M6 Mac mini vs Entire M5 Lineup — Transcript

Full transcript

1. 0:02Apple's entire current gen lineup is out
2. 0:05and into the world from the M5 base
3. 0:06model all the way to the M5 Ultra with
4. 0:09the curious early entrance of the M6
5. 0:11model in the Mac Mini. Today, I hope to
6. 0:13answer the question, where does this M6
7. 0:15fit in this sea of M5 competitors?
8. 0:18Because that M6 iMac is probably just
9. 0:21around the corner. I just went out and
10. 0:23got myself a prop for that video. I've
11. 0:24already compared the M6 Mac Mini to its
12. 0:26previous model, but because Apple
13. 0:28skipped a whole generation in between
14. 0:30these two, this thing got a giant leap
15. 0:32in terms of its capabilities. I wanted
16. 0:35to say in terms of its value, but its
17. 0:37bottom price went way up. Then again, so
18. 0:39did every other computer on the planet,
19. 0:41but the M4s are gone everywhere. No,
20. 0:43that's not true. They're still in the
21. 0:44iMac for like another 2 weeks. And the
22. 0:46way I look at it, there's kind of just a
23. 0:48BNA and an ANA, before neural
24. 0:51accelerators and after neural
25. 0:52accelerators. Thank you. The neural
26. 0:54accelerators and the GPUs, when
27. 0:56accessed, give such a boost that the M6
28. 0:59base model will even outperform the M4
29. 1:02Max in certain GPU-based things. But,
30. 1:05all the M5s have that, too. So, where
31. 1:07does this chip sit in this set of
32. 1:09choices? We've got an M5 MacBook Pro, an
33. 1:11M5 Pro MacBook Pro, an M5 Max MacBook
34. 1:14Pro, and this may seem a little silly
35. 1:16until you see that it is indeed not at
36. 1:19the top of every chart, the 96 GB 30 CPU
37. 1:22core 64 GPU core M5 Ultra Studio. And
38. 1:26this little engineering marvel will be
39. 1:28getting its own video very soon. I have
40. 1:30set up a battery of tests, and a guy
41. 1:32could point out that this isn't the most
42. 1:34apples-to-apples comparison. Get it?
43. 1:36Just because I've got different amounts
44. 1:38of RAM on almost every computer here. I
45. 1:40got 24 gigs on the M5, 48 gigs on the M5
46. 1:44Pro, 36 gigs on the M5 Max, 24 gigs
47. 1:47again on the M6, 96 gigs on the Ultra.
48. 1:50They've got different size SSDs, and
49. 1:52there truly will be tasks where the RAM
50. 1:54is is holding something back and not
51. 1:56necessarily its core count or its
52. 1:58cooling capacity. But at least both of
53. 2:00the base models have 24 gigs of RAM. And
54. 2:02honestly, I think we're going to have to
55. 2:03consider 24 gigs the baseline RAM moving
56. 2:06forward. I'm so sorry for what that
57. 2:08costs. I'm going to start out by
58. 2:09explaining and vastly simplifying what
59. 2:12happened between M5 and M6. Welcome to
60. 2:15CPU core education. Feel free to go to
61. 2:18this point in the video if you don't
62. 2:19want to follow along on this little
63. 2:20tangent. The M5 CPUs were made using
64. 2:23what Apple and TSMC are calling the
65. 2:253-nanometer process. Specifically, N3P
66. 2:29using FinFET transistors. The
67. 2:31transistors on a CPU are what do the
68. 2:33calculations, and there are billions of
69. 2:35them in every single chip. These
70. 2:36represent the actual ones and zeros when
71. 2:38people talk about digital things. And
72. 2:40again, to completely simplify how
73. 2:42computers work, the billions of
74. 2:44transistors in here are either on one or
75. 2:47off zero. And the biggest change they
76. 2:49came up with when moving to the
77. 2:502-nanometer process from the 3-nanometer
78. 2:53process was moving from FinFET a
79. 2:55field-effect transistor to GAAFET, a
80. 2:58gate-all-around transistor. If you zoom
81. 3:01in a thousand times and look at the
82. 3:02surface of a CPU, each transistor used
83. 3:04to be connected on three sides by
84. 3:06copper. And the oh-so-tiniest amount of
85. 3:09voltage could leak out on that fourth
86. 3:11side. Well, they plugged the leak. Now,
87. 3:13gate-all-around transistors have
88. 3:15connections all around the transistor.
