1
00:00:00,000 --> 00:00:03,840
Happy birthday me! Here's my present.

2
00:00:03,840 --> 00:00:10,400
Previously on LTT, I showed you guys this incredible $30,000 CPU cooler

3
00:00:10,400 --> 00:00:14,160
that is designed for one reason and one reason only.

4
00:00:14,160 --> 00:00:21,280
To chase PC performance records. But while you guys might have expected me to turn a CPU blue,

5
00:00:21,280 --> 00:00:27,480
instead, it was your balls. Because I teased it, but never actually hooked it up to any hardware.

6
00:00:27,480 --> 00:00:31,440
Well that changes today. And our goal is simple.

7
00:00:31,440 --> 00:00:36,600
Break a world record. An overclocking legend Charles Fuggerworth here

8
00:00:36,600 --> 00:00:40,120
gave me his personal guarantee that we will do it.

9
00:00:40,120 --> 00:00:43,880
So what are we waiting for? Well, we've got to prep the motherboard,

10
00:00:43,880 --> 00:00:47,880
we've got to insulate the socket, we've got to update the BIOS, we've got to upgrade the VGA BIOS,

11
00:00:47,880 --> 00:00:51,280
we've got to install the Windows, we've got to tune the Windows,

12
00:00:51,280 --> 00:00:55,760
and we've got to prep the VGA driver. And we've got to get the power cords.

13
00:00:55,760 --> 00:00:59,280
Oh, oh no, we've got those, we've got the power cords.

14
00:00:59,320 --> 00:01:03,320
We've got seven circuits to work with. Just seven? What, is that not enough?

15
00:01:03,320 --> 00:01:07,160
I think seven will work. Okay, as for the rest of it, well,

16
00:01:07,160 --> 00:01:12,160
it's a good thing we've got the whole day booked for this. And we've got the next 20 seconds booked for our sponsor.

17
00:01:20,960 --> 00:01:26,120
Before we can begin, one of the key enemies of subzero overclocking is condensation.

18
00:01:26,120 --> 00:01:29,160
See, the air around us is full of moisture.

19
00:01:29,160 --> 00:01:35,320
And it will gather on any cold surface, turn to liquid, and drip all over your sensitive electronics.

20
00:01:35,320 --> 00:01:38,760
Now, there are a number of different ways of approaching this problem.

21
00:01:38,760 --> 00:01:43,320
The board I'm holding in my hand uses a very similar approach to what you guys saw recently

22
00:01:43,320 --> 00:01:46,920
when we were chasing world records with our RTX 5090 Astral.

23
00:01:46,920 --> 00:01:50,280
You basically take PlastiDip, suck it up into a syringe,

24
00:01:50,280 --> 00:01:53,560
and drip it all over the sensitive components.

25
00:01:53,560 --> 00:01:56,640
Being sure to keep it away from any contact pins,

26
00:01:56,640 --> 00:02:00,200
say in PCI Express sockets, or in your power plugs.

27
00:02:00,200 --> 00:02:04,520
Next, you find a high quality neoprene-like closed cell foam,

28
00:02:04,520 --> 00:02:07,840
cut it up into nice little shapes, and fill in any big gaps.

29
00:02:07,840 --> 00:02:11,080
Of course, that's a ton of work, makes your board super ugly,

30
00:02:11,080 --> 00:02:16,540
and in the event that anything goes wrong, ASUS certainly isn't taking this back for an RMA.

31
00:02:16,540 --> 00:02:19,720
So Charles has a technique that he likes better.

32
00:02:19,720 --> 00:02:23,000
And it's to use nothing? Absolutely not.

33
00:02:23,000 --> 00:02:27,560
I just use a kingpin socket heater. Oh, so that goes on the back here?

34
00:02:27,560 --> 00:02:31,040
Yeah, it goes on the back of the board. It's a couple hundred watts of heating element.

35
00:02:31,040 --> 00:02:35,600
So counter-intuitively, you're heating the CPU area?

