AMD Zen 4 Ryzen 9 7950X and Ryzen 5 7600X Review: Retaking The High-Endby Ryan Smith & Gavin Bonshor on September 26, 2022 9:00 AM EST
Socket AM5: The New Platform for Consumer AMD
While the star for most CPU launches is typically the titular CPU, there’s a good argument to be had that for the Ryzen 7000 desktop launch, AMD’s overall platform warrants just as much attention, if not a bit more. After all, the company’s up-to-now current AM4 platform was introduced back in 2016, alongside Bristol Ridge, the last of the Bulldozers. A lot has changed for AMD in the last 6 years – not the least being the movement to the Zen core family – and now their underlying platform is getting a major overhaul as well.
The first new consumer platform and associated socket designed by AMD since the start of the Zen era, AM5 is designed to bring three major updates to AMD’s desktop ecosystem. These can be roughly broken down as improving power delivery, adding DDR5 memory support, and boosting the amount of bandwidth and flexibility for I/O devices downstream of the CPU. All three of these things require (or at least are best accomplished) with CPU pin changes, so this is AMD’s first chance in 6 years to overhaul everything that flows in and out of their desktop CPUs.
Physically, the most obvious change is of course the socket itself. AMD held on to using pin grid array (PGA) sockets for its mainstream consumer platforms for over a decade after Intel made the switch to land grid array (LGA), but now with AM5 the company is making the switch as well.
As was the case when Intel switched, the shift to LGA brings a couple of benefits, and a couple of trade-offs. Overall LGA allows for greater pin density; whereas socket AM4 was 1331 pins, socket AM5 is 1718 pins in the same package size. This switch also means that the relatively vulnerable pins are moved from the CPU package to the motherboard, which is modestly safer overall. The good news is that this means the days of trying to fix pins on an AMD CPU with a mechanical pencil are coming to an end; the bad news is that if LGA pins are damaged, the motherboard is essentially done for.
This also means that AM5 gets the vastly improved retention mechanisms required by LGA setups, as opposed to the zero-insertion force (ZIF) sockets of the last 30+ years. So CPUs are now much more securely locked down into their sockets, and the errant cases where trying to pull off a cooler and having the CPU come with it is no longer a possibility.
It’s notable here that AMD’s new socket actually has slightly more pins than Intel’s current LGA-1700 socket, at 1718 vs. 1700. Though sharp-eyed power users will note that Intel and AMD have pretty distinct pin configurations (never mind package sizes), with Intel reserving a decent amount of the of the pin pad for placing capacitors and other electronics on the underside of the CPU package, whereas AMD has placed everything on top.
But despite the significant changes in CPU sockets, AMD has worked to preserve backwards compatibility with standard socket AM4 coolers. This is part of the reason for AM5 chips’ odd heatspreader layout, and while I have some personal misgivings about keeping the clips around, there’s no arguing that it will help get the AM5 ecosystem immediately bootstrapped, as it won’t require entirely new coolers. With that said, this compatibility does not extend to coolers that need to replace the stock AMD backplate, so the compatibility rate is not 100%. Builders looking to reuse an old cooler in a system upgrade will want to look into matters to make sure that their cooler is indeed compatible with AM5.
AM5 Power Delivery: Higher Limits & SVI3 Power Management
A big portion of AMD’s new pin budget for AM5 has, in turn, been invested into power delivery. The AM4 platform was designed for CPUs at TDPs up to 105W – or more specifically, a maximum socket power of 142W. As seasoned overclockers already know, this isn’t a hard limit and it’s long been possible to drive more power than this through AM4. But doing so does exceed the design parameters of the socket, the trade-offs of which aren’t advisable for commercial off-the-shelf systems that may need to be supported for several years.
So for AM5, AMD has significantly increased the maximum power they can drive through their socket. The official TDP limit is 170 Watts, which maps out to a total socket power of 230 Watts, a 62% increase in stock power limits. And AMD is going to be taking full advantage of these higher power limits right out of the gate, with the Ryzen 9 7950X and other high-end processors slated to ship with 170W TDPs. In fact this is where a significant chunk of the multi-threaded performance gains from the new Ryzen chips are coming from, as previously AMD’s highest core count CPUs were TDP-constrained when trying to light up all of those cores.
|Stock AMD Socket AM5 TDP Groups|
|65W TDP||105W TDP||170W TDP|
|Socket Power (PPT) Watts||88W||142W||230W|
|Peak Current (EDC) Amps||150A||170A||225S|
|Sustained Current (TDC) Amps||75A||110A||160A|
Do note that this also means that cooling requirements are going up. For their 2 CCD parts with 170W TDPs, the 7900X and 7950X, AMD is outright recommending a 240mm (or larger) AIO cooler for keeping those chips under control. AMD only suggests tower coolers for the 1 CCD parts, as these have the same 105W TDP as AMD's former top-end AM4 parts. With that said, this doesn't mean an air cooled setup isn't going to be viable with the higher TDP parts, but you will definitely need a good air cooler if you want to keep ahead on dissipating all of the heat a fully loaded 16C processor can produce.
