Amd Ryzen 7000 Zen 4 CPU Series Detailed

AMD’s previous-generation Ryzen 5000 cpus accomplished a feat that was previously supposed to be extremely difficult: They outperformed Intel’s finest in every CPU test, including our ranking of the best CPUs for gaming, where AMD’s Rocket Lake was superior in every manner.

Then, though, there was Alder Lake. Intel’s new hybrid x86 architecture, which consists of large, powerful cores and compact, energy-efficient cores, propelled the corporation to the forefront in all areas of raw performance and assisted in resolving some of its most glaring issues with power consumption. But perhaps most critically, Alder Lake sparked an all-out price battle with Intel’s new “bare-knuckles” pricing strategy, particularly in the mid-range, giving gamers a choice.

Four “Raphael” CPUs from the new Zen 4 Ryzen 7000 series have been released by AMD. In this post, we have compiled the most significant performance benchmarks and facts to provide you with a broader perspective. AMD claims that the $699 16-core Ryzen 9 7950X is the world’s fastest CPU. The $299 Ryzen 5 7600X is the least costly member of the initial line of Zen 4 CPUs. These chips offer the finest performance on the market, a nice return to form based on our testing.

CPU	Price	Cores / Threads (P+E)	Base / Boost Clock (GHz)	Cache (L2+L3)	TDP / Max	Memory
Ryzen 9 7950X	$699	16 / 32	4.5 / 5.7	80MB	170W / 230W	DDR5-5200
Ryzen 9 7900X	$549	12 / 24	4.7 / 5.6	76MB	170W / 230W	DDR5-5200
Ryzen 7 7700X	$399	8 / 16	4.5 / 5.4	40MB	105W / 142W	DDR5-5200
Ryzen 5 7600X	$299	6 / 12	4.7 / 5.3	38MB	105W / 142W	DDR5-5200

But AMD isn’t standing still. With its Ryzen 7000 CPUs, AMD has stepped up the race to be the best performer. Ryzen 7000 can go up to 5.7 GHz, which is an amazing 800 MHz faster than the last generation. The new Zen 4 microarchitecture can boost IPC by up to 13%. The chips also include cutting-edge features including an integrated Radeon RDNA 2 graphics engine and AVX-512-based AI instructions.

A brief comparison of the Zen 4 CPUs to Intel’s Alder Lake has been presented below, and it is based on the results of our own rigorous testing. In 1080p gaming, the Ryzen 9 7950X is 5% better than the Intel Core i9-12900K. In threaded workloads, the Ryzen 9 7950X is 44% faster than the Core i9-12900K, while in single-threaded tests, the two processors are practically comparable.

The AMD Ryzen 5 7600X is 12 percent quicker than the Intel i5-12600K while playing games at Full HD. The performance difference between the two CPUs narrows to 6% when both are overclocked. A new level of value has been added to the $300 price range as a result of the fact that the basic 7600X is 4% quicker at gaming than Intel’s top Core i9-12900K; however, this comes at the expense of increased platform expenses.

In gaming, both CPUs are better than Intel’s best chip. But as great as they are, they are not perfect. The Zen 4 Ryzen 7000 series starts at $300 and only works with expensive DDR5 memory, unlike Intel’s cheaper DDR4 options. The high total platform costs make it hard to argue that the platform is worth it. AMD also used a lot more power to improve performance, which caused the machine to get hotter and need more power. But you get more work out of each watt.

Ryzen 7000 is clearly in the lead, but its real competitor, Raptor Lake, won’t be out until next month. Intel says that its performance has improved by 15% for single-threaded tasks, 41% for multi-threaded tasks, and 40% for “overall” tasks. This shows that the competition for desktop PC supremacy will be fierce.

The company will add at least one 3D V-Cache model to the four regular desktop PC chips that are already on the market before the end of the year. Intel has already made its Raptor Lake CPUs, so when AMD’s Ryzen 7000 comes out on October 20, it will face tough competition. All the information we know about is on this page.

Series at a Glance

• The first four models with the codename Raphael were introduced on September 27 under the designation “Raphael.”
• TSMC’s 5nm technology supports up to 16 cores and 32 threads (N5 used for compute die)
• 5,7 GHz peak frequency
• I/O die at 6nm, DDR5 memory controllers, and PCIe 5.0 interface
• DDR5 only (no support for DDR4), up to 125% more memory bandwidth per core
• RDNA 2 built-in GPU (present on IOD, very basic and low power)
• Zen 4 architecture features a 13% improvement in IPC
• AM5 LGA 1718 Socket is backward compatible with AM4 cooling.
• Motherboards Chipset: X670E Extreme, X670, B650E Extreme, and B650
• Up to 170W TDP, 230W peak power
• Support for AVX-512, VNNI

Specifications and Features

The TSMC 5nm manufacturing node is used for the CCD in Ryzen 7000 CPUs, whereas the TSMC 6nm technology is used for the I/O Die (IOD). The architecture is broken out in further detail below. These CPUs will fit into motherboards with a Socket AM5 connector.

