I really like your style of review. No needless bashing and lots of helpfull information about powerscaling, actual power consumption in gaming and efficiency + tuning. This might be my new go-to channel!
Except he is yet again another reviewer recommending buying Intel without considering the fact that the AM5 platform already supports PCIe 5 and will be supported until 2025 at the very least for another one or two CPU generations. Moreover, there will be the X3D V-Cache versions 7800X3D, 8800X3D and 9800X3D most probably all of which may well be supported by any AM4 motherboard bought today. A shame really.
Absolutely fascinating video. It really tells a whole different story about the 13900K than nearly all other reviews. It seems the factory tuning is sacrificing a lot of efficiency in the name of one-upping the 7950X by a few percent. In reality, the data shows it's capable of being very efficient when max performance isn't the main priority.
Exactly, Intel had to pay a price to not lose performance crown. Until ~170W they seem to have equal or very close perf/watt, then AMD can keep it up until 230W but Intel not, they have to push to 280W to reach same performance and they choose to push even more to beat performance.
That was the same with their 12th and 11th gen. The last time any of their upper end CPUs stayed within their advertised TDP was with the 7th gen. Even the 8700K and 9700K can easily go 50+W over what is written on the box. That also explains why there is no overclocking headroom left. The chips are already running near their limit, way past optimal efficiency. Got an old 3rd gen i5 here and even with 15% over single, 25% over all core boost and 32% over base clock it doesn't even touch it's 77W TDP. And here we have chips that actively pull upwards of 250W by design.
Most of it is the motherboards, they don't come with stock settings and most use massive voltage and unlimited power, causing thermal throttling. Most review sites or channels now days are tech illiterate morons. This channel is awesome and informative. Even sites like Techpowerup have their review of the 13900k only scoring 36k in CB, when in fact is scores 40k. They also claim is uses 200-300 watts gaming, which is insanely BS unless they did something to it.
Really interesting to see how the CPU performs with the manual UC/UV. Most reviewers just focus on full power potential and instantly write off the good parts for the high power consumption. But this shows actually that its a really competent cpu when tuned correctly. Thanks for making the review with this viewpoint in mind
I mean... why buy the most expensive CPU to instantly undervolt it? If you need that power - good. If you dont, go buy i3 12100 and be happy. Its cheap. It sips power. And it gives you the same experience in games as any top CPU on the market. I know what i'm talking about. I've sold 12900k before 13900k launch. To buy new and best. As i've always did. And got myself i3 12100f as a temporary solution. Didnt expect much. But I've noticed zero difference in any game i played. None. Numbers might be lower. But actual gaming experience is the same.
@@PanaehaliTut Why buy the most expensive cpu? Because you can still obtain 90% of its performance with a simple good undervolt setup. That's still going to be leagues above the second-in-line cpus so why wouldn't you? There's no reason to keep stock levels of performance clock, that just exists for bragging rights in marketing, but doesn't benefit end users at all. Realistically now you can get 12900k levels of performance at half the power draw (and sometimes even less). That's incredible, especially if you live somewhere where electricity prices aren't exactly consumer friendly these days. I am strongly considering getting a 13900K for my company/personal use, and fully intend on undervolting immediately so I can both use an air cooler and keep my current power consumption levels equal, while gaining probably 3x the performance of my current cpu. It also usually doesn't make sense to swap cpus each generation, I don't even know why you got rid of your 12900K so quickly, what exactly were you expecting? Double the performance at lower power consumption with stock volting? That just sounds like copium. If I were you I would have looked into a similar undervolt setup on the 12900K and waited at least 2 gens before upgrading - that is, if you don't have a ton of money burning a hole in your pocket. Top end gpus are never sold as good deals that benefit from being replaced every new generation. That spot falls on the 2nd and 3rd in line cpus that tend to have lower margins and higher generational uplift from stock settings.
@@PanaehaliTut Where it does become relevant is cooling solution. Arguably the price of electricity is over emphasized in many reviews but in gaming both peak and sustained power draw will affect real world performance depending on your setup. If you already had a setup for say a 12700 and wanted to get the best, knowing that the 13900 performs so well in efficiency standards (under the right conditions and tuning) may make the transition more compelling. I'm not even talking undervolting, the comparison of running intel powerlimits and such vs motherboard manufacturer ones is relevant for any kind of limited aircooling esp with a large gpu in the case. Honestly, I agree overall. But some people will want to undervolt this for things like small formfactor stuff. Where they only care about performance under extremely limited cooling. An I guess some people might stress a 13100 out a little more than you :P
@@faithnfire4769 at which point does 7700x or 7950x becomes a really good efficient option? Neither of them require the tremendous power draw or cooling requirements of 13900k at full load, and consumes less power during gaming as well
Most reviewers heavily criticized the high power consumption of modern CPUs/GPUs (which is reasonable) but they don't show what you've presented here. I'm impressed with the performance you've achieved by optimizing for efficiency and I hope others realize this too. I'm looking forward for your tuning of a 13600k/13700k. Thank you for this new perspective!
I actually wouldn't mind if Intel would sell them around 4.2-4.5 GHz and with a 80-100W power budget, leaving the 4.8+GHz at 200+W for those who really want/need it. Basically how they used to do it.
@@HappyBeezerStudios Turbo Boost is not efficient or normal operation: entering Turbo Boost territory always means performance to power draw trade-offs. Normal operation is base frequency. So if you look into Intel's offers you see: T: 35W target. SFF high efficiency builds. Non configurable. Turbo boost capable, but constrained. Non-K: 65W target. Baseline, reasonable efficiency ground. Non configurable. Turbo boost capable, but constrained. K: 125W target. Performance tilted, they already give away efficiency. Fully configurable: you can make them be whatever you like within the physical limits. They're also the higher quality silicon which is an oddity in itself. If you really understand the offering you already see what you're paying for and what's available: but blaming Intel and AMD for the power budget is non-sense, the point in 'K'(Intel) and 'X'(AMD) branding is you pay for a fully configurable CPU high quality silicon to make it whatever you want. Being able to totally customize something in this world has always costed money in every field. For reasonable cases both them have way more reasonably priced products that are PnP, so the system is actually right: it's we consumers who misunderstand it.
@@deidian635 And the K and X chips used to be configurable upwards at own risk, now they're only downwards. Back then if you bought a K chip you could easily get 20-30% more clock out of it, at the cost of efficiency. Now they're already at those clocks out of the box.
@@HappyBeezerStudios Less work users need to do: you can run single core 5Ghz+ even with crappy cooling because it comes set that way out of the box. Or really light loads can absolutely fly. When a load overwhelms available cooling the CPU will adjust to it. All under warranty. You really don't know what you are asking by saying it would be better if they sold them like 10 years ago. Nowaydays a CPU can be blazing fast for everyday scenarios out of the box, whatever cooling it has. Obviously good cooling will enable more sustained performance and here is where the end user comes in a pay for play model. Also if concerned about power use setting a single knob(power target) will get anyone wherever they want to be in the efficiency curve. The CPU can still run 5Ghz+ as long as it's within the power budget. PS: my work laptop uses a quite modest CPU(non-K) as is so damn termally constrained that even something relatively trivial like opening a browser makes the CPU throttle, but hey, despite all it can run 4Ghz+(for a few seconds) on a brutally constrained cooling making it somewhat fast for the weak hardware it runs. This is for perspective, but this can only happen due to modern CPUs and how they come shipped.
I must have watched 3-4 other reviews from the main PC content creators.. this is by far my favorite. Straight to the point, comprehensive analysis, and you have the consumer/gamer/all-around perspective that everyone is here for. Thank you Der8auer
Best review so far. I'm saddened by how much PC part reviews have been dumbed down on other channels. Nobody considers tweaking settings anymore. If you're savvy enough to build a PC, you can tune some settings in BIOS. Well done, Der8auer.
because maybe reviewing a CPU is not tweaking it. in this video nowhere there is a full comparison with tunned CPUs. just 1 tweaked. it's not a good review if you don't compare the same things. well done? really? if you want really in depth tweaking of different CPUs, check the skatterbencher channel.
@@kwanro9172 To be fair, the 13600K gives 30% more performance than the 12600K at 30% more power draw, meaning performance/W is the same. And some people have to pay for their electricity. So every bit saved is money saved.
@@HappyBeezerStudios If you buy a 13900K and have to worry about the electricity bill maybe you should not buy a 13900K just saying. Not sure how expensive 1 kWh is for you but this should not make a huge difference for someone that can afford to spend this kind of money on a CPU like this probably paired with a 4090.
Amazing work, Roman. You see everyone talking about how much power this thing CAN draw and nobody talking about how much power it really needs to draw.
It's ok. I like the power consumption implementation but I'd like some real life benchmarks as well. Like if you have a 4090 you'll be playing at 4k on ultra settings, not 1440p or 1080p on esports settings. If that benchmark doesn't show a difference between the CPUs because it's GPU limited then I wanna know about it because then I know that buying it is pointless for normal gaming. Honestly could care less about this 600fps nonsense.