36
00:02:35,600 --> 00:02:39,240
Correct, correct, correct. It prevents moisture from building up

37
00:02:39,240 --> 00:02:45,200
on the backside of the board. Okay, and then what about the front? I use a little bit of insulation, just neoprene.

38
00:02:45,200 --> 00:02:49,360
Just kind of like this? Yeah, just a little bit of neoprene. Any glue or anything? No, no, no, I'll eat.

39
00:02:49,360 --> 00:02:52,520
And then I use WD-40, WD-40 specialist.

40
00:02:52,520 --> 00:02:55,840
Shut up. No, I- You put WD-40 in your CPU socket.

41
00:02:55,840 --> 00:03:01,000
I spray it directly in the socket, and I fill the socket, and I can actually fire the system up with the socket wet.

42
00:03:01,000 --> 00:03:04,120
Now, you told me that you thought a manufacturer

43
00:03:04,120 --> 00:03:08,440
like wouldn't be able to tell you'd put that in there. Is that true? Correct, it's undetectable.

44
00:03:08,440 --> 00:03:12,400
Completely undetectable. That's a tech tip.

45
00:03:12,400 --> 00:03:16,560
Little bit of RMA fraud tips. We're not actually advocating for that.

46
00:03:16,560 --> 00:03:18,400
Oh, you need your board clean.

47
00:03:20,320 --> 00:03:23,560
We're actually gonna be using the first board today though, and it looks great.

48
00:03:23,560 --> 00:03:28,880
Like, we could probably run Sub-Zero for hours, days at a time on this.

49
00:03:28,880 --> 00:03:34,080
But there is one small problem. Noctua probably doesn't make a secu-firm mounting bracket

50
00:03:34,080 --> 00:03:37,440
for your Cascade Chiller, so how do we get the thing on the thing?

51
00:03:37,440 --> 00:03:41,680
I make my own custom 3D printed brackets. These are for different sockets.

52
00:03:41,680 --> 00:03:46,720
That's for X299. Oh, cool, okay. And this is for AMD, AM5, AM4.

53
00:03:46,720 --> 00:03:50,400
Oh, cool. Hey, got the Extreme Systems logo on there.

54
00:03:50,400 --> 00:03:53,760
Now, it looks like you're using, oh, crap.

55
00:03:53,760 --> 00:03:57,800
Nylon. It's the, oh. That's the HP Multi-Jet Fusion.

56
00:03:57,800 --> 00:04:01,920
Wow, it feels, you can't break it. Feels great.

57
00:04:01,920 --> 00:04:05,800
It's the HP Multi-Jet Fusion 4200. Like, I don't think I could break this.

58
00:04:05,800 --> 00:04:08,960
No. Just in case I ever have to swap it. Can you show me how to do it?

59
00:04:08,960 --> 00:04:13,960
Sure. You said you wanted a screw? Yeah, yeah. So just pull it back a little bit.

60
00:04:13,960 --> 00:04:17,000
Expose the O-ring. Uh-huh. Pull the O-ring.

61
00:04:17,000 --> 00:04:21,640
And I just give it the old, ow, I stabbed myself. Well, we're off to a good start today.

62
00:04:21,640 --> 00:04:23,560
Okay, there we go. Okay, O-ring's off.

63
00:04:25,080 --> 00:04:29,160
And that's that. Zip-tie engineering. Now, have you ever shown anyone

64
00:04:29,160 --> 00:04:32,920
what's inside this evap-head? 20 years ago. Super cool.

65
00:04:32,920 --> 00:04:37,480
And can you kind of describe it for us? Like, is it a maze style or like, what are-

66
00:04:37,480 --> 00:04:41,000
This is a spiral evap-er. This is the prototype spiral evap-er.

67
00:04:41,000 --> 00:04:44,000
This is the first one I ever made. What's better about the newer ones?

68
00:04:44,000 --> 00:04:48,720
The newer ones are a thinner material. So better, faster heat transfer.

69
00:04:48,720 --> 00:04:52,520
Now, if I wanted to swap it out for AMD then, do you want to show us what that would look like?