CPU	Price	Cores / Threads (P+E)	Base / Boost Clock (GHz)	Cache (L2/L3)	TDP / PBP / MTP	Memory
Ryzen 9 7950X	$699	16 / 32	4.5 / 5.7	80MB (16+64)	170W / 230W	DDR5-5200
Ryzen 9 5950X	$546 ($799)	16 / 32	3.4 / 4.9	74MB (8+64)	105W	DDR4-3200
Ryzen 9 7900X	$549	12 / 24	4.7 / 5.6	76MB (12+64)	170W / 230W	DDR5-5200
Ryzen 9 5900X	$398 ($549)	12 / 24	3.7 / 4.8	70MB (6+64)	105W	DDR4-3200
Ryzen 7 7700X	$399	8 / 16	4.5 / 5.4	40MB (8+32)	105W / ?	DDR5-5200
Ryzen 7 5700X	$268 ($299)	8 / 16	4.6	36 (4+32)	105W	DDR4-3200
Ryzen 5 7600X	$299	6 / 12	4.7 / 5.3	38MB (6+32)	105W / ?	DDR5-5200
Ryzen 7 5600X	$199 ($299)	6 / 12	3.7 / 4.6	35MB (3+32)	65W	DDR4-3200

In general, the new range of Ryzen 7000 SKUs has the same number of cores as the older generations of processors, but they run between 16 and 17 percent faster than their predecessors. In addition, the capacity of the L2 cache is increased across the board while the L3 cache remains the same.

The Ryzen 9 7950X, which has 16 cores and 32 threads, can be purchased for $100 less than the initial launch price of the Ryzen 9 5950X. On the other hand, the Ryzen 9 7900X, which has 12 cores and 24 threads, has the same launch price as its predecessor, the 5900X.

However, AMD added an additional $100 to the launch price of the eight-core, sixteen-thread Ryzen 7 7700X in comparison to the 5700X model. In addition, AMD has maintained the entry-level pricing for the Ryzen 5 7600X at the same $299 price point; nevertheless, this is not a full victory for the company since the high entry-level cost for the Ryzen 5000 series was not well welcomed. There is no mention of a Ryzen 7 7800X that will succeed the now-retired 5800X. It’s possible that AMD is saving a slot on this list for their X3D model that supports V-Cache.

Year / Processor	Peak Frequency	Frequency Gain	Process, Architecture
2017 – Ryzen 7 1800X	4.1 GHz	–	14nm Zen 1
2018 – Ryzen 7 2700X	4.3 GHz	+200 MHz / +5%	12nm Zen+
2019 – Ryzen 9 3950X	4.7 GHz	+400 MHz / +9%	7nm Zen 2
2020 – Ryzen 9 5950X	4.9 GHz	+200 MHz / +4%	7nm Zen 3
2022 – Ryzen 9 7950X	5.7 GHz	+800 MHz / +16%	5nm Zen 4

As a result of both the 5nm manufacturing process and an architecture that has been optimized for higher clock rates, AMD has improved clock speeds across the board for all of its four new product variants.To put things in perspective, the flagship Ryzen 9 7950X will feature a clock rate that is 16% higher than the Ryzen 9 5950X and its 5.7 GHz boost clock rate. Its 800 MHz boost over its predecessor is the highest increase in frequency seen so far in the Ryzen generation. The Ryzen 9 7900X is also significant since it is 800 MHz faster and has 12 processing cores than its predecessor.

AMD has stated that the Ryzen 7000 will still include Precision Boost 2 in order to expose the greatest boost frequencies available at all times. This will only take place on two of the chip’s cores, as is the case with all of AMD’s most recent CPUs. In spite of this, we are aware that Intel’s Raptor Lake will have a boost speed of 5.8 GHz, and maybe even higher.

AMD has also upped the TDP ratings, by 65W for the Ryzen 9 models and 45W for the Ryzen 5 models respectively.There is no increase in thermal design power (TDP) for the Ryzen 7 7700X. In addition to this, the maximum power consumption (MPC) of the AM5 socket has been increased to 230W. When compared to the previous generation Ryzen 5000, which had a limit of 142W, this is a huge improvement.

Versus Intel 13th-Gen Raptor Lake

CPU	Price	Cores / Threads (P+E)	P-Core Base / Boost Clock (GHz)	E-Core Base / Boost Clock (GHz)	Cache (L2/L3)	TDP / PBP / MTP	Memory
Core i9-13900K / KF	$589 (K) – $564 (KF)	24 / 32 (8+16)	3.0 / 5.8	2.2 / 4.3	68MB (32+36)	125W / 253W	DDR4-3200 / DDR5-5600
Ryzen 9 7950X	$699	16 / 32	4.5 / 5.7	–	80MB (16+64)	170W / 230W	DDR5-5200
Core i9-12900K / KF	$589 (K) – $564 (KF)	16 / 24 (8+8)	3.2 / 5.2	2.4 / 3.9	44MB (14+30)	125W / 241W	DDR4-3200 / DDR5-4800
Ryzen 9 7900X	$549	12 / 24	4.7 / 5.6	–	76MB (12+64)	170W / 230W	DDR5-5200
Core i7-13700K / KF	$409 (K) – $384 (KF)	16 / 24 (8+8)	3.4 / 5.4	2.5 / 4.2	54MB (24+30)	125W / 253W	DDR4-3200 / DDR5-5600
Core i7-12700K / KF	$409 (K) – $384 (KF)	12 / 20 (8+4)	3.6 / 5.0	2.7 / 3.8	37MB (12+25)	125W / 190W	DDR4-3200 / DDR5-4800
Ryzen 7 7700X	$399	8 / 16	4.5 / 5.4	–	40MB (8+32)	105W / 142W	DDR5-5200
Ryzen 5 7600X	$299	6 / 12	4.7 / 5.3	–	38MB (6+32)	105W / 142W	DDR5-5200
Core i5-13600K / KF	$319 (K) – $294 (KF)	14 / 20 (6+8)	3.5 / 5.1	2.6 / 3.9	44MB (20+24)	125W / 181W	DDR4-3200 / DDR5-5600
Core i5-12600K / KF	$289 (K) – $264 (KF)	10 / 16 (6+4)	3.7 / 4.9	2.8 / 3.6	29.5MB (9.5+20)	125W / 150W	DDR4-3200 / DDR5-4800
Core i5-13400 / F	?	10 / 16 (6+4)	3.4 / ?	?	24MB	65W / ?	DDR4-3200 / DDR5-5600
Core i5-12400 / F	$199 – $167 (F)	6 / 12 (4+0)	4.4 / 2.5	–	25.5MB (7.5+18)	65W / 117W	DDR4-3200 / DDR5-4800