@@IskanderVFX Its total shame really, honest is very rare these day in this trash website so called RUclips. A lot of reviewer nowadays is just people who keep overhyping AMD garbage like they still "cares about consumer" while in facts AMD got caught few times doing sh*tty shady business move and selling overpriced CPU. Guess what? Intel is "bad" but AMD is "good", that's always stupid people mentality which allowed AMD to selling overpriced garbage products like what happened with zen 3 and zen 4 but people can forgive AMD for pulling that BS thanks to all normies and those TrashTube popular channel like LTT, AMD Nexus, AMD Unboxed, Jayz2AMDCents and any other trash channel !!!
The power draw overlays is SO USEFUL, thank you! This is really intuitive data, you're the only channel I've seen do this. Also, adding the 90W PL was brilliant, this adds even more depth to the chip's performance review. Top notch, this review is amazing.
I typically only watch you and GN when it comes to technicals and stuff but with the comparisons this time yours was way easier to pick out differences between gaming and multi thread with their power draws. Overall really liked this video and the tips on what you can do to better power draw along with what happens when you do so. Thanks for the great video!
Yeah I agree, GN seemed to have only looked at power efficiency at multicore loads (full load) but not during games (from what I recall). It's nice to see the bigger picture here because you get to see that they're not only throwing power at the 13900k and it's extremely competitive in terms of efficiency when gaming. That being said, all this work takes a lot of times and makes for messy graphs so I can understand why some things are omitted from channels.
@@zaku2939 Yep and GN also didn't go into low power settings either to find more realistic power/watt sweetspots for both cpus, so all you get is the over-the-top stock overclock that makes zero sense and only exists to fight an arbitrary max performance war, when in reality 90% of the performance can be achieved at half the power.
@@Real_MisterSir Because nobody spends $6-700 on a high end CPU only to essentially "cripple" it. Gamers will want max FPS. People buying it for production want max performance as that's how you make money. Only a few rare enthusiasts with money to burn will do things like that.
@@FluffyWuffy17 Getting a better performance/watt at still high performance levels is not "crippling it". Losing at maximum 10% performance in situations where you're strictly gpu bound, is not crippling by any stretch of the imagination what you even on about?? Gamers don't buy flagship GPUs for max fps anyways, that money goes to the gpu first. Not a flying fuck gives any care to whether you get 500 or 600 fps in csgo, and below that the cpu matters less and less anyways so you might as well undervolt and not lose any noticeable performance whatsoever. These cpus are bought mostly by people who spend a lot of time on their computers in mixed workloads and gaming, someone who doesn't have a dedicated work computer with enterprise hardware, someone who isn't satisfied with pure gaming performance. Some people get these cpus just to be happy about having the best hardware currently available, sure, but realistically that's not the point of these components. If it was, then we would see a lot more focus on pure gaming performance in general at the top tier - stuff like the X3D would be the norm, and any multithreading potential would be a tag-on at a premium cost on top. But reality is that it's the complete opposite. Didn't used to be like this 4-6 years ago, but it is now. And workloads don't automatically pay for themselves just because you can complete a 3D render 10% faster - it doesn't equal 10% more profit. But what IS guaranteed, is a doubling in your electricity consumption, which most certainly can have an effect, especially on smaller companies where you don't just have long queues of render work waiting to be completed. That's enterprise work, and for that you'd be buying enterprise hardware anyways if that amount of time saving truly makes a difference in your earnings to justify the disregard of power consumption. Especially now where electricity prices keep rising. Plus you also need to invest in expensive cooling to even operate at the peak performance output levels, which you may not have space for in your current case, etc etc. Cutting down unnecessary variables is typically much more beneficial to a workload-based usecase, than what the extra 10% performance uplift would net you in time saved and potential money earned.
You absolutely crushed this exploration of power usage and eco mode performance! In this new age of self-inflicted energy production strains and high prices, these companies should be screaming about their eco performance per watt, especially Intel. They're getting hammered over their choice to use ALL the watts. It looks like the 13900k is a really great cpu at 90 watts.
Really insightful. I was considering doing a water cooled build and you showed that air cooling is fine since max power isn't important, this will save a lot of money and simplify the build
Same here, it's great to know that you can run a 13900K at very effective performance while still air cooling it. Thought air cooling would be a thing of the past, but seems like it's still got some years left.
Nice to see the power limit values again and the curve, however as mentioned can you use better liquid cooling to see how a custom water cooler does at high watts ? Of course you go a lot colder with LN2 etc but good watercooling results will be useful to a lot of folk. Thanks for the good work.
This is a really informative and best review that i have seen... Most reviewers do balls to the wall crank it up to the max reviews that do not tell the true story.. So thanks for this am sure gonna do the same tuning when i get my 13900k and 13600k for infotainment.. 👍
I really like the sensible direction your taking with your tuning videos with how brands are getting in a power war with their products now days. Personally i have a down clock script on linux when i run a game and i just figure out what clock a game needs and auto restrict the games along with using the low latency performance scheduler. Then in the desktop i keep full (sensible power) clocks for good response and productivity inside VM's etc. Keep up the good work :) I suppose its kind of good that these things all come hyper clocked and we just turn them down a bit rather than having to mod them etc. I feel like theyre missing a beat in showing people their efficiency a bit more and letting people toggle the power rather than perma run it.
99% of people won't even look at the power draw, they don't care how much power it uses, just the fact that it's too hot but they still just get a more beefy cooler and go with it. Only when you are very conscious about power usage in general would you consider how much your pc draws
Thank you for highlighting what literally no other review has done, namely showing power consumption whilst gaming. All the other reviews are just showing peak power consumption figures during all-core workloads such as in productivity apps, which 99% of people don't care about. Most people want to know about power consumption and thermals when gaming, especially with what's happening to energy prices in Europe right now.
I'm very much interested in seeing 13600K vs 13700K vs 13900K compared at different power points. Will the MT performance even out? Will ST performance be event per watt? Here's me hoping we can get something like that ♥
I think you have the best and most easily digestible charts and I love how you sort them by minimum fps. I prefer to have at least some visual reference for 1% lows.
Yes I think the 1% and 0.1% lows are the most important. Gamers Nexus showed that the 13900k has 0.1% lows as bad as the i3-12100 for eSports games like CSGO. Derbauer is the only major RUclipsr that includes eSports PUBG in the results, so it would have been really good to have 1% and 0.1% lows in the results
I've seen plenty talk about how inefficient the 13900k was in benchmarks...first time I see it tested in game... And showing better efficiency than the 7950x. This was great!
You easily have the best charts and benchmarks then any other tech reviewer. As much as I love Linus his charts are terrible to understand at a glance. Love your dedication
I love the power efficiency tests! Would be cool to see total system power draw with stock 13900K and RTX 4090 vs. power optimized 13900K and RTX 4090 and how their performance stacks.
Wow. You are the best! I'm happy someone mentions your channel somewhere. This is the real information that people should know. Many channels mislead people to think the new gen intel is bad because of the high power consumption. your information told totally a different story.
Appreciate the outside box approach compared to just taking one power figure and driving it into the ground repeatedly memeing “furnance”. Like you said for the 4090 Re:power/tuning…it’s hard to understand why intel and AMD both went with such higher power target on their flagship chips for the 10% gains vs more acceptable powers, but at least buyers know thanks to your clear videos it can be “fixed” with a few tweaks to your liking.
10:32 I appreciate a lot that you care about chart clarity. I always loved that you sort the charts based on min fps, so I kind of missed them a little in here, I hope they'll come back in the future. In any case, fantastic video.
having upgraded from a 12700k to a 13900k this thing is hot and power hungry. cooling it with a deepcool ls520 and it gets the job done. i just limited P1 to 125w and P2 to 250w. thinking of custom loop but for now everything seems to be working fine. overall impressed with this beast of a cpu.
I would love to see this on the 13600k. It still pulls a lot of power for a chip of its class, but with a power limit it would make it even more ideal for a small form factor build.
As far as RUclips goes your reviews have rapidly become my favorite, because you set out to do something specific and just analyze based on that. No pointless back and forth stoking responses from company fans (which there should never be in the first place) and good, well explained data. Personally I'd love to see what could be done with undervolting to see what sort of efficiency gains might exist with default turbo behavior. I greatly appreciate the wattage scaling chart especially because it gives a straightforward look at where performance may plateau (first saw this from Buildzoid and happy to see it here).
Finally a useful review. love how you actually review the performance of the cpus clocked down in power. Finding that sweet spot of efficiency and performance. Thank you for all you do.
Better video on the i9 13900K than Linus tbh. You helped me a lot with undervolting and getting my NH-D15 cooling it properly without thermal throttling
I'm well aware of your many successes outside of this channel but how does this channel only have 100k subscribers? So informative, innovative, entertaining and of course, pet friendly! Keep up the good work!