70
00:04:52,520 --> 00:04:58,480
This is a mounting plate. Yeah. Goes into the adapter, goes right on.

71
00:04:58,480 --> 00:05:03,040
Goes on like that. Install the O-ring. Oh, that's actually shockingly elegant.

72
00:05:03,040 --> 00:05:06,600
Now, are these O-rings one-time use? Did you just- No, no, no, they're multi-time use,

73
00:05:06,600 --> 00:05:10,000
but I got a lot of them. Oh, I see, okay. You were just worried I was going to ruin it.

74
00:05:10,000 --> 00:05:13,120
Yes, I was. Being honest. I've watched your videos, come on.

75
00:05:13,120 --> 00:05:17,120
Why does he have the sickest burns? For the coldest machine. Yeah.

76
00:05:17,120 --> 00:05:20,120
Now, is there a reason that we're going Intel today,

77
00:05:20,120 --> 00:05:25,000
specifically the 285K? The 285K is great for the benchmark

78
00:05:25,000 --> 00:05:28,880
that we're going to run. An NPU boost is extremely strong in the benchmark.

79
00:05:28,880 --> 00:05:33,840
Last question for you. Did you choose nylon for its ruggedness or for its insulating properties or both?

80
00:05:33,840 --> 00:05:37,360
Both. I went with PLA at first and the PLA broke right off the bat.

81
00:05:37,360 --> 00:05:40,880
Before I put this on, maybe dumb question, but how flexible is this?

82
00:05:40,880 --> 00:05:46,480
Um, okay. The hose itself, this part is flexible.

83
00:05:46,480 --> 00:05:50,560
That part is not. Okay. So is everything I'm doing okay?

84
00:05:52,280 --> 00:05:53,240
Okay, yep.

85
00:05:55,280 --> 00:05:58,920
Okay. Oh, great. We don't even have any thermal compound,

86
00:05:58,920 --> 00:06:02,840
but here's the thing. You can't use just any thermal compound.

87
00:06:02,840 --> 00:06:06,440
What would you go with? Probably some thermal grizzly cryonaut. And the reason for that

88
00:06:06,440 --> 00:06:12,500
is that certain thermal compounds might perform great at ambient temperature or hot,

89
00:06:12,500 --> 00:06:15,680
like a liquid metal, for instance, highly thermally conductive,

90
00:06:15,680 --> 00:06:18,940
but here's the thing. We're not gonna be at room temperature.

91
00:06:18,940 --> 00:06:22,920
We're gonna be way below and that liquid metal would become solid metal.

92
00:06:23,900 --> 00:06:27,800
Not too great. Also liquid metal is a bit messy. You're not into the mess.

93
00:06:27,800 --> 00:06:33,040
No, not at all, actually. I guess you do this for enough years. You just, you become not into the mess

94
00:06:33,040 --> 00:06:36,040
and it just goes on like this. Hold on one more second here. We got one more piece.

95
00:06:36,040 --> 00:06:40,280
We got a new piece of neoprene here. Right, for around the bottom here.

96
00:06:40,280 --> 00:06:43,440
Correct, correct. I need to cut it so it needs to clear the memory slots

97
00:06:43,480 --> 00:06:47,040
because these are dual dim board. Oh, sure.

98
00:06:47,040 --> 00:06:50,240
And then that just goes on there or a little something like that. Once you get to there,

99
00:06:50,240 --> 00:06:53,880
then we'll bring it down another, about a centimeter. About a centimeter, okay.

100
00:06:53,880 --> 00:06:59,920
Now it's funny. You were making fun of me the other day for using a rando mixture of imperial and metric units,

101
00:06:59,920 --> 00:07:04,400
but I just heard you use a centimeter. Well, it was less than an inch, so I had to use something else.

102
00:07:04,400 --> 00:07:07,420
You're measuring things like a Canadian now, boys.

103
00:07:07,420 --> 00:07:11,840
Well, I'm in Canada. And what's special about this particular Viper kit?

104
00:07:11,840 --> 00:07:16,240
That's the new ultra sets from Patriot. That's the 6000 C28.