Here’s how Ryzen 7000 compares to Intel’s Alder Lake and Raptor Lake CPUs. 

Intel’s Alder Lake Core i9-12900K is quicker in most applications than Ryzen’s flagship, but its next-gen 13900K adds eight extra E-Cores for better performance in heavily-threaded workloads. Intel’s 13700K and 13600K have four extra e-cores.

Intel has added E-cores to its 13400 SKU for the first time, making it a more competitive processor at the low end of the market, where AMD isn’t as strong. Pricing and performance are unknowns, and Intel hasn’t said anything. Intel will employ faster clock rates and more E-Cores to counter Ryzen 7000.

Zen 4 against Intel Raptor Lake is in many ways a frequency battle, with both chipmakers pushing their consumer CPUs to the highest frequencies we’ve seen. Increasing frequency also increases power consumption and TDP from both chipmakers. Higher peak power levels will benefit threaded workloads, thus we should anticipate more from each Zen 4 core.

Raptor Lake supports DDR4 and DDR5, whereas Zen 4 Ryzen 7000 only supports DDR5. Intel has a cost advantage since DDR5 is more expensive. We no longer experience DDR5 shortages, and prices continue to fall as supply increases and demand decreases.

According to early testing, the Zen 4 Ryzen 7000 CPUs offer great overclocking headroom. Ryzen 9 7950X defeated liquid nitrogen-cooled processors to establish four world records with ordinary liquid cooling. When subzero, the chips achieved 7.2 GHz on a single core and 6.5 GHz on all cores using LN2.

We grabbed a close-up snip of a bare Ryzen 7000 chip at AMD’s Computex event. The processor has two 8-core gold 5nm chiplets. AMD claims they are built on TSMC’s high-performance 5nm N5 manufacturing technology and are closer together than earlier Ryzen core chiplets. A shim emerges between the two core chiplets, possibly to maintain an equal surface. This tight alignment may also be owing to enhanced packing between the chips.

We also observe a distinct outline surrounding each CCD’s top, although we’re not sure whether this is a new metallization approach. Backside Metallization (BSM) incorporates an Au covering to avoid oxidation, improve TIM adhesion, and minimize thermal impedance. We also observe several vacant spaces for capacitors, which might mean bulkier future designs.

The new I/O device utilizes the 6nm technology and has PCIe 5.0 and DDR5 memory controllers and AMD’s RDNA 2 graphics engine. The new 6nm I/O die includes a low-power design based on AMD’s Ryzen 6000 processors, thus it has improved low power management and an extended palette of low-power states. AMD claims this processor uses 20W, less than Ryzen 5000, and will offer most of Ryzen 7000’s power reductions.

The new I/O die is around the same size as the 12nm die. Given that the 6nm die is denser than the 12nm die from GlobalFoundries, it’s fair to infer the integrated GPU devoured a large percentage of the transistor budget (possibly due in part to onboard iGPU cache). The huge 6nm I/O die will contribute to the cost of the chips since it’s more costly than the mature 12nm silicon used in Ryzen 5000 processors.

AMD has revealed that at least one Ryzen 7000 series processor will include 3D V-Cache this year, allowing amazing L3 cache capacity using their unique 3D-stacked SRAM innovation. This technology gives the Ryzen 7 5800X3D 96MB of L3 cache, boosting gaming performance. TeamGroup highlighted Raphael-X processors in a recent news release. AMD hasn’t named the 3D V-Cache Ryzen 7000 processors Raphael-X, although it follows the same naming scheme as the Milan-X server chips. It’s possible TeamGroup made a mistake, but many believe this is the codename for Zen 4 3D V-Cache consumer chips.

Ryzen 7000’s base memory is DDR5-5200, although the firm promises excellent overclocking. AMD EXPO is an alternative to Intel’s XMP. AMD will offer pre-defined memory profiles with dialed-in frequencies, timings, and voltages for one-click overclocking. Several rumors claim that AMD will feature ‘high-bandwidth’ and ‘low-latency’ EXPO profiles, similar to Intel’s Gear 1 and Gear 2 memory modes. New BIOS updates offer a 3GHz fabric frequency, previously Ryzen 5000 capped out at 2 GHz. We won’t know whether they are pre-assigned parameters until the chips arrive in our laboratories.

AMD says DDR5-5600 is’sweet spot’ for Zen 4 Ryzen 7000 CPUs. TeamGroup’s Vulcana modules are adjusted for this pace.

