During AMD’s ‘together we advance_PCs’ event at the end of August, the company unveiled its Ryzen 7000 series of desktop processors, with four SKUs aimed at the mid-range and high-end market segments. After whetting the audience's appetites with that announcement, tomorrow AMD will be officially releasing their long-awaited next-generation CPUs.

The launch of the Ryzen 7000 series brings a lot to digest, for casual fans and hardcore hardware enthusiasts alike. For their newest lineup of chips, AMD has given their desktop CPU platform a top-to-bottom update, not only releasing new CPUs, but releasing an entirely new platform, socket AM5 around it. As a result, for the first time in a few generations these chips are not drop-in compatible with existing AMD motherboards. But at the same time it has allowed AMD to deliver on a collection of platform improvements, ranging from DDR5 and PCIe 5.0 support to improved power management capabilities. AMD has even managed to sneak an entry-level Radeon RDNA2 architecture-based iGPU into the chip.

The Ryzen 9 7950X: 16 Cores, 32 Threads, New 170 W TDP: $699

We'll start, as always, with the CPUs themselves. AMD's flagship for this generation is the Ryzen 9 7950X, a 16 Zen 4 core CPU that AMD is looking to top the charts with for both single-threaded and multi-threaded workloads. The Ryzen 9 7950X has a base frequency of 4.5 GHz and a peak turbo clockspeed of 5.7 GHz, which makes it the highest clocked desktop x86 CPU to hit the market yet.

But don't think AMD's Zen 4 architecture is just about clockspeeds. AMD has also improved the IPC of their CPU architecture by an average of 13% – primarily relying on the addition of AVX-512 instruction support and comfortably larger caches and buffers throughout the CPU – which means that the Ryzen 7000 chips can deliver some significant performance improvements in a variety of single-threaded workloads.

As for multi-threaded workloads, AMD has been able to improve performance there as well, albeit with a reliance on both architecture improvements and higher TDPs to allow for higher sustained clockspeeds. One of the enabling factors here is that the AM5 platform allows for higher chip TDPs – up to 170W in the case of the 7950X – which is some 65W higher than the max TDPs on AMD's fastest 16 core Ryzen 5000 parts. As a result AMD is in a good position to deliver on the "leadership" class performance that the company is after, but not entirely for free.

The Ryzen 9 7900X, Ryzen 7 7700X, and Ryzen 5 7600X

Moving one down the stack is the Ryzen 9 7900X, which is a 12C/24T and 170W TDP part; it has a higher base frequency than the 7950X of 4.7 GHz, but with a slightly lower boost frequency of up to 5.6 GHz.

Below that, AMD has launched one Ryzen 7 part designed for mid-range desktop computing, the Ryzen 7 7700X. This is an 8C/16T SKU, with a boost frequency on a single core of up to 5.4 GHz, and a base frequency of 4.5 GHz. Notably, unlike the Ryzen 9 parts, this part has a more typical-for-AMD TDP of 105W.

Finally, also aimed at the mid-range market and the cheapest member of AMD's new product stack, we have the Ryzen 5 7600X. Offering 6C/12T with a TDP of 105W, the 7600X is Zen 4 at a more reasonable price point. The chip runs at a base frequency of 4.7 GHz, with a modest (compared to Ryzen 9) boost frequency on a single core of 5.3 GHz.

AMD Ryzen 7000 versus Ryzen 5000
AnandTech Cores
Threads
Base
Freq
Turbo
Freq
Memory
Support
L3
Cache
TDP MSRP
Ryzen 9 7950X 16C / 32T 4.5GHz 5.7GHz DDR5-5200 64 MB 170 W $699
Ryzen 9 5950X 16C / 32T 3.4 GHz 4.9 GHz DDR4-3200 64 MB 105 W $799
 
Ryzen 9 7900X 12C / 24T 4.7GHz 5.6GHz DDR5-5200 64 MB 170 W $549
Ryzen 9 5900X 12C / 24T 3.7 GHz 4.8 GHz DDR4-3200 64 MB 105 W $549
 
