AMD Ryzen 7000X3D Architecture Preview

Introduction

Last week at CES, AMD announced their Ryzen 7000X3D series CPUs that encompassed three CPUs, the Ryzen 7 7800X3D, the Ryzen 9 7900X3D and the Ryzen 9 7950X3D. Many of us expected the 7800X3D to arrive as a successor to the 5800X3D, thus making the 12 and 16 core counterparts a nice surprise.

As with most keynotes, AMD glossed over the specifications of the new chips, leaving us wondering what configuration the chiplets would hold and exactly how many 3D cache enabled cores would be provided on each. Luckily enough, we got to sit down for a half hour with AMD’s Scott Stankard, Director of Product Management for AMD Client and talk about the Ryzen 7000X3D Series to fill in some of the blanks left from the keynote.

7950X3D

Starting off with the Ryzen 9 7950X3D, we have a 16 core, 32 thread chip that sports 144MB of total cache, a boost clock of up to 5.7GHz and a base clock of 4.2GHz (a 300MHz decrease from the 7950X). This is running at 120W TDP, which is a reduction from its 2D counterpart by 50 watts.

The 7950X3D utilizes a chiplet configuration sporting two CCDs (core complex dies) that contain 8 cores each. One of the CCDs contains 3D cache while the other CCD does not, leaving us with an 8 + 8X3D configuration.

In our conversation with Scott, we learned that AMD wanted to provide a blend of performance for the processor so a user could get all of the benefits of the standard Zen 5 platform, but also have the ability for top of class gaming performance.

This is to be accomplished through AMD directing specific games or applications to primarily use a particular CCD. For example, most games will default to using the X3D CCD as their default while lightly threaded business applications will use the non-X3D CCD as their default. Scott stated that this type of core direction has been in the works for a while on the Zen platform and he does not expect to be as complex as the Alder Lake launch that has the whole big/little configuration to worry about. Scott also stated that these optimizations should work equally well in Windows 10 and Windows 11.

From a trade off perspective, the boost clock on the X3D CCD will be lower than the boost clock on the standard CCD. Therefore, the 5.7GHz boost specification only applies to the standard CCD. From a heavy multithreaded workload performance perspective (i.e. Cinebench), the overall chip will be slower than the 7950X due to the lower TDP and the X3D CCD running at a lower clock. However, the difference in performance will not be as pronounced as it was when comparing the 5800X to the 5800X3D.

7900X3D

Moving on to the Ryzen 9 7900X3D, we have a 12 core, 24 thread chip that sports 140MB of total cache, a boost clock of up to 5.6GHz and a base clock of 4.4GHz (a 300MHz decrease from the 7900X). This is running at 120W TDP, which is a reduction from its 2D counterpart by 50 watts.

The 7900X3D utilizes a chiplet configuration sporting two CCDs (core complex dies) that contain 6 cores each. One of the CCDs contains 3D cache while the other CCD does not, leaving us with an 6 + 6X3D configuration.

Other than the differences in the core count and the configured frequencies, everything written about above in the 7950X section should apply to the various tradeoffs discussed.

7800X3D

Finally, we arrive at the Ryzen 7 7800X3D. This chip is the most familiar to us, as quite frankly, it looks like a Zen 5 version of the 5800X3D. We have a 8 core, 16 thread chip that sports 104MB of total cache, a boost clock of up to 5.0GHz and a base clock that we’ll find out with we get ahold of the product to review. This is running at 120W TDP, which is a increase from its 2D counterpart by 50 watts (and by counterpart, I’m loosely saying the 7700X is its match because the 7800X didn’t show up to the game).

The 7800X3D utilizes a chiplet configuration sporting a single CCDs (core complex dies) that contains 8 cores. The lone CCD contains 3D cache for all of the cores leaving us with an 8X3D configuration.

From a performance perspective, we are expecting to see a comparable differential between the 7700X and the 7800X3D as we saw between the 5800X and the 5800X3D. It will likely lead on virtually all gaming performance, but fall behind on productivity and content creation as a result of the lower boost and base clocks.

Overclocking

Much like the 5800X3D, do not expect overclocking to be a thing with the Ryzen 7000X3D Series. When discussing this topic with Scott, he stated that we will gain access to using the Curve Optimizer, but the primary expected use case is to find ways to reduce power consumption and heat to run at the already configured speeds. Essentially, Curve Optimizer will allow you to reduce power by using some of the factory margin built into the chips.

Conclusion

The Ryzen 7000X3D Series of chips looks very interseting to us due to the design choices made on the 7900X3D and the 7900X3D to have a blend of capabilities by having one CCD with 3D Cache and the other without. A lot of bold performance claims were made about giving the X3D buyers a CPU that will bring the best of both worlds to them.

That being said, there’s a lot of risk in the implementation of thread assignment functionality which has been echoed by many folks out on the forums. Beyond what we learned in our meeting, we don’t have anything else to go on until we get the chips in hand and that’ll be about a month from now with an expected launch date of February 14, 2023 whenever they are launched they’ll let us know (see update from AMD at the link).

In the meantime, jump on into our forums at the link below and let us know your thoughts on AMD’s design choices with the Ryzen 7000X3D series of processors – will you be clicking add to cart next month?

Join the discussion in our forums...