Ryzen 7000 CPUs offer 24 PCIe 5.0 lanes from the socket (further details in the motherboard section). AMD is working with Phison, Micron, and Crucial to enable PCIe 5.0 SSDs. Crucial and Micron will have their first PCIe 5.0 SSDs, while third-party SSDs will utilise Phison’s E26 PCIe 5.0 SSD controllers. AMD claims a 60% performance boost for Zen 4 Ryzen 7000 PCs using PCIe 5.0 SSDs. October brings the first PCIe 5.0 SSDs.

Microsoft’s DirectStorage depends primarily on read throughput to minimize game loading times to a second. AMD said Ryzen 7000 will feature Smart Access Storage (SAS), which is based on the same APIs as DirectStorage. AMD’s PCIe 5.0 SSD support is explained here. Micron’s leading-edge flash doesn’t work at full speed, limiting SSD performance, but that will be fixed early next year. Then, we anticipate speedier models.

AMD’s Ryzen 6000 ‘Rembrandt’ processors enable Microsoft’s Pluton for improved security against physical assaults, encryption key theft, and firmware intrusions. Pluton debuted in the Xbox and AMD’s EPYC data center CPUs and complements existing security technologies like AMD Secure Processor and Memory Guard. Pluton is predicted in Ryzen 7000, although AMD hasn’t confirmed it.

Ryzen 7000 CPUs feature AVX-512 instructions, which may be utilized for VNNI for neural networks and BFLOAT16 for inference. AMD’s AVX-512 implementation ‘double-pumped’ 256-bit wide instructions to reduce frequency penalty.

Intel’s AVX-512 CPUs. This might reduce throughput per clock compared to Intel’s technique, although faster rates will mitigate part of the loss. More about the new implementation must wait.

The hybrid design of Intel’s Alder and Raptor Lake CPUs disables AVX-512. AIDA64 just implemented AVX-512 and y-cruncher support for Zen 4 CPUs.

New Integrated Graphics

All Ryzen 7000 CPUs will support some type of graphics, thus it does not seem that there will be graphics-less choices like Intel’s F-series. The RDNA 2 engine resides on the I/O Die and supports up to four display outputs, including DisplayPort 2 and HDMI 2.1 ports. Ryzen 7000 has the same video (VCN) and display (DCN) engine as the Ryzen 6000 “Rembrandt” processors.

The RDNA 2 iGPU has two computing units, four Accelerated Compute Engines (ACE), and one Hardware Streaming Engine (HWS).

In the few games that did run, we were treated to a slideshow, much like with Intel’s graphics.

The iGPU supports AV1 and VP9 decoding, H.264 and HEVC encoding and decode, USB Type-C with DisplayPort Alt Mode, DisplayPort 2.0 (adaptive sync, DSC, UHBR10, HDR), and HDMI 2.1 (HFR, 48Gbps FRL, DSC, HDR10+, and VRR). Additionally, 4K60 and hybrid graphics are supported.

Benchmarks and Zen 4 IPC

The $699 Ryzen 9 7950X is 5% quicker than Intel’s fastest gaming CPU, the Core i9-12900K, in its standard configuration. After overclocking the cores and RAM, the 7950X is nearly tied with the overclocked 12900K. This is a significant generational increase, since the Ryzen 9 7950X is 17% quicker than its predecessor, the Zen 3-powered Ryzen 9 5950X, which also has 16 cores. However, Intel only needs a 5% increase in gaming performance with Raptor Lake to equal the 7950X, setting the scene for a fierce rivalry next month.

The $299 Ryzen 5 7600X is 12% quicker than the $289 Core i5-12600K, with the gap closing to 6% when both processors are overclocked. More significantly, the basic 7600X is 4% quicker than Intel’s top Core i9-12900K, offering a new level of value at the $300 price range — although at the expense of higher platform expenses.

The Ryzen 5 7600X is 18% more powerful than its predecessor, the Ryzen 5 5600X, which was formerly the benchmark for mid-range gaming systems. Raptor Lake is alluring from a distance in the mid- and low-end price categories, but the 7600X will go a long way toward bolstering AMD’s defenses. You may also adjust the 7600X to squeeze out an additional 3% of performance, but as usual, performance increases will vary by game and chip quality.

The Ryzen 5 7600X costs $300 and is 2% quicker in gaming than the Ryzen 9 7950X, which costs more than twice as much and costs more than twice as much. As a result, the 76000X is an outstanding bargain for gaming, but, as is typical for Ryzen 9 processors, the 7950X is intended for users who want professional-grade application performance.

AMD’s $430 Ryzen 7 5800X3D remains the market’s fastest gaming processor by a significant margin, but this highly specialized chip comes with drawbacks – its 3D V-Cache does not improve performance in all titles. In addition, the 5800X3D is especially built for gaming, but cannot compete with comparably priced processors in office applications. AMD will release at least one Zen 4-powered Ryzen 7000 CPU with 3D V-cache this year, so if you’re interested in a specialist gaming chip of this kind, you may want to wait a few more months.

Naturally, the disparities between the chips diminish when we go to 1440p and introduce a GPU bottleneck, but the pattern stays basically the same, with little variances between the top-performing CPUs. Now that the battle between Intel and AMD is even more intense, it is essential to base your choice on the games you play most often.

The Ryzen 9 7950X completely destroys our overall assessment of multithreaded performance, which should come as no surprise, but the size of the performance gap is still astounding. The 7950X is 44% quicker than the Core i9-12900K, and overclocking does nothing to narrow the performance gap. Also, the 7950X is 45% quicker than the 5950X, proving that AMD’s performance promises are accurate.

