Intel Core i9-10980XE Review: Winning the middle
- 25 November, 2019 19:01
Intel’s new 18-core Core i9-10980XE isn’t reaching for the stars. It’s reaching for the middle. Squeezed by AMD’s consumer 16-core Ryzen 9 3950X on one end, and the 32-core Ryzen Threadripper 3970X on the other, it can’t win on raw performance. But with Intel’s aggressive pricing at $1,000, it actually competes well on bang for buck, a tempting deal for the content creators who are the prime market for this chip.
Why the Core i9-10980XE is almost a deal
The Core i9-10980XE takes Intel pricing in a new direction—down. While the two previous 18-core CPUs were released at $1,999, Intel will be shipping this new generation at $1,000.
For Intel, which has been loath to chase AMD on price wars, this fat discount could potentially change the fortunes for the new CPU. In the craziest turn of events, Intel is actually less per core than AMD’s upcoming Threadripper 3000-series of CPUs (see chart below).
Who’da thunk we could put “Intel” and “deal” in the same sentence. but here we are.
Why Core i9-10980XE isn’t exactly a 10th-gen chip
Before we get to performance, we’re going to try to explain where the Core i9-10980XE falls among Intel’s chip generations, because... it’s confusing.
Obviously, the new Intel Core i9-10980XE is a 10th-gen part because of the ‘10’ in its name right? Well, not exactly. It bears no relation to the 10th-gen Ice Lake CPUs based on the new 10nm process. Nor is it related to the 10th-gen Comet Lake U CPUs, which are basically upgraded versions of the 8th-gen Whiskey Lake U.
The Core i9-10980XE actually doesn’t look all that different from Intel’s previous two 18-core chips, the Core i9-7780XE and the Core i9-9980XE. If you look at all three 18-core Core i9 chips on Intel’s ARK, they’re all 14nm chips, with marginal clock speed improvements every generation.
Core i9-10980XE presents one major change over its predecessors. Intel uses its newer Cascade Lake X cores, introduced in its Xeon line earlier this year. Cascade Lake increases validated memory to DDR4/2933 and officially supports up to 256GB of RAM.
The chip also supports Intel’s DL Boost instructions, which accelerate deep learning and inference performance. The CPU should technically work with Intel’s Optane Persistent Memory DIMMs, too (although there are no announced plans to support it yet). And In a move certain to give some AMD fans jealous rage, Core i9-10980XE will fit into existing X299-based motherboards.
Some of performance promises of Cascade Lake X are still a little hard to quantify, though. You’ll see as we delve into the benchmarks, coming up next.
How we tested
A good price isn't a deal unless good performance comes with it. Can the new Core i9 keep up with its Ryzen rivals?
For this showdown, we installed the Core i9-10980XE in an Asus X299 Prime Series 30 motherboard with a Founders Edition GeForce RTX 2080 Ti GPU, a 512GB M.2 Samsung 960 Pro SSD, and 32GB of DDR4/3200 CL15. The CPU was cooled with a Corsair Hydro H115i with its fans manually set to 100 percent and iCue software unloaded. The latest BIOS was used in the X299 Prime Series 30, and the graphics card driver was the same among all the systems.
And yes, we left MCE on. We’ve lately decided not to turn off Multi-Core Enhancement manually, as we feel that no consumers actually do that. While that can induce slight variances as motherboard makers use MCE slightly differently, the results a consumer will see are closer with MCE on, rather than off.
3D Modelling performance
We’ll kick this off in the area that an 18-core CPU lives for: 3D modelling and rendering. Among the tasks run on the PC, 3D modelling and visualization typically offers the biggest return on investment for multi-core CPUs.
First up is Maxon’s Cinebench R20 benchmark. This updated benchmark uses the same engine you’ll find in newer versions of the company’s Cinema 4D application and now supports AVX2 and AVX512 as part of its workload.
Intel has typically done fairly well on the older Cinebench R15 and Cinebench R20, but as you can see from the results below, Ryzen 3000-chips have flipped that on its head. Despite having two more cores than the Ryzen 9 3950X, the Core i9-10980XE still loses. The only answer that explains this surprise is that Ryzen 9’s Zen 2 cores are simply faster in Maxon’s rendering engine.
To try to gauge the performance of a single core, we also run Cinebench R20 using a single thread. The 18-core Core i9-10980XE is nudged aside by the 5GHz Core i9 consumer chips, as well as both the 16-core and 12-core Ryzen 9 CPUs. Oof.
The visualization world isn’t totally focused on Cinema4D performance, though, so we also run the “proudly CPU-based” Corona renderer. It’s an unbiased photorealistic renderer (“unbiased” basically means it doesn’t take any shortcuts in work). Here, the 18-core Core i9-10980XE rallies and lands in front of the other CPUs we tested. In fact, it’s about 8 percent in front of the 16-core Ryzen 9 3950X.