As a person that tries to undervolt pretty much everything, I'm so glad that a big RUclipsr is finally diving into this topic - the voltage headroom on some of this new gear is absolutely ridiculous😁😁
This is definitely the best computer tech RUclips channel because i dont see such detailed reviews in any other channels about efficiency. It must be done this way rather than onlyshowing game fps numbers
Amazing video. Roman, as engineer myself (simmilar field 2 :) ), i thank you to do some tuning and research. Alot of influencers just put stock settings and pull wrong conclusions. My current and last intel chip i7 8700k takes stupid amounts of voltage stock. If you limit it and tune it takes so much less voltage for stock freq. I think this is behavior on all motherboards for last few gens. Motherboard manufacturers just put voltage to the limit of chip ( i guess becouse marketing, and sales). And chip runs 2 hot.
I have the same setup (Aorus Master + i9 13900KF) at home, and I found that 150W was the sweet spot for me. I do video rendering/transcoding (i.e., Handbrake) and if I leave it at full voltage, I can transcode an h.265 movie in 30 minutes. If I set the wattage down to 150W, then the same movie renders in about 32 minutes. If I tried to decrease the wattage more (i.e., 105W), then the render time went up to 36 minutes. At 150W, the CPU temperature remains about about 87C - 91C, using a 360mm AIO from OCool Eisbaer. This keeps the temps just below where it wants to start throttling performance. I do say I'm not happy about this. End-users shouldn't have to be under-volting their motherboards. I ended up having to buy a new case to accommodate that 360mm radiator, and it still wasn't enough to keep it cool at full voltage.
Best raptor lake review on youtube right here ! You did great with the 4090 too.Seems undervolting/PL is the new overclocking these days with hardware being pushed so hard past peak efficiency.
Thanks for the focus on power and efficiency. I always go for the top tier cpu for gaming anyway, so power and efficiency are a really important differentiator. With competition so great now, I hope intel and AMD really focus on that as a value.
Hi Roman, could you please make a undervolted comparison of the 13900K and the 7950X with single-core and multi-core performance, putting them head to head through the same tests as in this video?
Thanks for comparing performance per watt with different settings. A lot of people need to save electricity dollars, others are in SFF cases and need to limit heat, and the rest of us need some apples to apples comparisons.
Man you are the KING! All reviewers take CPU and just show us Cinebench or Aida power draw which we will never use!!!!! You however showing what matters and real use case scenario and how much power these will draw..... I don't care which is better but I care about charts that shows my use case not Cinebench or Aida only
@@MaxIronsThird "lock[ing] the fps" doesnt magically make 30fps look like 60fps. I think you misunderstand what I mean by "minimum fps". I mean the lowest fps observed - ie the fps during the largest lag spike. if that doesnt drop below my monitor's refresh rate I won't notice any lag at all, because I'm still seeing the same fps. you can lock fps lower than what you're capable of producing, but not higher (i mean, you can, but youre not going to magically hit that fps now if you didnt before)
@@MaxIronsThird i'm talking about lag spikes. ie 0.1% lows. those don't last long enough for that to make a huge difference and even if it does you can only do so much to make a lag spike not feel like a lag spike if that lag spike dips your performance noticeably lower than your refresh rate
@@xdMatthewbx Those aren't common, and if they were, he would've said something, just like he mentioned in the video, there was nothing abnormal with the 1 and .1 %.
I've been digesting almost every review since 13th gen release and I have to say your review is by FAR the most informative and sheds some real light on the subject a lot of people have just talked about but nobody explained really well.. You showed a way to make this cracker of a CPU (and hopefully the other 13th gen as well) into very power efficient power houses!!! This makes the 13900K far more interesting then before.. Our power prices are through the roof as you know (neighbor from left side).. Love your thoroughness of reviewing.. Must have taken you ages to do.. Make sure to get some sleep soon! :)
I dont know much about tuning my PC in my bios. But i had problems with my 13900k spiking at 90C with my AIO. I just tuned it to work at 90W as you did in your vid. my games and everything is running as good as before AND my max C is at 64C! Thank you SO SO SO SO much for this Video!
Solid review! Could you please share what the best optimal undervolting values can be used for good cooling with only a small performance hit? The explanations you said were above my head. Maybe a guide for the average user whenever you have a chance :) Love your content otherwise, hope to see more.
What Intel could do about it is to set 90 W as default limit and add a special Creative/Development mode into Tuning Utility that will automatically kick in only with selective set of apps, that you can change, running. So 300 W would be kicking by your will when your render task is running, or compilation, or compression task etc.
Yep this would be great, they could just have a hybrid power setting that automatically detects what a given workload demands from the cpu, and then apply whichever power limit that corresponds with the task the cpu is performing. If it's strictly P-core bound, it's going to be single-core heavy, like gaming. If it's multithreaded with all cores at work, it's a professional workload. Should be fairly easy to set up. Even for render tasks I feel like the 300w should only kick in during intervals where it optimizes workflow - for example if you're using an interactive render then the 300w could kick in whenever you're repositioning the camera and the render adjusts real-time to your inputs - or it kicks in at the beginning of a render where various photon calculations are made, but then lowers the power level once the actual rendering of the photon path traces kicks in.
I'm sure you've seen the reports that the 7950x gaming performance on Win11 improves when you disable 1 CCD likely due to Windows scheduling issues. Would you consider looking at this for some of the charts you have?
Games only use 1-8 cores depending on age, at least they only scale performance that far, so either way you are wasting your power budget running that second CCD, as far as gaming goes anyways. If only one CCD is active it can have all the power avaliable. Also latency between CCDs has always been an issue with Ryzen, especially when the CCDs only had 4 cores.
the 7700X is the best Zen 4 chip if you're gaming, and its the best because of price to performance. All the Zen 4 CPUs operate within 3-4% of each other in gaming, so if your are gaming don't bother with anything above 7700X imo! any amount of production load in your daily life will be much faster as you go up the tiers tho, while giving similar but about 1-2% less performance in gaming at each incremental tier. (That's right less performance as you go up the tiers.)
Love your review. This review makes you different from other reviewers. I would like to see this type of comparison on 13700K as well. With current increasing trend of energy price, I believe more user will consider to undervolt the cpu. Not all the user want 100% performance, why say no to 90% performance 50% energy cost? For those 1080p gamer (who probably wont use i9 anyway), will look for i7 or even i5 for top performance with lowest cost. For those user who play 1440p/4k gaming, they might want to see how low power cpu can go before hitting the performance hard.
20:20 “Most people are just going to game with their 13900K.” I build two high end computers for an AI code developer using 13900Ks and 4090FEs. The first was in a Fractal Design Torrent case with Noctua’s most capable air cooler, NH-D15. The second used a Fractal Design Pop Air case and a Noctua NH-U12A case. The Pop Air case was slightly modified so that the front fan support would not block two Noctua 140 mm fans. The second computer had a much less capable cooling solution. Testing and Tuning the second computer was done with Cenibench R23, Furmark, Intel Extreme Tuning Utility, CPUID HWMonitor, Windows 10 Task Manager, and Kill A Watt P3. It had an 850 W Gold power supply, so the power at the Kill A Watt should not be over 950 watts. The 13900K ran Cenibench and the 4090FE ran Furmark. I think Furmark is an overkill for a GPU like Prime 95 is an over kill for a CPU. Temperatures were never a problem for the 4090FE. Turbo Boost is meant to take advantage of the thermal inertia for a short time if the CPU is not fully utilized. Thread swapping is not talked about much. You can watch it happen if you look at HWMonitor temperature and utilization if you CPU render a single thread with Cenibench. If the CPU is near its thermal limit it will swap threads with a cooler core. This thread swapping takes time and results in a series of “no op” states on both cores and no work gets done An assumption for tuning the CPU is that the load will be nearly constant and the CPU temperature will be near the 100 C limit with dust compromising the cooling solution. To achieve this I chose to limit the P and E cores maximum frequency and limit turbo boot power and minimize duration so that the CPU’s package temperature was no more than 82 C. The results were good. The CPU’s package temperature was no more than 82 C with a Cenibench R23 multi score just over 80,000. The Kill A Watt showed about 800 W which means the PSU had a load of about 720 W.
So basically if you have the cooling power to cool down the 13900k, it can be to the 7950x in cinebench R20 but just doing the second run on that thing makes it lose consistently as per your own results which is I believe under a minute at full power. That means even a 360 AIO is useless for this thing if you're going to run it at Full Tilt for productivity and it's just better to go with the 7950x
This is the best 13th gen review I've seen so far. Thank you so much for making this efficiency video. It gave me a completely different perspective about the 13th gen. I hope you can further explore all three new CPUs. With 13700k and 13600k, the price equation would enter the picture if the performance is still there with power limitations. I'm looking forward to it if you do make more videos about it.
I hope we can forget "nanometer" naming schemes as X3D chips and other 3d stacked chips come out. I hope they name new processes like "le epic stack 2.0+"
Great job by Intel gen 13th delivering very decent price to performance, meanwhile Amd has been a lot disappointing with Zen 4, not only Amd is overheating but also the Cpu itself is overpriced, r9 7950x is $150-200 more than i9-13900K still Amd can't beat Intel in full benchmark, not to mention Intel still compatible with DDR4. Intel totally destroyed Amd here.