The Threadripper Pro 32-core Ryzen 9 5975WX, which costs $3,299, performs 17% slower than the 7950X overall. However, it is 5 times more expensive, illustrating how remarkable this level of threaded processing power is on a standard PC architecture. This likely has something to do with AMD’s decision to not release a Zen 3-powered HEDT platform.

This is a difficult act to follow for the 7600X, particularly when the Core i5-12600K is a few percentage points quicker at threaded tasks. This is due to the 12600K’s e-cores, and overclocking increases the 12600K’s advantage by 10%. In any event, the Ryzen 5 7600X represents a significant generational increase for Ryzen, since it is 34% quicker than its predecessor, the Ryzen 5 5600X.

Alder Lake processors from Intel outperformed Ryzen 5000 processors from AMD’s previous generation in terms of single-threaded performance, but the Zen 4 architecture makes a significant leap ahead, just edging out Alder Lake chips of equivalent price. In our cumulative evaluation of single-threaded performance, the Ryzen 9 7950X and 7600X tie the Core i9-12900K and 12600K, leveling the playing field.

As predicted, the overclocked setups provide little to no advantage for the Ryzen CPUs, whilst both Intel chips see a performance loss in this measure. The 12900K’s loss is predictable, since its all-core overclock of 5.1 GHz is lower than its single-core high of 5.2 GHz, while the 12600K’s loss is rather unexpected, as its overclock equals its maximum speed. Thread targeting seems to be less successful on overclocked Intel setups, resulting in a significant amount of thread migration during single-threaded workloads. This was determined after extensive study and repeated testing. This problem is currently being investigated, but the findings are reproducible, so we’re including them.

Measuring IPC is difficult due to the fact that it changes depending on the workload. AMD derived their 13% IPC boost from 22 distinct workloads, including gaming, which seems an odd inclusion given the possibility of graphics-induced bottlenecks. AMD also provided tasks with multiple threads. AMD’s data indicate that IPC increases range from 39% for the wPrime test to 1% for the CPU-z benchmark.

We examined a restricted subset of single-threaded workloads to determine clock-for-clock improvements, locking all processors to a static all-core frequency of 3.8 GHz and setting the memory transfer rate to the maximum officially supported value. Clearly, Zen 4 provides substantial IPC increases across a variety of workloads. As a consequence of Zen 4’s support for AVX-512, the y-cruncher and Geekbench 5 crypto scores suffer rather disproportionate improvements. As seen by the single-threaded and multi-threaded y-cruncher benchmarks, however, this performance does not scale linearly with increasing core loads.

Power Consumption

A specialized 15-layer N5 process node was created as a result of AMD and TSMC modifying the 5nm process to meet their design objectives. Unfortunately, we don’t know the specifics of the custom node. But compared to AMD’s 7nm process, which was used for its Ryzen 5000 chips from the previous generation, TSMC’s N5 typically offers 15% more performance at a given power level or 20% less power consumption at any given clock rate. Ryzen 7000, according to AMD, offers up to 40% more performance per watt at its standard TDP levels when combined with Zen 4 architectural advancements and SoC improvements. 

Specs	65W TDP	105W TDP	170W TDP
Socket Power (PPT) Watts	88W	142W	230W
Peak Current (EDC) Amps	150A	170A	225A
Sustained Current (TDC) Amps	75A	110A	160A

AMD has specified a new 170W TDP range, a new peak for the Ryzen family’s mainstream processors. Additionally, AMD has upped the basic TDP for Ryzen 9 by 65W and for Ryzen 5 by 45W. In addition, the AM5 socket’s peak power consumption (PPT) is now 230W. That is a big boost above the 142W limit of the previous-generation Ryzen 5000.

AMD also drastically raised the TDC and EDC amperage, increasing EDC by 60A and 30A, respectively, for the 65W and 105W TDP levels. The EDC rises by 15A for both the 65W and 105W TDP categories.

AMD claims that it has upgraded the platform power interface from SVI2 to SVI3, allowing it to shift from two variable power rails to three, hence providing for more precise control over the power supply to the socket. The SVI3 interface enables more precise and continuous telemetry for voltage, current, power, and temperature for multiple onboard voltage regulators, while the SVI2 interface does not provide monitoring of power and temperature. SVI3 also permits increased power states that save energy, such as phase shifting (shutting off phases when not needed).

Increasing the TDP and PPT will enable AMD to provide more performance, especially for its higher-core-count models, under severe multithreaded loads. In many circumstances, AMD’s prior 142W restriction with the previous-generation AM4 socket hindered performance, thus the extra 88W of power will be especially beneficial for the 12- and 16-core machines.

In addition, AMD has indicated that it will utilize the usual TDP and PPT calculations for chips that drop into the AM5 socket; to calculate the maximum power consumption of the chip, multiply the TDP by 1.35. (PPT).

According to our power assessments, the Ryzen 7000 chips use much more energy than their predecessors, the Ryzen 5000 processors. This is by design. Despite Ryzen’s generation-on-generation power consumption increase, Intel’s Alder Lake CPUs continue to use more power while providing less performance.

The renders-per-day-per-watt figures demonstrate that the Ryzen 5 7600X is much more power efficient than the Ryzen 5 5600X in the x264 HandBrake workload, although the base Ryzen 9 7950X is about equivalent to the previous-generation Ryen 9 5900X. As seen below, however, we know from our testing that the 7950X is much quicker than the 5950X in this exact application.

Obviously, the higher power consumption must be seen through the lens of the processor’s performance-per-watt, and here AMD claims to have achieved significant progress thanks to architectural and process node improvements.

