Intel Core i9-9900KS Special Edition Review: More power, less point

We don’t know whether Intel was listening to Taylor Swift when it created its all-5GHz Core i9-9900KS Special Edition chip, but maybe you know the line about Haters gonna hate, hate, hate, hate, hate. Because yes, haters, no matter what the reason was for the Core i9-9900KS SE to exist, we know you’re gonna hate.

After testing this unlocked CPU against its predecessor the Core i9-9900K and the Ryzen 9 3900X, both fine CPUs in their own right, the truth is, as always, more complex. The Special Edition’s all-core Turbo Boost indeed delivers impressive performance. But when you look at bang for buck, Ryzen 9 3900X still has a lot to offer. Read on for the details. 

What is the Core i9-9900KS Special Edition?

Intel’s Core i9-9900KS Special Edition was first announced at Computex. The CPU is built on Intel’s mature 14nm++ process. It is, for all intents and purposes, the very best of the well known Core i9-9900K we reviewed.

While the 8-core Core i9-9900K had an official base clock of 3.6GHz and officially would push one or two two cores to 5GHz on Turbo Boost, the 8-core Core i9-9900KS Special Edition can push all eight cores to 5GHz. 

Note the key words there: Turbo Boost. The Core i9-9900KS does not run full-time at 5GHz all day, every day. It just goes there an awful lot. This isn’t an easy feat to do as a product you can offer with a warranty. Yes, most of the original Core i9-9900K CPUs would run happily at 5GHz, but that was technically overclocking. The Core i9-9900KS Special Edition is sanctioned to do it and will do so automatically. Other than that, it is the same as the original Core i9-9900K. 

It is essentially a “binned” or sorted chip, where Intel takes the very best of its Core i9’s and makes them into the Core i9-9900KS Special Editions. All for the low cost of about $25 over the $489 of a Core i9-9900K CPU. For those who must see them side by side, we’ve rounded up all of the 9900-series parts here on Intel’s ARK.

Because you’ve known about the CPU for months and read our review of the Core i9-9900K as well as our review of the Ryzen 9 3900X, let’s just get into the testing.

How we tested

For this test, we dusted off the same setups that we used in our Ryzen 9 3900X review but updated them with the latest BIOSes, drivers and OS updates that were available as of October 20. Both systems used Nvidia GeForce RTX 2080 Ti cards. For the Intel side, we first retested the Core i9-9900K, then swapped it for the Core i9-9900KS and re-ran the tests. Both setups also used Corsair H80i V2s with their fans and pumps set to 100 percent.

We rose above all of the nuances and politics of running Intel’s Multi-Core Enhance and AMD’s Precision Boost Overdrive. We just left them as they were set on the motherboards, with the latest BIOSes.

3D Modelling and Rendering performance

First up is Maxon’s Cinebench R20. This test uses the 3D rendering engine Maxon sells in its commercial Cinema4D modelling application (and also licenses to third parties such as Adobe to use in Premiere CC). It’s long been a poster child of multi-core and multi-threaded efficiency. Generally, CPUs that have more cores and threads do better.

No surprise, then: While the Core i9-9900KS is a little faster than the Core i9-9900K, 8 cores just can’t beat 12 cores. The Ryzen 9 3900X walks away with test. 

Because most mainstream and popular applications (including Photoshop) run on only one thread, we also ran Cinebench using a single CPU core. Intel typically has the lead here, but it’s actually a mixed bag, with the Core i9-9900KS slightly faster than the Ryzen 9 3900X. The Core i9-9900K is actually third, which surprised us, but the reality is this is mostly a tie. 

We also like to gauge mult-core performance using the Corona Render engine. It’s a third-party, unbiased photorealistic renderer, which means it doesn’t take any shortcuts when rendering a 3D model. Like any 3D rendering engine, it’s also very efficient with core and thread count. For comparison, we left in the result from a Ryzen 7 2700X CPU, though that result was not obtained with the latest BIOS, OS or drivers. The winner, no surprise, is the 12-core Ryzen 9 3900X.

Again, if you were expecting more out of an “all-5GHz CPU,” you shouldn’t. In our experience clock speed often doesn’t trump core count in multi-threaded tasks. We see nothing here to change that view.

We’ll cut our 3D modelling tests a little short, because the results are likely all the same, with the 12-core battering both 8-core chips. We’ll end on V-Ray Next, which is an updated version of the Chaos Group’s V-Ray Next renderer. V-Ray Next is a physically based renderer and, well, like the others here, more cores equal more performance. Again, we left the score derived from the Ryzen 7 2700X for reference, though it was run with an older BIOS. But yes, the last 3D rendering test again proves that 8<12.

Video encoding performance

Like 3D rendering and modelling tasks, video encoding usually leans heavily on CPU cores and threads. However, we also find most video encoders don’t scale as efficiently as 3D rendering engines do. Other factors play a role, such as memory bandwidth, dedicated hardware for encoding, and support for special instructions of a CPU.

Our first test is Cinegy’s Cinescore benchmark. The test is designed to measure performance of commercial off-the-shelf hardware (that’s the PC, folks) for media and broadcast work. Tasks focus on SD, HD, UHD, and 8K resolutions with various codecs (including Cinegy’s own Daniel codec). The test works by loading samples into system RAM, where the performance is measured mostly independent of storage limitations.

The winner is AMD’s Ryzen 9 3900X, but it’s not the crushing victory the CPU saw in 3D modelling. In fact, the Ryzen 9 3900X is only about 12 percent faster in Cinescore than the Core i9-9900KS, even though the Ryzen has 50 percent more cores. As we said earlier, video encoding typically doesn’t scale with core and thread count, and it also hits diminishing returns far quicker than 3D rendering and modelling.

Cinescore is aimed a bit higher than most consumer video needs, so we also run a test more people can relate to: Adobe’s Premiere Creative Cloud. It’s probably the most popular non-linear editor and is used by the majority of small- to medium-sized businesses for video editing. (Eat your heart out, Final Cut Pro.)

For our test, we take a real-world video project shot on a Sony Alpha camera at 4K resolution. We take the three-minute video and output it using the Blu-ray preset, with the maximum render option checked. We also set Premiere CC to do the encode using the CPU rather than the GPU. This may sound odd, but there are indeed video snobs who consider CPU-based encoding as the gold standard.

We’ve found that Premiere CC typically scales with core count up into the low teens. The Ryzen 9 3900X again finds itself in front—but not by much. While 16 percent faster is great, it also doesn’t quite map out with the fact that it has 50 percent more cores to throw at the work. 

Before we finish off Premiere CC, we realize that a lot of people prefer the stupidly good speed of GPU encoding. So we take the exact same project we used above and encode it using the GPU.

You’d think that with the GPU supposedly doing all of the heavy lifting the CPU would no longer matter, but we’ve found it does—and apparently clock speed matters, too. As you can see, when doing a GPU encode both Intel CPUs suddenly take a commanding lead in Premiere CC.

Generally, the Ryzen 9 3900X leads in multi-threaded heavy workloads, and the Core i9-9900KS leads in single-threaded workloads. That’s about all you need to know. We’ll again summarize the situation with Cinebench R15 down below but that’s pretty much the story so far.

Keep reading for the lowdown on gaming performance. 

Source: Read Full Article