Intel's High-End 10th-Gen Desktop Processors Have Hit the 14nm Limit

Illustration for article titled Intels High-End 10th-Gen Desktop Processors Have Hit the 14nm Limit
Photo: Joanna Nelius (Gizmodo)

How much can Intel squeeze out of its 14nm architecture? A lot, surprisingly, especially since things didn’t go as planned for its 10nm process. After six years on the same process, and five on the same architecture, this latest round of desktop processors still manages to make a few changes, most notably a thinner die STIM (solder thermal interface material) that helps improve thermal performance, and a boost clock of up to 5.3GHz on its highest-end CPU, the Core i9-10900K. But when you run the benchmarks and compare it to other processors on the market, it’s clear that Intel has reached its limit with 14nm. After all this time Intel really will need to find a way to move on to a smaller transistor architecture (probably 7nm at this point) if it wants to compete with AMD.

Advertisement

Intel has been using the same 14nm process for the last six years. During that time AMD has made huge strides with its desktop processors, notably increasing both clock speed and core count, while moving to a 7nm process. The current adage is Intel still has the better processors for single-core applications like gaming, while AMD has the better processors for multi-core tasks like 3D rendering. But AMD is still making solid gaming CPUs too. Especially the 3rd generation Ryzens which have comparable performance to Intel’s processors when paired with the same graphics card. And AMD is generally cheaper, making them a better overall value than Intel’s CPUs, depending on the model and what you need a certain CPU for.

While Intel might excel at single-core performance—which is what gaming typically relies on—but its also using a practically antique architecture and process to do that. Intel’s 10nm, which finally launched last year with its Ice Lake mobile architecture, has struggled. Intel CFO George Davis acknowledged that 10nm “isn’t going to be as strong a node as people would expect from 14nm” during a March 2020 Morgan Stanley conference.

Advertisement

But while Intel bills itself as having the fastest gaming processor, I was expecting much better performance from Intel’s the Core i9-10900K, especially compared to the Core i9-9900K.

From what I experienced, if you just bought a two-year-old Intel Core i9-9900K or one-year-old Ryzen 9 3900X, you shouldn’t regret your purchase. Between transcoding, rendering, and gaming, the Core i9-10900K isn’t a mind-blowing improvement over the Intel Core i9-9900K, especially to warrant spending $488. Intel also sent me its Core i5-10600K, and for the price and performance that one is a much more reasonable purchase for $262. It has up to a 4.8GHz boost clock, although I generally reached speeds of 4.9GHz. The Core i9-10900K topped out at a max speed of 5.0GHz in my testing, even with Intel’s Thermal Velocity Boost and Turbo Boost Max Technology, which temporarily give the processor a 10o-200MHz boost at lower core temperatures.

But regardless if you go with a high-end option or a mid-tier option, you’ll need to shell out more money for a new motherboard since Intel redesigned its socket again. That’s a hard swallow for some when 14nm is nearing end of life and with 7nm on the horizon. Getting a new motherboard is more justifiable with the Core i5-10600K, as it gets stellar 1080p ultra gaming performance, but if you’re an enthusiast looking to squeeze out every last frame you can, you’ll only see a small performance bump for twice the cost of the mid-tier Core i5, even though the Core i9-10900K has four more cores.

Advertisement

But let’s get to the results so you can see why I’m struggling to be impressed by Intel’s latest high-end CPU. Our testbed included: RTX 2080 Ti, Asus ROG Maximus XII Extreme, G.Skill Trident Z Royal 16GB (2 x 8GB) DDR4-3600, Samsung 970 Evo NVMe M.2 SSD 500GB, Seasonic Focus GX-1000, and a Corsair H150i Pro RGB 360mm for cooling.

One of the things I noticed right away was how cool both processors stayed. Temperatures were generally between 70-75 degrees Celsius while under load. Granted, I used a hefty cooler, but this seems to back up Intel’s claims that thinning out the die STIM improved thermal performance. However, the power draw never went above 250W, while other reviewers have seen it go as high as 330W. This could be the reason why I saw lackluster performance in comparison: not enough power, although 330W is a bonkers amount of power to draw and most people won’t be building PCs designed to sustain that kind of power draw.

Advertisement

Another problem was the cores on the i9-10900K never hit 5.3GHz, even after I went with Intel’s suggestion of switching the cooling solution from Notcua’s NH-D15 to Corsair’s H150i Pro RGB 360mm, or air to all-in-one liquid cooling. It seems like even the H150i wasn’t enough to bring the temps down so the Thermal Velocity Boost and Turbo Boost Max Technology could kick in and give me the height of the power promised by Intel.

Our benchmarks showed the Core i9-10900K doesn’t beat the Ryzen 9 3900X in multi-core workloads, according to my testing and Gizmodo’s testing. (I, unfortunately, did not have any current-gen Ryzens on-hand, so I compared with our benchmark records.) Where the Core i9-10900K earns a respectable 37650 in Geekbench 4, the Ryzen 9 3900X more than edges it out with a 39850.