Architecture

The Zen 4 architecture, according to AMD, is an incremental improvement over the Zen 3 architecture, but the Zen 5 architecture, due in 2024, will be a complete overhaul. Zen 4 includes several enhancements, though. As you can see in the above images, AMD has made a number of advancements; nevertheless, extending the front end to better feed the execution units and enhancing branch prediction account for sixty percent of the IPC increase. In addition, AMD expanded the op-cache by 1.5 times, shifted to a two-branch-per-cycle prediction, enhanced load/store units, and doubled the L2 cache capacity. The increased L2 cache capacity adds two cycles to the L2 delay and four cycles to the L3 latency. AMD claims that this additional latency is not too damaging since the greater cache capacity results in higher hit rates, which more than compensate for the increase.

AMD has enabled support for AVX-512 instructions, giving it an odd edge over Intel, which pioneered SIMD instructions but disabled them in Alder Lake. AMD characterizes the execution of 256-bit wide instructions in its AVX-512 version as “double-pumped.” This implies that an AVX-512 instruction requires two clock cycles to execute. Nevertheless, this enables compatibility with AVX-512 while enhancing performance. This method also saves die space and mitigates the frequency and thermal penalties generally associated with AVX-512 workloads on Intel CPUs. AMD’s solution results in a poorer throughput per clock than Intel’s technique, but the higher clocks more than compensate for the disadvantage. AMD claims that AVX-512 delivers a 30% boost in multi-core FP32 workloads and a 2.5X increase in multi-core int8 operations compared to Zen 3. As seen by our own benchmarks, the strategy gives a substantial performance boost.

AMD claims that Zen 4 architectural changes result in a 13 percent boost in IPC over Zen 3. AMD also claims that its CPUs provide superior power efficiency in a much smaller packaging than Intel’s Alder Lake processors. The business contrasted its Zen 4 core with Intel’s Golden Cove to demonstrate that it is half the size at 3.84mm2 and 47% more power efficient.

Larger L2 caches on Intel’s CPUs have mostly benefited data center applications. Larger L2 caches often lower L3 cache accesses (theoretically by 40% in this example), which reduces congestion on the fabric and enables improved scalability and performance in all-core applications — as opposed to offering large performance gains in single-threaded workloads. Therefore, there is a probability that Zen 4’s expanded L2 capacity will be more beneficial for EPYC Genoa server CPUs than for the majority of desktop PC workloads. However, this does not mean that AMD will not get advantages for other sorts of work, such as gaming and desktop PC programs – every improvement in hit rates contributes to an increase in IPC.

Die size for the new 6nm I/O die (IOD) is 122mm2, which is comparable to the 124.94mm2 12nm IOD seen in Ryzen 5000 CPUs. It contains 3,400,000,000,000 transistors.

In addition, the Zen 4 compute die (CCD) measures 70mm2, which is slightly smaller than the Ryzen 5000 processor’s 83.74mm2 die. Given that the N5 technology for Ryzen 7000 is substantially denser than the 7nm process for Ryzen 5000, the smaller Zen 4 CCD die includes 6.5 billion transistors compared to 4.15 billion transistors for Zen 3 CCDs (a 36% increase for Zen 4).

AMD Codenames	Core	CCD (Core Compute Die)
Zen 2	Valhalla	Aspen Highlands
Zen 3	Cerberus	Breckenridge
Zen 4	Persephone	Durango
Zen 5	Nirvana	Eldora

Processors and AM5 Socket

The Ryzen 7000 looks like it has a lot going on at the top, where all of the capacitors are spread out over the PCB. This reduces the requirement for capacitors that face the socket, such as the vast arrays of capacitors that are distributed across the LGA pads on Intel’s CPUs. As a result, the bottom of the chip is left unpopulated save for the LGA pad array.

Ryzen’s capacitor arrangement necessitated the aesthetically pleasing heatspreader — the firm could not place the capacitors beneath the IHS owing to heat concerns. At launch, Zen 4 chips will include a maximum of 16 cores and 32 threads, same to the Ryzen 5000 series. AMD has informed us that AM5 will have a lifespan comparable to that of AM4, therefore it is feasible that future generations of Ryzen processors may have greater core counts in this socket.

A motherboard manufacturer posted a video of a Ryzen 7000 CPU being inserted into the new AM5 socket before removing it. Fortunately, we captured several screenshots before they removed the video. AMD is moving away from its long-standing Pin Grid Array (PGA) AM4 sockets and adopting a Land Grid Array (LGA) AM5 architecture with this new socket. The AM5 socket will continue to support AM4 coolers despite the radically changed LGA1718 socket interface (1718 pins).

The AM5 socket is 40x40mm in size, and the Ryzen 7000 chips adhere to the same length, width, Z-height, package size, and socket keep-out pattern as the previous-generation CPUs, allowing for backward compatibility with AM4 coolers. AMD claims that 95% of current coolers will function without the need for new brackets (which cooler manufacturers generally offer for free), and some of the biggest names in cooling, such as Noctua, have already declared backward compatibility with the socket.

Intel’s LGA 1700 socket has a greater pin density than the AM5 socket, which is an intriguing aside.

This is mainly due to the fact that Intel’s sockets have a huge empty area in the middle to accommodate the capacitors on the bottom of the chip. In contrast, AMD has positioned all capacitors on top of the PCB, maximizing socket space.