Ryzen 7 7700X 8C / 16T 4.5GHz 5.4GHz DDR5-5200 32 MB 105 W $399
Ryzen 7 5800X 8C / 16T 3.8 GHz 4.7 GHz DDR4-3200 32 MB 105 W $449
 
Ryzen 5 7600X 6C / 12T 4.7GHz 5.3GHz DDR5-5200 32 MB 105 W $299
Ryzen 5 5600X 6C / 12T 3.7 GHz 4.6 GHz DDR4-3200 32 MB 65 W $299
 

Comparing apples to apples, so to speak, between the new Ryzen 7000 series parts to the previous-generation Ryzen 5000 series parts, Ryzen 7000 has made some big overall improvements to the chips' capabilities. All of the Ryzen 7000 chips offer significant increases in both base and boost frequencies, which bodes well for overall performance. The worst we can say is that AMD hasn't increased their core counts at any price point/market segment, so all of the performance gains we'll see here today are entirely from architecture and clockspeeds, rather than the more immediate MT gains of throwing more silicon at the matter.

AMD's performance gains have been made possible in part through the Zen 4 architecture's superior power efficiency. While the Zen 4 architecture is modest refinement of Zen 3, delivering a 13% IPC improvement, it also gets the big advantage of being produced on TSMC's 5 nm process node, a full node's shrink from the TSMC 7nm process that was used for Ryzen 5000/3000. This efficiency has allowed AMD to boost clockspeeds without breaking the power bank, with the 105W TDP 7700X seeing a 700MHz improvement for no change in TDP. And multi-threaded performance is not left out in the cold, either; by increasing their top TDP to 170W, AMD is able to keep the CPU cores on their 12C and 16C parts at higher sustained turbo clocks, delivering much better performance there as well.

Of course one of the key arguments here is that more power calls for more cooling, which is very much true for the Ryzen 7000 series. Ryzen 7000’s TjMax for its Precision Boost Overdrive technology stands at 95°C, which means that the CPU will use all of the available thermal headroom right up to that point in order to maximize performance.

Although this can be overridden when manually overclocking, none the less the top-end Ryzen 7000 chips call for better cooling than their Ryzen 5000 counterparts. Users will need to employ more premium and aggressive coolers to squeeze every last drop of performance from Zen 4, as most of us are wont to do. AMD for their part has accounted for all of this with their design choices and product marketing, clearly advising Ryzen 9 79x0 owners to use a liquid cooler with these chips. Still, this does mean that AMD is not bundling their own CPU coolers with their retail SKUs, instead directing buyers to fairly powerful third-party coolers.

New AM5 Socket: AM4 Coolers will Support AM5 Too

AMD has also transitioned to a new platform for Ryzen 7000, named AM5. Along with AM5 also comes a new socket, the titular socket AM5, a LGA-1718 socket that is AMD's first use of the LGA form factor for mainstream desktop CPUs. Now what’s interesting is AMD has specified that most AM4 coolers will support the new AM5 socket, which is great for keeping compatibility with existing coolers.

This also means that AM4 is slowly on its way to becoming a thing of the past. While AMD is still (many) months away from replacing their complete Ryzen 4000/5000 stack with Ryzen 7000 parts, today is the first day and the first step to doing so. None the less, AM4 does offer some incredible deals right now (e.g. 5800X3D), as well as support for cheaper DDR4 memory. This sits in contrast to the AM5 platform, which is entirely DDR5-only. Though when it comes to memory AMD does have a small advantage over Intel; whereas Intel's 12th Gen Core chips only support a maximum (JEDEC) speed of DDR5-4800, the Ryzen 7000 chips are officially rated for DDR5-5200.