Excellent Review 👍🏻👍🏻 🙌🏻🙌🏻 It's great to see performance vs wattage and your explanation of the most efficient use of each of the top performing processors! Thank you !! Looking forward to your delidding video!!
This is the best video on raptor lake chips if you want power efficiency i've seen. It make complete sense. you are quite good at presenting the data as well in terms of scaling and watts to perf.
that's Intel's fault really, they could have do the same Eco Mode switch in the BIOS as AMD does and there were no confusion about how good/bad Intel's efficiency, you cannot expect a regular user tweaking in the BIOS themselves
Filling out the graph at 15:12 with other CPUs would be fascinating and give a true comprehensive comparison of efficiencies at different power levels. The 5800x3d then might challenge 13900 at different power levels. Currently looking at upgrading my PC and have no affinity with either brand, so content like this is invaluable.
Very interesting reviews! I ran into your channel while in Germany last week (thank you RUclips algorithm I guess) and I really enjoy your review style. I am looking to upgrade to a brand new PC coming from an almost ancient 6700k and 1070 (impossible to properly run my flying/driving sims), very curious to see the upcoming reviews of the 13600k/13700k and 4080! And of course AMD.. but never used AMD before.
hey man have to say best 13900k review period. I'm interested on to replace my 12700k, how well do you think it will perform limited about 200 to 260 watts? those are the defaults limits on my bios for the 12700k
I think realistically we need to see if all this is still viable on a 13700k, does it still maintain efficiency or does it play different (as in a more drastic loss in performance). I would only listen to der8auer in this regard. Then again that's up to whether he has time for that. Amazingly helpful video!
TL;DR: This applies to any CPU of any gen. Running more Ghz has a steep power draw cost: anything over 3Ghz will throw efficiency though the balcony until you can't simply sustain it any longer due to going out of spec in temperature/current/voltage. Long version: my 11900K can do 4.5Ghz all core on a 125W power budget, running 3.5Ghz just costs 70W and running 5.1Ghz costs 250+W, all core stress test type loads factoring out vectorial extensions: with them power figures go up to a maximum of +25% power draw depending on workload and the extensions used, but on return the CPU crunches numbers significantly faster(this is wild terrain: algorithm dependent, may be +40%, +100%, +200% execution speed). There's afaik many details with performance and efficiency which are common across every hardware. Those looking for functionality should stick with manufacturer already customized hardware for a easy life.
Excellent review as always. Like some already wrote i like how you dive a bit deeper than most other channels with optimizing efficiency. It shows that for just a little less performance we can save a lot of power and heat output too. Thank you for all the hard work and please keep it up 🙂😁
I'm a competitive Call of Duty player, the 1% and 0.1% are probably more important to me than the average abd certainly maximum frame rates. Therefore, I would strongly encourage you include those numbers, particularly when you're reviewing gaming CPUs. Also, at least in my mind when you're looking at a bleeding edge CPUs, the first people buying them, including the RTX 4090 are looking at high resolution and/or high refresh rate monitors in the competitive multiplayer games. Meaning while I'm certain Cyberpunk, Microsoft Flight are good to demonstrate DLSS, ray tracing, etc I would tend to believe the competitive first person gamer is looking at these CPUs and GPUs above all others. Therefore, it is wise to include those games. Trust me, I love your stuff and your information is always very pertinent, sometimes critical information is lacking for their depth in the relevancy of the products. In this case, top tier GPUs and top tier CPUs. Also, I'm assuming that the so-called 12th generation contact frames are also useful here .. I ordered one.
ive never paid attention to those numbers, what are they and what do they mean for gaming? I admit currently I mostly play Warframe, So as long as I am above 90fps(at 4k HDR, which my 3090 does do in that game) I can usually blast around the maps like jackrabbit rabbit fresh out Walter White's Winnebago no problem.
@@filanfyretracker .. I upgraded from my RTX 3090 to an RTX 4090 because @ 3440x1440 even with DLSS Quality I was capped around 180 FPS. Since I'm replacing my 175Hz to a 240-260Hz monitor the 4090 made sense. I only play multiplayer Call of Duty.
You turned to my favorite tech RUclipsr, very useful videos. Want to try this beast in Unraid server, with limited power consumption. Probably will be nice idle power draw
The 13900k undervolted using less power while having higher FPS than a 5800x3D is impressive. Neat stuff. Looking forward to upgrade my 12700k, DDR4 RAM, and 3080 Ti to a 13900k, DDR5 RAM, and 4090 in a couple months.
I've stopped watching other obsolete pc tech channels because der8auer is on the top of the game. No one can compete and even Jay2cents is just half qss copying this guy 😂😂
You're the man, been hoping a reviewer would do a power examination while gaming, been asking other reviewers in their comments, and then I got to yours and you actually did it. Next time, I do yours first. Yes the 13900k is way power inefficient in benchmarks and workloads, but in gaming, it's actually very efficient.. Also the artificial watt limitation? Beautiful work. Thank you sir. I'm quite happy with my i5-12600k and hopefully won't be upgrading for a few years, but I do build systems as a side hustle, and it's good to know what practical power/thermal gaming loads look like. But honestly, thank you for not ranting about the power consumption, every other review I've seen so far has pretty much ignored performance/power usage in games to rant about them under max loads, which I get, sucks, but when will a gamer ever see those? So does it really matter?
yep and same with Nvidia's new Ada architecture. This upcoming gen of laptops will be beastly if they can manage cooling slightly better than last gen ultrathins that didn't care one bit for cooling despite throwing power hungry hardware in them and thermal throttling constantly.
performance per watt comparisons are tricky. higher performance on per core basis, is achieved at lower perf/watt. Perhaps we could find SKUs for each architecture at 8-core, adjust frequency to match performance, then compare perf/watt
I really like your style of review. No needless bashing and lots of helpfull information about powerscaling, actual power consumption in gaming and efficiency + tuning. This might be my new go-to channel!
He's the only reviewer thus far that is technically correct and running on ideal setup and tweaks.
Mine is this as well as Optimum tech for his gaming opinions because I think Ali is only one who can actually aim & do great reviews
Might be? This guy should be watch maker cuz he’s got the clocks son, strait up Grandfathering
Except he is yet again another reviewer recommending buying Intel without considering the fact that the AM5 platform already supports PCIe 5 and will be supported until 2025 at the very least for another one or two CPU generations. Moreover, there will be the X3D V-Cache versions 7800X3D, 8800X3D and 9800X3D most probably all of which may well be supported by any AM4 motherboard bought today. A shame really.
@@lagarttemido thanks for giving us an example of a shill, great example, scarily accurate
Adding the power consumption to every chart was genius.
Especially useful for 139HX, we need to reconcile consumption and performance.
Absolutely fascinating video. It really tells a whole different story about the 13900K than nearly all other reviews. It seems the factory tuning is sacrificing a lot of efficiency in the name of one-upping the 7950X by a few percent. In reality, the data shows it's capable of being very efficient when max performance isn't the main priority.
Exactly, Intel had to pay a price to not lose performance crown. Until ~170W they seem to have equal or very close perf/watt, then AMD can keep it up until 230W but Intel not, they have to push to 280W to reach same performance and they choose to push even more to beat performance.
That was the same with their 12th and 11th gen. The last time any of their upper end CPUs stayed within their advertised TDP was with the 7th gen. Even the 8700K and 9700K can easily go 50+W over what is written on the box.
That also explains why there is no overclocking headroom left. The chips are already running near their limit, way past optimal efficiency.
Got an old 3rd gen i5 here and even with 15% over single, 25% over all core boost and 32% over base clock it doesn't even touch it's 77W TDP. And here we have chips that actively pull upwards of 250W by design.
that's true of basically every processor, the last 10% or so of performance requires WAY more energy than the other 90%
Most of it is the motherboards, they don't come with stock settings and most use massive voltage and unlimited power, causing thermal throttling. Most review sites or channels now days are tech illiterate morons. This channel is awesome and informative. Even sites like Techpowerup have their review of the 13900k only scoring 36k in CB, when in fact is scores 40k. They also claim is uses 200-300 watts gaming, which is insanely BS unless they did something to it.
Really interesting to see how the CPU performs with the manual UC/UV. Most reviewers just focus on full power potential and instantly write off the good parts for the high power consumption. But this shows actually that its a really competent cpu when tuned correctly. Thanks for making the review with this viewpoint in mind
Thanks! Happy to hear
I mean... why buy the most expensive CPU to instantly undervolt it? If you need that power - good. If you dont, go buy i3 12100 and be happy. Its cheap. It sips power. And it gives you the same experience in games as any top CPU on the market.
I know what i'm talking about. I've sold 12900k before 13900k launch. To buy new and best. As i've always did. And got myself i3 12100f as a temporary solution. Didnt expect much. But I've noticed zero difference in any game i played. None. Numbers might be lower. But actual gaming experience is the same.
