1
00:00:00,000 --> 00:00:04,320
We just finished shooting this video yesterday, but there's a lot more that we can talk about

2
00:00:04,320 --> 00:00:09,440
that didn't make it into the cut. In fact Linus cut like a whole bunch of it, even after I cut

3
00:00:09,440 --> 00:00:15,680
a whole bunch of it. So what we're looking at here is the system itself. We took the tower heat sink

4
00:00:15,680 --> 00:00:21,440
off so we could see the CPU. One of the problems with prototype hardware like this is that it's

5
00:00:21,440 --> 00:00:28,800
not really built to be robust. In this case it basically died right after we shot.

6
00:00:28,880 --> 00:00:36,000
Wait, really? Yeah, like I went to like take it over and like we put it back together because

7
00:00:36,000 --> 00:00:39,840
Linus had taken the part to hold it up and all that kind of stuff, but plugged it in and like

8
00:00:39,840 --> 00:00:44,800
the RAM did up. I'm like, huh, that's weird. And there was this high-pitched buzzing. I'm like,

9
00:00:44,800 --> 00:00:51,200
huh, that's weird. I didn't hear that before. What's going on? Spent the afternoon yesterday after the

10
00:00:51,200 --> 00:00:56,960
shoot trying to troubleshoot what's going on with this thing. And the TLDR is that either the motherboard

11
00:00:57,600 --> 00:01:05,120
or the CPU is dead. As it turns out, this little chip right here was doing like 100 degrees celsius

12
00:01:06,080 --> 00:01:12,640
and this little coil next to it, that was screaming. So unfortunately this is going

13
00:01:12,640 --> 00:01:20,400
to have to go back to its owner in a non-working state. Now that being said, they actually kind

14
00:01:20,400 --> 00:01:26,160
of expected this. Apparently this stuff actually just dies all the time. Like I said, it wasn't made

15
00:01:26,160 --> 00:01:31,600
to be robust. It was meant to be a quick and dirty solution for testing. It was never meant to be like

16
00:01:31,600 --> 00:01:40,320
a production sample. So they didn't beef up the PCB. They didn't use over spec components. They

17
00:01:40,320 --> 00:01:44,160
literally just wanted to get it running and see what they could do with it. And the fact that

18
00:01:44,160 --> 00:01:50,080
they got this far at all is actually really impressive because this is a whole platform.

19
00:01:50,400 --> 00:01:57,280
I had platforms previously, but the last one they made was like in 2011 or even earlier like 2008

20
00:01:57,920 --> 00:02:03,200
for the Isaiah cores. What we're looking at here is like a brand new thing based on a socket that

21
00:02:03,200 --> 00:02:09,120
they shouldn't be using. And they built it up from, well, just the stuff that they had on hand,

22
00:02:09,120 --> 00:02:14,720
including their IP, their development efforts over the years. Like we said in the end of the video,

23
00:02:15,120 --> 00:02:20,480
they're gone now. Centaur, for the most part, like I think the name still exists and their

24
00:02:20,480 --> 00:02:26,960
intellectual property still exists under VA's control. But their entire R&D division, their

25
00:02:26,960 --> 00:02:32,400
developers, their engineers, that was all purchased by Intel back in 2021. Unfortunately, this is

26
00:02:32,400 --> 00:02:37,040
going to be the last of these that we see. And that being said, all of the intellectual property

27
00:02:37,040 --> 00:02:42,880
has been mostly transferred to Jiaoshin at this point, who, you know, if Intel decides to revoke

28
00:02:43,600 --> 00:02:51,040
the x86 license, probably won't care. So, yeah. Do you think there'll be another like

29
00:02:51,680 --> 00:02:56,800
CPU maker, basically? Like that's not Intel or AMD? Oh, well, Qualcomm is starting to pick up quite

30
00:02:56,800 --> 00:03:04,320
a bit. Oh, really? And Apple Silicon is also like, you know, undeniably successful. So like, it's not

31
00:03:04,320 --> 00:03:13,520
like x86 is the only game in town, nor will it be forever. As far as x86 goes, Intel, AMD, and I

32
00:03:13,520 --> 00:03:19,360
guess Jiaoshin, and is there anybody else right now? I don't think there is. Because I think VA

