인텔 코어 울트라 20000: 성능 및 효율성 검토

### 인텔의 코어 울트라 20000 리뷰: 게이머를 위한 혼합 가방
* 💻 최근 기술 적용 범위에서 긍정적인 방향으로의 전환이 강조되고 있지만, **인텔의 코어 울트라 20000 프로세서**는 이러한 철학에 도전장을 내밀고 있습니다.
* ⚡ AMD가 더 낮은 전력 소비로 비슷한 성능을 달성했음에도 불구하고, 인텔의 새로운 Arrow Lake 프로세서는 **5%의 성능 향상**보다는 즉각적인 FPS 향상을 원하는 많은 게이머에게 성능이 부족한 것으로 보입니다.
* 📉 코어 울트라 5 245K 및 코어 울트라 9 285K는 **타일 기반 아키텍처를 도입**하고 캐시를 늘렸지만 게임 성능에서 **AMD의 제품에 비해 고전**하고 있으며, **285K는 종종 최신 칩에도 뒤처지는 경우**가 많습니다.
* 📊 게임 벤치마크에서는 **시티 스카이라인 2와 같은 타이틀에서 일부 개선**이 있었지만, **레드 데드 리뎀션이나 쉐도우 오브 더 툼레이더와 같은 고사양 게임에서 전반적인 성능은 실망**스러웠으며 AMD의 칩과 비교했을 때 **'학살'이 두드러짐**.
* 💡 특히 전력 소비가 **최대 40%까지 감소**하는 등 전력 소비가 크게 줄었지만, 성능에 대한 대가로 게이머들의 반응이 엇갈리고 있습니다.
* 생산성 테스트 결과, 특히 렌더링 작업에서 **새 칩이 이전 칩보다 성능이 뛰어난 것**으로 나타났지만 전반적인 경쟁력, 특히 AMD의 Ryzen 시리즈에 대한 경쟁력은 여전히 의문입니다.
* 📉 코어 울트라 9 285K의 가격은 **589달러**로, 새로운 마더보드와 DDR5 메모리가 필요하기 때문에 AMD의 제품을 더 쉽게 이용할 수 있는 시기를 정당화하기 어렵다는 점이 또 다른 장애물입니다.
* 🔄 궁극적으로 인텔은 Arrow Lake를 통해 "재도약"을 목표로 하고 있지만, 그 결과는 **데스크톱 CPU 시장에서 "회복을 위한 먼 길"**을 시사하며 게이머들은 이러한 새로운 프로세서에 대한 투자에 **"회의적"인 태도**를 보이고 있습니다.

인텔의 새로운 CPU는 이전 세대와 AMD의 성능을 뛰어넘지 못해 전력 효율이 개선되었음에도 불구하고 게이머들에게 실망감을 안겨주고 있습니다.

Part of my new philosophy lately has been to bring more positivity to our coverage. I mean, certainly, we have a responsibility to acknowledge and educate about problems, but we also want to highlight the good. That's going to be hard today when we watched Intel's presentation on their new Core Ultra 20000 desktop processors. Our first thought was, Boy, do these ever look familiar. And while we were impressed that AMD was able to achieve similar performance at significantly less wattage, not many of you were gamers. Don't want 5% improvement; they want FPS, and they want it now. And, well, kind of kills me to have to be the one to tell you this, but Arrow Lake is more like arrow through the heart because you're not even going to get 5%, at least not across the board. The good news is that your power bill is going to go down, but in many cases, so will your performance. So, strap in for a very bumpy ride with me as I try to navigate what on Earth is going on with CPUs in 2024.

but first, a short layover to visit my aunt's Segway to our sponsor Ridge. Their MagSafe wallets are back in stock. It's the Ridge wallet you know and love, but this time it attaches to your phone. Crazy! Check it out in the link in the description and pick one up today. AOL Lake is full of changes, some of which are pretty radical. I mean, for starters, gone is SMT or hyper-threading, along with Intel's traditional monolithic die approach on the desktop. Instead, we get a tile-based architecture with the FOS packaging that we first saw in Meteor Lake. And they're laying down clusters of Skymont efficiency or E-course between our now single-threaded Lion Cove performance or P's. We also get four XE Graphics cores on the iGPU tile that should be a lot stronger than Intel's previous attempts at integrated graphics. And while it clearly isn't on the level of AMD's 3D V-Cache chips, they have upped the level two cache per piece core by 50% from two Megs to three Megs. Wow.