@@PanaehaliTut Why buy the most expensive cpu? Because you can still obtain 90% of its performance with a simple good undervolt setup. That's still going to be leagues above the second-in-line cpus so why wouldn't you? There's no reason to keep stock levels of performance clock, that just exists for bragging rights in marketing, but doesn't benefit end users at all.
Realistically now you can get 12900k levels of performance at half the power draw (and sometimes even less). That's incredible, especially if you live somewhere where electricity prices aren't exactly consumer friendly these days. I am strongly considering getting a 13900K for my company/personal use, and fully intend on undervolting immediately so I can both use an air cooler and keep my current power consumption levels equal, while gaining probably 3x the performance of my current cpu.
It also usually doesn't make sense to swap cpus each generation, I don't even know why you got rid of your 12900K so quickly, what exactly were you expecting? Double the performance at lower power consumption with stock volting? That just sounds like copium.
If I were you I would have looked into a similar undervolt setup on the 12900K and waited at least 2 gens before upgrading - that is, if you don't have a ton of money burning a hole in your pocket. Top end gpus are never sold as good deals that benefit from being replaced every new generation. That spot falls on the 2nd and 3rd in line cpus that tend to have lower margins and higher generational uplift from stock settings.
@@PanaehaliTut Where it does become relevant is cooling solution. Arguably the price of electricity is over emphasized in many reviews but in gaming both peak and sustained power draw will affect real world performance depending on your setup. If you already had a setup for say a 12700 and wanted to get the best, knowing that the 13900 performs so well in efficiency standards (under the right conditions and tuning) may make the transition more compelling.
I'm not even talking undervolting, the comparison of running intel powerlimits and such vs motherboard manufacturer ones is relevant for any kind of limited aircooling esp with a large gpu in the case.
Honestly, I agree overall. But some people will want to undervolt this for things like small formfactor stuff. Where they only care about performance under extremely limited cooling. An I guess some people might stress a 13100 out a little more than you :P
@@faithnfire4769 at which point does 7700x or 7950x becomes a really good efficient option?
Neither of them require the tremendous power draw or cooling requirements of 13900k at full load, and consumes less power during gaming as well
I have found THE channel. This is the one.
Hats off to you & your legit testing & power utilization/under volting manual tuning.
Excellent review, thank you for showing the performance & power consumption in eco mode.
Most reviewers heavily criticized the high power consumption of modern CPUs/GPUs (which is reasonable) but they don't show what you've presented here. I'm impressed with the performance you've achieved by optimizing for efficiency and I hope others realize this too. I'm looking forward for your tuning of a 13600k/13700k. Thank you for this new perspective!
I actually wouldn't mind if Intel would sell them around 4.2-4.5 GHz and with a 80-100W power budget, leaving the 4.8+GHz at 200+W for those who really want/need it. Basically how they used to do it.
@@HappyBeezerStudios Turbo Boost is not efficient or normal operation: entering Turbo Boost territory always means performance to power draw trade-offs.
Normal operation is base frequency. So if you look into Intel's offers you see:
T: 35W target. SFF high efficiency builds. Non configurable. Turbo boost capable, but constrained.
Non-K: 65W target. Baseline, reasonable efficiency ground. Non configurable. Turbo boost capable, but constrained.
K: 125W target. Performance tilted, they already give away efficiency. Fully configurable: you can make them be whatever you like within the physical limits. They're also the higher quality silicon which is an oddity in itself.
If you really understand the offering you already see what you're paying for and what's available: but blaming Intel and AMD for the power budget is non-sense, the point in 'K'(Intel) and 'X'(AMD) branding is you pay for a fully configurable CPU high quality silicon to make it whatever you want. Being able to totally customize something in this world has always costed money in every field.
For reasonable cases both them have way more reasonably priced products that are PnP, so the system is actually right: it's we consumers who misunderstand it.
@@deidian635 And the K and X chips used to be configurable upwards at own risk, now they're only downwards. Back then if you bought a K chip you could easily get 20-30% more clock out of it, at the cost of efficiency. Now they're already at those clocks out of the box.
@@HappyBeezerStudios Less work users need to do: you can run single core 5Ghz+ even with crappy cooling because it comes set that way out of the box. Or really light loads can absolutely fly. When a load overwhelms available cooling the CPU will adjust to it. All under warranty.
You really don't know what you are asking by saying it would be better if they sold them like 10 years ago. Nowaydays a CPU can be blazing fast for everyday scenarios out of the box, whatever cooling it has. Obviously good cooling will enable more sustained performance and here is where the end user comes in a pay for play model.
Also if concerned about power use setting a single knob(power target) will get anyone wherever they want to be in the efficiency curve. The CPU can still run 5Ghz+ as long as it's within the power budget.
PS: my work laptop uses a quite modest CPU(non-K) as is so damn termally constrained that even something relatively trivial like opening a browser makes the CPU throttle, but hey, despite all it can run 4Ghz+(for a few seconds) on a brutally constrained cooling making it somewhat fast for the weak hardware it runs. This is for perspective, but this can only happen due to modern CPUs and how they come shipped.
My 13700k is pulling 60w at PGA 2k23 all maxed out 1440p and 165fps stable. Its running 45-48c 💪💯👍
I must have watched 3-4 other reviews from the main PC content creators.. this is by far my favorite. Straight to the point, comprehensive analysis, and you have the consumer/gamer/all-around perspective that everyone is here for.
Thank you Der8auer
Best review so far. I'm saddened by how much PC part reviews have been dumbed down on other channels. Nobody considers tweaking settings anymore. If you're savvy enough to build a PC, you can tune some settings in BIOS.
Well done, Der8auer.
because maybe reviewing a CPU is not tweaking it.
in this video nowhere there is a full comparison with tunned CPUs.
just 1 tweaked.
it's not a good review if you don't compare the same things.
well done? really?
if you want really in depth tweaking of different CPUs, check the skatterbencher channel.
So you're telling me you'd buy a $700+ cpu just to undervolt it and get the performance of 13600k or last gen? okay buddy
@@kwanro9172 no he said tweaking it.
@@kwanro9172 To be fair, the 13600K gives 30% more performance than the 12600K at 30% more power draw, meaning performance/W is the same.
And some people have to pay for their electricity. So every bit saved is money saved.
@@HappyBeezerStudios If you buy a 13900K and have to worry about the electricity bill maybe you should not buy a 13900K just saying. Not sure how expensive 1 kWh is for you but this should not make a huge difference for someone that can afford to spend this kind of money on a CPU like this probably paired with a 4090.
I like how you and your team showed the PPW stuff with power limits. A nice perspective not normally explored. Steve, take note lol.
I still think what is most impressive is the eco modes and their efficiency
Exactly! I was surprised by 7950x efficiency but this video persuaded me, that even 13900k can be power efficient. That's great!
@@fybyfyby AMD know what they are doing, even the last gen Zen was efficient but they go cought up in Intels mhz game.
12900T at 2.2Ghz had power efficiency comparable to AMD laptop chips. Every chip has a VF curve and you can decide where on that curve you go.
Ye, this is what it looks like when silicone is EOL lol.
@Coffee Fueled Curmudgeon but wouldn't that be the same as getting a 12900K or 12900KS ?
Amazing work, Roman. You see everyone talking about how much power this thing CAN draw and nobody talking about how much power it really needs to draw.
Absolutely THE best review out there of 13900k.
It's ok. I like the power consumption implementation but I'd like some real life benchmarks as well. Like if you have a 4090 you'll be playing at 4k on ultra settings, not 1440p or 1080p on esports settings. If that benchmark doesn't show a difference between the CPUs because it's GPU limited then I wanna know about it because then I know that buying it is pointless for normal gaming. Honestly could care less about this 600fps nonsense.
You're doing much better testing than the bigger channels bro, keep it up
night and day, most other channels got me guessing and confused with the 13900k but this review was brilliant
Exactly and much better delivering the info
@@IskanderVFX people just trashed the 13900k as a " 350 W" inefficient CPU
@@IskanderVFX Its total shame really, honest is very rare these day in this trash website so called RUclips. A lot of reviewer nowadays is just people who keep overhyping AMD garbage like they still "cares about consumer" while in facts AMD got caught few times doing sh*tty shady business move and selling overpriced CPU. Guess what? Intel is "bad" but AMD is "good", that's always stupid people mentality which allowed AMD to selling overpriced garbage products like what happened with zen 3 and zen 4 but people can forgive AMD for pulling that BS thanks to all normies and those TrashTube popular channel like LTT, AMD Nexus, AMD Unboxed, Jayz2AMDCents and any other trash channel !!!
The power draw overlays is SO USEFUL, thank you! This is really intuitive data, you're the only channel I've seen do this. Also, adding the 90W PL was brilliant, this adds even more depth to the chip's performance review. Top notch, this review is amazing.
I typically only watch you and GN when it comes to technicals and stuff but with the comparisons this time yours was way easier to pick out differences between gaming and multi thread with their power draws.
Overall really liked this video and the tips on what you can do to better power draw along with what happens when you do so.
Thanks for the great video!
Thanks!
