AMD Ryzen 9 9900X in detail
We’re wrapping up tests of the new Ryzen 9000 desktop CPUs. This with a model that once increased the number of cores compared to its predecessor, or a pilot model (Ryzen 9 3900X) from eight to twelve on AMD’s mainstream platform. Even compared to the Ryzen 9 5900X, the new Ryzen 9 9900X is more than twice as fast in some practical situations. This model is also the best when it comes to efficiency.
AMD Ryzen 9 9900X in detail
The fourth Ryzen with 12 cores for mainstream platforms is built on the AMD Zen 5 architecture. While the first two models (R9 3900X and R9 5900X) still supported the AMD AM4 platform (and DDR4 memory), the Ryzen 9 9900X, like the older Ryzen 9 7900X, comes with support for the AMD AM5 platform and DDR5 memory.
The intergenerational difference is mainly in the architecture used (Zen 5 vs. Zen 4), while the basic structure of the processor remains unchanged. One of the three chiplets is even the same – the I/O chip is manufactured the same way (N5P) – what’s different (N4P) are the chiplets with the processor cores. In addition to the new architecture (Zen 5), there is a higher density of transistor integration (than in Zen 4) on a comparable area (approx. 71 mm²/chiplet). Cooling is to be improved, as indicated by the reported temperatures from the diagnostic tools. This should be due to the restructuring of the elements of the core with a view to reducing the concentration of high temperatures in one place. More details on this can be found in the subsection of our Ryzen 5 9600X test.
Of course, we don’t know if the situation with the reduced temperatures is as optimistic as the internal sensors report. In any case, the processor’s internal management works with these lower values and as a result, higher clock speeds of, for example, single-core boost are achieved. Its amount strictly depends on the temperature.
The processor cores are spread between two chiplets (CCD), symmetrically (6+6 cores). The claimed baseline clock speed of the all-core boost is 300 MHz lower for the Ryzen 9 9900X compared to the Ryzen 9 7900X. This is because of both lower TDP and PPT, which means the power limit should kick in sooner. The PPT of the Ryzen 9 9900X is the same as that of the Ryzen 9 7900X TDP – 170W. And the PPT, as you know, is 1.35 times the TDP. That’s roughly the ratio by which the Ryzen 9 9900X could/should be more power efficient. The overall power consumption also depends on the nature of the power management of the motherboard itself or the (in)efficiency of its VRM.
The list price of the Ryzen 9 9900X is lower by 50 USD (AMD’s suggested 499 USD) than it was for the Ryzen 9 7900X, but the newer model is naturally more expensive in stores. The discontinued Ryzen 9 7900X is often significantly cheaper, sometimes the difference in the final price is up to around 150 EUR.
Please note: The article continues in the following chapters.
Manufacturer | AMD | AMD | Intel | |
Line | Ryzen 9 | Ryzen 9 | Core i7 | |
SKU | 9900X | 7900X | 14700K | |
Codename | Granite Ridge | Raphael | Raptor Lake Refresh | |
CPU microarchitecture | Zen 5 | Zen 4 | Golden Cove (P) + Gracemont (E) | |
Manufacturing node | 4 nm + 6 nm | 5 nm + 6 nm | 7 nm („Intel 7 Ultra“) | |
Socket | AM5 | AM5 | LGA 1700 | |
Launch date | 08/15/2024 | 09/26/2022 | 10/17/2023 | |
Launch price | 499 USD | 549 USD | 409 USD | |
Core count | 12 | 12 | 8+12 | |
Thread count | 24 | 24 | 28 | |
Base frequency | 4.4 GHz | 4.7 GHz | 3.4 GHz (P)/2.5 GHz (E) | |
Max. Boost (1 core) | 5.6 GHz (5.62 GHz unofficially) | 5.6 GHz (5.75 GHz unofficially) | 5.6 GHz (P)/4.3 GHz (E) | |
Max. boost (all-core) | N/A | N/A | 5.5 GHz (P)/4.3 GHz (E) | |
Typ boostu | PB 2.0 | PB 2.0 | TBM 3.0 | |
L1i cache | 32 kB/core | 32 kB/core | 32 kB/core (P), 64 kB/core (E) | |
L1d cache | 48 kB/core | 32 kB/core | 48 kB/core (P), 32 kB/core (E) | |