33
00:03:19,360 --> 00:03:25,600
actually had like most of the stragglers. Centaur and they also had Cyrix. They purchased them both

34
00:03:25,600 --> 00:03:31,200
at the same time and in 1999, I think, and they discarded the work that Cyrix was working on. But

35
00:03:31,200 --> 00:03:37,280
they used Cyrix as brand because they were more recognizable for the Cyrix 3, which was a Centaur

36
00:03:37,280 --> 00:03:45,840
design. Yeah. Poor Cyrix and poor Centaur. Intel like kind of really messed with them. They messed

37
00:03:45,840 --> 00:03:50,160
with a lot back then. Like, what you have to realize is that back in the 90s, it was still

38
00:03:50,160 --> 00:03:56,720
relatively common for like IBM and ST Micro and a whole bunch of other companies to like just

39
00:03:56,800 --> 00:04:01,680
straight up make Pentium chips. We're not, some Pentiums actually, yes. IBM actually made some

40
00:04:01,680 --> 00:04:10,000
Pentiums, I think. I distinctly remember seeing some Pentiums in like a drawer and like one was

41
00:04:10,000 --> 00:04:20,400
Intel and one was IBM and I'm like, okay. I know for a fact that 486s were like that and they

42
00:04:20,400 --> 00:04:26,320
continued to make 486s for a long time. In fact, you can still buy 486 computers that are embedded

43
00:04:26,400 --> 00:04:31,600
with like brand new 486 processors on modern technology like probably like 14 nanometer or

44
00:04:31,600 --> 00:04:36,240
something like that at this point. So like, yeah, that's kind of how it used to be. But Intel started

45
00:04:36,240 --> 00:04:41,360
cracking down on it and around the Pentium. AMD brought out their K6 and in this case,

46
00:04:41,360 --> 00:04:49,200
IDT, Integrated Device Technologies, spun off Centaur or they created Centaur as a subsidiary

47
00:04:49,920 --> 00:04:57,920
with the specific goal of creating a low-cost CPU. Because at the time, from what I understand,

48
00:04:57,920 --> 00:05:05,760
I think the founder of Centaur, Glenn Henry, was a IBM Fellow. Previously to this was the

49
00:05:05,760 --> 00:05:12,000
Senior Vice President of Products at Dell and he noticed that you can't buy a CPU at a cost

50
00:05:12,000 --> 00:05:16,640
that would allow computers to be sold below $1,000. And he's like, okay, well this market

51
00:05:16,640 --> 00:05:22,400
segment basically is where the computing industry is going to like heat up. So if we can get in there,

52
00:05:23,200 --> 00:05:28,000
we'll actually have a really good position. So they got to work. They built the wind chip,

53
00:05:28,000 --> 00:05:32,160
which I think they called the C6 and then the wind chip too. Neither were particularly

54
00:05:32,160 --> 00:05:36,880
successful in the West, but in the overseas market, they were more successful because they were low

55
00:05:36,880 --> 00:05:42,800
cost, very low power and like half the size of a Pentium in terms of the core dye. Their floating

56
00:05:42,880 --> 00:05:48,160
point unit wasn't great, but in terms of integer, which is what most desktop apps at the time were

57
00:05:48,160 --> 00:05:55,520
doing, it was perfectly fine. While that killed Cyrix for the enthusiast market, the whole floating

58
00:05:55,520 --> 00:06:01,680
point is the future because 3D and all that kind of stuff, it didn't really hurt IDT's wind chip or

59
00:06:01,680 --> 00:06:07,040
Centaur's wind chip in the same way because it was marketed as like a lower end, lower cost

60
00:06:07,040 --> 00:06:12,320
thing. So it had adequate performance, but it was good for power consumption and it was good for

61
00:06:12,400 --> 00:06:17,360
the wallet. So it actually did do a pretty good job, I think, early on. Ultimately, they were

62
00:06:17,360 --> 00:06:22,400
sold on to Via in 1999. Did they do anything else at IDT? I don't remember. I know that part of the

63
00:06:22,400 --> 00:06:27,600
reason why they went with the like super efficient design was because they had actually previously

64
00:06:27,600 --> 00:06:31,840
worked with MIPS, which is like a, they used to make processors that were kind of like risk,