To go with the new AM5 platform and provide motherboards for their new CPUs, AMD has unveiled four(ish) new chipsets. These are the B650 and X670, as well as their "Extreme" variations, the B650E and X670E. The top-end X670E series will feature both PCIe 5.0 lanes to the top PEG slot and support for PCIe 5.0 NVMe storage devices. The regular X670 chipset, on the other hand, foregoes the mandatory PCIe 5.0 speeds for the PEG slot in favor of easier-to-implement PCIe 4.0. In either case, both versions of X670 are intended to offer a plethora of I/O options, and in keeping with the general tiered structure of AMD's AM5 chipsets, X670 boards will generally offer better designs, better controllers, and better specifications.

The B650 chipsets, meanwhile, are designed to be more affordable, doing away with some of the I/O lanes and overall I/O flexibility the X670 chipsets enjoy. Like the Extreme X670, the B650E is intended for boards that will offer PCIe 5.0 to the PEG slot and NVMe storage. Otherwise, the lowest-tier B650 chipset dials that back to PCIe 4.0 for the PEG slot as well.

For this week's launch, only the X670/X670E boards will be available. Buyers looking for the cheaper B650/B650E boards will need to hold out until October.

New I/O Die: TSMC 6nm For Ryzen 7000

Last, as has been the case for the last couple of Ryzen desktop generations, for the Ryzen 7000 series AMD is constructing their CPUs out of chiplets. All Ryzen 7000 desktop chips are built from an I/O Die (IOD) as well as either one or two core complex dies (CCDs) depending on the SKU. The IOD hosts all of the PCIe 5.0 lanes, the DDR5 integrated memory controller (IMC), and new for Ryzen 7000, an integrated GPU based on AMD's Radeon RDNA2 GPU architecture. All things considered, the IOD used for the Ryzen 7000 is a pretty significant overhaul compared to AMD's previous IOD, with AMD implementing several new performance and power-saving features, as well as further cutting down on power consumption thanks to TSMC's 6nm process.

It’s time to dive deep into all of AMD’s new improvements and changes for its Zen 4 microarchitecture. Over the following pages we’ll, be going over the following:

  1. Ryzen 7000 Overview: Comparing Ryzen 7000 to Ryzen 5000 specifications
  2. Socket AM5: The New Platform For Consumer AMD
  3. More I/O For AM5: PCIe 5, Additional PCIe Lanes, & More Displays
  4. AM5 Chipsets: X670 and B650, Built by ASMedia
  5. DDR5 & AMD EXPO Memory: Memory Overclocking, AMD’s Way
  6. Ryzen 7000 I/O Die: TSMC & Integrated Graphics at Last
  7. Zen 4 Architecture: Power Efficiency, Performance, & New Instructions
  8. Zen 4 Execution Pipeline: Familiar Pipes With More Caching
  9. Test Bed and Setup
  10. Core-to-Core Latency
  11. SPEC2017 Single-Threaded Results
  12. SPEC2017 Multi-Threaded Results
  13. CPU Benchmark Performance: Power, Web, & Science
  14. CPU Benchmark Performance: Simulation and Encoding
  15. CPU Benchmark Performance: Rendering
  16. CPU Benchmark Performance: Legacy Tests
  17. Gaming Performance: 720p and Lower
  18. Gaming Performance: 1080p
  19. Gaming Performance: 4K
  20. Conclusion
Socket AM5: The New Platform for Consumer AMD
Comments Locked

205 Comments

View All Comments

  • Tomatotech - Friday, September 30, 2022 - link

    Nice idea but you’re swimming against the flow of history. The trend is always to more tightly integrate various components into smaller and smaller packages. Apple have moved to onboard RAM in the same package as the CPU which has bought significant bandwidth advantages and seems to have boosted iGPU to the level of low-end dGPUs.

    The main takeaway from your metaphor of the 650w dGPU with a 55w mainboard and 100-200w CPU is that high-end dGPUs are now effectively separate computers in their own right - especially as a decent one can be well over 50% of the cost of the whole PC - and are being constrained by having to fit into the PC in terms of physical space, power supply capacity, and cooling capacity.

    It’s a shrinking market on both the low end and high end for home use of dGPU, given these innovations and constraints and I don’t know where it’s going to go from here.

