Maquinari MacBook Neo Apple A18 Pro Refrigeració Cinebench 3DMark Overclocking Salem Techsperts

MacBook Neo refrigerat a 11 W: 2.076 punts a Cinebench i un 3DMark imprevisible

Salem Techsperts desmunta un MacBook Neo amb A18 Pro i prova dissipadors, aire fred, massilla tèrmica i una cèl·lula Peltier. El processador sosté prop d’11 W i millora molt a Cinebench, però els resultats gràfics són erràtics.

Fins on pot arribar el MacBook Neo si se n’elimina gairebé tota la limitació tèrmica? Salem Techsperts converteix el portàtil d’entrada d’Apple en un banc de proves obert: desmunta la pantalla i diversos perifèrics, manté la bateria connectada, col·loca dissipadors a banda i banda de la placa i hi afegeix aire exterior i una cèl·lula Peltier. L’objectiu declarat és superar els registres d’un altre creador i recuperar el títol informal de «MacBook Neo més ràpid del món».

L’experiment aconsegueix sostenir uns 11 W durant una càrrega llarga i arriba a 2.076 punts multinucli a Cinebench, però no produeix una victòria neta. Les proves de 3DMark varien molt entre execucions idèntiques i el rival conserva quatre dels sis millors resultats gràfics comparats. És una demostració molt visual de com la temperatura condiciona un xip sense ventilador, però també de per què més potència no sempre es converteix en més puntuació.

Un A18 Pro dins d’un portàtil sense ventilador

Apple equipa el MacBook Neo de 2026 amb un A18 Pro de sis nuclis de CPU —dos de rendiment i quatre d’eficiència—, GPU de cinc nuclis, 8 GB de memòria unificada i un xassís passiu. No hi ha cap ventilador intern que expulsi la calor quan una càrrega sostinguda manté tots els nuclis ocupats.

Segons les mesures mostrades al vídeo, la configuració original acaba reduint el consum fins a prop de 4 o 5 W quan s’escalfa. La preparació extrema arriba a 11 W, un increment de fins al 175% respecte del valor de referència que dona el canal. Això no és una especificació oficial d’Apple ni una nova configuració permanent: és telemetria observada en una unitat profundament modificada.

L’equip conserva la bateria connectada perquè, segons el creador, desendollar-la fa que el processador redueixi dràsticament el rendiment. Alhora, retira pantalla, antenes, altaveus i altres peces per exposar la placa i poder accedir a la part inferior del xip.

El muntatge: coure, PTM7950, massilla i aire fred

El pla inicial combina un dissipador superior, làmines de coure, material de canvi de fase PTM7950 i massilla tèrmica Upsiren. A sota de la placa hi situa una cèl·lula Peltier, que transfereix calor d’una cara a l’altra, refrigerada al seu torn per un circuit i aire fred. Un bufador industrial introdueix aire de l’exterior per mantenir baixa la temperatura del conjunt.

Durant el desmuntatge descobreix que unes peces que semblaven simples separadors també condueixen calor cap a la carcassa superior. Les substitueix per massilla, aplana el contacte amb molta cura i utilitza el mateix xassís com a dissipador en una prova posterior. El procés inclou connexions i desconnexions amb la placa encesa, risc de curtcircuit, pressió directa sobre el silici i possibilitat de condensació o xoc tèrmic.

No és, per tant, un tutorial segur per reproduir a casa. Apple especifica un rang ambiental d’ús de 10 a 35 °C, mentre que una part de l’assaig aprofita aire exterior molt més fred. El muntatge invalida qualsevol comparació amb un portàtil normal i pot causar danys irreversibles.

Cinebench puja fins a 2.076 punts

Abans d’aplicar el fred exterior, el primer muntatge marca 1.849 punts multinucli a Cinebench. Amb el sistema complet, el processador manté aproximadament 10,8–11,1 W durant minuts. Una execució arriba a 2.051 punts; després de revisar el contacte tèrmic, el millor resultat que es veu al vídeo és de 2.076 punts. En mononucli obté 631 punts, molt a prop, segons la comparació del creador, d’un M4 Max en aquella mateixa escala.