L2 cache | 1 MB/core | 1 MB/core | 2 MB/core (P), 3× 4 MB/4 cores (E) | |
L3 cache | 2× 32 MB | 2× 32 MB | 1× 33 MB | |
TDP | 120 W | 170 W | 125 W | |
Max. power draw during boost | 170 W (PPT) | 230 W (PPT) | 253 W (PL2) | |
Overclocking support | Yes | Yes | Yes | |
Memory (RAM) support | DDR5-5600 | DDR5-5200 | DDR5-5600/DDR4-3200 | |
Memory channel count | 2× 64 bit | 2× 64 bit | 2× 64 bit | |
RAM bandwidth | 89.6 GB/s | 83.2 GB/s | 89.6 GB/s/51.2 GB/s | |
ECC RAM support | Yes (depends on motherboard support) | Yes (depends on motherboard support) | Yes (with vPro/W680) | |
PCI Express support | 5.0 | 5.0 | 5.0/4.0 | |
PCI Express lanes | ×16 + ×4 + ×4 | ×16 + ×4 + ×4 | ×16 (5.0) + ×4 (4.0) | |
Chipset downlink | PCIe 4.0 ×4 | PCIe 4.0 ×4 | DMI 4.0 ×8 | |
Chipset downlink bandwidth | 8.0 GB/s duplex | 8.0 GB/s duplex | 16.0 GB/s duplex | |
BCLK | 100 MHz | 100 MHz | 100 MHz | |
Die size | 2× 70.6 mm² + 118 mm² | 2× 71 mm² + 118 mm² | ~257 mm² | |
Transistor count | 8,16 + 3,37 mld. | 2× 6,57 + 3,37 bn. | ? mld. | |
TIM used under IHS | Solder | Solder | Solder | |
Boxed cooler in package | No | No | No | |
Instruction set extensions | SSE4.2, AVX2, FMA, SHA, VAES (256-bit), AVX-512, VNNI | SSE4.2, AVX2, FMA, SHA, VAES (256-bit), AVX-512, VNNI | SSE4.2, AVX2, FMA, SHA, VNNI (256-bit), GNA 3.0, VAES (256-bit), vPro | |
Virtualization | AMD-V, IOMMU, NPT | AMD-V, IOMMU, NPT | VT-x, VT-d, EPT | |
Integrated GPU | AMD Radeon | AMD Radeon | UHD 770 | |
GPU architecture | RDNA 2 | RDNA 2 | Xe LP (Gen. 12) | |
GPU: shader count | 128 | 128 | 256 | |
GPU: TMU count | 8 | 8 | 16 | |
GPU: ROP count | 4 | 4 | 8 | |
GPU frequency | 400–2200 MHz | 400–2200 MHz | 300–1600 MHz | |
Display outputs | DP 2.0, HDMI 2.1 | DP 2.0, HDMI 2.1 | DP 1.4a, HDMI 2.1 | |
Max. resolution | 3840 × 2160 px (60 Hz)? * | 3840 × 2160 px (60 Hz) | 7680 × 4320 (60 Hz) | |
HW video encode | HEVC, VP9 | HEVC, VP9 | HEVC, VP9 | |
HW video decode | AV1, HEVC, VP9 | AV1, HEVC, VP9 | AV1, HEVC, VP9 |
* We have no certainty on this parameter. AMD does not specify the maximum resolution and maximum refresh rate in publicly available materials. However, it is possible that it will be the same as for Ryzen 7000, i.e. 3840 × 2160 px (60 Hz).
- Contents
- AMD Ryzen 9 9900X in detail
- Methodology: performance tests
- Methodology: how we measure power draw
- Methodology: temperature and clock speed tests
- Test setup
- 3DMark
- Assassin’s Creed: Valhalla
- Borderlands 3
- Counter-Strike: GO
- Cyberpunk 2077
- DOOM Eternal
- F1 2020
- Metro Exodus
- Microsoft Flight Simulator
- Shadow of the Tomb Raider
- Total War Saga: Troy
- Overall gaming performance
- Gaming performance per euro
- PCMark and Geekbench
- Web performance
- 3D rendering: Cinebench, Blender, ...
- Video 1/2: Adobe Premiere Pro
- Video 2/2: DaVinci Resolve Studio
- Graphics effects: Adobe After Effects
- Video encoding
- Audio encoding
- Broadcasting (OBS and Xsplit)
- Photos 1/2: Adobe Photoshop and Lightroom
- Photos 2/2: Affinity Photo, Topaz Labs AI Apps, ZPS X, ...
- (De)compression
- (De)encryption
- Numerical computing
- Simulations
- Memory and cache tests
- Processor power draw curve
- Average processor power draw
- Performance per watt
- Achieved CPU clock speed
- CPU temperature
- Conclusion
I am wondering what is going on with certain tests where 9900x is inferior to 7900x. In particular, AIDA64 FPU Julia and Mandel. For instance, in Julia test 9900x only achieves 28% of 7900x’s performance…!? Does it have something to do with test versions (7900x was tested almost 2 years ago).
This also makes me to ask if there is a way to tell which cases are actually comparable? Is it possible to select only the exact same versions?
I assume that this is a consequence of non-optimization of the application (in this case Aida64) for given tests with specific processors, or rather their architectures.