Discussion Speculation: Zen 4 (EPYC 4 "Genoa", Ryzen 7000, etc.)

[Vattila]

Except for the details about the improvements in the microarchitecture, we now know pretty well what to expect with Zen 3.

The leaked presentation by AMD Senior Manager Martin Hilgeman shows that EPYC 3 "Milan" will, as promised and expected, reuse the current platform (SP3), and the system architecture and packaging looks to be the same, with the same 9-die chiplet design and the same maximum core and thread-count (no SMT-4, contrary to rumour). The biggest change revealed so far is the enlargement of the compute complex from 4 cores to 8 cores, all sharing a larger L3 cache ("32+ MB", likely to double to 64 MB, I think).

Hilgeman's slides did also show that EPYC 4 "Genoa" is in the definition phase (or was at the time of the presentation in September, at least), and will come with a new platform (SP5), with new memory support (likely DDR5).

What else do you think we will see with Zen 4? PCI-Express 5 support? Increased core-count? 4-way SMT? New packaging (interposer, 2.5D, 3D)? Integrated memory on package (HBM)?

Vote in the poll and share your thoughts!

 

[Saylick]

Not a surprise with the prices, especially the 7800x3d. I'm guessing we'll have to wait until at or near April's release date to get data to see if its worth it over a 13700k.

I think we might get data from the 7950X3D since the V$ CCD should be roughly the same as the one in the 7800X3D. If the BIOS let's reviewers turn off the non-V$ CCD, we'd be able to get something very close to the 7800X3D and you'd gleam some insight right then.

 

[Joe NYC]

I think we might get data from the 7950X3D since the V$ CCD should be roughly the same as the one in the 7800X3D. If the BIOS let's reviewers turn off the non-V$ CCD, we'd be able to get something very close to the 7800X3D and you'd gleam some insight right then.

Hopefully, AMD will make all 3 chips available to reviewers at the same time, and release all results at the same time, to save them some work...

 

[In2Photos]

Hopefully, AMD will make all 3 chips available to reviewers at the same time, and release all results at the same time, to save them some work...

Would be nice, but I doubt it.

 

[biostud]

That was my first thought as well.

But my second thought was that with the track record AMD marketing has over last year, 7900x3d and 7950x3d will be released with broken Windows 11 scheduler, to complete the trifecta of bungled launches.

Releasing it the 28th and only the 7950X3D smells very much like, they promised February, but something did not go as planned, so they had to push the release day as far as possible and deliver as little as possible without making Lisa Su look too bad.

 

[blckgrffn]

Releasing it the 28th and only the 7950X3D smells very much like, they promised February, but something did not go as planned, so they had to push the release day as far as possible and deliver as little as possible without making Lisa Su look too bad.

Nah, that's February for the marketing department and you start with only offering the priciest one. A non-trivial number of fence sitters waiting for the 3D chips will say "YOLO" and get the monster chip ASAP vs the least expensive one. The margins are better on the big boy so you get maximum exposure for it and then quietly release the other options, or create dedicated launches for them lined up at the competition.

Social media wise, you are in the news cycle for much longer this way too, maximum coverage.

Hopefully, AMD will make all 3 chips available to reviewers at the same time, and release all results at the same time, to save them some work...

That would likely cost them money in lost sales of higher end SKUs. That's against their duty to the shareholders.

This smells like they grew up and decided to make money.

 

[eek2121]

Going to wait for a few reviews, but I might pick a 7950X3D up at launch.

 

[blckgrffn]

Going to wait for a few reviews, but I might pick a 7950X3D up at launch.

It’s likely to be the big dog for quite a while. If you are spending on AM5 and you can budget it in, might as well imo.

 

[Hougy]

Did I understand this right? In the 7900X3D and 7950X3D looks like only one one the CPU dies will receive the extra cache die. So that must be why their top frequencies are the same as the non-3D version: one of the dies doesn't have the extra cache die so it can boost higher.

How will the scheduler know if it's better the pick the core that boost higher or the cores with more cache? Is Windows 11 already prepared?

 

[maddie]

Did I understand this right? In the 7900X3D and 7950X3D looks like only one one the CPU dies will receive the extra cache die. So that must be why their top frequencies are the same as the non-3D version: one of the dies doesn't have the extra cache die so it can boost higher.

How will the scheduler know if it's better the pick the core that boost higher or the cores with more cache? Is Windows 11 already prepared?

View attachment 75812

We're all travelling to the future together.

 

[Markfw]

We're all travelling to the future together.

My 7950x3d will be running linux and probably primegrid.

 

[DrMrLordX]

How will the scheduler know if it's better the pick the core that boost higher or the cores with more cache? Is Windows 11 already prepared?

Hard to say for sure. I think the day-one buyers are gonna be beta testers for this thing.

 

[leoneazzurro]

IIRC AMD said there would have been new software for correctly routing the threads on the 7900X3D/7950X3D.

 

[Vattila]

How will the scheduler know if it's better the pick the core that boost higher or the cores with more cache? Is Windows 11 already prepared?

In the worst case, if the OS does a lousy job, you can manually set the processor affinity for the process (e.g. game).

See Post #11516 for a tip on how to launch a program with the desired affinity.

 

[MadRat]

I don't see this as anything but raw promotion of the new standards, to dry up AM4 holdouts.

 

[MrTeal]

I guess we'll have to wait another couple quarters to get a proper 7950X3Dx2 XTX chip.

 

[Mopetar]

Hard to say for sure. I think the day-one buyers are gonna be beta testers for this thing.

Hard to say. There's nothing that would have prevented AMD from internally testing this kind of configuration with Zen 3 and working out potential issues years ago. Whether they did this or not is another matter.

 

[biostud]

Hard to say. There's nothing that would have prevented AMD from internally testing this kind of configuration with Zen 3 and working out potential issues years ago. Whether they did this or not is another matter.

They didn't...

 

[Hail The Brain Slug]

They didn't...

The original "5900X3D" that was demoed back in the day was a hybrid layout like the 7900X3D and 7950X3D. To pretend AMD hasn't been testing this layout and working on their driver for it would be folly.

 

[biostud]

The original "5900X3D" that was demoed back in the day was a hybrid layout like the 7900X3D and 7950X3D. To pretend AMD hasn't been testing this layout and working on their driver for it would be folly.

That is true, I forgot about that one.

 

[Joe NYC]

Hard to say for sure. I think the day-one buyers are gonna be beta testers for this thing.

The review web sites should be getting their samples by mid February and they will have at least a week to fiddle with the chips before posting their results.

 

[Joe NYC]

A couple of things from the Tom's podcast on the new V-Cache chips:
- the SRAM chips will be N6 (confirming MLID leak stating this)
- the 79x0X3D chips will not be relying on Windows scheduler but on a "white list" in the driver to direct the games to the V-Cache CCD.

https://twitter.com/x/status/1621222268994224129

 

[gdansk]

In which driver? The chipset driver?

 

[Abwx]

In which driver? The chipset driver?

The CPU driver...

 

[Joe NYC]

The CPU driver...

Means, there is going to be a lot of driver updates,,,

 

[gdansk]

The CPU driver...

That's weird, I go to AMD's site and there are no CPU drivers.

So they'll make a new driver bundle to install for the two asymmetrical CPUs instead of bundling it in the chipset drivers? Yuck.