    Since I got optic fibre, I’ve started renting cloud based high-end dGPU and it has been amazing albeit the software interface has been frustrating at times. With symmetric gigabit service and 1-3ms ping, it’s like having it under my desk. I worked out that for unlimited hours and given the cost of electricity, it would take 10 years for my cloud rental costs to match the cost of buying and running a home high end dGPU.

    Not everyone has optic fibre of course but globally it’s rolling out year by year so the trend is clear again.
  • Castillan - Wednesday, September 28, 2022 - link

    "

    clang version 10.0.0
    clang version 7.0.1 (ssh://git@github.com/flang-compiler/flang-driver.git
    24bd54da5c41af04838bbe7b68f830840d47fc03)

    -Ofast -fomit-frame-pointer
    -march=x86-64
    -mtune=core-avx2
    -mfma -mavx -mavx2
    "

    ...and then later the article says:

    "The performance increase can be explained by a number of variables, including the switch from DDR4 to DDR5 memory, a large increase in clock speed, as well as the inclusion of the AVX-512 instruction set, albeit using two 256-bit pumps."

    The problem here being that those arguments to Clang will NOT enable AVX-512. Only AVX2 will be enabled. I verified this on an AVX512 system.

    To enable AVX512, at least at the most basic level, you'll want to use "-mavx512f ". There's also a whole stack of other AVX512 capabilities, which are enabled with "-mavx512dq -mavx512bw -mavx512vbmi -mavx512vbmi2 -mavx512vl" but some may not be supported. It won't hurt to include those on the command line though, until you try to compile something that makes use of those specific features, and then you'll see a failure if the platform doesn't support those extensions.
  • Ryan Smith - Friday, September 30, 2022 - link

    Correct. AVX-512 is not in play here. That is an error in analysis on our part. Thanks!
  • pman6 - Thursday, September 29, 2022 - link

    intel supports 8k60 AV1 decode.

    Does ryzen 7000 support 8k60 ??
  • GeoffreyA - Monday, October 3, 2022 - link

    The Radeon Technology Group is getting 16K ready.
  • yhselp - Thursday, September 29, 2022 - link

    I'd love to see you investigate memory scaling on the Zen 4 core.
  • Myrandex - Thursday, September 29, 2022 - link

    The table on page four mentions "Quad Channel (128-bit bus)" for memory support. Does that mean we could have a 4 memory slot solution, with one memory module per channel, with four channel support? This way to drastically increase memory bandwidth all while maintaining those fast DDR5 frequencies?
  • Ryan Smith - Friday, September 30, 2022 - link

    No. That configuration would be no different than a 2 DIMM setup in terms of bandwidth or capacity. Slotted memory is all configured DIMMs; as in Dual Inline Memory Module.
  • GeoffreyA - Friday, September 30, 2022 - link

    All in all, excellent work, AMD, on the 7950X. Undoubtedly shocking performance. Even that dubious AVX-512 benchmark where Intel used to win, Zen 4 has taken command of it. However, lower your prices, AMD, and don't be so greedy. Little by little, you are becoming Intel. Don't be evil.

    Thanks, Ryan and Gavin, for the review and all the hard work. Much appreciated. Have a great week.
  • Footman36 - Friday, September 30, 2022 - link

    Yawn. I really don't see what the big fuss is about. I currently run 5600X and was interested to see how the 7600X compared and while it does look like a true uplift in performance over the 5600X, I would have to factor in cost of new motherboard and DDR5 ram! On top of that, the comparison is not exactly apples to apples in the testing. 7600X has a turbo speed of 5.3, 5600X 4.6. 7600X runs with 5200 DDR5 and 5600X 3200 DDR4, 7600X has TDP 105W, 5600X 65W. If you take a look at the final page where the 7950X is tested in ECO mode which effectively supplies 65W instead of 105W you lose 18% performance. If we try to do apples to apples and use eco mode with 7600X, to get apples to apples with 65W of 5600W, then lower boost to 4.6ghz then the performance of the 2 cpu's looks very similar. Perhaps not the way I should be analyzing the results, but just my observation....

Log in

Don't have an account? Sign up now