there's also been some big changes to core turbo behavior. More on that now. You thought I was going to say later, didn't you? We've got two chips with us today: the Core Ultra 5 245K, that's a 6P core, 8E core chip that boosts to 5.2 and 4.6 GHz respectively. This is Intel's intended gamer sweet spot, at least for now. Then we've got the Core Ultra 9 285K, that gets 8P cores and double the E cores for a total of 24 cores. This one should be the productivity and also just do anything monster. It'll do 5.5 GHz on the P cores and 4.6 on the E cores, at least it sort of will, because the 285K has five, yes five, different boost frequencies depending on what you're doing. It technically can actually turbo up to 5.7 GHz, but that's only going to be a couple select cores in lightly threaded workloads. All right, thanks.

Intel, as for the last CPU, the one we don't have today, it gets eight P cores but only 12 e-cores and a slightly slower clock speed. So far, so interesting, but how's it going to play out for the folks shopping today? We wanted to compare Intel's own last-gen AMD's Ryzen 9000, and for the folks gaming on the top chips from the last couple of generations, we included AMD's most popular X3D chips. A couple of important points, though. We decided to roll the latest Windows 11 24 H2, and our 96 and 9700X chips are tuned to the new 105-watt TDP that AMD unlocked last month. And finally, we didn't bother with the 5800 X 3D for productivity; that's a gaming chip. So, with that in mind, let's talk gaming performance. All of our tests were run at 1080p low to all but eliminate GPU bottlenecks and put as much pressure on these CPUs as possible. In City Skylines 2, not only do we see a

decent improvement over 14th gen in both averages and 1% lows, but the 285k is beating everything in 1% lows. The only problem is it's crushing our 245k as well. It turns out in heavily multi-threaded games, this current generation gaming chip is looking decidedly last gen. It's not all bad news; we found some generational uplift in Total War: Warhammer 3. It's not much, and also the 285k gets kind of stomped on by the 9700x in both averages and 1% lows, and also the 245k near the bottom of the pack. And that's it; that's the positive news for gaming. As we always say, look to multiple sources to paint the most complete picture, but for the rest of our suite, the results are very disappointing. Shadow of the Tomb Raider has the 245k in second last place, beating the 14600 K only in 1% lows, with the 285k not doing a whole lot better. Our x3d chips crush

it in averages, and even the vilified 9,000 Series chips from AMD come out ahead. But AMD is far from Intel's only enemy here. Even Intel's own 14700 K is decently ahead of our new Core Ultra 9 in average FPS. Fortunately, that's not the case in Rocket League, but the 285K is still losing to its predecessor. It seems that this game doesn't care much about 3D vCache but loves lots of cores and lots of threads, and is the only game we tested that placed the 9950 X on top. That's bad news for our Core 5, which trade places depending on whether you're looking at averages or 1% lows. Yikes. And in Red Dead Redemption, we see another case where not only is AMD beating Intel into the ground, but 14th gen is right alongside getting in some kicks where it can. This is a massacre, ladies and gentlemen. An F123 doesn't look a whole lot better. I mean sure, they're beating last gen here, but not by a whole lot, and

you know what else beats last gen? Our new commuter backpack. It's got all the tech-forward design you love from our OG backpack, but in a smaller, tighter package. Sign up for a notification at lm.g/c Commuter or at the link down below. Moving on to Counter-Strike 2, both new chips are solidly in the bottom three, with no improvements to write home about, and the 285k is down about 7% in averages from last gen. Cyberpunk sees a boost, but only to averages; it actually drops in 1% lows for our 285k. And as for the poor Core Ultra 5, it's at the bottom of the chart, behind a 2-year-old chip running in an 8-year-old socket. Truly, this is a sad day for gaming. And therefore, the W, or is it? We asked Intel to tell us their top two and bottom two games that showcased their new power efficiency gains, and we tested them ourselves with the 285k. And while it's clear you're going

to see some huge swings in performance depending on what you play with these new chips. It is equally clear that you're going to see a reduction in power consumption across the board, something that Intel, unlike AMD, really needed to get under control with this generation. We'll touch a bit more on power and thermals later, but you can see here that yes, in Final Fantasy 14 Dawn Trail, we're losing 20% of our performance, but we're also dropping 40% of our power consumption. In Hitman, that was a similar story with a 45% decrease in power for only a 133% loss in FPS. And those are the worst two. What about the best? In CV 6, we've got a 35% power reduction while gaining almost 9% performance on the turn timer. And in F123, nearly 40% less power with 12% more average FPS. Don't get me wrong, guys, these are truly small slivers of silver on some very large cumulonimbus clouds.