La dada més interessant no és només el pic, sinó la potència sostinguda. En una càrrega de renderització llarga, el muntatge evita la baixada que provocaria l’acumulació de calor al xassís passiu. La descripció del vídeo calcula una millora multinucli del 41,71% respecte de l’estat original.

Cinebench 2024 utilitza el motor de renderització Redshift de Maxon i separa proves de CPU mononucli i multinucli. Les xifres són útils per comparar execucions del mateix test, però no representen automàticament la velocitat de totes les aplicacions ni el consum habitual d’un usuari.

El contacte tèrmic importa més que la sofisticació

El primer sandvitx amb PTM7950, làmina de coure i més PTM produeix resultats pitjors dels esperats. Salem Techsperts sospita que les capes no pressionen prou bé sobre el xip. En retirar el coure i posar massilla directament entre la superfície i el dissipador, Cinebench puja fins al millor registre del dia.

La prova il·lustra una limitació habitual de la refrigeració improvisada: afegir materials no garanteix menys resistència tèrmica. Cada unió pot deixar espais d’aire, una pressió desigual o una capa massa gruixuda. El fet que la Peltier tampoc millori sempre el resultat reforça aquesta conclusió. Un sistema més complex només ajuda si el camí complet de la calor està ben resolt.

El creador també prova una modificació menys aparatosa, aprofitant la carcassa superior com a dissipador amb coixinets de silici. Informa d’un 28% més en multinucli i un 25% més en mononucli respecte de l’equip original, tot i que el teclat s’escalfa. És menys extrema que la configuració de finestra, però continua sent una modificació no avalada per Apple.

3DMark no respon com Cinebench

La part gràfica desmunta l’expectativa que 11 W assegurin tots els rècords. Amb temperatures baixes i el mateix muntatge, els resultats oscil·len entre valors molt diferents: el vídeo ensenya puntuacions aproximades de 1.371, 1.449, 1.451 i 1.490 en una de les proves. Salem Techsperts obté el primer lloc a Steel Nomad per només dos punts, però acaba per sota del rival en quatre de sis tests i només entra al top 12 de Solar Bay.

UL Solutions descriu Steel Nomad com una càrrega gràfica multiplataforma sense ray tracing, mentre que Solar Bay mesura ray tracing en dispositius lleugers. Són càrregues diferents de Cinebench i poden topar amb límits de GPU, memòria, controlador, freqüència o distribució de potència que la refrigeració de la CPU no elimina.

El vídeo no aïlla la causa de la variabilitat. El mateix autor planteja si el silici s’ha degradat, si treballa massa fred o si hi ha un altre límit del sistema, però no ho demostra. Tampoc publica prou repeticions controlades per atribuir la dispersió a un únic factor. La conclusió honesta és que les puntuacions són inconsistents, no que el xip s’hagi espatllat.

Cyberpunk a 50 FPS i una conclusió amb matisos

Com a prova complementària, el MacBook Neo modificat executa Cyberpunk 2077 a 1080p i el canal mostra uns 50 FPS amb la refrigeració tipus sandvitx. És un resultat cridaner per a un equip bàsic i passiu, però el vídeo no detalla tots els ajustos gràfics ni ofereix una comparació completa amb l’estat original.

L’experiment sí que demostra que el marge tèrmic de l’A18 Pro és considerable: una dissipació externa permet sostenir més del doble de potència que el valor al qual pot caure sota càrrega. També demostra el límit d’un titular basat només en watts. Cinebench millora de manera clara; 3DMark no ho fa de forma estable, i la cèl·lula Peltier no supera necessàriament un bon contacte amb aire fred.

El balanç final és més interessant que un rècord absolut. Salem Techsperts aconsegueix la seva millor puntuació de CPU i alguns primers llocs, però admet la derrota parcial en GPU. El vídeo converteix el MacBook Neo en un laboratori de gestió tèrmica, no en una recepta pràctica per obtenir un portàtil diari més ràpid.

