Overclocker Claims Intel X299 VRM Temps are a ‘Disaster’ With Skylake-X
One of the major features Intel has claimed with for its new X-Series of CPUs (both the Skylake-X and Kaby Lake-X) is that they’re all unlocked, with higher TDP ratings and a top-end platform to pair with the chips, dubbed X299. However, according to overclocker der8auer, there’s a serious set of problems with every X299 board he’s tested to date.
The first issue is that at least some X299 boards are only shipping with a single 8-pin power connector for the CPU. An overclocked Skylake-X pushing above 4.5GHz can pull more than 300W. According to his tests, the 8-pin cable temperature on his open testbed is hitting 65C in an air conditioned room.
That’s significant, because most people don’t run their PCs in open testbeds, may not have air conditioning, and may not have proper cooling at the bottom of the chassis (I confess, until this video, I’d never actually considered how hot an 8-pin power cable gets under load.) He suspects that in a closed-case configuration an overclocked Skylake-X with just one 8-pin cable could be hitting 90C-95C and recommends against any Skylake-X motherboard limited to just one 8-pin connector.
But the larger problem is the VRM design. Every motherboard der8auer tested — and he mentions three by name: the Gigabyte Aorus X299 Gaming 3 motherboard, the Asus Prime X299-A, and the MSI X299 Pro Gaming Carbon — were tested with the same CPU, a Skylake-X overclocked to 4.6GHz at 1.25v. According to der8auer, this chip is known to hit 5GHz and was chosen for these tests precisely because its behavior had been well-characterized. The cores were tested in the non-AVX version of Prime95 for 10 minutes and temperature readings were taken from the front and back of the motherboard.
Data by der8auer, graph by ExtremeTech. Click to enlarge.
Note that der8auer isn’t certain his temperature probe was in the proper spot on the back of the Asus motherboard, as the gap between front and back temperatures is usually larger than what he observed. But either way, these temperatures show extremely high VRM heatsink temps and it has a direct impact on the CPU’s ability to hold its overclock. As the VRM temps rise, the CPU starts underclocking to 1.2GHz. The MSI board’s better temperatures are apparently explained by more aggressive throttling when the CPU is under load.
In at least two cases, der8auer simply removed the VRM heatsinks altogether, aimed a 120mm fan at the motherboard, and cut the ambient VRM temperature by up to 40C. We’d expect passive heatsinks to be enhanced by adding an air cooler. But the performance of the passive VRM cooling clearly isn’t up to snuff if your goal is to overclock beyond stock speeds.
What manufacturers promise vs. what they deliver
There’s a nasty tangle of expectations and reality when it comes to high-end overclocking, motherboard performance, and what consumers expect. Der8auer puts the blame for this problem on Intel’s decision to pull its X299 launch in by several months, and thinks that boards with better cooling should be shipping soon. But at the same time, it’s not as if motherboards just added VRMs, or as if the X299 platform is the first Intel chipset to require good cooling to overclock well.
The bottom line is this: When enthusiasts invest in overclocking, they often want that investment to pay off with something like a guarantee of performance, stability, and longevity for the parts in question. People pay for larger power supplies, better cooling, and what they believe are higher-quality components out of a belief that said components are required for safe overclocking — and to be clear, sometimes they very much are.
The problem is, from the viewpoint of pretty much everyone, overclocking components voids your warranty. This holds true even if Gigabyte or Asus sells you an overclocked card themselves. If the box says the GPU will hold a 1.6GHz stock clock, they guarantee the chip at that frequency, even if they also advertise overclocking performance. Even the boutique OEM system builders only offer warranties on the overclocks they perform in-house.
The final problem is that it’s a lot less expensive to advertise high-end overclocking than it is to use the components that absolutely guarantee it. Motherboard companies are typically penny pinchers, and their high-end X299 boards aren’t going to move huge volumes compared to mainstream products. I’m not calling out any single company here (or justifying the practice), just pointing out that there’s a nasty collision between what users want, what companies are willing to support, and what their financial incentives are.
Ultimately, I agree with der8auer that this VRM situation is, at best, going to require some replacement of VRM coolers, better thermal paste, or an active fan to resolve. Customers who want to aggressively overclock should be aware of this as a potential issue. If you don’t plan to overclock, there may be no problem here — but if you do, be aware that better VRM cooling may be required.