65
00:06:31,840 --> 00:06:36,480
so reduced instruction set. That's kind of the philosophy they went at with x86. You know,

66
00:06:36,480 --> 00:06:46,240
they went at it from the idea that they were half-man, so like cost-effective and power efficient

67
00:06:46,240 --> 00:06:54,400
and half-beast x86. x86 is like a big monolithic beast even today, although there's like some

68
00:06:54,400 --> 00:07:03,920
risky stuff in it. It's just weird. CPUs are weird. I can tell. I don't know. As far as the

69
00:07:03,920 --> 00:07:09,200
motherboard goes, I have no idea what the cost would be. Probably going to be similar to other

70
00:07:09,200 --> 00:07:14,480
motherboards on the market, especially considering the CPU is considered to be like a value budget

71
00:07:14,480 --> 00:07:20,960
option. You know, they said that you needed like 24 of Intel's Xeon Platinum cores in order to

72
00:07:20,960 --> 00:07:27,360
match the performance of their AI accelerator, which unfortunately, to this day, nobody really

73
00:07:27,360 --> 00:07:34,320
knows how it works. Like there's there's specifications and like details about like the

74
00:07:34,320 --> 00:07:39,440
inner workings, but like I think the only publicly available information about its

75
00:07:39,440 --> 00:07:45,360
performance is from ML Perf from like 2019 or something like that. It wasn't great.

76
00:07:46,640 --> 00:07:52,080
The NVIDIA T4, I mean like we said in the video, it was not even a contest. It was

77
00:07:52,160 --> 00:07:57,520
way more expensive at $2,000 just for the card, but you could slot that into a machine that costs

78
00:07:57,520 --> 00:08:03,280
a lot less and then get roughly equivalent performance per dollar between this and that,

79
00:08:04,000 --> 00:08:10,320
but you could also add another one of those or even like a consumer GPU to get more AI oomph,

80
00:08:10,320 --> 00:08:16,480
right? Like more machine learning. Whereas this, I mean, you could, you could add them,

81
00:08:17,200 --> 00:08:22,560
but like by the time you're done, like it only has 44 PCI Express lanes. You're not going to add

82
00:08:22,560 --> 00:08:29,040
that many. I don't know. It's it's a, what was what the platform even looked like? It wouldn't

83
00:08:29,040 --> 00:08:38,080
use this socket purely, right? Like so I don't actually know what their plan was once like stuff

84
00:08:38,080 --> 00:08:44,640
like the T4 started to like come down in price once Intel's processors started coming out with

85
00:08:44,640 --> 00:08:49,680
more and more deep learning acceleration. I think that's probably what mostly put them off.

86
00:08:49,680 --> 00:08:56,000
That and Epic. What was Epic Rome? Did Epic Rome have 64 cores? I think it was 64 and they have eight.

87
00:08:56,000 --> 00:09:01,040
And is it cheaper? Epic? No, no, not at all, but Epic has way more PCI Express lanes. So you can

88
00:09:01,040 --> 00:09:05,920
just slot in way more stuff. So like it's a more scalable. Whereas I don't see this scaling. That's

89
00:09:05,920 --> 00:09:10,880
kind of the big problem and I think that's why nobody really bought it. Presumably if they had

90
00:09:10,880 --> 00:09:15,520
customers lined up, then somebody would have ended up with like a one off or something.

91
00:09:16,480 --> 00:09:22,160
But no, that's unfortunately the way that went. And now, of course, NVIDIA is swimming in cash

92
00:09:22,160 --> 00:09:28,320
because AI accelerators, deep learning, all that kind of stuff is just like the current hotness.

93
00:09:28,320 --> 00:09:33,840
Was it super micro stock is just going stonks right now because they sell the servers that

94
00:09:33,840 --> 00:09:38,720
this stuff goes in. So like, I'm kind of waiting for all that to die down a little bit before making

95
00:09:38,720 --> 00:09:47,280
any final calls on whether or not that's reasonable. But yeah, I don't think Santor would have been in

96
00:09:47,280 --> 00:09:52,560
the running for that. Do you think it's important that there's more like chips or do you think like

97
00:09:52,560 --> 00:09:59,360
AMD and Intel are like, yeah, we got this figured out and we don't need more CPUs in the market.