Yeah I agree, GN seemed to have only looked at power efficiency at multicore loads (full load) but not during games (from what I recall). It's nice to see the bigger picture here because you get to see that they're not only throwing power at the 13900k and it's extremely competitive in terms of efficiency when gaming. That being said, all this work takes a lot of times and makes for messy graphs so I can understand why some things are omitted from channels.
@@zaku2939 Yep and GN also didn't go into low power settings either to find more realistic power/watt sweetspots for both cpus, so all you get is the over-the-top stock overclock that makes zero sense and only exists to fight an arbitrary max performance war, when in reality 90% of the performance can be achieved at half the power.
@@Real_MisterSir Because nobody spends $6-700 on a high end CPU only to essentially "cripple" it. Gamers will want max FPS. People buying it for production want max performance as that's how you make money. Only a few rare enthusiasts with money to burn will do things like that.
@@FluffyWuffy17 Getting a better performance/watt at still high performance levels is not "crippling it". Losing at maximum 10% performance in situations where you're strictly gpu bound, is not crippling by any stretch of the imagination what you even on about?? Gamers don't buy flagship GPUs for max fps anyways, that money goes to the gpu first. Not a flying fuck gives any care to whether you get 500 or 600 fps in csgo, and below that the cpu matters less and less anyways so you might as well undervolt and not lose any noticeable performance whatsoever.
These cpus are bought mostly by people who spend a lot of time on their computers in mixed workloads and gaming, someone who doesn't have a dedicated work computer with enterprise hardware, someone who isn't satisfied with pure gaming performance. Some people get these cpus just to be happy about having the best hardware currently available, sure, but realistically that's not the point of these components. If it was, then we would see a lot more focus on pure gaming performance in general at the top tier - stuff like the X3D would be the norm, and any multithreading potential would be a tag-on at a premium cost on top. But reality is that it's the complete opposite. Didn't used to be like this 4-6 years ago, but it is now.
And workloads don't automatically pay for themselves just because you can complete a 3D render 10% faster - it doesn't equal 10% more profit. But what IS guaranteed, is a doubling in your electricity consumption, which most certainly can have an effect, especially on smaller companies where you don't just have long queues of render work waiting to be completed. That's enterprise work, and for that you'd be buying enterprise hardware anyways if that amount of time saving truly makes a difference in your earnings to justify the disregard of power consumption. Especially now where electricity prices keep rising.
Plus you also need to invest in expensive cooling to even operate at the peak performance output levels, which you may not have space for in your current case, etc etc. Cutting down unnecessary variables is typically much more beneficial to a workload-based usecase, than what the extra 10% performance uplift would net you in time saved and potential money earned.
thank you for showing both power and performance in the same chart!
Yess, finally someone talks about "standarized power draw" between 13900k vs 7950x.. thanks roman..
You absolutely crushed this exploration of power usage and eco mode performance! In this new age of self-inflicted energy production strains and high prices, these companies should be screaming about their eco performance per watt, especially Intel. They're getting hammered over their choice to use ALL the watts. It looks like the 13900k is a really great cpu at 90 watts.
Really insightful. I was considering doing a water cooled build and you showed that air cooling is fine since max power isn't important, this will save a lot of money and simplify the build
Same here, it's great to know that you can run a 13900K at very effective performance while still air cooling it. Thought air cooling would be a thing of the past, but seems like it's still got some years left.
Nice to see the power limit values again and the curve, however as mentioned can you use better liquid cooling to see how a custom water cooler does at high watts ? Of course you go a lot colder with LN2 etc but good watercooling results will be useful to a lot of folk.
Thanks for the good work.
This is easily the most insightful, forward thinking review.
Incredible work, I always enjoy your rational approach!
This is a really informative and best review that i have seen... Most reviewers do balls to the wall crank it up to the max reviews that do not tell the true story.. So thanks for this am sure gonna do the same tuning when i get my 13900k and 13600k for infotainment.. 👍
Regarding lowering wattage, there seems to be a huge difference in data, compared to hwu testing with limited power draw.
I mean they have different methodology, testing different hardware... Why would the data be same?
HUB’s number are wrong but of course are the ones being splattered everywhere now.
You seem to be the only one including power usage, nowadays with ever increasing energy bills, it can be a deciding factor.
Other major reviewers (Hardware Unboxed and Gamers Nexus) also include power usage
I really like the sensible direction your taking with your tuning videos with how brands are getting in a power war with their products now days.
Personally i have a down clock script on linux when i run a game and i just figure out what clock a game needs and auto restrict the games along with using the low latency performance scheduler. Then in the desktop i keep full (sensible power) clocks for good response and productivity inside VM's etc.
Keep up the good work :) I suppose its kind of good that these things all come hyper clocked and we just turn them down a bit rather than having to mod them etc.
I feel like theyre missing a beat in showing people their efficiency a bit more and letting people toggle the power rather than perma run it.
99% of people won't even look at the power draw, they don't care how much power it uses, just the fact that it's too hot but they still just get a more beefy cooler and go with it.
Only when you are very conscious about power usage in general would you consider how much your pc draws
Man love this channel. everyone else is doing normal reviews. This channel goes straight to the interesting efficiency stuff.
Thank you for highlighting what literally no other review has done, namely showing power consumption whilst gaming. All the other reviews are just showing peak power consumption figures during all-core workloads such as in productivity apps, which 99% of people don't care about. Most people want to know about power consumption and thermals when gaming, especially with what's happening to energy prices in Europe right now.
I'm very much interested in seeing 13600K vs 13700K vs 13900K compared at different power points. Will the MT performance even out? Will ST performance be event per watt? Here's me hoping we can get something like that ♥
0:37 Thank you that you've included i9-12900KS, which about all other channels forgot.
I think you have the best and most easily digestible charts and I love how you sort them by minimum fps. I prefer to have at least some visual reference for 1% lows.
Yes I think the 1% and 0.1% lows are the most important. Gamers Nexus showed that the 13900k has 0.1% lows as bad as the i3-12100 for eSports games like CSGO. Derbauer is the only major RUclipsr that includes eSports PUBG in the results, so it would have been really good to have 1% and 0.1% lows in the results
I've seen plenty talk about how inefficient the 13900k was in benchmarks...first time I see it tested in game... And showing better efficiency than the 7950x. This was great!
You easily have the best charts and benchmarks then any other tech reviewer. As much as I love Linus his charts are terrible to understand at a glance. Love your dedication
I love the power efficiency tests! Would be cool to see total system power draw with stock 13900K and RTX 4090 vs. power optimized 13900K and RTX 4090 and how their performance stacks.
Wow. You are the best!
I'm happy someone mentions your channel somewhere.
This is the real information that people should know.
Many channels mislead people to think the new gen intel is bad because of the high power consumption.
your information told totally a different story.
Appreciate the outside box approach compared to just taking one power figure and driving it into the ground repeatedly memeing “furnance”. Like you said for the 4090 Re:power/tuning…it’s hard to understand why intel and AMD both went with such higher power target on their flagship chips for the 10% gains vs more acceptable powers, but at least buyers know thanks to your clear videos it can be “fixed” with a few tweaks to your liking.
10:32 I appreciate a lot that you care about chart clarity. I always loved that you sort the charts based on min fps, so I kind of missed them a little in here, I hope they'll come back in the future. In any case, fantastic video.
Charts are on point and perfect.
Great review. I think this is the best price / performance review I've seen so far compared to others that I've seen.
having upgraded from a 12700k to a 13900k this thing is hot and power hungry. cooling it with a deepcool ls520 and it gets the job done. i just limited P1 to 125w and P2 to 250w. thinking of custom loop but for now everything seems to be working fine. overall impressed with this beast of a cpu.
I would love to see this on the 13600k. It still pulls a lot of power for a chip of its class, but with a power limit it would make it even more ideal for a small form factor build.
As far as RUclips goes your reviews have rapidly become my favorite, because you set out to do something specific and just analyze based on that. No pointless back and forth stoking responses from company fans (which there should never be in the first place) and good, well explained data.
Personally I'd love to see what could be done with undervolting to see what sort of efficiency gains might exist with default turbo behavior. I greatly appreciate the wattage scaling chart especially because it gives a straightforward look at where performance may plateau (first saw this from Buildzoid and happy to see it here).
Your reviews are standing head and shoulders above the others, at least for what interests me. Well done!
Finally a useful review. love how you actually review the performance of the cpus clocked down in power. Finding that sweet spot of efficiency and performance. Thank you for all you do.