89. 3:17And I was going to 3D print a couple of
90. 3:19models to further clarify how this
91. 3:21works, but my P2S said that would take
92. 3:2316 hours. So, you get this 3D model. And
93. 3:26hopefully, if I do the editing right,
94. 3:28it's like it's in this room with me.
95. 3:29See, with the FinFET, the copper's
96. 3:31making contact with the channel on three
97. 3:33sides, the top and both sides, where the
98. 3:35bottom is sitting on the substrate. With
99. 3:37the GAAFET, the copper is fully wrapped
100. 3:39around this whole channel. And there are
101. 3:41more channels shooting through the thing
102. 3:43also. Good luck with all that, future
103. 3:45editor Nicholas. So, controlling the
104. 3:46voltage better stole back some
105. 3:48efficiency. And also, there are more
106. 3:50packed in here. More efficiency means
107. 3:52more transistors with less heat
108. 3:53produced. Heat is the enemy of CPU
109. 3:56speed. Boom, Bob's your uncle, M6.
110. 3:59Additionally, M5 had introduced the
111. 4:01super core. They also then created a
112. 4:03mid-tier core, confusingly named the
113. 4:06performance core, not to be confused
114. 4:08with the performance core from the M4
115. 4:10lineup. And they still kept the
116. 4:11efficiency cores, although only in the
117. 4:13lower tier base model of the M5s. So,
118. 4:16the M5 base model has super cores and
119. 4:18efficiency cores, no performance cores.
120. 4:20Where the M5 Pro, Max, and Ultra chips
121. 4:22have a mixture of super cores and
122. 4:24performance cores, foregoing the
123. 4:25efficiency cores altogether. But, with
124. 4:27the introduction of the M6, they decided
125. 4:29there was a place for all types of
126. 4:31cores. The two super cores tackle heavy
127. 4:33single-threaded tasks, performance cores
128. 4:36do the hard multi-threaded activity, and
129. 4:38the efficiency cores strictly handle the
130. 4:40background administrative work. And that
131. 4:42is the design differences between the M5
132. 4:44and the M6. Now, let's run a bunch of
133. 4:46tests and see what happens. Whoa, my
134. 4:48camera's going crazy. How does the
135. 4:49next-gen 2-nanometer, 170 GB per second
136. 4:53RAM bandwidth M6 processor set hold up
137. 4:56against the increasingly wide lanes as
138. 4:58we move up the M5 lineup? First, we'll
139. 5:00start out by boosting the M6's
140. 5:01confidence with the single-core
141. 5:03benchmark. The rest of the computers
142. 5:05here should score about the same. M5, M5
143. 5:08Pro, Max, and Ultra all have the same
144. 5:10super core. There'll be slight
145. 5:12variations from core to core as I test
146. 5:14them, but I expect these will all be in
147. 5:15the same neighborhood. Not the M6,
148. 5:17though. As I ran through these, and I
149. 5:19even tossed in the M5 Air just for good
150. 5:21measure, that scored 730 points. M5 Max,
151. 5:24724. M5 base model, 730. Strangely, the
152. 5:28M5 Pro was the lowest-scoring. That was
153. 5:30odd. And then the M5 Ultra, 751. This
154. 5:33test serves to show that for single or
155. 5:36dual-threaded operations, which honestly
156. 5:38is 95% of what 95% of the public ever
157. 5:41does on a computer, but not you, dear
158. 5:44viewer. Some of you are real power users
159. 5:46like I am now. But, for most
160. 5:48applications, all of these computers
161. 5:49will be about the same speed. Because
162. 5:51your mail, your message apps, Telegram,
163. 5:53Teams, WhatsApp, your Excel, and web
164. 5:55browsers, all of that runs on one or two
165. 5:58super cores. And in the Ultra, for
166. 6:00instance, if you're just using those
167. 6:01apps, you're going to have 28 CPU cores
168. 6:04sitting there doing nothing most of the
169. 6:05time. That also means that the M6, which
170. 6:08finished with the score of 786, beating
171. 6:10out all of these competitors, will
172. 6:12actually be a faster computer if you
173. 6:14don't use heavily parallel processing
174. 6:16apps. If you've got a white-collar
175. 6:18computer job, the M6 sits at the head of
176. 6:20the pack. The M6 is a faster computer
177. 6:23than the M5 Ultra. So, after one test,
178. 6:25the M6 is in the lead. The multi-core
179. 6:27test was a bit of a different story,
180. 6:29though. From worst to best, we had the
181. 6:31base model M5, 4489. The base model M6,
182. 6:355542. The M5 Pro, 6978. The M5 Max,
183. 6:398550. And then the Ultra, 15,085.