98
00:09:59,360 --> 00:10:04,640
The fact that Ryzen came along and forced them to do something different made it so that the

99
00:10:04,640 --> 00:10:10,160
market right now is so competitive like Intel had no competition in the data center,

100
00:10:10,160 --> 00:10:17,680
like maybe a little bit of power PC for a little while in the like 2010s. But like now ARM was

101
00:10:17,680 --> 00:10:22,720
starting to become a bit of an issue for them. But like, how much software is developed for ARM on

102
00:10:22,720 --> 00:10:28,240
the server market, right? So like, you'd have to spin that up. And obviously it has been at this

103
00:10:28,240 --> 00:10:33,440
point. But, you know, at the time, they didn't really have to worry that much about it. Whereas

104
00:10:34,160 --> 00:10:40,800
now AMD with Epic, like they're taking a significant chunk out of Intel's market share in the like

105
00:10:40,800 --> 00:10:45,600
the data center, as well as the desktop, like Intel is losing a lot of ground. That being said,

106
00:10:45,600 --> 00:10:51,680
they're still strong. It's just kind of unthinkable. Like as of 2015, as early as 2015 or late, sorry,

107
00:10:51,680 --> 00:10:56,480
people were calling that AMD was going to go under. Basically, the only thing keeping them alive was

108
00:10:56,480 --> 00:11:03,680
the console contracts that they had supplying Microsoft and Sony. At least that's my perception.

109
00:11:03,680 --> 00:11:10,320
Like basically, nobody was buying their CPUs or even their GPUs at that point, I don't think.

110
00:11:10,320 --> 00:11:16,480
They just weren't competitive, right? Especially the FX series. Like, oh, don't get me started on the

111
00:11:16,480 --> 00:11:22,800
FX series. You know, get started. Just because you have eight integer units and four floating

112
00:11:22,880 --> 00:11:29,440
point units doesn't mean you have eight cores. Like most stuff that we're doing today requires

113
00:11:29,440 --> 00:11:34,320
floating point. I thought we learned this lesson in the 90s with Cyrix, but apparently not. AMD

114
00:11:34,320 --> 00:11:41,040
relearned the lesson and they suffered for it, I think rightly. But we also suffered as consumers.

115
00:11:41,040 --> 00:11:46,320
Intel quad core with slightly more performance, slightly more performance, a little bit more clocks.

116
00:11:47,280 --> 00:11:56,640
Nothing new. Whereas now, like my Ryzen 3000, 3900X that I use at home, it's like real old now.

117
00:11:56,640 --> 00:12:01,680
And that's not something that would have happened over the course of like three years or four years.

118
00:12:02,560 --> 00:12:10,160
Like back in the 20, like the 2010s, like CPUs had stagnated completely. There was no development

119
00:12:10,160 --> 00:12:17,200
whatsoever. It was just like, yep, a little bit more, 10%, 15% in these workloads asterisk.

120
00:12:24,400 --> 00:12:30,240
Competition is good and more chips on the market means more innovation. And that's why I'm kind

121
00:12:30,240 --> 00:12:36,720
of excited to see like ARM takeoff in Windows, because it's taken off on macOS pretty well.

122
00:12:36,720 --> 00:12:43,280
Powers, you know, Android and iOS. And we had x86 on Android for a short period of time, but that

123
00:12:43,280 --> 00:12:48,960
didn't really do much. Actually, if anything, more chips in the smartphone market would actually

124
00:12:48,960 --> 00:12:52,560
be a good thing. You know, just because they're all ARM doesn't mean that they're all the same,

125
00:12:52,560 --> 00:12:58,800
right? Like Apple Silicon is very different from, you know, Snapdragon 8cx2 or whatever. You know,

126
00:12:58,800 --> 00:13:05,760
more players in that game would be great. Like, say NVIDIA made a new SoC, or AMD got in the game,

127
00:13:05,760 --> 00:13:11,440
or even Intel, right? Like, that could shake up the mobile industry and make phones exciting

128
00:13:11,440 --> 00:13:16,880
again, right? Like, hey, we can do all sorts of like really cool stuff on our phones now. Whereas

129
00:13:16,880 --> 00:13:22,720
now it's like, okay, the camera's better. They say it's faster, but like, is it really? Like,

130
00:13:22,720 --> 00:13:27,200
it's not actually appreciably faster. Like, it's kind of like that Intel stagnation again.

