AMD’s “together we advance PCs”l Livestream presentation has just concluded, with AMD CEO Dr. Lisa Su laying the groundwork for the release of the next generation of AMD Ryzen desktop CPUs. Based on AMD’s Ryzen 7000 announcement from Computex 2022, the eagerly awaited presentation detailed AMD’s launch plans for their first family of Zen 4-architecture-based CPUs, which will begin with four enthusiast-oriented chips. AMD’s Ryzen 7000 chips will debut on September 27th, topped by the 16-core Ryzen 9 7950X, with the company expecting to reclaim the performance crown in virtually all PC CPU categories, including gaming and content creation.
The latest generation of AMD’s desktop CPUs is propelled by a combination of architectural enhancements underlying the Zen 4 architecture and the production of CPU core chiplets on TSMC’s cutting-edge 5nm process. This combination will enable AMD to deliver a 13% increase in IPC over its Zen 3 architecture, up from an 11% increase claimed at Computex, as well as a significant increase in CPU clock speeds. The top-tier Ryzen 9 7950X will have a maximum turbo frequency of 5.7GHz, which is 800MHz (16%) faster than the equivalent Ryzen 9 5950X. AMD anticipates a generational increase of 29% in single-threaded performance and even more in multi-threaded workloads.
Launching alongside the new Ryzen 7000 series chips is AMD’s AM5 platform, which will serve as the foundation of AMD’s consumer desktop platform until at least 2025. AM5 introduces DDR5 and PCIe 5.0 support as well as a new LGA socket for AMD’s chips. Four chipsets are currently planned: X670 Extreme, X670, B660 Extreme, and B660. X670 series boards will be available in September, with B660 series boards to follow in October.
Ryzen 7000 Family: 4 Zen 4 SKUs
AMD’s product stack for the launch next month should look very familiar to regular AnandTech readers and AMD enthusiasts. The Ryzen 7000 generation continues AMD’s strategy of launching enthusiast-oriented desktop processors based on a group of four high-end processors, which has served the company well in the past.
|AMD Ryzen 7000 Series Processors
Zen 4 Microarchitecture
|Ryzen 9 7950X||16C / 32T||4.5GHz||5.7GHz||16 MB||64 MB||170 W||$699|
|Ryzen 9 7900X||12C / 24T||4.7GHz||5.6GHz||12 MB||64 MB||170 W||$549|
|Ryzen 7 7700X||8C / 16T||4.5GHz||5.4GHz||8 MB||32 MB||105 W||$399|
|Ryzen 5 7600X||6C / 12T||4.7GHz||5.3GHz||6 MB||32 MB||105 W||$299|
As established by AMD at Computex, the maximum number of cores for Ryzen 7000 processors is 16. Therefore, across AMD’s product line, the core counts for this generation are the same as for the previous generation. The top-tier AMD product will feature 16 CPU cores, followed by 12, 8, and finally 6 CPU cores. As in the past, AMD constructs its chips with up to two Zen 4 CPU core chiplets (CCDs), each containing eight Zen 4 CPU cores.
As an aside, AMD has not disclosed any additional information regarding the GPU’s integrated RDNA2 architecture at this time. Even though all Ryzen 7000 processors include an integrated graphics processing unit (iGPU), we do not yet have any information regarding its specifications. AMD has made it abundantly clear that the iGPU for these desktop chips is a relatively small configuration designed for basic desktop tasks and is not intended to be a high-performance GPU like the GPUs on AMD’s APUs.
The Ryzen 9 7950X is the world’s fastest processor
The leading processor is the Ryzen 9 7950X. The top-tier implementation of AMD’s Zen 4 architecture will utilize two fully enabled CCDs to provide a total of 16 CPU cores. Due in large part to TSMC’s 5nm process, it also features extremely high clock speeds; the base clock speed is 4.5GHz, and the turbo clock speed can reach up to 5.7GHz for a single thread. The latter is 200MHz faster than AMD’s presentation at Computex, where the company only reached 5.5GHz.
This chip is advertised by AMD as having an 80MB cache. This corresponds to 64MB of L3 cache, 32MB on each CCD, and 1MB of L2 cache for each Zen 4 CPU core. This is the same amount of L3 cache as the Ryzen 5000/Zen 3 family, but the L2 cache per core has doubled since the previous generation.
Additionally, the 7950X will be AMD’s most power-hungry mainstream desktop Zen processor to date. According to previous AMD disclosures, the Power Package Tracking (PPT) limit for these chips will be 230W. The TDP for the chip is 170W. Previous Ryzen 5000 chips were frequently TDP-limited when all of their CPU cores were under heavy load, preventing them from reaching the higher clock speeds they’d otherwise be able to support. This is one of the primary reasons why AMD is seeing such significant MT performance gains, which exceed their average ST performance gains.
