This is a review of the AMD’s current top-of-the-line CPU, the Ryzen 9 7950X3D. We also discussed the difference between X3D and non-X3D processor. Thank you AMD for lending us a review unit for a few weeks. We were able to collect enough data to create this review but they didn’t touch anything we say in this review.
Table of Contents
Specification Sheet
Here’s an overview of the current processors in the Ryzen 9 7000 Series Lineup.
Processor | 7950X3D | |
Cores / Threads | 16/32 | |
TDP | 120W | |
Max. Boost Clock | 5.7GHz | |
Base Clock | 4.2GHz | |
Included Cooler | None | |
Max Temp. (Tjmax) | 89°C | |
L3 Cache | 128 MB | |
Price PCHub (Php) | 39,995 | |
Price DataBlitz (Php) | ⠀⠀⠀ | 39,990 |
Unboxing
Inside the box is the CPU, an AMD Ryzen 9 7000 Series sticker badge, an AMD Red Team card, and the warranty document. Filling up the remainder of the space is a solid foam.
The box of the Ryzen 9 7950X3D is very big when you consider that there is no included cooling solution.
What is X3D?
The abbreviation of the X3D is Extended 3D Technology (EXT) feature which we saw first last year. The Ryzen 7 5800X3D is the first X3D model and it become the best gaming CPU in the market.
What makes it better compared to the non-X3D version is that it has more L3 cache that it has. How they did it is that they stacked an additional layer of cache on top of the existing L3 cache which is called 3D V-cache technology and AMD named it Ryzen X3D for its CPUs that have that technology.
It stores frequently used data and instructions, making it quicker for the CPUs to access them, which improves overall system performance. L3 cache helps reduce the time the CPU spends waiting for data from the main memory, making tasks run faster and smoother.
Getting more L3 cache improves its performance in intensive applications like multitasking, data-intensive tasks, and gaming are some of the examples that will have significant benefits.
Testing Methodology
Parts Used:
Motherboard | ASRock B650M PG Riptide WiFi (BIOS 1.21) | |
Cooler | Thermalright Peerless Assassin 120 SE | |
RAM | 2x Patriot Signature Premium 8GB 5200 CL42 | |
PSU | Seasonic Prime Gold 1200W | |
GPU | Zotac RTX3060Ti 8GB Dual OC LHR | |
OS | ⠀⠀⠀ | Windows 10 22H2 |
Note: All of the items used aside from the CPU were personally bought.
Software Used:
HWInfo
Benchmate 10.12.2
Blender Benchmark Launcher 3.1.0
PCMark 10 (Steam)
Apex Legends
Civilization VI
CS:GO
Cyberpunk 2077
Overwatch 2
Shadow of the Tomb Raider
Valorant
Metrics Observation
In the motherboard, the profile used was PBO Enabled with 85C TjMax. SOC Voltage including everything was set to Auto. The memory did not have EXPO so XMP was used for 5200 CL40. For the RAM OC, this was set manually. The exact timings were determined through trial and error. The final timing that was stable was 6000 40-40-40-32-96 1T.
Here are the readings on an all-core run:
CPU (Tctl/Tdie) Temperature | 86.5°C | |
CPU Die (Average) Temperature | 85.0°C | |
CPU CCD1 (Tdie) Temperature | 79.3°C | |
CPU CCD2 (Tdie) Temperature | 68.4°C | |
CPU Package Power | 125 W | |
Bus Clock Speed | 99.8 MHz | |
Core 0-7 Clock Speed | 5.125 GHz | |
Core 8-15 Clock Speed | 5.540 GHz | |
Core VID Voltage | 1.225 V | |
CPU VDDCR_VDD Voltage | 1.236 V | |
CPU VDDCR_SOC Voltage | 0.994 V | |
CPU VDD_MISC Voltage | ⠀⠀⠀ | 1.100 V |
Something to note is that the room temp is at 32-34°C over the course of all the runs. To note, Core 0-7 is CCD1, the CCD with the 3D V-Cache, meanwhile Core 8-15 is CCD2, the CCD without 3D V-Cache. We can see that the temperature of the CCD1 is higher compared to the CCD2. The clock speed of the cores on the CCD with the 3D V-Cache are also 415 Mhz slightly lower than without the 3D V-Cache.
Manual overclocking is disabled on this CPU on the ASRock motherboard and this might be a concern for some users who want to squeeze more performance from the CPU. With CCD1 being warmer as well, the CCD2 can’t boost more even while at 68°C as the CPU die will go over 85°C TjMax limit set by PBO on our motherboard.
Benchmarks
Here is the comparison of the DDR5-5200 CL42 and DDR5-6000 CL40 for the 7950X3D for synthetic benchmarks.
Here’s an overview of all the synthetic benchmarks tested.
We can see that a lot of the performance is not utilized when running a slower set of memory.
Here is the comparison of the DDR5-5200 CL42 and DDR5-6000 CL40 for the 7950X3D for gaming benchmarks.
Here’s an overview of all the gaming benchmark for comparing the two RAM speeds.
In gaming, average FPS improves in almost all of the games with the 1% low showing more improvement. Better 1% low is very important as it tends to show better consistency and reduced perceived lag.
Shadow of the Tomb Raider’s built-in benchmark shows data for both CPU and GPU.
The graph clearly shows that the CPU can clearly run much faster GPUs than the RTX3060Ti that was tested.
AMD’s 7000 Series processors, different to most previous higher-end CPUs, all have integrated graphics. Here are some benchmarks for the integrated graphics incase anyone wants to play on it.
Ryzen 9 7950X3D Conclusion
The Ryzen 9 7950X3D is a really fast CPU if you are able to supplement it with a fast RAM. For tasks that takes advantage more of the additional L3 Cache than higher clock speed, the 7950X3D will be for you. Most tasks with this priority tend to mostly be gaming. The integrated GPU is usable on some games, averaging more than 60 fps. For an additional Php 5,000 over the 7950X for the additional L3 cache, this CPU is not for everyone.
Currently, as of writing the price of the AMD Ryzen 9 7950X3D is priced at Php 39,950 in some online stores we saw, and for non-X3D is priced at Php 34,450.
Notes/Disclaimers
- Benchmarks are currently limited to only this CPU as the team is yet to benchmark other systems
- 5200 CL40 and 6000 CL40 are both relatively slow compared to most DDR5 memory, so more performance can be attained with other faster memory kits
- The RTX3060Ti is not the fastest GPU currently available, so you should be able to get more fps in gaming scenarios
- Some games only have average fps data due to our current limitation in data collection