Up next is the Persistence of Vision Ray tracer that got its start on the Amiga. Luckily, it’s been updated along the years for more modern hardware. Like all 3D modelling apps, more cores means more performance—normally. But as with Cinebench R15, we see the 16-core Ryzen 9 3950X slightly outpace the 18-core Core i9 10980XE chip.
All is not lost as we get to the Chaos Group’s V-Ray Next renderer, which was used in such movies as Avengers: Endgame. The Core i9-10980XE slightly pulls ahead of the 16-core Ryzen 9 3950X.
It’s clear to us that 3D modelling is mostly a give-and-take between the Ryzen 9 and the Core i9-10980XE, with certain rendering engines favoring Ryzen and others favoring Core. Overall, it feels like a tie in 3D modelling—which, if you’re AMD, isn’t bad, because its top Ryzen 9 has two fewer cores and costs about 25 percent less.
Moving on to video encoding performance, we want to point out that video encoding usually doesn’t scale with core count as efficiently as 3D rendering. A platform’s memory bandwidth, cache performance, special instructions, and microarchitecture can easily affect the outcome.
Our first test tasks the free and popular HandBrake 1.2 to convert a 4K video using the HEVC encoding to 1080p resolution. The results put the Ryzen 9 firmly in front—but we have to say—we wonder if we didn’t misrecord this performance run, as we expected it to be closer to the Core i9. We’ll rerun our test and update this section as needed.
Even so, if you look at the performance of the 18-core Core i9-10980XE against the 8-core Core i9-9900KS, it’s a win, but not exactly what you might expect with the core count.
Besides encoding our 4K file to 1080p HEVC, we also did runs at 2160p with a 60-fps frame rate. We’ve found this workload seems to scale better with core count, as the 12-core Ryzen 9 3900X is now handily in front of the Core i9-9900KS. We also see the scores of the Core i9-10980XE and Ryzen 9 3950X close up nicely. The win, however, still handily goes to the Ryzen 9 3950X. With two tests in a row going to Ryzen 9, we’re wondering if maybe Ryzen 9 is actually faster after all.
Up next is Cinegy’s Cinescore 10.4 test. This is a free benchmark intended to let broadcast professionals gauge performance of commercial off-the-shelf hardware (that’s the PC, folks) at handling broadcast-focused codecs. You’ll notice that we again don’t see that huge a leap as we go from 8 to 12 to 16 and 18 cores. While it doesn’t outright lead, the Ryzen 9 3950X has a slight edge over the Core i9-10980XE.
Our next test shifts gears from the broadcast industry to Adobe’s immensely popular Premiere Pro 13.5. For this test we take a short 4K video shot on Sony Alpha cameras and output it using the Blu-ray preset in Premiere, with the Maximum Render option checked. We also direct Premiere to use the CPU for its encoding rather than the default of the GPU. Such a test is useful because despite the encoding performance of GPUs, some video nerds will turn their nose up at anything except CPU-based encoding for the utmost in quality.
The result? It’s essentially a tie between the Core i9-10980XE and the Ryzen 9 3950X.
We wanted to dial it up a notch though, so we took the same workload and exported it using HEVC to 2160p, using the High Profile. The result finally gives the 18-core Core i9 some breathing room, as we see it finish the encode about 15 percent faster.
Just like the 3D rendering section, the good news for Core i9-10980XE is it can be faster. The bad news is it can be slower too. Depending on how you look at it, it’s not that bad, or it’s a disaster, because the Ryzen has fewer cores and costs less, too.
Keep reading for gaming performance and more.
Did you really buy a 16-core or 18-core CPU to play games?! said that old time gamer who doesn’t understand that besides playing games, you might also stream it to multiple platforms at once and then also re-edit it, add in 3D graphics and other core-count-intensive actions that old-school gamers just don’t get.
So yes, gaming can be important to someone who also needs a big, fat, hairy CPU. We’ll also say that we essentially skip reporting performance at 2566x1440 resolution, because you’re just GPU-bottlenecked most of the time and the performance gap between any of the CPUs is fairly marginal. And if you’re playing at 4K resolution, even with the mighty GeForce RTX 2080 Ti, it’s not going to matter which CPU you have 9 out of 10 times.
The first game is Gears of War 5, which is fairly GPU-bound even at 1080p. What we can see is that higher clocks matter, and it doesn’t look like the Core i9-10980XE has quite enough. The fastest is Intel’s all-5GHz Core i9-9900KS chip. The Core i9-9900K follows next, and then the pair of Ryzen chips. The Core i9-10980XE is in last place but in the grand scheme of things, the game is well-playable on all five CPUs. So, we wouldn’t make a federal case out of it.