but they are there. Justin Intel has informed us that despite claims of APO or application optimization being enabled by default on AOL Lake, it wasn't in fact off on the motherboard they sent us for our review. We've rerun the tests and found about a 7% average frame rate increase in CS2, about 6.5% increase in Shadow of the Tomb Raider, and a decrease of about 6% in Red Dead Redemption 2's 1% lows. The rest of our games only saw a 1 to 2% change, which could be attributed to run variant. Some of our productivity scores even went down slightly with the feature enabled. For the record, APO should be enabled by default on all boards, either in time for launch or shortly thereafter. I can't believe I'm still doing this character. It also appears that Windows 1124 H2 can have lower performance in games compared to other Windows versions. This is the version our testing used, but since 24 H2 is publicly available, you may even be

using it right now. We're standing by our results. If you want to learn more about the wide variety of performance you can expect depending on which versions of Windows you're running, check out Hardware Unbox coverage as they tested the 285K on multiple Windows versions and power modes, showing some pretty big changes in FPS. Your mileage may vary. We're still not impressed with a lake when it comes to gaming. Back to you [Music] Lonus. I don't expect almost any gamer to look at this and think, I am very compelled to buy this product. But what I can hope is that, like we saw with AMD Zen one, that maybe this is the start of a long journey to recovery for Intel. Because if they can figure out what's going on with the games where they're getting absolutely dumpstered, productivity looks pretty okay in Blender. There's a solid decrease in time taken to render for the 285K against the 14900K, and we get about a 10% improvement between the core fives as

well, unfortunately for Intel, the 950x is still on top. But considering that the 245k beats both the 9700x and 9600x, I'm seeing this as a bright spot for Intel, a surprising one given the lack of hyp threading on this new generation. Compiling with GDau sees more improvements for a lake 2 with our 245k once again beating the AMD 9700x. It may only have six cores, but those ecores are really starting to show their strength now that we see them do more heavy lifting in multi-threaded workloads. Intel has really invested in making these new Skymont ecores beefy to the point where they're really more like underclocked Raptor Cove P-cores. In Handbrake, we see more improvements in AV1 encoding. The Core Ultra 9 sees a 22% uplift over last gen, with about a 15% uplift for the Core Ultra 5. Neither of them can crack the behemoth that is the 32-thread 950x, but it is nice to see.

that the 245k isn't completely useless. Trailing just behind the 9700x, x264 encoding is a similar story with smaller gains over previous gen on both chips, but hey, now our 245k is beating the 9700x by a hair. And we see very similar results in Prime seeve, with our 285k in second place and the 245k admirably beating its 9600 X counterpart. Especially interesting in our results is Cinebench. The 285k finally lands on top of a chart in multicore, and both A-Lake chips comfortably beat their 14th gen counterparts. The lack of P-core multi-threading and the extra E-core clusters really hurt the Accor Ultra 5, but at least in single core, it puts on a respectable show, trailing just behind the 14900 K. Just like in games, though, the story here heavily depends on the workload, and we got just abysmal results in Puget bench's Photoshop, with both new chips performing worse than last gen and

straggling behind AMD by a wide margin in Premier. Intel is better than AMD, but this was already their fight to lose, and unfortunately, they managed it. AOL Lake doesn't even manage to beat their last gen 14th Gen processor, so pck. Victory and seven zip is another disappointing loss in both compression and decompression. Wamp wamp. Can we maybe find some more silver? How about AI? I mean, after all, unlike AMD's most recent desktop offerings, these new Intel chips have NPUs, or neural processing. And at least when it comes to open Veno, we really see them shine with our 285k, netting an 82% increase over the 14900 K CPU score in the same test. And since our 245k has the same NPU tile, it gets nearly the same score there as well, walloping the 14600 K. Here we start to see what all the fuss is about with Intel's new iGPU as well. Those four XC cores are putting in work with more than double the performance of last gen.