184. 6:44I think something's wrong with my M5
185. 6:45Pro. That has the same CPU as the M5
186. 6:48Max, exactly. Their score is too far off
187. 6:50for that. That's odd. Anyway, GPU tests,
188. 6:52same story. M5 base, 24,000. M6 base,
189. 6:5631,000. M5 Pro, 54,000. The Max, 74,000.
190. 7:01And the M5 Ultra coming in with 132,000
191. 7:04Cinebench bucks. That's a lot higher.
192. 7:06So, with raw benchmarking, the M6 now
193. 7:09takes its position between the base
194. 7:10model M5 and the M5 Pro, which I guess
195. 7:13is where you'd expect it to be. But, raw
196. 7:15benchmarking doesn't tell the whole
197. 7:16story, now, does it? For my whole life,
198. 7:18looking at the speed of a computer
199. 7:19really just meant looking at the speed
200. 7:21of the CPU, the GPU, the RAM, and the
201. 7:23drive. But, now we've got an NPU and
202. 7:26media engines that actually do a lot of
203. 7:28stuff. One of those stuffs is exporting
204. 7:30video projects for you other YouTubers
205. 7:32and videographers out there. Live
206. 7:34streamers also use the media engines.
207. 7:35Plex servers use them. Playing YouTube
208. 7:37videos in the background while you do
209. 7:39other stuff on your computer is using
210. 7:41the media engines. And the M6 media
211. 7:43engine looks like it got a redesign. I
212. 7:45sat here and re-exported the last video
213. 7:47I made for YouTube on each one of these
214. 7:48computers over and over. This is a
215. 7:5014-minute long video shot in 8K and
216. 7:53rendered into 4K, which means 4K to 4K
217. 7:55exports will be a lot faster than this.
218. 7:57So, the actual time that you're seeing
219. 7:59here isn't representative of what you
220. 8:01might experience if you shoot in 4K.
221. 8:02But, the comparison between these two
222. 8:04will stay the same as a percentage. The
223. 8:06M5 base model took 5 minutes and 26
224. 8:08seconds to do this export. Next, there
225. 8:10was the M5 Pro with 4 minutes and 55
226. 8:13seconds. The M6 did it in 4 minutes and
227. 8:158 seconds. So, the M6 beats the M5 Pro
228. 8:18at video exporting by a whole minute
229. 8:20almost. M5 Max, 2 minutes and 45
230. 8:22seconds. This one has double media
231. 8:24engines, that's why. And the M5 Ultra
232. 8:26can export a 14-minute video from 8K
233. 8:29footage in a minute and 34 seconds. Shut
234. 8:32up. This has quad media engines, which
235. 8:34means if you are a video editor, the Mac
236. 8:36mini M6 is a better piece of hardware
237. 8:39than the M5 Pro in some circumstances,
238. 8:41if you don't use a ton of effect layers
239. 8:43and animated titles, color grading to
240. 8:45where you're going to overwhelm the
241. 8:46GPUs. But, it takes a lot to do that
242. 8:48now. Like, no normal-looking video would
243. 8:50put you in that position. For less than
244. 8:52half the price, the M6 Mac mini can be a
245. 8:55more capable video editing machine than
246. 8:57the M5 Pro. And if you do use a lot of
247. 8:59magnetic mask for whatever reason, I
248. 9:01showed in the video that I just made
249. 9:02that this computer will beat the M5
250. 9:04Ultra at that specific task. And that's
251. 9:07NPU bottlenecked. So, there must have
252. 9:09been some minor breakthrough in NPUs
253. 9:11between M5 and M6. Another interesting
254. 9:13thing I noted while this was taking
255. 9:14place, the RAM usage. Like I said, all
256. 9:17these computers have different amounts
257. 9:18of RAM. And for instance, the Ultra when
258. 9:21exporting video is using 52 gigs of RAM.