Gaming, by and large, does like Intel, and as we move to Shadows of the Tomb Raider at 1920x1080 set to Highest Quality and with DXR turned off, the Core i9-10980XE pulls ahead of both Ryzen chips and is outpaced only by the 5GHz Core i9 chips. But again, just as its slower performance in Gears of War 5 wasn’t a huge deal, this isn’t really either.
Sometimes with testing, nothing makes sense. Far Cry 5 is one of those games that runs better on Intel despite the Ryzen and Radeon logos splashed in your face at launch. We handily see the high-clock Core i9 chips lead the way. And yes, Ryzen chips just do poorly in Far Cry 5, as we expect. In fact, this is what we’ve come to call the “Ryzen gap,” where games simply run 15 to 20 percent slower on AMD.
But what explains the performance of the Core i9-10980XE? It should arguably be running at a slightly higher clock speed than the Ryzen chips, yet the results are essentially a tie. We’re simply at a loss here, and in a few benchmarks it’ll get even weirder.
Our next game is Deus Ex: Mankind Divided. Again, it’s at 1920x1080 set to Ultra quality. We use this game in particular because it was one of the games that was optimized in the days of first-gen Ryzen, which plain dragged in gaming against Intel. The result? Let’s call it a wash. Yes, technically the 5GHz Core i9-9900KS is out in front, but not enough to really matter. And yes, the Core i9-10980XE is technically slower, but not enough to really matter.
Our next game is the DirectX12 poster child: Ashes of the Singularity. The game was one of the first optimized games for multi-core CPUs under DX12. The reality is it doesn’t scale up to the core counts we have today. The results are again mostly a wash: The Core i9-10980XE is last, but not by enough to matter.
Most of the games we ran are GPU-intensive. The reality is most of the games played today aren’t like that at all. To look at that category of eSports game, we use Tom Clancy’s Rainbow Six Siege at 1080p ad, with quality set to Ultra. In a better look for the Core i9-10980XE, it’s basically dead even with the Core i9-9900K. All three Core i9 chips have decent margins over the Ryzen 9 chips, but we’re already running in the 360-fps range, so it’s not like anyone’s hurting for frame rates.
Our conclusion on gaming is that the Core i9-10980XE pulled out respectable, but not stellar scores. But then, so do the Ryzen 9 3950X and Ryzen 9 3900X. Basically you can’t make a bad choice, so don’t chose on gaming performance, rather choose on the activities surrounding the gaming. If you do indeed need to render 3D graphics to drop in a game video, more cores will help—and also cost more money.
We’ll close off by looking at the performance of the Core i9-10980XE running from 1 thread to 36 threads, using Maxon’s older Cinebench R15. If you look at the graphic below, you can immediately see the issue for Core i9-10980XE: At no point does it ever hold a commanding lead over Ryzen 9.
Presented as percent difference, it looks far more drastic for Core i9, which is essentially under water against the 16-core Ryzen 9 3950X the entire time.
Granted, Ryzen 3000-series chips simply rock Intel in Cinebench these days. And if we could run V-Ray Next or another renderer this way, the Core i9 situation would likely improve. The only problem with that is while the situation is likely to improve, it’s only going to improve a little. Even if the Core i9-10980XE held a 7-percent advantage, it’s supposed to, as it has two more cores and costs 25 percent more.
What about DL Boost?
One cool new feature of the Intel’s new Cascade Lake X chips is the support for DL Boost which accelerates AI and Inference work loads. You’re not seeing the results here because they include an unreleased CPU. We can tell you that yes, it looks like applications and tasks that take advantage of Intel’s DL Boost give it a big leg up over the competition. So yes, if you’re an AI or machine language researcher or developer, just having access to this in an 18-core, $1,000 chip is a reason to consider Core i9.
When Intel slashed prices a month ago on its new Core i9 CPUs, we thought the move was a sign that Intel was getting serious about AMD and was ready for the long fight.
Seeing the performance of the Core i9-10980XE though, we now realize the move was really more about survival. If Intel had launched the Core i9-10980XE at the same $2,000 price as its predecessors, it would have been immediately irrelevant in the face of the Ryzen 9 3950X. And yes, at $2,000 it would be crossing the path of the 32-core Ryzen Threadripper 3970X.
In the end, while some might say $1,000 is still too expensive, we think it’s actually not too horrible of an ask in the grand scheme of things. Yes, $800 would make us feel better but considering the CPU it replaces started at $2,000, we’re not expecting miracles.