Interestingly, according to IHS, Ryzen 7000 chips were manufactured and distributed in Taiwan, but Ryzen 5000 chips were distributed in Taiwan but manufactured in the United States.

A photo showing the backside of Ryzen 7000’s integrated heat spreader (IHS) was posted to a Facebook group by an anonymous contributor. The picture reveals quite a deal about the chip, including the fact that AMD’s Zen 4 Ryzen 7000 CPUs will continue to employ solder thermal interface material (TIM). The IHS also looks to be rather thick, which aids in heat dissipation and reduces cooling needs.

We can also see adhesive at each mounting point on the eight ‘arms,’ which is a difference from AMD’s Ryzen 5000 chips, which were sealed all around. The two compute dies ride a heat spreader’s edge. As shown, there is no space for a third die inside the packaging unless AMD dramatically alters the die locations.

On the top of the PCB, the cutouts that accommodate the surface mount devices (SMDs) are visible (these are mostly capacitors). These top-facing SMDs will add a substantial amount of risk to delidding, although this would have little appeal because AMD employs solder TIM. The design does provide a possibility of excess thermal paste leaking onto the capacitors, however non-conductive thermal pastes eliminate this worry. If you use a conductive paste, it may be prudent to apply a sealer, such as clear nail paint, to the exposed capacitors closest to the heat spreader.

Motherboards: X670E Extreme, X670, B650E and B650

Five years, five CPU generations, four architectures, four process nodes, 125+ CPUs, and 500+ motherboard designs have been supported by socket AM4 from AMD. It all began with the modest A-Range ‘Carrizo’ processors that predated even Ryzen, therefore it’s time for a new socket, AM5, and a new series of motherboards. AMD has now committed to maintaining AM5 until 2025+, and it anticipates that the Socket AM4 ecosystem will also endure for a longer period of time.

The X670 and X670E chipsets will be available at launch, while the B650E and B650 will become available in October (AMD will showcase the B-Series boards on October 4, 2022). AMD has said that the price of motherboards would decrease to as low as $125, but has not specified whether this applies to X- or B-series boards.

In terms of connection, Raphael processors will use a new AM5 socket that supports PCIe 5.0 and DDR5 interfaces, mirroring Alder Lake. Socket AM5 motherboards will expose up to 24 PCIe 5.0 lanes to the customer, the most PCIe 5.0 lanes straight from the socket in the industry, and use four extra PCIe 5.0 lanes to connect to the chipset (less expensive motherboards can use a PCIe 4.0 connection to the chipset, AMD recently qualified the interface for Ryzen 7000). Additionally, AMD seems to allow PCIe lane bifurcation, which divides a connection into many allocations (turning an x8 slot into two x4 interfaces, for instance, which is very helpful for SSDs). Alder Lake does not allow PCIe lane bifurcation, and it is unclear if Raptor Lake will support it. Moreover, AMD’s new 600-series chipsets are completely fanless.

In addition to supporting up to 14 SuperSpeed USB 2.0 ports at 20Gbps and Type-C, the chipset also supports Wi-Fi 6E with DBS and Bluetooth LE 5.2. Additionally, AMD has introduced support for USB 4 to their Ryzen platform, thus this feature will also be enabled. Wi-Fi 6E capability is a distinct chip resulting from AMD’s partnership with MediaTek. AMD has not defined (some MSI boards, for instance, use the RZ616). We anticipate to see more discrete chips for support for 40 Gbps USB 4.0 Type-C ports, which we’ve already seen on certain top AM5 motherboards – it’s evident that AMD has substantially increased connection possibilities. The Ryzen 7000 CPUs have the SVI3 power architecture, which allows more power phases from the motherboard and provides quicker voltage response.

Extreme will feature PCIe 5.0 for two GPU slots and one M.2 NVMe SSD connector. This chipset is intended for motherboards that prioritize severe overclockability and connection, establishing a new tier above AMD’s normal range. X670E is the premium category, and although we’ve seen a few inexpensive variants offered at retailers for less than $300, we’ve also seen fully-equipped boards with astronomically high MSRPs that top at $800 for standard models.

The Standard high-end motherboards are powered by the X670 chipset, which will be available in different versions with differing PCIe capabilities. The M.2 port will support the PCIe 5.0 interface, but the first graphics slot may support PCIe 4.0 or PCIe 5.0, depending on the motherboard. This provides a lower-priced sub-tier of PCIe 4.0 X670 motherboards, enabling enthusiasts to avoid the additional expenditures of PCIe 5.0.

Chipset B650 will enable PCIe 5.0 for a single NVMe port, but only PCIe 4.0 for the graphics slot. This chipset also enables overclocking, although as expected, the power provisions are not as strong as they are on more costly boards. (For more information on overclocking, see our tutorial on How to Overclock a CPU.)

The B650E motherboards will feature PCIe 5.0 in both the M.2 and GPU slots, while the B650 will only have PCIe 5.0 in the M.2 slot and PCIe 4.0 everywhere. The only difference between the two kinds of B-series motherboards, according to AMD, is the PCIe connector.

Socket AM5 motherboards enable up to four display outputs through HDMI 2.1 Fixed Rate Link (FRL) and Displayport 1.4 High Bit Rate 3 (HBR3) thanks to the RDNA 2 graphics engine aboard the 6nm I/O die of Ryzen 7000 CPUs.