Contrast i context

Fonts consultades

5 fonts
  1. 01
  2. 02
  3. 03
  4. 04
  5. 05

Font de treball

Transcripció amb marques de temps

12 fragments
Consulta la transcripció
  1. 0:00 , obre el vídeo en una pestanya nova

    WE'RE ON THE TOP. WE SET A 3D MARK world record, baby. All right, Jay's two cents. Come at me, bro. All right, Jay's two cents. Come at me, bro. Hello, Jake. Hello, Alex. I know you were expecting Jay's two cents, but you're getting Andy no cents. Now, when you released that video, I had the fastest MacBook Neo in the world. And after you released that video, I still had the fastest MacBook Neo in the world. Now, you may have had a higher single core score and a higher 3D Mark score, but I broke that 3D Mark score the next day. It is about a week after, and I finally have the time to break that record and set a world record for every single 3D Mark score that you did and every Cinebench score that you did, because I am determined to have the fastest MacBook Neo in the world, and no one is going to beat it after today. I really hope Jay's two cents doesn't do a liquid nitrogen build. Now, you may ask, "How am I going to do this?" And the answer is PTM 7950, 30° air, an industrial air blower, a heat sink, and a Peltier cooler. Pel- Peltier Peltier Pel- the the the water cool thing that makes chips cold, and UpSiren thermal putty. And this is my game plan. I've already loaded up these random power

  2. 1:31 , obre el vídeo en una pestanya nova

    capacitor IC chip thingies with UpSiren thermal putty. Does it do anything? I have no idea. We're going to find out. So, this is UpSiren thermal putty. The sticker is going to go over like this, and then I'm going to have a copper shimskees on the sticker right here. So, probably two copper shims right here. And then, we're going to have this heat sink with PTM 7950 directly on the A18. And the idea is to keep everything cool. Everything cool. And then, we are going to take this cooler and put it under the chip with some thermal putty to sandwich it. Heat sink on top, Peltier on the bottom. We're going to cool it from both sides. And we're going to see what happens. Now, it's going to be a challenge to get my previous score without traveling to somewhere colder, but thankfully if I do travel, I won't lose access to the content I enjoy right here at home thanks to today's sponsor, NordVPN. NordVPN allows you to change your virtual location so you can access your favorite shows and sites from anywhere without sacrificing security. Because even on an open network, it secures your connection against cyber threats. And the threat protection pro feature adds an extra layer of security. It scans downloaded files for malware, blocks intrusive or harmful pop-ups, and prevents you from visiting malicious sites. Even better, the dark web monitor continuously scans hacker hangouts and immediately alerts you if your details are up for sale. And if so, they include easy steps to follow to help you resolve it. A single NordVPN account protects up

  3. 3:03 , obre el vídeo en una pestanya nova

    to 10 devices simultaneously, covering your entire household. Use the link in the description or visit nordvpn.com/salem to get an exclusive deal including four extra months. It's completely risk-free with a 30-day money-back guarantee, so you can try it out with confidence, which is something I lack with this experiment. Now, there are a few tricks to this MacBook because you can't just take the logic board off and use it like a desktop, unfortunately. And the only reason you can't do that is because the battery. As soon as you disconnect this battery, the CPU throttles itself to unmanageable speeds. It It becomes a slug. It's very slow. You can't use it. So, what I have to do, I have to completely disconnect the logic board, and then I have to remove the display. Now, I have to do that because I can't get the Peltier underneath it. I still You'll to keep that battery connected. Thankfully, we do have some play on the battery cable. There's a few millimeters weep that we can play with, so we don't have to worry about stretching any of the cables. So, I'm going to disconnect the display, remove the Wi-Fi antenna, remove the display, remove this thing, remove the speakers because we don't need the speakers. I can remove the trackpad as well because I don't need the trackpad. That's not the trackpad. Basically, I'm going to remove as much as I possibly can and then just connect it to the monitor, so it still functions. And let's see how that sandwich goes. And as for the chip itself, the chip is recessed underneath the sticker. So, I have to put a copper shim. It's a 0.5 mm copper shim in order to have it make contact with this cuz this is a little thicker than this right