And while AMD’s peak power consumption is going higher, AMD is also implementing additional technologies to better manage power consumption, and to keep their chips from fully ramping up unless they need to. The big change here is that AMD has ported a lot of their power savings technologies from the Ryzen 6000 Mobile family over to the AM5 platform (and Ryzen 7000 IOD), most notably Scalable Voltage Interface 3 (SVI3), which as in the case of the mobile parts, is designed to allow faster and more discrete control over the voltage required from the chip.
SVI3 is a high-speed, 2-way communications protocol between the CPU and the voltage regulators, allowing the CPU to not only better direct the increasingly large number of VRM phases on high-end motherboards, but to keep an eye on VRM health, monitoring things such as power delivery and VRM temperature. This allows an AM5 CPU to make better and more aggressive decisions about power management, as opposed to having to make more conservative assumptions about the state of the VRMs.
Of particular note here, these upgrades allow for a larger number of overall power states the CPU can be in, which in turn allows for more optimal use of the VRMs. For example, this allows automatic phase shedding to shut off some of the VRM phases feeding the CPU when they’re not needed, which boosts the overall power efficiency of the system (extra phases are great when you need a lot of power, but waste power themselves at low loads).
Altogether, AM5 supplies CPUs with 3 variable power rails, which is up from the 2 rails supplied by AM4. The CPU cores and graphics cores get a shared rail, and there’s a separate SoC rail to power most of the rest of the IOD. Finally, there’s a miscellaneous VDD rail that provides power to the GMI/Infinity Fabric interconnects, along with some other, smaller IPs. Separate from that, there are still fixed rails for things like the memory, 3.3v rail, etc.
Native BIOS Flashback Support
With AMD slated to support the AM5 platform through at least 2025, this time around the company is more actively planning around the future of the platform. In particular, AMD is taking a stab at motherboard forward compatibility, which was increasingly an issue for AM4 over its six-year reign.
The big breaker there was that, even when supported by the chipset, older boards required newer BIOSes to support newer generations of CPUs. Thus the only way to make an old board work with a newer CPU was to flash it with a newer BIOS, which in turn required a working CPU to begin with – a catch 22 situation that ultimately resulted in AMD developing a loaner kit program which rented out older AM4 CPUs that customers could use to flash their older boards if they didn’t already have a new CPU for it (e.g. they just bought it off the shelf).
Ultimately, the more elegant solution to the problem is to allow BIOS flashing without a working CPU (or BIOS), which does an end-run around the whole problem. That exists today in the form of USB BIOS Flashback features; however flashback has been a motherboard-level feature that’s normally only found on select high-end motherboards and has to be implemented by the motherboard vendor itself. So for the AM4 generation, it was not a universal (or even widely available) option.
For the AM5 platform, AMD is taking matters into their own hands to make USB BIOS Flashback a universal feature. Ryzen 7000 chips will be able to support USB BIOS Flashback mode across the board and regardless of the BIOS currently installed. As a result, users will always be able to flash an updated BIOS to their AM5 boards, regardless of the CPUs supported by or the operational status of the current motherboard.
The net impact is that when AMD releases future chips on the AM5 platform – say, a Zen 5 chip in 2024 – it will be possible to install a compatible BIOS without first having to use a Zen 4 chip. Though this does assume, of course, that Zen 5 chips will work in current X/B-600 series boards, which is likely but should not be taken as a given. Otherwise, this change also offers a moderate improvement to the reliability of the AM5 platform, by making it so that the BIOS on any board can be restored should it end up corrupted for whatever reason.
Post Your CommentPlease log in or sign up to comment.
View All Comments
Tomatotech - Friday, September 30, 2022 - linkNice 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. Reply
Castillan - Wednesday, September 28, 2022 - link"
clang version 10.0.0
clang version 7.0.1 (ssh://email@example.com/flang-compiler/flang-driver.git
-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. Reply
Ryan Smith - Friday, September 30, 2022 - linkCorrect. AVX-512 is not in play here. That is an error in analysis on our part. Thanks! Reply
pman6 - Thursday, September 29, 2022 - linkintel supports 8k60 AV1 decode.
Does ryzen 7000 support 8k60 ?? Reply
GeoffreyA - Monday, October 3, 2022 - linkThe Radeon Technology Group is getting 16K ready. Reply
yhselp - Thursday, September 29, 2022 - linkI'd love to see you investigate memory scaling on the Zen 4 core. Reply
Myrandex - Thursday, September 29, 2022 - linkThe 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? Reply
Ryan Smith - Friday, September 30, 2022 - linkNo. 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. Reply
GeoffreyA - Friday, September 30, 2022 - linkAll 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. Reply
Footman36 - Friday, September 30, 2022 - linkYawn. 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.... Reply