131
00:13:27,200 --> 00:13:30,400
Yeah. And I don't know how we got onto this topic from the center.

132
00:13:31,200 --> 00:13:33,760
But yeah, we're just talking about this and then we're just talking about

133
00:13:34,800 --> 00:13:38,960
it. It's okay. It's okay. Casual talks. Yeah. Like, I don't know. It's just, it's one of those

134
00:13:38,960 --> 00:13:41,920
things where I think that more competition in the market is always going to be a good thing.

135
00:13:42,560 --> 00:13:49,840
And whenever there's consolidation, so like Centaur no longer exists. I mean,

136
00:13:49,840 --> 00:13:53,600
they were mostly only doing things for the embedded market anyway. But like,

137
00:13:53,600 --> 00:13:57,520
Intel buying them just basically means that Intel has more of a monopoly now.

138
00:13:58,320 --> 00:14:03,680
Right. Whereas, I guess it's not, it's technically a duopoly. We got Intel and we got AMD.

139
00:14:04,960 --> 00:14:09,680
There is also ARM. ARM is getting, and RISC-5 actually. RISC-5 is getting bigger in the embedded

140
00:14:09,680 --> 00:14:14,400
space, which is good because that's another new architecture that's out and it's open actually.

141
00:14:14,400 --> 00:14:22,080
So like, people can use it with, I'm not sure if it's royalty free, but like, you can contribute

142
00:14:22,080 --> 00:14:26,640
to the standard and you can adjust it the way you want to. In kind of a similar way to ARM,

143
00:14:26,720 --> 00:14:33,360
but like, I think it's more, more liberal. I'd have to look it up. I don't remember off the

144
00:14:33,360 --> 00:14:38,080
top of my head. Is there more things we won't talk about but this chip? Yeah, actually, there is

145
00:14:38,080 --> 00:14:45,360
one more thing. You might be wondering why we went with Trident Z Royal RAM. The reason for this

146
00:14:46,400 --> 00:14:52,560
is not because we thought we'd bling it out. Although it does kind of match the laser etching.

147
00:14:52,560 --> 00:14:57,600
The reason we used this is because this is actually surprisingly our only kit of four

148
00:14:57,600 --> 00:15:04,800
Samsung B-Die. Oh, really? BIMS. And the reason we needed that is because the IMC, the memory

149
00:15:04,800 --> 00:15:11,200
controller on this thing, it was definitely an afterthought. Like, definitely, definitely. Like,

150
00:15:11,200 --> 00:15:16,320
it only goes up to 3,200 megatransfer per second for one. So like, this RAM is rated for, what is

151
00:15:16,320 --> 00:15:23,760
it rated for? Okay, it's 3,200. Fine. But you have to input the manual, like the timing is

152
00:15:23,760 --> 00:15:28,800
manually. There's no XMP, no nothing like that. So you just go in, you go manual, and then you

153
00:15:28,800 --> 00:15:34,160
have to like set every individual timing. Thankfully, the person who lent us this board

154
00:15:34,160 --> 00:15:39,280
actually already had timing styled in that were stable and it worked great. Maybe that was why

155
00:15:39,280 --> 00:15:44,880
the board blew up. I don't know. But yeah, like you can't even punch in directly some of the values.

156
00:15:45,440 --> 00:15:49,520
You punch in a value and it's like, oh, it's too high. And it's like, oh, okay, well, I'll punch in

157
00:15:49,520 --> 00:15:54,400
as close as I can get and then I'll put in like an offset because there's another offset function

158
00:15:54,400 --> 00:15:58,640
further down. And it's like, why didn't they just, like, this is super weird development hardware

159
00:15:58,640 --> 00:16:05,280
stuff. Like, why not just make it and the offset automatically based on what you type in?

160
00:16:05,280 --> 00:16:12,080
Just the last one. Well, I mean, that is fine. Yeah, the the BIOS is actually really weird too.