Since AMD is the first CPU manufacturer to release this generation of parts, the 7950X has no direct competition; however, if you’re considering an Intel Core i9 12900K, AMD would be happy to upsell you to a 7950X. According to AMD, the fully developed Zen 4 chip should be the world’s most powerful desktop CPU. We’ll have to confirm this at review time, but it’s not hard to believe given the architecture updates and improvements brought about by TSMC’s 5nm process. Nonetheless, it will eventually compete with Intel’s Raptor Lake CPUs once they become available.
Surprisingly, AMD has priced its new flagship chip at $699. This is $100 less than the launch price of the 5950X, which was $799 in late 2020. This chip’s higher price was not well received, but for a considerable amount of time, AMD sold every unit it could produce. Nevertheless, the chip shortage has subsided and AMD now has greater access to wafers and substrates; the company will not be without competition for the duration of the chip’s life.
Ryzen 9 7900X
Stepping down from the 7950X is the Ryzen 9 7900X. This is another two CCD part, but only six cores are enabled on each CCD, for a total of twelve CPU cores and twenty-four threads. This part retains the same 64MB of L3 cache as the 7950X, which is identical to how the L3 cache scales down (or rather, does not scale down) on the Ryzen 5000 series.
With a TDP of 170W, the same as the 7950X, the 7900X has a slightly faster base clock speeds. This component will offer a base frequency of 4.7GHz, 200MHz higher than the 7950X. However, the maximum turbo frequency is 5.6GHz, which is slightly lower than that of the 7950X. Therefore, if you want the best single-threaded performance, the 7950X is positioned and configured to be the fastest option for both ST and MT workloads.
The Ryzen 9 7900X will cost $549, the same price as AMD’s previous 12-core processor, the 5900X. This will be a consideration for a true apples-to-apples comparison of generational performance gains at a constant cost.
Ryzen 7 7700X
At the bottom of AMD’s Ryzen 7000, product stack is the Ryzen 7 7700X. This is the 8-core Zen 4 equivalent to the AMD 5800X from the previous generation. Similar to its predecessor, AMD employs a single Zen 4 CCD, but this one is fully enabled. The base clock speeds will be 4.5GHz, with a maximum turbo clock speed of 5.4GHz.
Since there is only one CCD, the L3 cache is also reduced to 32MB. In conjunction with the 8MB of L2 cache distributed across the Zen 4 CPU cores, AMD advertises this chip as having a total of 40MB of cache.
In contrast to the 12- and 16-core processors, the 7700X will have a more conventional TDP of 105 watts. This makes the chip easier to power (and cool), but it remains to be seen how this affects turbo clocks under heavy MT loads. With fewer CPU cores overall, there is less silicon to illuminate, but this also means AMD has less electrical headroom to increase clock speeds.
The 7700X will be available for $399, which, like the 7950X, represents a slight price reduction over the previous generation (5800X, which was priced at $449).
Ryzen 5 7600X
The Ryzen 5 7600X is the only Ryzen 5 class processor included in the Ryzen 7000 launch lineup. Similar to its predecessor, this component has 6 cores and 12 threads. It is powered by a single Zen 4 CCD with two CPU cores disabled.
The chip’s base clock speed will be 4.7GHz (similar to the 7900X), while its turbo clock speed will be the lowest of all the initial Ryzen 7000 processors at 5.3GHz. AMD boldly asserts that the 7600X, despite being the slowest of the Zen 4 processors, should be 5% faster in gaming than Intel’s Core i9 12900K. (never mind the cheaper chips).
This chip’s TDP is increasing, however. In the previous generation, AMD’s top six-core chip drew 65W, whereas now even the bottom of the stack begins at 105W. Therefore, this should not be considered a small chip; AMD anticipates that it will both consume and deliver a great deal of power.
+13% Desktop IPC and AVX-512: A Zen 4 Architecture Teaser
These new chips are supported by AMD’s Zen 4 architecture. With only thirty minutes, AMD’s presentation was insufficient to delve deeply into the Zen 4 architecture (AMD has to save something for the launch). However, CTO Mark Papermaster did take the stage to provide additional technical information on the Zen 4 architecture.
AMD has revised their IPC estimate for Zen 4 first and foremost. As they have continued to tune the chips and firmware, they are now observing a 13% increase in IPC in “typical desktop applications” compared to Zen 3.
AMD claims that when they lock things down at an iso-frequency of 4GHz, they see gains ranging from 1% to 39%, depending on the workload. As a result, as is typically the case, the precise performance gains vary depending on the task. Still, several of AMD’s benchmarks come in close to the 13% threshold.