  4. 4:35 , obre el vídeo en una pestanya nova

    here, so it touches everything. So, I have to have the shim to to bring it up. I have PTM 7950 on the heat sink. I have the other heat sink attached with Conductonaut. That's in the middle of there, right there. And I have a little bit of PTM 7950 on here as well as some thermal paste on here. That way, everything makes contact with the heat sink. I'm going to have the industrial air blower blowing in from the window. The cold air intake is going to be taking cold air from the window as well. And the Peltier cooler is going to be in between the two, so that the cold air is going through the radiator in here, cooling this thing down as much as possible. Because remember, the way a Peltier cooler works, one side of this needs to be warm and the other side needs to be cold. So, this side's going to be cold. This side needs to be warm. This is going to be underneath. So, I'm going to have to be very careful to make sure that this isn't blowing air directly onto that. A few minutes later. Our board is out and this is where it gets a little tricky. You see these two little pads right here? Those two pads? This cooler is not going to fit with those two pads in the way. It's not going to have a good connection. If you see like this, it's not going to have a good connection. Can't make a connection. So, I have to remove those two pads. Those two pads act as a stability like little mounts right here so the board doesn't bend. Now a lot of people said why don't you just use this as a heat sink? I don't think this would make a very good heat sink. Maybe I'll do it in the long run and just see how it goes. Because I mean there nothing can really hurt by just putting some upsilon right here and using this as a heat sink. But

  5. 6:05 , obre el vídeo en una pestanya nova

    obviously if I want to extract as much power as possible from this thing, I want to use that on the bottom of this. So I'm going to remove these stickers cuz that's all they are little stickers here. And then I'm going to slap some thermal putty on here and let's see how that goes. Breaking news, this served dual purposes. It's a pad and it's a heat sink. So this is graphite, graphene, whatever it is. So they serve as a heat sink and they serve as little pads to help stabilize the board and keep it straight. So that means we are going to have to replace these guys. Probably going to stick with thermal putty honestly. I think thermal putty would be the better option right here, right here, uh and most importantly right on the CPU. So obviously Apple thought the same thing. They can cool down the chip just a little by using the top case as a heat sink. 15 minutes later. Man, this is super sketchy because I don't want to expose the chips in the middle right there cuz I can short out the CPU, but I need as thin of a layer as possible because the less material the Peltier has to move through the more cooler it can get. Yeah, that was a little sketchy one. This was as thin as I could make it. As you can see it's super thin without risking damage by shorting out the chips. So all the chips are covered. We have the two blobs on each side to try to help out with the where the heat sinks little boosters were. Let's see how it goes. Oh, some complications here. So I couldn't get the monitor to display unless I entered the password in because this is an accessory and Apple doesn't trust it so it won't it won't take input from external keyboard I guess or I'm

  6. 7:36 , obre el vídeo en una pestanya nova

    just an idiot. I'm probably just an idiot. So I had to plug in the keyboard temporarily and all this stuff. Look at this. This is how sketchy this is. USB-C's plugged in. Uh Batteries connected. So, now what I'm going to do is something you should never do. I'm going to disconnect the keyboard, display, trackpad, and whatever that is at the all this Jeez, get out of there. At the same time while it's on. And I hope this thing doesn't short out on me. That was by far the sketchiest thing I've ever done in the history of Salem Techsperts. Sweating. Sweating right now. Oh god, this is so sketchy. I actually ended up throwing some extra putty right there on the end. Right here because that's where all the PMIC, the power management chips are, I think. I don't know. Again, I'm not that smart. Also now, at least when I put the heat sink on, it's not going to bend the board. So, I put the heat sink on right here. It's just going to be directly on the Peltier right here. So, let's set this bad boy up. Now, obviously the PTM 7950 and all that other stuff, the thermal paste and stuff, it's all cold right now. So, in order for me to get a good connection, I need to warm it up so that they can get into a state where they can actually fuse. So, that's when the conductivity is at its highest when it's Obviously, you apply it, it's cold, and then it needs to get warm, and then cool down again, and that's when it fuses up. It changes phases, and we get a good connection. So, that's what I'm doing right now. I'm just going to warm this thing up. Okie dokie. Just having this set up, we got 1849