Now I'll have to wait to find out why the reviewers have so big differences when power limiting the CPU.
yeah its pretty weird, like if bioses were bugged for some and setting higher voltages or something
XTU bug
Better video on the i9 13900K than Linus tbh. You helped me a lot with undervolting and getting my NH-D15 cooling it properly without thermal throttling
I'm well aware of your many successes outside of this channel but how does this channel only have 100k subscribers? So informative, innovative, entertaining and of course, pet friendly! Keep up the good work!
this is a newer channel just for english viewers i think
Location, he need move from de to eng speaker country, Sweden for example you can get view from eu & us
As a person that tries to undervolt pretty much everything, I'm so glad that a big RUclipsr is finally diving into this topic - the voltage headroom on some of this new gear is absolutely ridiculous😁😁
This is definitely the best computer tech RUclips channel because i dont see such detailed reviews in any other channels about efficiency. It must be done this way rather than onlyshowing game fps numbers
Amazing video. Roman, as engineer myself (simmilar field 2 :) ), i thank you to do some tuning and research. Alot of influencers just put stock settings and pull wrong conclusions. My current and last intel chip i7 8700k takes stupid amounts of voltage stock. If you limit it and tune it takes so much less voltage for stock freq. I think this is behavior on all motherboards for last few gens. Motherboard manufacturers just put voltage to the limit of chip ( i guess becouse marketing, and sales). And chip runs 2 hot.
I have the same setup (Aorus Master + i9 13900KF) at home, and I found that 150W was the sweet spot for me. I do video rendering/transcoding (i.e., Handbrake) and if I leave it at full voltage, I can transcode an h.265 movie in 30 minutes.
If I set the wattage down to 150W, then the same movie renders in about 32 minutes.
If I tried to decrease the wattage more (i.e., 105W), then the render time went up to 36 minutes.
At 150W, the CPU temperature remains about about 87C - 91C, using a 360mm AIO from OCool Eisbaer. This keeps the temps just below where it wants to start throttling performance.
I do say I'm not happy about this. End-users shouldn't have to be under-volting their motherboards. I ended up having to buy a new case to accommodate that 360mm radiator, and it still wasn't enough to keep it cool at full voltage.
Did you lower the voltage a bit not wattage?
Best raptor lake review on youtube right here ! You did great with the 4090 too.Seems undervolting/PL is the new overclocking these days with hardware being pushed so hard past peak efficiency.
Thanks for the focus on power and efficiency. I always go for the top tier cpu for gaming anyway, so power and efficiency are a really important differentiator. With competition so great now, I hope intel and AMD really focus on that as a value.
Hi Roman, could you please make a undervolted comparison of the 13900K and the 7950X with single-core and multi-core performance, putting them head to head through the same tests as in this video?
Thanks for comparing performance per watt with different settings. A lot of people need to save electricity dollars, others are in SFF cases and need to limit heat, and the rest of us need some apples to apples comparisons.
One of the best reviews. Also, I think it was a good idea to leave out the lows in fps.
Man you are the KING! All reviewers take CPU and just show us Cinebench or Aida power draw which we will never use!!!!!
You however showing what matters and real use case scenario and how much power these will draw..... I don't care which is better but I care about charts that shows my use case not Cinebench or Aida only
please include minimum fps in future charts, i always care way more about that than average. i get used to low average fps. i feel frame drops though
Just lock the fps then
@@MaxIronsThird "lock[ing] the fps" doesnt magically make 30fps look like 60fps. I think you misunderstand what I mean by "minimum fps". I mean the lowest fps observed - ie the fps during the largest lag spike. if that doesnt drop below my monitor's refresh rate I won't notice any lag at all, because I'm still seeing the same fps. you can lock fps lower than what you're capable of producing, but not higher (i mean, you can, but youre not going to magically hit that fps now if you didnt before)
@@xdMatthewbx If you don't have a 120Hz VRR/FreeSync display, locked 30fps will 100% feel better than hovering between 35 to 55 fps.
@@MaxIronsThird i'm talking about lag spikes. ie 0.1% lows. those don't last long enough for that to make a huge difference and even if it does you can only do so much to make a lag spike not feel like a lag spike if that lag spike dips your performance noticeably lower than your refresh rate
@@xdMatthewbx Those aren't common, and if they were, he would've said something, just like he mentioned in the video, there was nothing abnormal with the 1 and .1 %.
I've been digesting almost every review since 13th gen release and I have to say your review is by FAR the most informative and sheds some real light on the subject a lot of people have just talked about but nobody explained really well.. You showed a way to make this cracker of a CPU (and hopefully the other 13th gen as well) into very power efficient power houses!!! This makes the 13900K far more interesting then before.. Our power prices are through the roof as you know (neighbor from left side)..
Love your thoroughness of reviewing.. Must have taken you ages to do.. Make sure to get some sleep soon! :)
I dont know much about tuning my PC in my bios. But i had problems with my 13900k spiking at 90C with my AIO. I just tuned it to work at 90W as you did in your vid. my games and everything is running as good as before AND my max C is at 64C! Thank you SO SO SO SO much for this Video!
Solid review! Could you please share what the best optimal undervolting values can be used for good cooling with only a small performance hit? The explanations you said were above my head. Maybe a guide for the average user whenever you have a chance :) Love your content otherwise, hope to see more.
Before watching :
13900k + 4090 needs a nuclear power plant to run
After watching :
Efficiency master race
What Intel could do about it is to set 90 W as default limit and add a special Creative/Development mode into Tuning Utility that will automatically kick in only with selective set of apps, that you can change, running. So 300 W would be kicking by your will when your render task is running, or compilation, or compression task etc.
Yep this would be great, they could just have a hybrid power setting that automatically detects what a given workload demands from the cpu, and then apply whichever power limit that corresponds with the task the cpu is performing. If it's strictly P-core bound, it's going to be single-core heavy, like gaming. If it's multithreaded with all cores at work, it's a professional workload. Should be fairly easy to set up.
Even for render tasks I feel like the 300w should only kick in during intervals where it optimizes workflow - for example if you're using an interactive render then the 300w could kick in whenever you're repositioning the camera and the render adjusts real-time to your inputs - or it kicks in at the beginning of a render where various photon calculations are made, but then lowers the power level once the actual rendering of the photon path traces kicks in.
This is very insightful compared to a lot of other reviews.
I'm sure you've seen the reports that the 7950x gaming performance on Win11 improves when you disable 1 CCD likely due to Windows scheduling issues. Would you consider looking at this for some of the charts you have?
Games only use 1-8 cores depending on age, at least they only scale performance that far, so either way you are wasting your power budget running that second CCD, as far as gaming goes anyways. If only one CCD is active it can have all the power avaliable. Also latency between CCDs has always been an issue with Ryzen, especially when the CCDs only had 4 cores.
Disabling one of the CCDs in Ryzen breaks S3 and S4 sleep.
the 7700X is the best Zen 4 chip if you're gaming, and its the best because of price to performance. All the Zen 4 CPUs operate within 3-4% of each other in gaming, so if your are gaming don't bother with anything above 7700X imo! any amount of production load in your daily life will be much faster as you go up the tiers tho, while giving similar but about 1-2% less performance in gaming at each incremental tier. (That's right less performance as you go up the tiers.)
Love your review. This review makes you different from other reviewers. I would like to see this type of comparison on 13700K as well.
With current increasing trend of energy price, I believe more user will consider to undervolt the cpu.
Not all the user want 100% performance, why say no to 90% performance 50% energy cost?
For those 1080p gamer (who probably wont use i9 anyway), will look for i7 or even i5 for top performance with lowest cost.
For those user who play 1440p/4k gaming, they might want to see how low power cpu can go before hitting the performance hard.
20:20 “Most people are just going to game with their 13900K.”
I build two high end computers for an AI code developer using 13900Ks and 4090FEs. The first was in a Fractal Design Torrent case with Noctua’s most capable air cooler, NH-D15. The second used a Fractal Design Pop Air case and a Noctua NH-U12A case. The Pop Air case was slightly modified so that the front fan support would not block two Noctua 140 mm fans. The second computer had a much less capable cooling solution.
Testing and Tuning the second computer was done with Cenibench R23, Furmark, Intel Extreme Tuning Utility, CPUID HWMonitor, Windows 10 Task Manager, and Kill A Watt P3. It had an 850 W Gold power supply, so the power at the Kill A Watt should not be over 950 watts.
The 13900K ran Cenibench and the 4090FE ran Furmark. I think Furmark is an overkill for a GPU like Prime 95 is an over kill for a CPU. Temperatures were never a problem for the 4090FE.
Turbo Boost is meant to take advantage of the thermal inertia for a short time if the CPU is not fully utilized. Thread swapping is not talked about much. You can watch it happen if you look at HWMonitor temperature and utilization if you CPU render a single thread with Cenibench. If the CPU is near its thermal limit it will swap threads with a cooler core. This thread swapping takes time and results in a series of “no op” states on both cores and no work gets done
An assumption for tuning the CPU is that the load will be nearly constant and the CPU temperature will be near the 100 C limit with dust compromising the cooling solution. To achieve this I chose to limit the P and E cores maximum frequency and limit turbo boot power and minimize duration so that the CPU’s package temperature was no more than 82 C.
The results were good. The CPU’s package temperature was no more than 82 C with a Cenibench R23 multi score just over 80,000. The Kill A Watt showed about 800 W which means the PSU had a load of about 720 W.