For this generation, Papermaster has reintroduced one of our favorite diagrams, the IPC breakdown, which illustrates how the various Zen 4 architecture changes contributed to the overall IPC gain.
According to AMD, the biggest contributors to this generation are their new front-end, load/store improvements, and then an improved branch predictor, execution engine, and the larger 1MB L2 cache per CPU core. Papermaster did reveal that the execution engine’s operand cache size has been increased by 50 percent, so it’s clear that AMD has been tinkering with their various caches at the CPU core level for Zen 4.
Papermaster also confirmed for the first time that Zen 4 will support AVX-512 instructions, including the Ryzen 7000 series. In addition to the foundation (AVX-512F) instructions, it is still unclear which subsets of AVX-512 AMD will support it, as AVX-512 is a bit of a mess of standards. However, Papermaster did specifically mention Vector Neural Network Instructions (VNNI) as one of the supported additional subsets.
Intel built a 512-bit wide, true SIMD machine for executing AVX-512 instructions, whereas AMD did not. AMD will instead execute these instructions in two cycles. This means that AMD’s implementation will still benefit from the additional instructions, register file space, and other technical enhancements introduced by AVX-512, but they will not experience the inherent doubling in SIMD throughput.
As the primary reason for keeping AMD’s SIMD design at 256 bits, Papermaster cited the extreme power requirements for a true 512-bit SIMD block when explaining the company’s decision. As we’ve already observed in Intel chips with AVX-512 support, the massive throughput of a 512-bit SIMD combined with its high-density results in a hard spike in power consumption when using it, requiring Intel’s chips to downclock on AVX-512 workloads (sometimes significantly) to maintain power and thermals. Using a narrower 256-bit SIMD will allow AMD to illuminate fewer transistors simultaneously, making it easier to maintain consistent clock speeds and power consumption. Nonetheless, I don’t believe AMD minds that the die space requirements for a 256-bit SIMD are significantly less than those for a 512-bit SIMD; a full 512-bit SIMD requires a large number of transistors to construct and a large number of transistors to activate during heavy workloads.
The impact of TSMC’s 5nm process on AMD’s CPU cores and their resulting chiplets should not be underestimated. AMD’s CCD chiplets have utilized TSMC 7nm for both the Ryzen 3000 (Zen 2) and Ryzen 5000 (Zen 3) generations; therefore, AMD’s desktop CPUs have been on 7nm for more than three years. As a result, the transition to 5nm has been a long time coming and marks the first time in a considerable amount of time that AMD has been able to enjoy a complete node shrink for their desktop CPUs.
The benefits of 5nm are evident in all aspects, including clock speeds, power consumption, and die size. Particularly, the improvements to clock speeds represent a major shift for AMD, which a couple of generations ago had a clock speed disadvantage compared to Intel, which limited AMD’s overall single-threaded performance. Now, AMD’s top chips are expected to exceed 5 GHz, delivering a significant clock speed increase on top of their 13% IPC gains – for a total of 29% more single-threaded performance.
In contrast, a Zen 4 CPU core and its associated cache are smaller than a Zen 3 core and its 512KB L2 cache, despite the architectural improvements made to Zen 4 by AMD, such as the doubling of the L2 cache size to 1MB. The resulting Zen 4 core is 18% smaller than its predecessor, measuring 3.84mm2 according to Papermaster. In addition, he threw in comparison to the size of Intel’s Golden Cove core, which is found in their Alder Lake processors, in an attempt to be cheeky. According to AMD, a Zen 4 core is exactly half the size of a Golden Cove core, giving AMD a significant die size advantage (and thus at some level, manufacturing costs).
We’ve known for some time that TSMC’s 5nm process has a significant density advantage over Intel’s 7nm process, and this unquestionably plays a significant role in AMD’s size advantage. Nonetheless, as AMD strives to promote the notion that Zen 4 is a design of the highest caliber, area efficiency is yet another aspect in which AMD competes with Intel – and can use it to its advantage.
In a concluding aside, Papermaster disclosed the height of the metal stack that AMD employs. Overall, AMD employs a 15-layer telescoping stack.
Zen 4 vs. Zen 3: A Performance Preview
What does this mean for AMD in light of their architectural changes and their access to TSMC’s 5nm process? The company boldly asserts that they have the world’s fastest desktop CPUs and that the Ryzen 7000 chips will deliver significant improvements over both Intel’s current chips and AMD’s previous-generation Ryzen 5000 components.
And while all first-party benchmarks should be taken with a grain of salt, the following is what AMD claims for performance and power.
