that being said, AMD has been great with socket compatibility so far.
Yeah I would be very surprised if AMD changed socket for just 1 generation, its best to keep Zen 3 on AM4 and then change the socket when they go up to DDR5 and PCI 5.0
 

arrowflash

TS Booster
The unified cache and 'native' 8 core design should be enough by itself to improve Ryzen's performance in games to at least parity. Right now if you run 1440p or higher with good RAM the difference is mostly academic between a 3950x and a 9900k anyway. 10 percent at 1080p, less the more resolution you pile on. That's the trend as it is.

AMD still need more though, they need to be able to say we can match or beat Intel in every area for the consumer. Gaming is the last bastion of Intel supremacy and I think it will finally fall to Zen 3. Although we probably won't be seeing it for another 9 months.

Needless to say game developers are already wrestling with Zen 2 and Navi in new consoles. That's another bonus AMD have in their pocket. When most devs become familiar with the Zen architecture it will surely only help, even just a little, as those games drop on PC.
I hope so, but afaik another achilles heel of AMD CPUs in gaming performance is latency - let's hope Zen3 improves significantly on this area as well, or manages to at least reach gaming performance parity with Intel by brute forcing their way through IPC alone...

While in GPU bound games the difference might be negligible nowadays, Intel still has a significant advantage in some CPU bound games, and also in emulation.
 

Daniele 00

TS Enthusiast
I hope so, but afaik another achilles heel of AMD CPUs in gaming performance is latency - let's hope Zen3 improves significantly on this area as well, or manages to at least reach gaming performance parity with Intel by brute forcing their way through IPC alone...

While in GPU bound games the difference might be negligible nowadays, Intel still has a significant advantage in some CPU bound games, and also in emulation.
Intel has an advantage in name lenght aswell. You see... I N T E L... 5 char. A M D... only 3 char. Intel won.
Man, in general Ryzen 1 was already faster than intel, except for these who only see at pc as a gaming mashine.

About latency, the article is being positive about it ("Testing has shown that Zen 3 apparently has "significantly enhanced" latency across the chip."), so I hope for the good.
 

Vulcanproject

TS Evangelist
I hope so, but afaik another achilles heel of AMD CPUs in gaming performance is latency - let's hope Zen3 improves significantly on this area as well, or manages to at least reach gaming performance parity with Intel by brute forcing their way through IPC alone...

While in GPU bound games the difference might be negligible nowadays, Intel still has a significant advantage in some CPU bound games, and also in emulation.
The latency that hinders Ryzen against Intel is addressed here, moving data across the core complexes (one CCX reading the other's cache) incurs a latency penalty. The changes being made for Zen 3 eliminates that, at least for anything up to 8 core parts.

I would say the advantage averages out to less than 10 percent at the high end. Techspot themselves did a recent article with tuned memory and showed that was the case. This was fairly impressive in the face of the high clocks of the 9900ks. I would expect that gap to reduce just by the passage of time alone, as new consoles employ 8 core Zen 2 CPUs.

In terms of emulation that's mostly a software thing. With most emulation software being built on and for Intel designs the past 15 years and AMD's inferior single threaded performance (up til at least Zen 3) it's not going to change too quickly.

If AMD manage to surpass the IPC of Intel then it'll be much better than older Zen parts at emulation. But I wouldn't expect it to be as good unless the market sustains a heavy shift in favour of AMD.
 

JimboJoneson

TS Addict
The unified cache and 'native' 8 core design should be enough by itself to improve Ryzen's performance in games to at least parity. Right now if you run 1440p or higher with good RAM the difference is mostly academic between a 3950x and a 9900k anyway. 10 percent at 1080p, less the more resolution you pile on. That's the trend as it is.
...
Its only 4% if you tune your ram up a bit ... and its not just about 1080p -- its 4% @1080P with a 2080ti ... any less video card, and that 4% shrinks to ~0% at any resolution.

So for 99.99% of gamers that don't own a 2080ti, and the ones that don't use theirs to play at 1080p ... the difference is entirely "academic" and not experienced in the real world.

Who really cares about 4% in a hardware and setup scenario that less than 0.1% of all gamer's use where its less than 4% and most likely 0% in the setup you actually have?

What if AMD sneaks out the win by a tiny 4%? It'll be claimed as "insignificant" by a certain some, you can be assured, but currently a 4% advantage with a 2080ti at 1080p is ABSOLUTELY DESTROYING! AMD in gaming.

The differences are all already entirely "academic"; that's why an artificial bottleneck is induced with a 2080ti and low resolution for the reviews ... because the graphs would all be the same if they didn't.

Also keep in mind some of us use our CPUs for adult stuff, like work, video making, multitasking, design, etc. ... gaming is just one facet.
 
Last edited:

JimboJoneson

TS Addict
So I haven't seen anything concrete, and AMD has stated AM4 has a lifespan of up to 2020, but with major architectural changes, you can't be 100% sure about that. With that being said, AMD has been great with socket compatibility so far.
Rob Hallock from AMD stated in a PCWorld interview that "AMD won't ever change sockets "just because" or to force more money out of its customers. We'll keep to the same sockekt as long as they can fit their newer CPU designs into them and the technology doesn't demand otherwise --- like a moving to using DDR5 for example" - paraphrased.

Zen3 is DDR4 and Zen4 is DDR5 so it seems he may have been hinting that AM4 would still be good with Zen3.

My guess would be that Zen3 may be restricted to X570 boards though ... that is a possibility.
 
Last edited:

cfbcfbcfb

TS Rookie
Zen3 is Am4 compatible where did you read it was not?

AMD won't be changing sockets until DDR5 and Zen 4 most likely.
It depends on motherboard bios support. If a motherboard supplier doesn't release a bios update to support the cpu, having an AM4 socket isn't going to matter.
 

cfbcfbcfb

TS Rookie
Yes but that should be obvious didn't think I had to mention that.
It might have been obvious to YOU, but not to everyone. Thing is, very few OEM boards and very few older boards will support newer cpu's. Believe it or not, most people DO NOT know anything about that.
But I'm sure the line of people who do will make me want to never post anything helpful ever again.
 

Lionvibez

TS Evangelist
It might have been obvious to YOU, but not to everyone. Thing is, very few OEM boards and very few older boards will support newer cpu's. Believe it or not, most people DO NOT know anything about that.
But I'm sure the line of people who do will make me want to never post anything helpful ever again.
Ok!
 

Maxiking

TS Booster
Lol the only way that can manage 50% more performance in scientific workloads is avx512 support, so this leak is complete bs and no, you don't give away pre es samples to your partners to test them and how they perform, another bs. Motherboards aren't ready, bioses aren't ready, microcode isn't ready, who is going to test them and how?