According to the release plan on Computex 2022, AMD officially released four CPUs based on the Zen 4 architecture today, positioned as desktop enthusiast level, the flagship model is 16 cores and 32 threads Ryzen 9 7950X, with a maximum frequency of 5.7GHz, with outrageous performance, and is confirmed to be officially released on September 27. This means that AMD will win back the CPU performance champion in gaming and content creation in the PC field in a short period of time. The biggest change in the Zen 4 architecture is that the CPU process is transferred to TSMC's N5 process, and IPC has increased by 13% compared to the Zen 3 architecture, higher than the 11% increase promoted on Computex. The CPU frequency has been greatly improved, with the highest frequency gradually increasing from 4.9GHz of Ryzen 9 5950X to 5.7GHz of Ryzen 9 7950X, with a frequency increase of up to 16%, a single thread is expected to increase by 29%, and the performance improvement of multi-threaded will be even more.
paired with the Ryzen 7000 series is the AMD AM5 platform. The new platform will become the basic platform of consumer-grade product lines before 2025. AM5 has introduced support for DDR5 and PCIe 5.0, using a brand new LGA slot. According to the plan, four chipsets will be launched, namely X670 Extreme, X670, B660 Extreme and B660. The X670 series will be launched in September and the B660 series will be launched in October.
In terms of core count, Ryzen 7000, like Zen 3, maintains the configuration of up to 16 cores and 32 threads. According to positioning, 12, 8 and 6 CPU cores will be allocated downwards. AMD will use up to 2 Zen 4 core chiplet CCD to build processors, each chiplet contains 8 Zen 4 CPU cores. The specific division is as follows:
By the way, although the Ryzen 7000 series are equipped with iGPUs, the details of the iGPU specifications have not been disclosed so far, so the details of the RDNA2 architecture iGPU cannot be determined. However, for the enthusiast-grade high-performance desktop CPU, everyone basically doesn't care.
Next, let's interpret the details of the Ryzen 7000 series.
The fastest CPU in history: Ryzen 9 7950X
Ryzen 9 7950X is the complete body of the Ryzen 7000 series. The two small chip CCDs are fully enabled, and a total of 16 CPU cores are provided. Thanks to TSMC's N5 process, although the basic frequency is only 4.5GHz, the maximum frequency of a single thread can accelerate 5.7GHz. Obviously, the 5.5GHz displayed by AMD on Compute 2022 is still a handful.
Ryzen 9 7950X has 80MB cache, each chip small CCD is equipped with 32MB L3 cache, 2 CCDs add up to a total of 64MB, and each Zen 4 core is equipped with 1MB L2 cache. Compared with the Ryzen 5000, the L3 cache capacity has not changed, but the L2 cache has doubled.
This processor will also be the most power-consuming CPU in AMD's history. TDP is up to 170W, and the chip's power package tracking (PPT) limit is 230W. The power consumption increase caused by multi-threading performance is one of the problems. Previously, the Ryzen 5000 series received TDP limit due to its power consumption, so it could not obtain higher frequencies.
all performance will mean that the Ryzen 7950X will outperform the Intel Core i9-12900K, becoming the most powerful CPU in the world, and will compete with the upcoming Raptor Lake CPU.
The point is that the price of this flagship chip is $699, equivalent to RMB 4,836, which is $100 lower than the Ryzen 5950X of that year. Such pricing strategy may be based on TSMC's destocking warning and the general market environment. Ryzen 9 7900X
Ryzen 9 7900X with the core of
Ryzen 9 7900X also uses a dual CCD configuration, but only 6 cores are enabled for each CCD, which ultimately constitutes 12 cores and 24 threads, but retains the 64MB L3 cache of Ryzen 9 7950X, which is the same as the previous generation Ryzen 5000 series.
In terms of power consumption, the Ryzen 9 7900X has 170W TDP power just like the 7950X. At the same time, the basic frequency is 200MHz higher than the 7950X, reaching 4.7GHz, but the maximum power is 5.6GHz, which is different from the 7950X flagship.
Ryzen 9 7900X is priced at US$549, equivalent to RMB 3798, and the price is the same as 5900X.
8 core 7700X and 6 core 7600X
Next is Ryzen 7 7700X, the positioning is the same as the previous generation 5800X, and only 1 complete CCD is used, with a basic frequency of 4.5GHz and a maximum frequency of 5.4GHz.Since there is only one CCD, the L3 cache dropped to 32MB, plus the 8MB L2 cache in the kernel, reaching the advertised 40MB cache.
Ryzen 7 7700X's TDP is the traditional 105W, and it should be better in terms of heat dissipation. The price is US$399, equivalent to RMB 2,760.
