Thursday, January 5th 2023
AMD Confirms Ryzen 9 7950X3D and 7900X3D Feature 3DV Cache on Only One of the Two Chiplets
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.
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.
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.
164 Comments on AMD Confirms Ryzen 9 7950X3D and 7900X3D Feature 3DV Cache on Only One of the Two Chiplets
7000X
7000X + 3DCache = 7000X3D
The only way in which (example) 7950X3D can consume less or like 7950X is to reduce the frequency.
This reality is officially confirmed!
7950X: 4.5GHz base clock
7950X3D: 4.2GHz base clock (300MHz less)
7900X: 4.7GHz base clock
7900X3D: 4.4GHz base clock (300MHz less)
What we don't know, yet, is how much amd will push the limits of these processors. Only the reviews will show us the real consumption in applications, but it is certain that X3D will be weaker in all applications that do not react to the extra cache memory, such as rendering and encoding.
I don't know if it has been discussed until now, but it is a problem with the AM5 platform if you want to use all the memory slots.
In the video: 3090Ti with i7 860 "13 year old". With a video card from 13 years ago + the most powerful processor in gaming in 2023, you can't play anything even in 1080p.
But I'm sure many of you will blame this when the X3D models fail to live up to the extreme hype, even though having extra L3 on both CCDs would result in the same problem with inconsistent latencies across L3. (when in reality many have shown unrealistic expectations)I don't think there are many (or any) AAA PC-games for the past 10 years that have been single threaded, it's actually not possible to obtain rendering that's independent on IO and simulation without at least two threads, which is why this has been common since the early 2000s. But this doesn't mean all threads have significant load though, so this may be the source of your confusion?
Still, games do use relatively "few" threads if that's what you mean, but they still need top performance and low latency between certain threads. This is more important than whether the extra L3 is on the same die or not, but even core-to-core latency variances doesn't matter a whole lot.
The price would probably go up by another $100 or so, for no actual benefits.
Scheduling is a problem, but in a worst case scenario, you can always assign affinity manually.
The problems for the user are a bit more delicate, when choosing an X3D instead of an X, the following must be taken into account:
1. Higher price compared to non-X3D brothers
2. Less aggressive frequency curve (altered performance in many applications that are not affected by extracache)
3. Only 89 degree Tmax.
4. Very limited overclocking (manual overclocking is impossible).
5. How much it helps me, practically speaking.
In short, a lot of compromise.
CCD A = x cores, x cache
CCD B = y cores, y cache
Because I can't see a way it works simply (the CCD's are either both 6 core or 1x 8 and 1x 4)
Sounds like a 7600X3D is definitely an option now (or should be)
Wonder if the 25% ish more cache per core at the same boost clock will help
A full CCX 8 cores with 3D cache + a CCX with 4 (7900X3D) and 8 cores (7950X).
For 7600X3D the price is not justified. More expensive than 13600K, with some results in gaming, but effectively destroyed in the other applications. Let's not forget that, except for gaming, a possible 7600X3D will perform below the 7600X in other applications.
www.phoronix.com/review/amd-5800x3d-linux6/4
inb4: This doesn't count, since it doesn't fit your workload useage...
Puget System:
Overall, the Ryzen 5800X3D is a great proof of concept product for using 3D V-Cache on desktop CPUs, but we would recommend following AMD's advertising of the product, and using it primarily for gaming. For content creation, it can be on par with the standard 5800X in many cases, but in others, you could encounter a decent drop in performance depending on the application you are using
Also, result for 5800X vs 3D same thing, it's another product in their sku though. Except for gaming the 5800X3D has limited use cases
By the way thank you for the Puget Systems assessment of what every owner of a X3D chip already knows. What they should have added is you will have so much fun Gaming and pushing your 1% lows that it will be a fading memory that the 5800X3D is not as fast in productivity vs the non cache brother or sister.
Have you seen AMD's new laptop lineup. 30 hour battery life is nothing to sneeze at for business purposes. Especially given the fact that a large portion of the Work force has had to buy new tech recently and the motto in business is as cheap as possible. That meant that when we transitioned to working from home everyone in my Company got laptops. Guess what kind of laptops they were? Yep that's right AMD, because they have better APUs period. With the efficiency that AMD has achieved it will bode them well for Corporate customers going forward.
Did you mention manual OC and AM5 together (for a regular user)? If you draw back, all AMD has done is increased their stack. If you don't want a X3D chip you can still enjoy AM5. If you don't want a X chip you can get a non X chip and be happy. There will probably be refreshes on the MBs for both launches too like B620 with non X chips and a revised B650 and X670 lineup but the X3D but the current lineup is also fine. More boards means more choice, what it also hopefully means is that boards will come back to reality in pricing .
lmao thank you for proving my point:Those are not (potential, unlikely) 7600X3D reviews.... did you have those....?
And even in some of puget's application, the 3DX is performing higher.
This is what you said sir. My highlightThis is your words. You are the one that claims it will perform bellow. Not that it will be mixed or application dependent. but in non-gaming workload it will perform bellow.
W1zzard's review also have 3DX sometimes being better than non-3D in gaming workload.
and workload doesn't need to be rendering or photoshop
www.igorslab.de/en/amd-ryzen-7-5800x3d-against-the-rest-workstation-test-with-some-shadow-but-also-some-light/2/
By the way, on the next page (Inventor) the 5800X3D loses on all fronts to the 5800X. And their conclusion is: "Those who rather want to work should therefore better keep their hands off the Ryzen 7 5800X3D in most cases, because what is offered is simply too little in relation to the other products.".
PS: 5800X3D costs as much as 5900X or 13600K and most likely a 7800X3D will cost as much as a 7900X. I wish you success in gaming for a lot of money, gentlemen.