259. 9:23I think that's because since it has four
260. 9:25media engines, it's as if it's exporting
261. 9:27four videos simultaneously. So, it's
262. 9:29doing four projects and then it combines
263. 9:31them at the end. That's how I think that
264. 9:33works. So, it needs four times as much
265. 9:35RAM. Or that was my theory until I
266. 9:36looked at the M5 Pro, which this one has
267. 9:3948 gigs of RAM, and it was somehow
268. 9:40making use of 34 gigs of it to do the
269. 9:43export. But this only has one media
270. 9:45engine. So, if you have more, it'll just
271. 9:47use more. The M5 base model was only a
272. 9:49little slower than the M5 Pro. They both
273. 9:51just have the one media engine, so maybe
274. 9:53the RAM is the difference there. This
275. 9:55guy was only using 16 gigs of RAM for
276. 9:57the export. I don't know what to do with
277. 9:59that information, but there it is. And
278. 10:00with a whole day of testing, as you can
279. 10:02imagine, I had to make sure that all
280. 10:04these laptops had adequate power to do
281. 10:06their thing. And therefore, I'd like to
282. 10:07welcome you to the sponsored segment of
283. 10:09today's video, brought to you by Anker
284. 10:12Power. If you're a person who leaves
285. 10:13their laptop plugged in all the time,
286. 10:15but you're also a person who wants to
287. 10:17keep that laptop for as many years as
288. 10:19you can, then this ad placement may just
289. 10:21be for you. You may or may not know that
290. 10:23with lithium-ion batteries, your
291. 10:25charging profile makes a significant
292. 10:27difference in the overall longevity of
293. 10:29your device. Basically, the engineers
294. 10:31over at Anker have figured out a
295. 10:33charging curve, one that's specific to
296. 10:35your Mac. So, this charger knows which
297. 10:37computer's plugged into it, and it has a
298. 10:39care mode. Let's get that little icon
299. 10:41going. That care mode will be less harsh
300. 10:43than full-speed charging, and add up to
301. 10:46two additional years of life to your
302. 10:47computer. For instance, it's perfectly
303. 10:49safe to pump a whole lot of juice into
304. 10:51your computer between 0% and 45% battery
305. 10:55percentage. But then after that, it's
306. 10:56best for the battery's health for the
307. 10:58charger to back off the fuller it gets.
308. 11:00Anker has optimized that curve, and just
309. 11:02by changing how it charges your
310. 11:04computer, you can get up to an extra 40%
311. 11:07of lifespan out of it. There's also an
312. 11:09auto mode, where this charger will crank
313. 11:11out up to 100 watts for as long as your
314. 11:13computer will accept that much power.
315. 11:14And unlike other brands, it will
316. 11:16maintain that 100 watts without
317. 11:18overheating or slowing down. Different
318. 11:20size batteries will accept different
319. 11:22wattages for longer. For instance, the
320. 11:24MacBook Neo will only charge with like
321. 11:2645 watts, no matter what, no matter
322. 11:27which charger you're using. But the
323. 11:29MacBook Pro can pull that full 100 watts
324. 11:32and go from zero to 50% in like 30
325. 11:34minutes. Also, keeping your battery
326. 11:36charged at 100% so like keeping your
327. 11:38laptop plugged into the wall all the
328. 11:40time will degrade your battery quite a
329. 11:42bit faster than if you were to keep it
330. 11:44between 20 and 80% state of charge. And
331. 11:46Anker has a charging manager desktop
332. 11:48software now that can tell this brick to
333. 11:50stop once your computer gets to 80% cuz
334. 11:53they're communicating with each other.
