AMD Confirms Ryzen 9 7950X3D and 7900X3D Feature 3DV Cache on Only One of the Two Chiplets

[btarunr]

AMD today announced its new Ryzen 7000X3D high-end desktop processors to much fanfare, with availability slated for February 2023, you can read all about them in our older article. In our coverage, we noticed something odd about the cache sizes of the 12-core 7900X3D and 16-core 7950X3D. Whereas the 8-core, single-CCD 7800X3D comes with 104 MB of total cache (L2+L3), which works out to 1 MB L2 cache per core and 96 MB of L3 cache (32 MB on-die + 64 MB stacked 3DV cache); the dual-CCD 7900X3D and 7950X3D was shown with total caches of 140 MB and 144 MB, while they should have been 204 MB or 208 MB, respectively.

In our older article, we explored two possibilities—one that the 3DV cache is available on both CCDs but halved in size for whatever reason; and the second more outlandish possibility that only one of the two CCDs has stacked 3DV cache, while the other is a normal planar CCD with just the on-die 32 MB L3 cache. As it turns out, the latter theory is right! AMD put out high-resolution renders of the dual-CCD 7000X3D processors, where only one of the two CCDs is shown having the L3D (L3 cache die) stacked on top. Even real-world pictures of the older "Zen 3" 3DV cache CCDs from the 5800X3D or EPYC "Milan-X" processors show CCDs with 3DV caches having a distinct appearance with dividing lines between the L3D and the structural substrates over the regions of the CCD that have the CPU cores. In these renders, we see these lines drawn on only one of the two CCDs.



It shouldn't be hard for such an asymmetric cache setup to work in the real world from a software perspective, given that we are now firmly in the era of hybrid-core processors thanks to Intel and Arm. Even way before "Alder Lake," when AMD started shipping dual-CCD client processors with the Ryzen 3000 "Matisse" based on "Zen 2," the company closely collaborated with Microsoft to optimize OS scheduling such that high-performance and less-parallelized workloads such as games, are localized to just one of the two CCDs, to minimize DDR4 memory roundtrips.

Even before "Matisse," AMD and Microsoft confronted multi-threaded workload optimization challenges with dual-CCX architectures such as "Zen" and "Zen 2," where the OS scheduler would ideally want to localize gaming workload to a single CCX before saturating both CCXs on a single CCD, and then onward to the next CCD. This is achieved using methods such as CPPC2 preferred-core flagging, and which is why AMD highly recommends you to use their "Ryzen Balanced" Windows power-plan included with their Chipset drivers.

We predict that something similar is happening with the 12-core and 16-core 7000X3D processors—where gaming workloads can benefit from being localized to the 3DV cache-enabled CCD, and any spillover workloads (such as audio stack, network stack, background services, etc) are handled by the second CCD. In non-gaming workloads that scale across all 16 cores, the processor works like any other multi-core chip, it's just that the cores in the 3DV-enabled CCD have better performance from the larger victim cache. There shouldn't be any runtime errors arising from ISA mismatch, as the CPU core types on both CCDs are the same "Zen 4."

AMD Ryzen 7000X3D processors go on sale in February 2023.

View at TechPowerUp Main Site

 

[Space Lynx]

my guess is the next x3d chips in a year or two will haver both ccd, its probably just a backup plan for if Intel comes out swinging again soon, they can just add another x3d cache to the other ccd and swing back to take the crown again.

lol. dumb, they should have just swung all the way and slam dunked.

 

[Pumper]

Seeing as 3D cache benefits primarily games and lower thread applications, that sounds fine to me, unless they mess up the software side and the wrong cores (without 3DV) get priority usage for no reason.

Nice to see only 120W default TDP.

 

[Selaya]

epicfail.

 

[Camm]

Starting to think that we will see 8000 series in a Zen / ZenC layout with those Epyc bound low cache cores making up the low power "e" cores, and with 3D Cache on the full fat Zen cores, and that this is just a harbinger of such.

 

[TechLurker]

my guess is the next x3d chips in a year or two will haver both ccd, its probably just a backup plan for if Intel comes out swinging again soon, they can just add another x3d cache to the other ccd and swing back to take the crown again.

lol. dumb, they should have just swung all the way and slam dunked.

I wonder if part of it was simply a cost issue. Even the prototype 5900X3D only had the 3D cache on one die instead of both, yet EPYCs with X3D will supposedly have it across all CCDs.

That and probably keeping their trump card available as either another half-step up (X3D+; or "Black Editions"), or as an interim step up, like how Zen+ was just refined Zen1, bridging the period until AMD could release a major next step in Zen2 (basically, start off the Ryzen 8000 or 9000 series as single X3D and dual X3D chips, with non-X3D options available or omitted in favor of binning/scalability).

 

[Lionheart]

epicfail.