So basically if you have the cooling power to cool down the 13900k, it can be to the 7950x in cinebench R20 but just doing the second run on that thing makes it lose consistently as per your own results which is I believe under a minute at full power. That means even a 360 AIO is useless for this thing if you're going to run it at Full Tilt for productivity and it's just better to go with the 7950x
This is the best 13th gen review I've seen so far. Thank you so much for making this efficiency video. It gave me a completely different perspective about the 13th gen. I hope you can further explore all three new CPUs. With 13700k and 13600k, the price equation would enter the picture if the performance is still there with power limitations. I'm looking forward to it if you do make more videos about it.
I'm interested to see how limiting the 13600k performs.
And how it would stack up against the 7600x's power draw!
actually quite impressive how good raptor lake's power efficiency can be against zen 4 on 5nm
The name 7Nm (Intel 7) and TSMC 5Nm is not the actual size. It’s just marketing name.
@@erisium6988 regardless of the actual nanometer numbers TSMC N5 (5nm) still has higher transistor density than Intel 7 (10nm)
I hope we can forget "nanometer" naming schemes as X3D chips and other 3d stacked chips come out. I hope they name new processes like "le epic stack 2.0+"
@@XxXnonameAsDXxX they have a naming scheme for a reason tho
Great job by Intel gen 13th delivering very decent price to performance, meanwhile Amd has been a lot disappointing with Zen 4, not only Amd is overheating but also the Cpu itself is overpriced, r9 7950x is $150-200 more than i9-13900K still Amd can't beat Intel in full benchmark, not to mention Intel still compatible with DDR4. Intel totally destroyed Amd here.
Honestly the best and most informative comparison review out on youtube
Excellent Review 👍🏻👍🏻 🙌🏻🙌🏻
It's great to see performance vs wattage and your explanation of the most efficient use of each of the top performing processors!
Thank you !!
Looking forward to your delidding video!!
This is the best video on raptor lake chips if you want power efficiency i've seen. It make complete sense. you are quite good at presenting the data as well in terms of scaling and watts to perf.
that's Intel's fault really, they could have do the same Eco Mode switch in the BIOS as AMD does and there were no confusion about how good/bad Intel's efficiency, you cannot expect a regular user tweaking in the BIOS themselves
Really love your videos on efficiency. Those perf/watt was totally unexpected from intel
Filling out the graph at 15:12 with other CPUs would be fascinating and give a true comprehensive comparison of efficiencies at different power levels. The 5800x3d then might challenge 13900 at different power levels. Currently looking at upgrading my PC and have no affinity with either brand, so content like this is invaluable.
Very interesting reviews! I ran into your channel while in Germany last week (thank you RUclips algorithm I guess) and I really enjoy your review style. I am looking to upgrade to a brand new PC coming from an almost ancient 6700k and 1070 (impossible to properly run my flying/driving sims), very curious to see the upcoming reviews of the 13600k/13700k and 4080! And of course AMD.. but never used AMD before.
hey man have to say best 13900k review period. I'm interested on to replace my 12700k, how well do you think it will perform limited about 200 to 260 watts? those are the defaults limits on my bios for the 12700k
I think realistically we need to see if all this is still viable on a 13700k, does it still maintain efficiency or does it play different (as in a more drastic loss in performance). I would only listen to der8auer in this regard. Then again that's up to whether he has time for that. Amazingly helpful video!
Should be. power specks and clocks are around the same, so the same positioning way above peak efficiency should apply.
TL;DR: This applies to any CPU of any gen. Running more Ghz has a steep power draw cost: anything over 3Ghz will throw efficiency though the balcony until you can't simply sustain it any longer due to going out of spec in temperature/current/voltage.
Long version: my 11900K can do 4.5Ghz all core on a 125W power budget, running 3.5Ghz just costs 70W and running 5.1Ghz costs 250+W, all core stress test type loads factoring out vectorial extensions: with them power figures go up to a maximum of +25% power draw depending on workload and the extensions used, but on return the CPU crunches numbers significantly faster(this is wild terrain: algorithm dependent, may be +40%, +100%, +200% execution speed).
There's afaik many details with performance and efficiency which are common across every hardware. Those looking for functionality should stick with manufacturer already customized hardware for a easy life.
Wow can't believe I'm only seeing this now. Excellent video! Bravo!
I think it would be interesting to see performance and power draw using only the P Cores, with E Cores disabled.
really tried most of undervolting for my 13900ks . Finally found your video and applied 13:32 ratios. all is fine now . thank you so much
Can you do the thermal tests again but with the contact frame? I wonder how much the contact frame helps this time around. Thanks!
Excellent review as always. Like some already wrote i like how you dive a bit deeper than most other channels with optimizing efficiency. It shows that for just a little less performance we can save a lot of power and heat output too.
Thank you for all the hard work and please keep it up 🙂😁
I'm a competitive Call of Duty player, the 1% and 0.1% are probably more important to me than the average abd certainly maximum frame rates. Therefore, I would strongly encourage you include those numbers, particularly when you're reviewing gaming CPUs. Also, at least in my mind when you're looking at a bleeding edge CPUs, the first people buying them, including the RTX 4090 are looking at high resolution and/or high refresh rate monitors in the competitive multiplayer games. Meaning while I'm certain Cyberpunk, Microsoft Flight are good to demonstrate DLSS, ray tracing, etc I would tend to believe the competitive first person gamer is looking at these CPUs and GPUs above all others. Therefore, it is wise to include those games. Trust me, I love your stuff and your information is always very pertinent, sometimes critical information is lacking for their depth in the relevancy of the products. In this case, top tier GPUs and top tier CPUs. Also, I'm assuming that the so-called 12th generation contact frames are also useful here .. I ordered one.
ive never paid attention to those numbers, what are they and what do they mean for gaming? I admit currently I mostly play Warframe, So as long as I am above 90fps(at 4k HDR, which my 3090 does do in that game) I can usually blast around the maps like jackrabbit rabbit fresh out Walter White's Winnebago no problem.
@@filanfyretracker .. I upgraded from my RTX 3090 to an RTX 4090 because @ 3440x1440 even with DLSS Quality I was capped around 180 FPS. Since I'm replacing my 175Hz to a 240-260Hz monitor the 4090 made sense. I only play multiplayer Call of Duty.
You turned to my favorite tech RUclipsr, very useful videos. Want to try this beast in Unraid server, with limited power consumption. Probably will be nice idle power draw
The 13900k undervolted using less power while having higher FPS than a 5800x3D is impressive. Neat stuff. Looking forward to upgrade my 12700k, DDR4 RAM, and 3080 Ti to a 13900k, DDR5 RAM, and 4090 in a couple months.
Heart you so much for the performance-per-power charts, electricity is expensive and per-watt factors heavily into my own purchasing decisions.
I've stopped watching other obsolete pc tech channels because der8auer is on the top of the game. No one can compete and even Jay2cents is just half qss copying this guy 😂😂
Adding the cooler in the system setup slide would be helpful, but you mentioned it at the end of the video. Thank you
Will remember for next time! Thanks for the feedback
I'd love for you to continue manually tuning all CPU's and GPU's you review to see what their optimal efficiencies are.
I love the cat picture with the motherboard great thumbnail
I like that his cats always want to be a part of the action. They're so beautiful.
Love the power draw with the FPS generated.
Green (low energy) gaming is becoming more relevant for me personally.
You're the man, been hoping a reviewer would do a power examination while gaming, been asking other reviewers in their comments, and then I got to yours and you actually did it. Next time, I do yours first. Yes the 13900k is way power inefficient in benchmarks and workloads, but in gaming, it's actually very efficient.. Also the artificial watt limitation? Beautiful work. Thank you sir. I'm quite happy with my i5-12600k and hopefully won't be upgrading for a few years, but I do build systems as a side hustle, and it's good to know what practical power/thermal gaming loads look like.
But honestly, thank you for not ranting about the power consumption, every other review I've seen so far has pretty much ignored performance/power usage in games to rant about them under max loads, which I get, sucks, but when will a gamer ever see those? So does it really matter?
so 13900k undervolted can probably be put in laptop and still powerful AF.
yep and same with Nvidia's new Ada architecture. This upcoming gen of laptops will be beastly if they can manage cooling slightly better than last gen ultrathins that didn't care one bit for cooling despite throwing power hungry hardware in them and thermal throttling constantly.
Looking forward to some more tinkering, deliding and maby some of the lower tire models in the future :)
The efficiency at lower power target is quite impressive! This combined with the efficiency of Ada makes me very excited about laptops next year
The use of cats in tech videos is supremely underrated.
cant wait to get my hands on the 13900K, will be my first I9 :D upgrading from the 9700k too so its going to be insane
Is it worth upgrading from the 9700k tho?
upgrade to AM5 much better
i'm running 4790k, when i can afford to upgrade next year, probably end of year i'm going to feel like i'm time travelling lmfao
nah 12th and 13th gen are exactly same architecture with higher clock
Me tooooo I game and stream on a single pc with 9700k and an rtx 3080ti but my cpu is so bottlenecking and running at 100% most of the time.
performance per watt comparisons are tricky. higher performance on per core basis, is achieved at lower perf/watt. Perhaps we could find SKUs for each architecture at 8-core, adjust frequency to match performance, then compare perf/watt