Unfortunately, AMD CPUs don't support open VOO; they use Windows ML, and in that race, they beat Intel's CPU scores. However, in most cases on Intel, you won't be using Windows ML. So, depending on your use case, these new NPUs might really matter to you because, I don't know, you can afford a brand new CPU and motherboard but not a GPU. I don't know; we'll have to see how developers use these things. What could actually be a factor for you is power. While our 285K peaked at 250 watts to top the charts earlier, it only averaged 218, which is a 30.3-watt drop from the 14900K. To be clear, Intel is still clearly juicing their chips harder than AMD, but at least it's finally coming down, and it looks even better when we compare our 245K to the 14600K. Okay, less power, higher score; you'll love to see it. It's just too bad that AMD twisted the knife on their most recent launch. It's also just not that

clear of an improvement. We looked at Intel's best case for performance per watt in gaming earlier, but in our usual process for power and thermal testing, we use F123 and Gen over Gen. It's more of a mixed bag here. Yeah, they're using fewer watts on both SKUs, but surprisingly, they're running a few hotter. So, it looks like unless they're using significantly less power, you can expect a pretty similar operating temperature. But hey, at least you can reuse your old cooler with the new LGA 1851 socket. Oh yeah, I only briefly mentioned this before. If you want these new chips, you're going to need a new motherboard. Sick, truly. Guys, I don't remember the last launch that left me this disappointed. I mean, there is a ton of cool stuff here to get excited about, like the potential prediction improvement thanks to removing SMT and the new knobs and dials for overclocking. Leaks have even popped up showing the 285K reaching over 350 watts for an alleged 82% performance gain. It's just

too bad that the majority of you guys watching this video are probably never going to try your hand at overclocking. Beyond enabling XMP on your RAM anyway, speaking of RAM, big changes there. We're getting CU dim support and a DDR5-only memory controller. Intel told us that these CPUs should see some decent game gains when cranking up the speed to 8,000 megat transfers per second or greater, which sounds pretty cool, except that ours completely refused to do it. So we weren't able to verify their claims. Stay tuned though, maybe now that both AMD and Intel have settled on DDR5, we'll have to test a wide range of memory speeds on both platforms in the not-too-distant future, maybe an update or two in the meantime will fix Intel's issues in time for that, which leaves us with one last kick in the teeth. When we started writing this review, the Ryzen 9 950X was scarce and prices were high, but as of the weekend before the embargo lift, the

950x went on sale for as low as $599. That didn't last long, and at the time of filming, they're already back to the $700 range. But it's a good way to get us started on the conversation about pricing. Today, you can buy a core Ultra 9 285k for $589. You're also going to need a new motherboard that supports the new 1851 socket, and realistically, probably some new DDR5 memory if you want to try to get faster speeds out of it. The Ultra 7 265k, probably the chip that actually matters in this launch, and unfortunately, the one we have zero benchmarks for, costs almost 200 bucks less at 3.94. And the Ultra 5 245k is 309. Considering that Arrow Lake isn't a drop-in replacement, it is really hard to get excited about this pricing when we look at the competition, especially when we consider that the Ryzen 7 9800 X3D is rumored to be released next month. Intel's calling

AOL Lake a reset on the desktop roadmap, and I guess I'm just hoping that that's true and that this is a Zen one moment. Like, that was a major change with some real improvements, but it wasn't that competitive. The real competitiveness came later with Zen 2 and especially Zen 3. So for Intel right now, unless you're super keen on an NPU on desktop, or you're one of those people who still thinks that AMD is Kirkland brand Intel, it's really hard for me to find a reason to buy these. If you do want to pick one up, we'll have it and all the CPUs that we tested linked in the description down below. As always, I'm your host, Luis Sebastian, and this is a segue to today's sponsor, Squarespace. Have you ever checked out our beautiful website? We built it from the ground up with Squarespace. They make building a website easy with Squarespace Blueprint, a tool designed to give you great layouts and styling options so your site is optimized for every device. And with

their fluid engine editor. You can easily customize those layouts thanks to a code-free drag-and-drop approach. They also have integrated and optimized SEO tools that help your site rank higher in search results. And if you plan on selling products, Squarespace accepts credit cards, PayPal, and even Apple Pay to give your customers flexibility. Start building your website today and use offer code LTT for 10% off at squarespace.com/LTT. If you guys enjoyed this video, you might want to check out our review of the 7800 X 3D. Who knows what the 9800 X 3D will have in store for gamers, but boy, has this line been a very exciting one to watch.