  7. 9:07 , obre el vídeo en una pestanya nova

    on Cinebench. So, our wattage was around like 6.4. Obviously, it's at nothing now. But, uh yeah, you can this thing is warm. It's warm to the touch, which is awesome. And as you can see, we're as flat as we're going to be. So, now, this is the setup. I'm going to put this in the window, but we have the cooler right here. So, the intake right here, exhaust is right here. Obviously, if it's intaking air that's as cold as balls, and is going to be as cold as balls as well. So, we're going to turn that on. This is nice and solid now, and it's going to take the air from outside, going to go through here, and we have it on the right side of the fan, because that's where the cold air is going to come in, right on this side, cuz there are two intakes. Ideally, you would have both going outside, but I don't I only have one tube, cuz we're cheap here at Salem Tech Squad. So, let's see how fast we can go here. Now, I don't want to thermal shock this thing, so I'm not going to just go balls out immediately. I'm going to let it cool down with the Peltier cooler first, and then we're going to put the thing in there, so that it gets really low. But, my biggest fear is that the CPU is literally going to shatter, because the temperature difference is going to be so drastic, it's just going to crack. So, I'm going to try to avoid that. One eternity later. So, this is what we get with the final setup. I blocked off this intake right here, because I want it to just intake from here. I also did end up removing the Peltier from over there, because it wasn't really making a difference for this one. So, I'm just going to put this in here. Cuz it's it stays nice and sturdy right there. So, now this thing's taking in cold air, as well,

  8. 10:38 , obre el vídeo en una pestanya nova

    uh and we're maximizing the uh intake. So, uh maximizing the whatever. So, this thing's taking air straight from there, blowing right here, nice and cold, and we are down to 10°. 10°, and the Peltier cooler is also running on the bottom. So, let's see what we can get now. All right, we're going to start our test. We click start. Let's go back to the sensors. See how much wattage we can crank out of this thing. 10.8. 10.9. Oh my god, can we hit 11? 11 W. Holy [ __ ] So, we're about 6 minutes in. This thing is pegged at like 10 Look at that, 11 W. We're still hitting 11 W 6 minutes in the test, which is absolutely insane. By this time, if we're doing it normally, it would have throttled to 105, we'd be at like 5 W if that. So, dude, I this score is going to be nuts. nuts Holy 11.01 11.01 Man, my arms are getting tired from just record four 04 Holy sh 11.10 11.10 [ __ ] W s I feel like I'm on [ __ ] crack right now. This is This is what meth feels like. Okay, hold on. Lower the export The cops are coming for me. They heard that the wattage was too high. All right, we're almost there.

  9. 12:09 , obre el vídeo en una pestanya nova

    2050 2050 which somehow is not higher than what I had before. What? Well, that score is amazing. I'm a little salty because I managed to hit 2051 but I didn't record it. I recorded the uh the result, but I didn't record the actual test. I managed to hit 2051 when it was colder. It's It's 40° out right now. Yeah. But, this is the coldest it's going to get until like 3:00 a.m. and I'm not trying to come here at 3:00 a.m. But, 2051 is good. It's just crap, man. 2051 without the Peltier just on cold air. Woo. All right, let's try that single core score cuz I need to beat that one, too. 631 which puts us right there. Look how close we are to an M4 max on single thread. That's bananers. bananers Now, time for 3DMark. So, this is interesting. I'm getting lower scores on 3DMark even though I have lower temperatures and I'm using both the Peltier on the bottom. 402, dude. What the That's garbage. So, what's going on? My temps are good. This thing is obviously staying cool. What's going on here? I had a higher score without the Peltier, just using the heatsink when it was connected to the chassis. So, what the hell is going on? All right, that's more like it. 2076 for our multi-core score. The only thing I changed was I replaced