Then there is the 6 core Ryzen 5 7600X, 6C12T configuration, that is, 2 of the cores are disabled in a single Zen 4 CCD. The basic frequency of the 7600X is maintained at 4.7GHz and the highest frequency is 5.3GHz. Although it is the lowest frequency among the released products, the frequency can still be hardened by the Core i9-12900K. Therefore, AMD also boldly stated that the average speed performance of the 7600X is 5% faster than that of the 12900K.
However, the price is that the power of the Ryzen 5 7600X also starts from 105W, rather than the previous 65W, so high-end cooling and power supply this year are unavoidable.
Zen 4's counterattack: 13% IPC improves
Ryzen 7000's excellent performance, of course, is attributed to the excellent architecture of AMD Zen 4. Although the press conference is not long, CTO Mark Papermaster still provides technical details related to Zen 4.
For example, AMD has modified its Zen 4 architecture IPC estimate, and by adjusting the circuit, chip and firmware, the average performance is improved by 13%. According to AMD, if the frequency is locked at 4GHz, the performance gain will be between 1% and 39% depending on the workload, depending on the actual workload, so on average, it is close to 13% of performance.
At the same time, AMD also provides an IPC breakdown diagram to analyze how the Zen 4 architecture brings IPC benefits. The biggest contribution is the front part of the new design, followed by improvements to loading and storage, followed by branch predictors, execution engines, and larger L2 caches that all improve IPC performance. In particular, the execution engine cache has achieved a 50% increase.
Zen 4 also confirms that AVX-512 instruction set is supported, but it does not explicitly indicate which AVX-512 subset is supported. It can only be determined that AVX-512 will be used in the vector neural network instruction VNNI. This operation method is completely different from Intel . Intel built a 512-bit SIMD to execute AVX-512 instructions, while AMD used 2 cycles for power considerations, 256-bit to implement AVX-512 instructions.
This means that AMD does not need to add too much transistor to AVX-512. Although it will be slower than Intel in terms of efficiency, reasonably, most of the time, most people don’t have many opportunities to use AVX-512.
Then don’t forget that TSMC N5 has had a significant impact on the AMD Ryzen 7000 series. Before that, the Zen 2 of the Ryzen 3000 and the Zen 3 of the Ryzen 5000 both used the TSMC 7nm process, and the service time has exceeded 3 years. This transformation will undoubtedly bring a significant leap to the CPU.
TSMC N5 advantages are throughout every detail of the Ryzen 7000 series. For example, the clock frequency of has increased significantly, the entire staff has exceeded 5.3GHz, IPC has gained 13% gain, and a single thread has increased by 29%, etc. The new
new process also makes the actual area of the 1MB L2 cache smaller than the 512KB L2 cache on Zen 3. The overall area of the AMD Zen4 core has also been reduced by 18%, almost half of the Golden Cove core of the Intel 7 process, and the cost has been further reduced. Given that TSMC N5 is better than Intel 7 processes, AMD undoubtedly gained the bonus of the process process.
Performance link
Next is the favorite performance link. Compared with the Zen 3 architecture, the Zen 4 improvement of the Ryzen 7000 is very comprehensive. In AMD's first-party test, AMD quoted three different TDP data: 65W, 105W and 170W. At the lowest TDP, the 7950X's Cinebench R23 multi-threading performance improvement was the highest, reaching 74%, followed by a 37% increase of 105W and a 35% increase of 170W. Under the TSMC N5 process, AMD will pay more attention to showing the benefits of CPU at the same power consumption.
Overall, Zen 4 can provide up to 49% performance improvement compared to Zen 3, and at the same performance level, the energy consumption of 7950X is 62% lower than that of 5950X.
vs. Core i9-12900K, the Ryzen 9 7950X has a 57% rendering performance in V-Ray Benchmark 5.x and a 47% performance per watt. The span is very large.
Even if it is a 6-core 7600X, it can be 5% higher than 12900K in the game load.
According to AMD's plan, the first batch of Zen 4 products will be officially launched on September 27, and AM5 of the same period will also be launched. From the current perspective, the Ryzen 7000 chip supply is sufficient, and AMD has invested in more wafer and substrate products to protect the new products.
AMD will also release a DDR5 overclocking memory set optimized for the Ryzen 7000 series, with built-in EXPO overclocking extended configuration, corresponding to Intel XMP, and a speed of DDR5-6400.
After the Ryzen 5000 series encountered the epidemic and chip out of stock, we were lucky enough to see AMD engineers ushering in the Jedi counterattack in the Ryzen 7000 series. For enthusiasts, especially high-end PC gamers, being able to buy higher performance CPUs and GPU at the original price or even discount is undoubtedly a good start.