Thank You for Trying, Intel - Core Ultra 200S Review

Linus Tech Tips is a passionate team of "professionally curious" experts in consumer technology and video production who aim to educate and entertain.


Linus Tech Tips

인텔의 새로운 코어 울트라 20000 데스크톱 프로세서

인텔의 새로운 코어 울트라 20000 프로세서는 타일 기반 아키텍처를 특징으로 하며 하이퍼스레딩 및 모놀리식 다이 설계에서 벗어나고 있습니다.

이 칩은 게이머를 대상으로 하는 코어 울트라 5 245K와 생산성을 목표로 하는 코어 울트라 9 285K를 통해 성능과 효율성의 조합을 제공합니다.

새로운 아키텍처에는 더 강력한 통합 그래픽과 코어당 레벨 2 캐시가 50% 증가했지만 일부 사용자에게는 성능 향상이 미미할 수 있습니다.

새로운 게임 칩 세대의 실망스러운 성능

새로운 게이밍 칩 세대는 14세대에 비해 미미한 개선만을 보여주며, 285k 모델은 멀티 스레드가 많은 게임에서 어려움을 겪고 있습니다.

토탈 워: 워해머 3는 약간의 향상을 제공하지만 285k는 9700x와 심지어 인텔의 14700 K보다 성능이 떨어짐니다.

전반적으로 새 칩은 여러 게임에서 낮은 순위를 차지하며, AMD의 9,000 시리즈와 14세대 인텔 칩이 지속적으로 새 모델을 앞섰음니다.

인텔의 새로운 칩: 성능 및 전력 소비

인텔의 새로운 칩은 상당한 성능 향상과 전력 소비 감소를 보여줍니다.

일부 게임에서는 성능 저하가 발생하지만, 전력 소비량도 눈에 띄게 줄어듭니다.

애플리케이션 최적화 기능이 성능에 미치는 영향은 미미했으며, 최신 Windows 버전은 게임 성능 저하의 원인이 될 수 있습니다.

전반적으로 새로운 칩은 생산성 작업에서 약간의 개선을 보였지만 인텔의 게임 성능은 여전히 많이 부족합니다.

인텔의 차세대 프로세서: 놀라운 턴어라운드

인텔의 245k 칩은 강력한 E-코어와 향상된 멀티스레드 성능 덕분에 예상외로 AMD 9700x를 능가하는 성능을 발휘하고 있습니다.

새로운 스카이몬트 E코어는 언더클럭된 랩터 코브 P코어와 비슷하게 견고하게 설계되었습니다.

인텔의 최신 칩은 일부 워크로드에서 어려움을 겪지만, 285k 칩은 14900K에 비해 CPU 점수가 82% 향상되는 등 AV1 인코딩 및 AI 작업에서 상당한 향상을 보여줍니다.

AMD CPU는 개방형 VOO를 지원하지 않고 Windows ML을 사용하므로 CPU 점수에서 인텔보다 성능이 떨어집니다.

인텔의 새로운 NPU는 새로운 CPU와 마더보드를 구입할 수 있지만 GPU는 구입할 수 없는 사용자에게는 중요할 수 있습니다.

인텔 칩의 전력 소비는 감소하고 있지만 작동 온도는 비슷하게 유지되고 있으며 최신 인텔 CPU를 사용하려면 새로운 마더보드가 필요합니다.

인텔의 새로운 CPU 가격 및 경쟁 제품

인텔의 950x 칩은 잠시 599달러까지 떨어졌지만 이후 다시 700달러대로 돌아왔습니다.

새로운 울트라 9 285k의 가격은 589달러이며, 울트라 7 265k와 울트라 5 245k는 더 낮은 가격으로 출시될 예정입니다.

인텔의 애로우 레이크 출시는 데스크톱 로드맵이 재설정된 것으로 보이지만, 곧 출시될 Ryzen 7 9800 X3D의 경쟁력에 대한 의문이 제기되고 있습니다.