161
00:16:12,240 --> 00:16:16,960
There's all kinds of dead ends and like stuff placed in locations you wouldn't expect. And

162
00:16:17,520 --> 00:16:22,960
it was clear that it wasn't really meant for prime time yet. That being said,

163
00:16:22,960 --> 00:16:26,160
running this at 2.5 gigahertz, which again, might have also killed it,

164
00:16:27,040 --> 00:16:33,040
versus the two gigahertz that is the standard. And with the 3200 megatransfer per second RAM,

165
00:16:33,040 --> 00:16:38,560
it actually did do a fairly decent bump in performance. Like, yeah, it's not good.

166
00:16:39,520 --> 00:16:46,320
Like, you're looking at like, Haswell level performance, maybe, but it's running at 2.5

167
00:16:46,320 --> 00:16:53,840
gigahertz. If it was running at like 3.5 or 4, like that might actually be pretty fast. And the fan

168
00:16:53,840 --> 00:16:59,360
on our cooler basically never spun. Like, it would spin when you like hit it with a load. I think

169
00:16:59,360 --> 00:17:06,640
I only ever got it to like 60 or 70 degrees, maybe 80 with like, like the most prime 95 I

170
00:17:06,640 --> 00:17:10,880
could throw at it. And then immediately the fan would stop after you're finished because it cooled

171
00:17:10,880 --> 00:17:15,760
down immediately. Like, it's actually pretty cool. That's another consideration, I guess,

172
00:17:15,760 --> 00:17:23,040
in the data center. You've got banks and banks of machines. This thing's sipping power. Right?

173
00:17:24,000 --> 00:17:31,040
I don't know. Either way, I know that Jiaoxin used a lot of the development that was behind this

174
00:17:31,040 --> 00:17:37,360
in their Silicon as well. And vice versa. There's actually the chip, like chipset parts of the system

175
00:17:37,920 --> 00:17:42,160
are all labeled Jiaoxin, even though this is not a Jiaoxin product. So I'm guessing that either

176
00:17:42,160 --> 00:17:49,280
Centaur supplied that too. And it's just known as Jiaoxin. Or Jiaoxin supplied it here. It's

177
00:17:49,280 --> 00:17:54,160
interesting to see like a kind of missing link, because we had that other Jiaoxin CPU from a

178
00:17:54,240 --> 00:18:01,280
little while ago that we knew then was based on Isaiah, but we didn't really know much about this.

179
00:18:02,080 --> 00:18:05,360
And I think that might have been because we, this might not have even been announced at the time

180
00:18:05,360 --> 00:18:11,600
we made that video. Come to think of it. So like, this is like what they were working on at that time.

181
00:18:12,240 --> 00:18:17,040
I don't think it would have been successful had they actually launched it. But it would have been

182
00:18:17,040 --> 00:18:23,360
interesting actually having another competitor, at least in some respects in the x86 market,

183
00:18:23,440 --> 00:18:28,080
even if it's in the server or whatever. Kind of like the Intel Arc. It's like, yeah, it's not good,

184
00:18:28,080 --> 00:18:34,640
but it's, it's a cool, it's a, it's like an interesting, I guess, product to begin, I guess.

185
00:18:34,640 --> 00:18:41,840
Yeah, like, I'm really rooting for Arc because we've reached kind of a stagnation point with GPUs too.

186
00:18:41,840 --> 00:18:46,560
It's not that the technology has slowed down in GPUs, it's that the pricing for performance

187
00:18:47,120 --> 00:18:52,080
has slowed down dramatically. And in fact, backslid, like that's, that's the major problem. And also

188
00:18:52,080 --> 00:19:02,560
the lack of like dedicated RAM, right? Like the 3060, the RTX 3060 came with 12 gigs of RAM.

189
00:19:03,120 --> 00:19:08,000
And Linus got tripped up by that yesterday because like, oh yeah, the 4060s only got eight. That was

190
00:19:08,000 --> 00:19:13,840
a big thing. Yeah, that was a big thing because the 3060 had 12. I don't know. It's, it's one of those

191
00:19:13,840 --> 00:19:20,160
things where like, competition is always good. It's, it's never not good. I guess no, there's,

192
00:19:20,160 --> 00:19:24,320
there's all, there are, I suppose, reasons why you wouldn't want competition.

193
00:19:26,800 --> 00:19:32,320
Like what's one example? I'm trying to think it's probably going to be like a super political thing,

194
00:19:32,320 --> 00:19:38,960
but like, you know, hospitals, I don't like that in the US. Oh, okay, that's fair.