Advertisement

The Ryzen 9 3900X also blows away the Core i9-10900K when it comes to transcoding and 3D rendering times. While the AMD processor transcoded a 4K video to 1080p in Handbrake in 264.2 seconds (4.4 minutes), the Intel processor transcoded the same video in 360 seconds (6 minutes). That disparity continues into 3D rendering times with Blender. The Ryzen 9 3900X can render a 3D image in 158.2 seconds (2.6 minutes), while the Core i9-10900K renders the same image in 279 seconds (4.7 minutes).

Advertisement

That’s not a surprise since the Ryzen 9 3900X is a 12-core/24-thread processor, while the Core i9-10900K is a 10-core/20-thread, and AMD is better at handling multi-core workloads. But here’s the surprising part: The Core i9-10900K didn’t outperform its predecessor, the Core i9-9900K, which has 8-cores/16-threads. I did reach out to Intel about this, but unfortunately don’t have any hard answers at the moment. (I’m currently awaiting a new CPU and motherboard to re-test.) Aside from different motherboards, our two test rigs were specced the same. It’s possible that the very expensive motherboard I’m using is not able to handle how much power the Core i9-10900K needs, but again I’m not sure if that is the issue here because the Asus ROG Maximus XII Extreme should be able to do that.

The Core i9-9900K transcoded that 4K video in 351 seconds (5.9 minutes) and rendered the same 3D image in 231 seconds (3.9 minutes). However, in Geekbench 4, the 9900K has a much lower multi-core score, 33912 compared to the 10900K’s 37650. In single-core performance, the 9900K slightly outran the 10900K, 6109 to 6015.

Advertisement

While the 10900K has a boost clock of up to 5.3GHz, I never saw those numbers, only a peak of 5.0GHz on occasion, but even with AI Optimization enabled, I saw 4.8-4.9GHz speeds. This suggests there’s not much gain from one CPU generation to the next on a per-core basis. The 10900K clearly offers better multi-core processing over the 9900K due to its two extra cores, but that doesn’t explain why the 9900K outperformed the 10900K in Blender and Handbrake.

Advertisement

Intel is still unmatched in single-core performance compared to AMD, 5981 to the Ryzen 9's 5376. In a game like Civ VI, where the Ryzen’s AI turn time is 9.08 ms at 4K on high, Intel’s AI turn time is 6.4 ms. However, going back to the Core i9-9900K, it was slightly slower in Civ VI at 6.9 ms.

Even the Core i5-10600K had Civ VI AI turn score close to both of those Core i9s: 6.5 ms. Its single-core score in Geekbench 4 wasn’t too far behind both either at 5836, but was way slower in multi-core workloads—unsurprising for a 6-core/12-thread processor.

Advertisement

Moving on to gaming performance, the Core i5-10600K is really one of the best mid-tier CPUs you can get, price versus performance. It’s within five frames per second compared to the Core i9-9900K on Shadow of the Tomb Raider and has near-identical performance on Far Cry 5. The i5-10600K can even pull in an average of at least 60 fps at 4K ultra in most games with the help of a RTX 2080 Ti.

The Core i9-10900K gets about 10 to 15 fps more in most games than the i5-10600K with the same GPU, breaking more than 130 fps in all games aside from Total War: Warhammer II. There’s no question that if you want the absolute best gaming performance, you go with the i9-10900K. But it’s value proposition isn’t as great as the i5-10600K because of its small gains over the i5-10600K and its high power demands.

Advertisement

Suspected and unsolved issues aside with the Core i9-10900K, Intel has consistently been able to make better processors every generation, even with the same 14nm process. The degree at which makes them worth upgrading to or putting in your next rig depends on what you want your PC to do, and how much money you’re willing to spend, of course.

If you had no need for multi-core workloads, the Core i5-10600K packs plenty of juice, and even if you don’t have an RTX 2080 Ti you’ll still get more than enough 1080p ultra performance with something like a RTX 2060 Super. What you’re saving on the CPU you can put toward the GPU. I can whole heartily recommend this processor. The i9-10900K is for a niche group, and if you want it, go for it. It’s value proposition doesn’t quite do it for me, though.

Advertisement

But given Intel should be moving to a new process next year it just doesn’t seem like the best idea to pick up what is, hopefully, a CPU from its last generation of the 14nm process. If you absolutely need an Intel CPU right now, that i5-10600K is a stunning little deal, but with AMD’s next generation of Ryzen processors expected later this year, and Intel’s move to a smaller (and likely much faster and efficient) process next year it might be better to wait, unless you need that stellar 1080p performance right this second.

README

  • Core i5-10600K is one of the best mid-tier CPUs on the market.
  • Unusual Core i9-10900K benchmarks possibly tied to motherboard issues.
  • Good, but still lower than expected performance for both CPUs.
  • Will need a brand new motherboard.
Advertisement

Staff Reporter, Reviews at Gizmodo. Formerly PC Gamer, Maximum PC.

Share This Story

Get our newsletter

DISCUSSION

It seems like even the H150i wasn’t enough to bring the temps down so the Thermal Velocity Boost and Turbo Boost Max Technology could kick in and give me the height of the power promised by Intel.

What thermal compound were you using between the CPU and the heat sync? You could have the best cooler in the world, but sub-par compound will always bottleneck you. Kryonaut tends to give you a huge boost over the garbage aftermarket compound that typically comes pre-applied to most coolers.