We’ll begin with iso (like-for-like) comparisons of iso-power between the Ryzen generations.
|AMD Desktop CPU Generations|
|CPU Architecture||Zen 4||Zen 3||Zen 2|
|CPU Cores||Up To 16C / 32T||Up To 16C / 32T||Up To 16C / 32T|
|CPU PCIe Lanes||24x PCIe 5.0||24x PCIe 4.0||24x PCIe 4.0|
|Manufacturing Process||CCD: TSMC N5
IOD: TSMC N6
|CCD: TSMC N7
IOD: GloFo 12nm
|CCD: TSMC N7
IOD: GloFo 12nm
AMD provided performance specifications for three different TDPs: 65W, 105W, and 170W. The 7950X demonstrated a 74% improvement in Cinebench R23 MT performance at the lowest TDPs. These advantages decreased as TDPs increased, falling to 37% at 105W and then 35% at 170W.
This is intended to highlight how AMD’s emphasis on energy efficiency, combined with TMSC’s 5nm process, has resulted in significant gains across the board, but particularly at lower TDPs. Zen 4 is significantly more power-efficient than its predecessor when AMD is not loading CPU clock speeds into the stratosphere, which is always well into the diminishing returns of the voltage/frequency curve. And even when AMD maxes out its clock speeds, Zen 4 still comes out on top.
AMD claims that Zen 4 will deliver up to 49% more performance at iso-power than Zen 3 overall. Note that this is not an average but an “up to” figure. However, this demonstrates what Zen 4 should be capable of when it puts its best foot forward.
AMD also provided iso-performance figures for Cinebench, which is the opposite of what was previously stated. According to the manufacturer, Zen 4 (7950X) consumes up to 62% less energy than Zen 3 at the same level of performance (5950X). And while AMD will reinvest the majority of these gains into improving the performance (clock speeds) of these Ryzen 7000 desktop components, this is another example of the efficiency gains that AMD is touting.
AMD offers similarly optimistic figures in comparison to archrival Intel, albeit with a greater degree of cherry-picking. AMD claims that the 7950X provides 47% greater performance per watt than the 12900K in Chaos V-Ray.
And in gaming workloads, even the (relatively small) 6-core 7600X will outperform the 12900K by an average of 5%, according to the company. A 6-core configuration is roughly equivalent to Zen 4 having one hand tied behind its back.
AMD asserts that the Ryzen 7000 series will provide 11% better gaming performance than the 12900K and 44% better performance in highly multithreaded content creation applications. Which, if confirmed by independent testing, would place the Ryzen 7000 at the head of the pack for the upcoming generation of PC processors.
Ryzen 7000: Due out on September 27
AMD’s first step into the Zen 4 era will begin on September 27th, bringing the current era to a close. On that date, all four initial Ryzen 7000 CPUs and the first wave of AM5 platform motherboards will be available for purchase.
According to AMD, even though initial launch logistics are always an interesting challenge (especially now that shipping is still a bit erratic), the overall supply of Ryzen 7000 chips should be adequate. While it’s not uncommon for launch-day allocations of new high-end hardware to sell out, AMD has indicated that Ryzen 7000 chips should be readily available. In contrast to the Ryzen 5000 launch, which occurred near the peak of the chip and substrate shortage, Ryzen 7000 chips should be readily available. AMD has invested in wafer and substrate capacity to ensure that this is the case.
In the meantime, the launch of AM5 motherboards will be spread out over two months. AMD’s board partners will have their high-end X670/X670E boards available for the September 27 launch. The mainstream B650 boards, including the newly announced B650E chipset, will be released one month later. Therefore, prospective purchasers will have to choose between purchasing an X670 board at launch or waiting to purchase cheaper B650 boards.
AMD estimates that motherboard prices will begin at $125. This is likely for entry-level B650 motherboards without PCIe 5.0 support. PCIe 5.0 will undoubtedly increase the price of B650E and other motherboards, but at this time it is unclear how much PCIe 5.0 will increase the production costs and retail prices of AM5 motherboards.
Alongside the release of AM5 motherboards, the 27th will see the introduction of the first factory-overclocked memory kits optimized for the Ryzen 7000 series. AMD will offer the EXtended Profiles for Overclocking (EXPO) standard for this generation, which is AMD’s equivalent to Intel’s XMP DIMMs in terms of pre-programmed profiles for factory overclocked memory. AMD expects at least 15 EXPO-enabled memory kits with speeds up to DDR5-6400 to be available at launch.
With that, the die for Zen 4 has been cast. AMD is aiming for nothing less than a leadership position in this generation, and on September 27th, we’ll get a firsthand look at how well the company’s engineers realize this goal. After a pandemic, a chip supply shortage, and other factors, it will be nice to return to a more regular PC CPU launch schedule.