335. 11:54There's even a little slider in the
336. 11:55software that can tell you how much
337. 11:57extra life you're getting out of these
338. 11:58settings. Look at that. add two years of
339. 12:00usable life to this battery just by
340. 12:02charging it differently. I'm a guy who
341. 12:04loves looking at data and even if I'm
342. 12:05not going to do anything with it. This
343. 12:07has a display on it that shows you
344. 12:08what's going on. Charging percent and
345. 12:10wattage, temperature, you can switch
346. 12:12between auto mode and care mode. It's
347. 12:13even got an accelerometer in it so it
348. 12:15auto orientates depending on which way
349. 12:17you have it plugged into the wall. It'll
350. 12:19receive firmware updates into the
351. 12:20future. This is a solid charger. There
352. 12:22is a link at the top of the description
353. 12:24to the Anker laptop smart display
354. 12:26charger. Let's put the CPUs and the GPUs
355. 12:29to work together. I always like to do a
356. 12:30nice Lightroom export test. This is sort
357. 12:33of a maximum load test because it'll
358. 12:35cause your computer to stress both the
359. 12:37CPUs and the GPUs together to export raw
360. 12:40photos into JPEGs. In this case, 1200
361. 12:42something photos from a couple of years
362. 12:44ago. The M5 base model runs through
363. 12:46those 1200 photos in just 3 minutes and
364. 12:4854 seconds. That's respectable. It's
365. 12:51also last place, not surprisingly. The
366. 12:53M6 comes in ahead by 50 seconds knocking
367. 12:55nearly 30% off that time at 3 minutes
368. 12:58and 4 seconds. And that's closer to the
369. 13:00M5 Pro's time than it is to the M5 just
370. 13:02for reference. The M5 Pro finishing in 2
371. 13:05minutes and 25 seconds. The M5 Max
372. 13:07speeds that up another 30 seconds
373. 13:09finishing in a minute 58. And the Ultra
374. 13:11almost cutting that time in half does
375. 13:13this export in a minute and 8 seconds.
376. 13:15That's 1200 photos processed in 68
377. 13:19seconds. AKA 17 photos per second 33
378. 13:22megapixel Sony raw to JPEG conversions.
379. 13:25Woofta. Next, onto some clumsy video
380. 13:28game comparisons. Games have so many
381. 13:30different graphic settings that I don't
382. 13:32know, I think we're just set now. You
383. 13:34can buy anything current and be fine.
384. 13:36You just have to be okay with
385. 13:37increasingly muddy graphics. But here we
386. 13:39are in Crimson Desert. I tested these on
387. 13:421440p on high graphic settings on an
388. 13:45Apple Studio Display, which does have a
389. 13:47max refresh rate of 60 frames per
390. 13:49second, which I understand the real
391. 13:51gamers out there consider this to be an
392. 13:53insult, but to my generation X eyes,
393. 13:55where I grew up with a 480 interlaced
394. 13:58picture at an effective 30 frames per
395. 14:00second in the 8-bit era, these look
396. 14:02great. Anyway, M5 base model was the
397. 14:05worst. In Crimson Desert, even with
398. 14:07frame generation on, although AI upscale
399. 14:09was turned off and that would be the fix
400. 14:11here if you're going to play with this
401. 14:12computer. You can play at 720 with AI
402. 14:14upscaling it to 1440p and it looks
403. 14:16convincing to me. Anyway, 1440 real
404. 14:19resolution at high settings, we're
405. 14:21getting less than 30 frames per second,
406. 14:22a very choppy 27. Gross. M6, I think
407. 14:26does really well here. Same settings,
408. 14:281440p, high settings, frame generation
409. 14:30on, hitting above 40 frames per second a
410. 14:33lot of the time. That's great. And what
411. 14:34you're seeing on the screen behind me,
412. 14:36the picture didn't actually look like
413. 14:37this. I shot this footage with my iPhone
414. 14:39in 24 frames per second, which means
415. 14:41pretty much every frame has multiple
416. 14:43frames in it, like the computer screen
417. 14:45frame changed while the iPhone shooting
418. 14:48frame was still open. Maybe that makes
419. 14:49sense, maybe it doesn't. So there's
420. 14:50multiple exposures per frame, so the
421. 14:52footage looks like ass. I will not
422. 14:54repeat this mistake in the future. I
423. 14:55just did not want to repeat loading the
424. 14:57game up in all these different computers
425. 14:58again. We're just going to have to go
426. 14:59off the data in the upper right corner
427. 15:01here. The game actually looks super good
428. 15:03on the screen, even at the crappy frame
429. 15:05rates. M5 Pro, a little over 50 frames
430. 15:07per second. And then we get into
431. 15:09overkill territory, where both the M5
432. 15:11Max and the Ultra just have the frame
433. 15:12rate pinned to the monitor's refresh
434. 15:14rate at 60 frames per second, no matter
435. 15:16what. The Ultra even does that at 4K.