Yup! It's definitely a epic fail even though there's no benchmarks yet, such a fail. SMH

 

[Lovec1990]

7900X3D and 7950X3D sound like AMD is experimenting with two diffrent CCDs on same CPU so basicaly they gonna do something similar than Intel did with Pcores+Ecores

 

[Theswweet]

There's simply no point in having the 3D cache on both of the CCDs for games; any games that actually tried to make use of cache on both dies would negate the actual performance improvement that comes with the cache at all. Instead allowing the other CCD to have the "normal" Zen 4 boost clocks means that AMD can have their cake and eat it too; 7950x performance in games that don't benefit from cache - ie, not a downgrade - while improving performance for games that do. At least, if the software works the way they hope it will.

 

[The Von Matrices]

The part I find more interesting is that the 7950X3D has a 120W TDP compared to 170W on the regular version.

 

[Guwapo77]

How we honestly say if we like this decision or not, we don't have any benchmarks. To me, these chips look like they will kick in some face. No speed decreases either...I can't wait for you and Hardware Unboxed to bench these.

 

[katzi]

The design makes sense/is logical - for gaming purposes.

Games mostly being single or lightly threaded, so you would want to keep all those threads on the one CCD for latency's sake. So it Also makes sense to only have the X3D cache on 1 CCD, the one with the fastest/best cores. I imagine having it on Both would create latencies that would hinder game performance, but in Epyc scenarios it's used totally differently.

 

[izy]

How we honestly say if we like this decision or not, we don't have any benchmarks. To me, these chips look like they will kick in some face. No speed decreases either...I can't wait for you and Hardware Unboxed to bench these.

I think that the 3DV cache CCD will boost up to 5Ghz and the non 3DV cache CCD will boost like the normal chip, a bit misleading, i dont think it will bost higher than 5Ghz in games, maybe 5.2ghz? , we will have to wait and see.

 

[Lei]

So how much is 7800x3d? 400$ or 450$

Looks like 7800x3d can beat 13900ks

 

[izy]

It was great if 7800x3d had 4.4- 4.6ghz base and 5.2 -5.4 ghz boost, maybe they are unlocked and could be oced? I doubt.

 

[Eykxas]

The design makes sense/is logical - for gaming purposes.

Games mostly being single or lightly threaded, so you would want to keep all those threads on the one CCD for latency's sake. So it Also makes sense to only have the X3D cache on 1 CCD, the one with the fastest/best cores. I imagine having it on Both would create latencies that would hinder game performance, but in Epyc scenarios it's used totally differently.

No... Games have been multithreaded for a decade now. Back to 2008, most games from codemasters, activision etc... was already multithreaded. And from 2013/2014, almost every game is multithreaded.

 

[Hxx]

it will come down to pricing. the current cpu lineup from both teams is pretty amazing and relatively affordable so its gonna take more than just a few % increases in fps to incentivize potential buyers

 

[sLowEnd]

Hopefully there's no weird scheduler-related shenanigans resulting from this kind of design.

 

[Dirt Chip]

As it mainly serve gaming, no real need for more than 8 3d high cache cores.
You just need to proper allocated any game load to them.

 

[Lei]

Wish they could sell 7800x3d without stock cooler

 

[umeng2002]

If you look at the max 7800X3D Boost clock, it is way lower than the 79xxX3D max boost clocks. This means the 3D cache adds too much heat or power requirements for that one CCD to boost as high as the non-3D cache CCD.

So if you get the 79xxX3D chips, only the CCD without the cache will go to the advertised max boost. Maybe if one or two cores were lit up on the 3D CCD, it might get close though.

 

[Deleted member 185088]

Maybe to keep price in check, as AM5 is very expensive already. Also could be more cache won't help so much beyond a certain level.

 

[Chaitanya]

7900X3D and 7950X3D sound like AMD is experimenting with two diffrent CCDs on same CPU so basicaly they gonna do something similar than Intel did with Pcores+Ecores

Low risk experiment.

 

[Marax]

For content creation X3D cache have no significant impact (at least till today), but what have - lower clock speeds...
Pudget review 5800x vs 5800X3D

So this product will be target for gamers who make/render content on the same pc.
And it make sense to slap 3D cache on one die only, as games won't heavily use more than 8 cores. In practical use it happens to few of them normally, other 14 cores will be employed with low utilisation.

 

[The Von Matrices]

So if you get the 79xxX3D chips, only the CCD without the cache will go to the advertised max boost. Maybe if one or two cores were lit up on the 3D CCD, it might get close though.

That's already the case with both 7950X CPU's I've had experience with. Only one of the dies (cores 0-7) is good enough to hit the max boost and the other die (cores 8-15) is of poorer quality and tops out at about 200 MHz slower than max boost even if there is only one thread pinned to the second die. I notice that the scheduler rarely puts threads on the second die except when the first die is entirely loaded, so this X3D CPU shouldn't be that different in having dissimilar die.