Five future flagship X670E motherboards from MSI, ASRock, ASUS, Gigabyte, and Biostar were introduced by AMD first. Later, the business had representatives from every major motherboard manufacturer demonstrate their entire product lines during a webcast. Gigabyte’s motherboards have new features such as latchless M.2 ports and PCIe slots. For a complete summary, including specs and images for a variety of AMD AM5 motherboards, see this page. In reality, Gigabyte is already providing overclockers with motherboards. Asus and MSI have also shown their complete lineups of X670E and X670P motherboards.

Despite the fact that AMD has not yet specified its dual-chipset orientation, the images of the motherboards have corroborated many of the previously revealed characteristics. According to our sources, AMD’s standard B650 platform will have a single chipset chip with a PCIe 4.0 x4 link to the Ryzen 7000 CPU. However, materials we’ve seen indicate that certain AM5 CPUs support PCIe 5.0 connections.

Meanwhile, the enthusiast X670 platform utilizes two of these ASMedia chips (our sources claim the chips are similar and not arranged in a north/southbridge fashion), thereby tripling the networking choices. In addition, these chipsets are linked together. In contrast to AMD’s current 500-series motherboards, which use distinct processors for the X- and B-series motherboards, Intel’s 500-series motherboards use the same chip for both the X- and B-series motherboards. Clearly, the new method will give cost and design flexibility benefits.

Another study on the 600-series chipset (codenamed Promontory 21 – PROM21) has corroborated our results and offered more insight into the design’s power- and cost-saving capabilities. The estimated size of the chipset is 40mm2 (19x19mm). If you enlarge the tweet above, you’ll see another intriguing aspect of AMD’s new chiplet-based chipset strategy: motherboard manufacturers may position the chipsets in a variety of orientations. We’ve also seen what seems to be a PCIe switch positioned between two chipsets, presumably for fanout connections, so it’s possible that we’ll observe a variety of implementations.

Finally, AMD has verified that only DDR5 memory will be supported by the AM5 socket. The manufacturer claims that DDR5’s increased performance justifies its higher price, but we’ll need to keep a careful eye on pricing. As previously reported, DDR5 remains more costly than DDR4, primarily because DDR5 is the first major memory generation with integrated power management ICs (PMICs) and voltage regulator modules (VRMs). These have been in chronic short supply because to the pandemic, but happily, DDR5 cost has decreased as PMIC and VRM supply has increased.

Regrettably, DDR5 is still more costly than DDR4 equivalents.

However, due to DDR5’s more complicated power circuits and design, these modules will continue to cost more than DDR4 modules. DDR5 also has ECC techniques for data at rest, necessitating extra dies for the same memory capacity as DDR4. Regardless of availability, DDR5 will remain more costly than DDR4.

Audio driver code has also been contributed to Linux in order to enable the audio co-processor (ACP) included on the chips. This patch arrived rather late, so there may be no out-of-the-box audio support for Ryzen 7000 CPUs under Linux when the chips are originally released; however, this will shortly be remedied.

Pricing

SKUs	Price	Cores / Threads (P+E)	Base / Boost Clock (GHz)	Cache (L2/L3)	TDP / PBP / MTP	Memory
Ryzen 9 7950X	$699	16 / 32	4.5 / 5.7	80MB (16+64)	170W / 230W	DDR5-5200
Ryzen 9 5950X	$546 ($799)	16 / 32	3.4 / 4.9	74MB (8+64)	105W	DDR4-3200
Ryzen 9 7900X	$549	12 / 24	4.7 / 5.6	76MB (12+64)	170W / 230W	DDR5-5200
Ryzen 9 5900X	$398 ($549)	12 / 24	3.7 / 4.8	70MB (6+64)	105W	DDR4-3200
Ryzen 7 7700X	$399	8 / 16	4.5 / 5.4	40MB (8+32)	105W / ?	DDR5-5200
Ryzen 7 5700X	$268 ($299)	8 / 16	4.6	36 (4+32)	105W	DDR4-3200
Ryzen 5 7600X	$299	6 / 12	4.7 / 5.3	38MB (6+32)	105W / ?	DDR5-5200
Ryzen 7 5600X	$199 ($299)	6 / 12	3.7 / 4.6	35MB (3+32)	65W	DDR4-3200

AMD’s price for the high-end 16-core Ryzen 9 7950X is reasonable, but the $100 premium for the Ryzen 7 7700X seems excessive. In addition, the persistent high price of the entry-level model may be a deterrent for aficionados. However, Intel has indicated that it would raise chip prices owing to supply chain interruptions and other causes, so we won’t know the complete pricing picture until Raptor Lake is released.

The X670 and B650 AM5 platforms only support DDR5 memory, which has cost implications for systems developed around the future Zen 4 CPUs from AMD. This indicates that Intel’s Raptor Lake will likely have a platform price advantage over commonly accessible DDR4 platforms, which might be advantageous at the middle and bottom of the product stack. AMD has a countermeasure in the form of less priced PCIe 4.0-only X670 motherboards, but we’ll have to wait and see how this plays out once these boards hit the market.

AMD informed us that it anticipates supply and pricing to improve until the end of the year, and that memory manufacturers are attempting to cut prices. AMD said that a price crossover between DDR4 and DDR5 might come as early as the middle of 2023. Notable is the fact that the most current market predictions indicate a future memory supply glut as a result of the recent unexpected decrease in PC demand, which could also aid price.

All of these variables suggest that you may have to pay more than you would for comparable Intel Raptor Lake systems, at least for the first Zen 4 ‘Raphael’ Ryzen 7000 processors for Socket AM5.