436. 15:18The Ultra really is some kind of
437. 15:19monster. I did run all these back
438. 15:21through a game in crossover real quick
439. 15:23because you do take a little bit of a
440. 15:24performance hit with the translation
441. 15:26layer when you're playing Windows games.
442. 15:28But honestly, everything stayed super
443. 15:30playable. They rank in the same order
444. 15:31with the M6 doing a little better than
445. 15:33the M5 base in Kingdom Come Deliverance
446. 15:352 getting a little more than 30 frames
447. 15:37per second. Again, at 1440p, but this
448. 15:39time on medium settings. M5 Pros in the
449. 15:4150s, M5 Ultra once again pinning
450. 15:44everything to 60 frames per second even
451. 15:46in 4K, even in crossover. What even
452. 15:48can't this thing do for the low low
453. 15:50price of a used car? And finally, with
454. 15:52local image generation. Just another
455. 15:54thing I started tossing in to test
456. 15:56computers. Although, I'm having a hard
457. 15:57time accounting for some of the results
458. 15:59I got here and today, so far, I don't
459. 16:02have all the tools I need to dig into
460. 16:04why. The MX Power Gadget gauge thing
461. 16:06that I normally use hasn't been updated
462. 16:08since April, so the M6 processors don't
463. 16:11work right in there. And actually,
464. 16:12neither do the Ultra. And I don't think
465. 16:14the Max do either. I don't think that
466. 16:15they've got the sensing the super core
467. 16:17versus performance core thing figured
468. 16:19out. have told me a little bit more
469. 16:20about what's getting used, what's not
470. 16:22getting used, why the results I'm seeing
471. 16:24happened. That said, the M5 base model
472. 16:27after the warm-up takes about 16 seconds
473. 16:29to create a picture of a man standing in
474. 16:31the middle of the street in Manhattan.
475. 16:33The M6, which has more GPU cores than
476. 16:35the M5 base model, and each core should
477. 16:38be a little faster based on literally
478. 16:40every other test, it should be quite a
479. 16:41bit faster. It only improved that time
480. 16:43by 1 second. I tried it over and over.
481. 16:45That's odd. And similarly odd, the M5
482. 16:48Pro pumps this photo out in 6 seconds,
483. 16:50and the M5 Max, which has way more CPU
484. 16:53cores, only beats that by 1 second. Then
485. 16:55of course, the Ultra can crank out these
486. 16:57lifelike photos, 1024 by 1024 resolution
487. 17:00in 2 seconds flat. This monument to
488. 17:03engineering. I'm guessing that neural
489. 17:05accelerators are the big player here,
490. 17:07and since that's the norm now and the M6
491. 17:09is so new, maybe this software isn't
492. 17:11optimized to whatever new thing it's
493. 17:13able to do. But what do I know about
494. 17:15diamonds? I'm just a boxing promoter.
495. 17:17So, after all of this, the M6 Mac Mini,
496. 17:20well, it depends. For Bill Morrison, a
497. 17:22guy in the ops department of my salaried
498. 17:24job, this computer is actually at the
499. 17:26front of the pack. Nothing he ever does
500. 17:28taps more than two super cores. Bill
501. 17:30having 10 super cores would actually be
502. 17:32slower than having two super cores
503. 17:34because these super cores are slower
504. 17:35than these super cores. Although in
505. 17:37office apps, this is not noticeably
506. 17:39faster. They're both instantaneous.
507. 17:41These computers are all overkill for
508. 17:43Bill. For the YouTube creator who makes
509. 17:45two hour long interviews but doesn't
510. 17:47have a whole lot of effects and has to
511. 17:48wait a lot of time for exporting, this
512. 17:50computer sits just ahead of the M5 Pro
513. 17:53but behind the M5 Max. Though, as a side
514. 17:55note, the M5 Pro does do proxy
515. 17:57transcoding just a little bit faster
516. 17:59than the M6. But it's a lot more
517. 18:00expensive, too. For everything else, you
518. 18:02can consider this like an M5 Plus,
519. 18:05significantly beating the base model M5
520. 18:07at everything but performing in the
521. 18:09space between the M5 and the M5 Pro for
522. 18:12$800 less. Goodbye.
523. 18:15What am I doing? I'm switching to this.