195
00:19:40,400 --> 00:19:44,400
You know, like, oh, I can't go to that hospital. It's outside of my network.

196
00:19:45,680 --> 00:19:48,960
You know, I know I'm bleeding out, but can you please go to the one a couple of miles away?

197
00:19:49,920 --> 00:19:53,600
Like, for products, competition is always good.

198
00:19:55,120 --> 00:19:57,680
I mean, but aren't your products to the hospital?

199
00:19:58,480 --> 00:20:06,320
Yes. No, like, like legitimately, like, unironically, yes. Hospitals are apparently selling

200
00:20:06,320 --> 00:20:14,480
customers, like, I say customer, wow, patient data for like AI training and like ad targeting

201
00:20:14,800 --> 00:20:18,800
Oh, I've heard this. I've heard this. This is crazy. Wow, this is so like.

202
00:20:18,800 --> 00:20:21,680
Yeah, and it's not, it's not a few of them. It's like a lot of them. Yeah.

203
00:20:22,560 --> 00:20:28,240
It's actually really bad. It's really tragic. I hate that, like, this happened under our watch,

204
00:20:28,880 --> 00:20:32,800
but I'm pretty sure we didn't kill it. And the reason why is because that's screaming

205
00:20:32,800 --> 00:20:37,760
from this coil. We heard that on set while it was still running. Oh, that was why,

206
00:20:37,760 --> 00:20:41,680
why Linus actually finally turned it off because he was like, oh yeah, like we can hear that on

207
00:20:41,680 --> 00:20:45,040
audio, right? And they get, yeah, that's coil wind. He just like flipped it off. He's like,

208
00:20:45,040 --> 00:20:50,560
I'll just take it. I wanted to pick it up anyway for the end of it, right? So he did that.

209
00:20:50,560 --> 00:20:55,920
And that was the last time it ever ran. In the small chance you get a fix, maybe you can do a

210
00:20:55,920 --> 00:21:02,640
pickup. Yeah, I don't think I'm gonna, I can't. I spent all afternoon yesterday performing CPR.

211
00:21:05,040 --> 00:21:08,720
There's, there's not a whole lot else I can do. Like I checked the socket for bent pins.

212
00:21:08,720 --> 00:21:14,240
There was one that was iffy. I fixed that. No, I cleaned the bottom of the CPU for the pads just

213
00:21:14,240 --> 00:21:22,480
in case Linus had finger grease on it. And no, I tried, you know, pulling out the battery.

214
00:21:24,320 --> 00:21:34,240
Whatever. I tried. Successfully pulled out the battery and held the scene off for clear,

215
00:21:34,240 --> 00:21:40,480
which apparently on other versions of this board are like doing that, like holding CMOS clear

216
00:21:40,480 --> 00:21:46,720
after pulling power and the battery for about 15 seconds. Brings them back to life. Not in this

217
00:21:46,720 --> 00:21:50,800
case. So like I went through a whole bunch of stuff. Check the jumpers, making sure that they

218
00:21:50,800 --> 00:21:57,440
were still okay. They're all the same as when we started shooting. So yeah, sadly, I don't think

219
00:21:57,440 --> 00:22:00,720
there's anything that I'm going to be able to do. It's going to have to go back to the owner and see

220
00:22:00,720 --> 00:22:07,440
if he can do some additional diagnostics. Hopefully you enjoyed watching this. I certainly enjoyed

221
00:22:07,440 --> 00:22:14,000
rambling for more like 45 minutes. Really? Yeah, it's like quarter after 12.

222
00:22:15,760 --> 00:22:20,560
Let us know if you really like this series and we'll keep doing it because I want to do more

223
00:22:20,560 --> 00:22:25,040
of these. I want to rant at the camera a bit. I don't get the opportunity to do that if I'm just

224
00:22:25,040 --> 00:22:29,920
doing an A-roll or like even if I put a rant in an A-roll Linus, it's just like, yeah, that's too

225
00:22:30,000 --> 00:22:38,000
ranty or that's too harsh or that's too off topic. And I'm just like, it's not that off.

226
00:22:39,760 --> 00:22:42,320
Leave a like.