Wednesday, January 4th 2023
AMD Ryzen 7000X3D Announced, Claims Total Dominance over Intel "Raptor Lake," Upcoming i9-13900KS Deterred
AMD today announced its Ryzen 7000X3D "Zen 4" desktop processors with 3D Vertical Cache technology. With these, the company is claiming to have the world's fastest processors for gaming. The company claims to have beaten the Intel Core i9-13900K "Raptor Lake" in gaming, by a margin it feels comfortable to remain competitive with against even the upcoming Core i9-13900KS. At the heart of these processors is the new "Zen 4" 3D Vertical Cache (3DV cache) CCD, which features 64 MB of L3 cache stacked on top of the region of the "Zen 4" CCD that has the on-die 32 MB L3 cache. The 3DV cache runs at the same speed as the on-die L3 cache, and is contiguous with it. The CPU cores see 96 MB of transparent addressable L3 cache.
3DV cache is proven to have a profound impact on gaming performance with the Ryzen 7 5800X3D "Zen 3" processor that helped it beat "Alder Lake" in gaming workloads despite "Zen 3" being a generationally older microarchitecture; and AMD claims to have repeated this magic with the 7000X3D "Zen 4" series, enabling it to beat Intel "Raptor Lake." Unlike with the 5800X3D, AMD don't intend to make gaming performance a trade-off for multi-threaded creator performance, and so it is introducing even 12-core and 16-core SKUs, so you get gaming performance alongside plenty of muscle for creator workloads.
The series consists of three SKUs, the 8-core/16-thread Ryzen 7 7800X3D, the 12-core/24-thread Ryzen 9 7900X3D, and the flagship 16-core/32-thread Ryzen 9 7950X3D. The 7800X3D comes with an unknown base frequency above the 4.00 GHz-mark, along with up to 5.00 GHz boost. The 7900X3D has 4.40 GHz base frequency, and up to 5.60 GHz boost. The flagship 7950X3D ticks at 4.20 GHz base, and boosts up to 5.70 GHz.
There's something interesting about the cache setup of the three SKUs. The 7800X3D has 104 MB of total cache (L2+L3), whereas the 7900X3D has 140 MB and the 7950X3D has 144 MB. The 8-core CCD in the 7800X3D has 64 MB of 3DV cache stacked on top of the 32 MB on-die L3 cache, resulting in 96 MB of L3 cache, and with each of the 8 cores having 1 MB of L2 cache, we arrive at 104 MB total cache. Logically, the 7900X3D and 7950X3D should have 204-208 MB of total cache, but they don't.
While we await more details from AMD on what's happening here, there are two theories—one holds that the 3DV cache for the 7900X3D and 7950X3D is just 32 MB per chiplet, or 64 MB L3 cache per CCD. 140 MB total cache for the 7900X3D would hence come from ((2 x 64 MB L3) + (12 x 1 MB L2)); and for the 7950X3D this would be ((2 x 64 MB L3) + (16 x 1 MB L2)).
The second more radical theory holds that only one of the two CCDs has 64 MB of 3DV cache stacked on top of the on-die 32 MB L3 cache, and the other is a conventional "Zen 4" CCD with just 32 MB of on-die L3 cache. The math checks out. Dating all the way back to the Ryzen 3000 "Zen 2" Matisse dual-CCD processors, AMD has worked with Microsoft to optimize Windows 10 and Windows 11 schedulers to localize gaming workloads to one of the two CCDs (using methods such as CPPC2 preferred-core flagging), so if these processors indeed have an asymmetric L3 cache setup between the two CCDs, the one with the 3DV cache would be preferred by the OS for gaming workloads.
In its presentation, AMD uses the term "the world's best gaming processor" with the 7800X3D and not the 7950X3D. This should mean that despite its lower maximum boost frequency, the 7800X3D should offer the best gaming performance among the three SKUs, and very likely features 96 MB of L3 cache for the CCD; whereas the 7900X3D and 7950X3D feature either lower amounts of 3DV cache per CCD, or that asymmetric L3 cache setup we theorized.
In terms of performance, AMD is claiming anywhere between 21% to 30% gaming performance gains for the 7800X3D over the previous-generation 5800X3D. This can be associated with the IPC increase of the "Zen 4" core, and faster DDR5 memory. AMD claims that the 7800X3D should particularly shine with CPU-limited gaming scenarios, such as lower-resolution high refresh-rate setups.
The 7950X3D is claimed to beat the Core i9-13900K in gaming performance by anywhere between 13% to 24% in the four tests AMD showed, while also offering big gains in multi-threaded productivity benchmarks. Especially in workloads involving large streaming data, such as file-compression and DaVinci Resolve, the 7950X3D is shown offering between 24% to 52% performance leads over the i9-13900K (which we doubt the i9-13900KS can make up for).
The Ryzen 7000X3D processors will be available from February 2023, and should be drop-in compatible with existing Socket AM5 motherboards, with some boards requiring a BIOS update. The USB BIOS Flashback feature is standardized by AMD across motherboard brands, so this shouldn't be a problem.
3DV cache is proven to have a profound impact on gaming performance with the Ryzen 7 5800X3D "Zen 3" processor that helped it beat "Alder Lake" in gaming workloads despite "Zen 3" being a generationally older microarchitecture; and AMD claims to have repeated this magic with the 7000X3D "Zen 4" series, enabling it to beat Intel "Raptor Lake." Unlike with the 5800X3D, AMD don't intend to make gaming performance a trade-off for multi-threaded creator performance, and so it is introducing even 12-core and 16-core SKUs, so you get gaming performance alongside plenty of muscle for creator workloads.
