Overclocker: Intel’s Core i9-9900K May Run Fine On Older Motherboards

One of the long-term differences between Intel and AMD has been their approach to motherboard sockets and upgrade cycles. Generally speaking, Intel switches between chipsets and socket iterations far more quickly than AMD does. This has always been a pain point for Intel enthusiasts. There have always been questions about how necessary these CPU socket upgrades actually were from generation to generation, and a new video from overclocker der8auer argues that Intel’s Z370 motherboard was simply unnecessary based on tests of how the CPU socket behaves under load.

First, some background. One of the changes Intel made with Z370 was to increase the number of pins devoted to providing power and ground. The company’s latest socket has 18 additional pins reserved for power and 14 for ground (this is also true for Z390). The reason Intel could make this change without changing the number of pins in the socket is that these pins were previously reserved and didn’t function.

For his second test, der8auer (real name: Roman Hartung) began taping off power delivery pins to see what the current per pin would be. At 69 pins taped off — meaning 69 power delivery pins out of 146 total not making contact with the motherboard — the current is just 1.92A/pin.

After six hours of testing with no ill effects, the implication seems to be clear. The power delivery in these sockets appears to be massively over-engineered. “69 pins are taped. 69 pins have been running under Prime 95 load with 200W pulling from the CPU,” Hartung says. “So that’s 1.92 amps across a single pin.”

(Note: The above is literally what der8auer says in the video, but the math doesn’t quite line up. Out of 146 pins total, taping off 69 of them leaves 77 pins delivering power, not 69).

Does This Prove Intel Motherboard Upgrades Are Pointless?

Headlines have already gone up claiming that both the power improvements to Z270 and even Z370/Z390 are pointless and unnecessary. This is undeniably possible. AMD’s ability to use the same chipset and socket across multiple product generations is proof that it’s possible to design a socket for longevity. AMD has pledged to maintain backward compatibility with first-generation Ryzen motherboards until at least 2020 (presumably this means through 2020), which means platforms that supported an eight-core or 16-core chip at launch (Ryzen and Threadripper respectively) may retire with 16-core/32-core support if AMD launches doubled-up core counts on both platforms on 7nm. Intel has always had a vested interest in selling motherboard chipsets because it manufactures them directly. While this was true for AMD as well for a brief time, the company hasn’t owned fabs since the GF spin-off 10 years ago.

But we’ve also seen long-term consequences of inadequate power delivery before, especially when overclocking. Back 10 years ago, systems based on Intel’s LGA1156 and P55 chipset had severe problems over time when pushed to the edge, due to physical differences in sockets manufactured by Foxconn. These chips and motherboards could literally burn out due to high current loads, as Anandtech discusses.

So in theory, I think it would have been fine to run the 9900K also on Z270 and it would also be compatible to Z170 of course. But you have to keep in mind that Z270 and Z170 boards are quite a lot worse when it comes to VRM and VRM cooling because manufacturers listened to our feedback and developed their boards much more. So I think it’s kind of good that we have Z390 boards now, with proper VRM cooling that are suitable for the 9900K, but I think it would also make sense to have Z270 and Z170 boards compatible to 9900K to people who want to upgrade their CPUs and don’t want to overclock, but want to keep their mainboard.

Relationship Status: It’s Complicated

Here’s how I personally split the difference on this topic. First, Intel treats its chipsets as a profit center, and it links new motherboard standards to new product iterations. This sometimes adds to the cost of being an Intel customer as opposed to an AMD customer, though the details and degree vary. The fact that AMD is able to take a different approach to this issue, with fewer replacement cycles over time, is proof that Intel is not literally required to switch socket standards as often as it does.

With that said, it is extremely difficult for any journalist or hardware reviewer to categorically declare which power improvements are or are not required. Intel engineers have previously represented to wfoojjaec that they design CPUs to last for at least a decade while operating above 90 percent load and in adverse thermal conditions. Engineers tend to be conservative by nature, and they have to design towards the worst-case scenario, not the best. Changes to a socket that don’t appear reasonable from a mainstream perspective might be made to solve a handful of edge cases that the average user is never even aware of, or to ensure robust operation in high-stress environments years from now, when the physical and electrical characteristics of the CPU are subtly different than they are at launch. There’s also the fact that new socket standards or tweaks to existing standards cover more than just the CPU socket. New design iterations can also make changes to DRAM routing, PCB layers, and board VRMs.

Again, it’s fair to argue that Intel could be more forward-looking than it is when it comes to these topics. The point is, the sum total of the differences between the Z370/Z390 and the Z170/Z270 chipsets may not fully be captured in the types of testing der8auer did. One major change between the Z170 and Z370/Z390 days is that back when Skylake launched, Intel CPU’s actual power consumption still closely conformed to their specified maximum TDPs. Today, that’s no longer the case — a Core i9-9900K is rated for 95W, but will typically draw far more unless you configure the motherboard to clamp the CPU’s performance to that power envelope. A Skylake-era Z170 capable of delivering a steady 140W to the CPU socket (over-engineered compared with the 91W TDP and real-world power consumption of the Core i7-6700K) might destabilize and crash if asked to handle the Core i9-9900K, which can pull up to 165W at stock according to third-party testing. As a reminder, TDP and power consumption are often treated as stand-ins for one another but are two different things. TDP refers to the amount of power a CPU cooling solution must be capable of dissipating over time in order to ensure adequate operation. Power consumption is the amount of power the CPU is actually using.

What all of this means in aggregate is that while the official CPU TDPs may not have changed much, the actual demands on the socket have evolved much more — and those demands may have led to other, more subtle changes in designs. This is to say nothing of the changes motherboard manufacturers often make to Intel’s default power management to increase their own performance by pushing CPUs harder (this type of shenanigan is a topic for its own article).

Rather than focus on trying to determine which of Intel’s CPU socket changes are strictly necessary and which are not, I’d recommend a different approach. When evaluating whether AMD or Intel is a better deal, consider the question in light of how often you like to upgrade your own hardware and how you feel about motherboard swaps. If you like to extract the maximum amount of value from a platform, AMD’s promise of longer-term support might be attractive. If you always tend to upgrade CPU and motherboard at the same time and have no issues with doing so, you may not mind Intel’s approach, even if it costs more money compared with just doing a CPU upgrade.

Continue reading

Intel’s Raja Koduri to Present at Samsung Foundry’s Upcoming Conference

Intel's Raja Koduri will speak at a Samsung foundry event this week — and that's not something that would happen if Intel didn't have something to say.

Ryzen 9 5950X and 5900X Review: AMD Unleashes Zen 3 Against Intel’s Last Performance Bastions

AMD continues its onslaught on what was once Intel's undisputed turf.

Leaked Benchmarks Paint Conflicting Picture of Intel’s Rocket Lake

Rumors about Rocket Lake have pointed in two opposite directions recently, but the more competitive figures are more likely to be true.

Intel’s Iris Xe Max Discrete GPU Is Slower Than the Integrated Version

Intel's Iris Xe Max has debuted, but the discrete GPUs performance is a bit odd, as new benchmarks show.