  10. 13:39 , obre el vídeo en una pestanya nova

    the PTM7950 with some Upsiren thermal putty. That's the only thing I changed. So, what I think was happening is the PTM7950 that I had on the chip itself wasn't making contact with the copper sheet that I had on the heatsink. So, if you remember, I had the heatsink, PTM7950, the copper sheet, PTM7950, and then the chip. So, what I did now is I removed the copper sheet, I removed the PTM7950, and I went back to just using the Upsiren thermal putty right on the chip and right on the heatsink. And then, I got a record score for myself. So, that's good news. Now, hopefully, I can break the 3DMark score now because what the hell was going on there? So weird. I was at some one point I was like, am I running too cool? Am I like breaking the chips? I didn't know what the hell was going on. So, we're going to check out the single-core, and then we're going to get back to 3DMark. Only 631 on single-core, though. So, it's a little lower than we had before. Let's try 3DMark. All right, I'm convinced something is happening. Either I'm I cooked my silicon or something. The score right before this was 1449. And I've hit 1448 as well. It's all over the place. No matter what I do, it's the exact same process. I just don't understand what's going on. Case in point, 1490. That's a huge difference. That's a 30-point difference from what I just got. Like, I don't I don't understand what's going on with this thing. Is it too cold? So, this is actually a perfect example. I didn't even know you could do this. So, 194929294847.

  11. 15:10 , obre el vídeo en una pestanya nova

    Like, what the hell was even the greatest technician that's ever lived? What's going on, dude? I'm so confused. And where's my 1450? There's my 1451 score. And I have not been able to hit that. But this is all from today with the same setup from here to here, the same setup. But like, what the hell? 1371, where what? What's going on? And then here we go. Number one on uh Steel Nomad, I beat Jake by two points. But like, why? Why? Why is this thing so inconsistent? So, this is where I officially admit my 3DMark defeat. Even though my Cinebench scores were significantly higher, I just can't come close to some of Jake's 3DMark scores for some reason. He ended up beating me in four out of six benchmarks with Solar Bay being the biggest difference where I could only get within the top 12 globally. Embarrassing. I even tried doing a direct on-chip Peltier cooling setup, but it still wasn't as good as the original cold air intake scores I got the first time. I would love to hear what you guys think the reason for this difference is. Even though my setup can consistently pull 10 watts, my score isn't as good as before. Have I just cooked the silicon in this chip too much? Let me know in the comments, but first, here's a few more experiments before we wrap up. Starting with the Cyberpunk results on 1080 using the sandwich cooling method, which resulted

  12. 16:40 , obre el vídeo en una pestanya nova

    in a solid 50 FPS. Not bad. One of the more interesting experiments was testing out Cinebench scores on a board using the top case as a heat sink with the silicon pad mod as well. The difference here was insane versus stock. With the top case heat sink getting a 28% higher score in multi-core and 25% higher score single-core. Obviously, the keyboard gets a little warmer, but it wasn't uncomfortable by any means. So, I would actually recommend doing these two mods if you want a little more power out of this thing. But that's all for today. I hope you guys enjoyed it, and I'll see you in the next one. All right. That looks good here. Did I forget to turn off the light last night? What the heck? She Lupe. Dude, what the heck is Yo, what what what are you What's on your shirt, man? Nothing. No Listen, Lupe. What the heck? I I can explain. I can explain. I'm This is just It's all It's all for a video. No, what video are you freaking doing, man? It's um Lupe. Lupe. Lupe. Lupe. Huh? What's going on, man? Is it morning already? Dude, I had a freaking nightmare, man, and it was crazy. Must have been what you ate last night. Dude, it must have been that Chinese. Yeah, causes nightmares. Yeah. All right. Uh let's let's let's get to the video. All right.