There's something interesting about the cache setup of the three SKUs. The 7800X3D has 104 MB of total cache (L2+L3), whereas the 7900X3D has 140 MB and the 7950X3D has 144 MB. The 8-core CCD in the 7800X3D has 64 MB of 3DV cache stacked on top of the 32 MB on-die L3 cache, resulting in 96 MB of L3 cache, and with each of the 8 cores having 1 MB of L2 cache, we arrive at 104 MB total cache. Logically, the 7900X3D and 7950X3D should have 204-208 MB of total cache, but they don't.
While we await more details from AMD on what's happening here, there are two theories—one holds that the 3DV cache for the 7900X3D and 7950X3D is just 32 MB per chiplet, or 64 MB L3 cache per CCD. 140 MB total cache for the 7900X3D would hence come from ((2 x 64 MB L3) + (12 x 1 MB L2)); and for the 7950X3D this would be ((2 x 64 MB L3) + (16 x 1 MB L2)).
The second more radical theory holds that only one of the two CCDs has 64 MB of 3DV cache stacked on top of the on-die 32 MB L3 cache, and the other is a conventional "Zen 4" CCD with just 32 MB of on-die L3 cache. The math checks out. Dating all the way back to the Ryzen 3000 "Zen 2" Matisse dual-CCD processors, AMD has worked with Microsoft to optimize Windows 10 and Windows 11 schedulers to localize gaming workloads to one of the two CCDs (using methods such as CPPC2 preferred-core flagging), so if these processors indeed have an asymmetric L3 cache setup between the two CCDs, the one with the 3DV cache would be preferred by the OS for gaming workloads.
In its presentation, AMD uses the term "the world's best gaming processor" with the 7800X3D and not the 7950X3D. This should mean that despite its lower maximum boost frequency, the 7800X3D should offer the best gaming performance among the three SKUs, and very likely features 96 MB of L3 cache for the CCD; whereas the 7900X3D and 7950X3D feature either lower amounts of 3DV cache per CCD, or that asymmetric L3 cache setup we theorized.
The 7950X3D is claimed to beat the Core i9-13900K in gaming performance by anywhere between 13% to 24% in the four tests AMD showed, while also offering big gains in multi-threaded productivity benchmarks. Especially in workloads involving large streaming data, such as file-compression and DaVinci Resolve, the 7950X3D is shown offering between 24% to 52% performance leads over the i9-13900K (which we doubt the i9-13900KS can make up for).
The Ryzen 7000X3D processors will be available from February 2023, and should be drop-in compatible with existing Socket AM5 motherboards, with some boards requiring a BIOS update. The USB BIOS Flashback feature is standardized by AMD across motherboard brands, so this shouldn't be a problem.
177 Comments on AMD Ryzen 7000X3D Announced, Claims Total Dominance over Intel "Raptor Lake," Upcoming i9-13900KS Deterred
Edit: fixed grammar
Or maybe single threaded thermal density hasn't changed all that much from regular Zen 3/4, and allows similar ST temps and per-core power? We never got a chance to properly find out on 5800X3D due to the enforced Fmax limit, but even under very heavy ST loads such as single core y-cruncher (e.g. BBT) the temp scaling based on per-core power doesn't look all that different.
Whereas MT thermal behaviour obviously looks very different from normal Zen 3. It's only when whole-CCD power draw reaches a certain point (closer to 100W cores power) that the 5800X3D becomes clearly overwhelmed and temps start scaling exponentially, and ST boost would never come close to that. I would assume that regular Zen 4 is more of the same right now.To be fair, who knows what power will actually look like. We both know the 5800X3D doesn't actually use all of its PPT like the other CPUs do, maybe it's just 162W in name.
I just hope that AMD doesn’t completely lockout the BIOS again. PBO2 was great for Windows users, but I would have been out of luck without KomboStrike using Linux.
www.overclock.net/threads/5800x3d-owners.1798046/post-29008800
iirc (don't quote me, I don't exactly remember the block diagram for voltage rails) as of Raphael there is still no decoupling of Vcache from Vcore. But on the flip side, there also shouldn't be a decoupling of CCD1 Vcore from CCD2 Vcore - judging from the high ST clocks of 7950X3D, maybe AMD is confident enough in the longevity of Vcache not to artificially constrain it to 1.35V anymore?
Though......assuming AMD didn't pull magic out of a hat and reinvent how CPPC works, the 7800X3D's clocks would just be market segmentation. Even +200 wouldn't do much in terms of pure ST performance.
Though, the initial skatterbencher writeup did say something interesting:However, at the same time skatterbencher also claims that Vcache is permanently bound to VDDM with no option of bypass like Vcore has, and VDDM can't be controlled. So while independent cache clock might already be a thing on 5800X3D, independent cache volts is not happening.
There is a segment where they try to prove that the dLDOs enable per-core voltage when below 1.25V, but I'm unsure what to think of it as they're trying to say that per-core VID (which can vary per core, not news) = per-core Vcore, which traditionally we know not to be true
I was hoping AMD would have surprised all of us with a 7600X3D. Not feeling the 7800X3D considering its definitely gonna be pitched for $450+
I get no fps drops even in Limsa with my 13900K at 4K.Yes, yes it is. The 13900K dominates all Ryzen 7000 chips at 4K (and all others) with a 4090.
MEH.
Did you watch the Live Show at all, or just go based off tech threads?
very small part of it was for gaming. AMD really did a great job tonight, Lisa Su was fantastic.