105
00:07:16,240 --> 00:07:19,560
So just super tight latencies and good rated speed.

106
00:07:19,560 --> 00:07:22,800
Correct. With CPUs, you know, you guys would take the time

107
00:07:22,800 --> 00:07:27,800
and you would hand bin. But in the case of memory, do you basically trust that the top kits

108
00:07:27,800 --> 00:07:31,560
from someone like a Patriot are already binned enough or do you bin those further?

109
00:07:31,560 --> 00:07:35,440
We've been them further. Okay, so this is a particularly nice kit of this.

110
00:07:35,440 --> 00:07:39,040
Correct, super binned. God binned.

111
00:07:39,040 --> 00:07:42,120
The last thing we're gonna need before we can fire up our bench is a GPU.

112
00:07:42,120 --> 00:07:46,520
Now, we're not gonna need RTX 5090 performance for the first record we're chasing

113
00:07:46,520 --> 00:07:49,940
because it's solely concerned with CPU benchmarks.

114
00:07:49,940 --> 00:07:54,760
But it's here already anyway. It's subzero prepped with plastic dip all over it,

115
00:07:54,760 --> 00:08:01,080
not to mention paper towel. And it's hooked up to Bruce Chillis,

116
00:08:01,080 --> 00:08:04,560
our industrial chiller, which we don't have to run at subzero speed.

117
00:08:04,560 --> 00:08:09,600
So we're probably just gonna run it at, you know, 10, 20 C for starters.

118
00:08:09,600 --> 00:08:13,680
Let's get this on here. Now that the bench is together, it's time to fire up the chiller.

119
00:08:13,680 --> 00:08:18,520
But there's a specific launch sequence starting with the first phase.

120
00:08:18,520 --> 00:08:23,240
If you haven't seen part one, this is a cascade system that essentially uses

121
00:08:23,240 --> 00:08:26,960
a refrigeration system to cool a refrigeration system,

122
00:08:26,960 --> 00:08:32,240
to cool a third refrigeration system in order to reach ungodly low temperatures.

123
00:08:32,240 --> 00:08:35,840
We're expecting negative 90 Celsius on our CPU,

124
00:08:35,840 --> 00:08:41,320
rock freaking solid. But we can't just turn on the compressors willy-nilly.

125
00:08:41,320 --> 00:08:45,040
We've got to turn on the first phase first, which is the bottom one.

126
00:08:45,040 --> 00:08:49,840
Correct, neat. Because Charles actually legitimately doesn't know

127
00:08:49,840 --> 00:08:52,860
how much power it draws. This will be fun for both of us.

128
00:08:52,860 --> 00:08:57,000
We got a few kilowatts and we're gonna hook them all up in line with,

129
00:08:57,000 --> 00:09:00,360
here we go, our power. Ah, it's alive.

130
00:09:02,520 --> 00:09:06,520
Oh, what do we got? What do we got? Oh, 116.7 volts.

131
00:09:07,000 --> 00:09:10,720
941, 941 watts. Just a few.

132
00:09:10,720 --> 00:09:15,160
So you can see why Intel's recent woes around power consumption of their chips

133
00:09:15,160 --> 00:09:20,160
doesn't bother him. His cooler, a third of it, draws 1,000 watts.

134
00:09:20,160 --> 00:09:25,840
Now that the first two phases are kind of settled in, they're at 700, 750 watts each,

135
00:09:25,840 --> 00:09:29,080
but we haven't even added the third phase yet.

136
00:09:29,080 --> 00:09:32,520
I can't read it. What does it say? Said 250.

137
00:09:32,520 --> 00:09:36,000
Oh, that's not bad. And then it jumped to 1,100. That's a lot more.

138
00:09:36,000 --> 00:09:40,080
And now we're settling in around the same. Now we know how much power it consumes.

139
00:09:40,080 --> 00:09:43,080
Now let's see how much power it can make. Oh, wow.

140
00:09:43,080 --> 00:09:46,520
We're already at 13 below. 17, 20.

141
00:09:46,520 --> 00:09:50,880
I guess we can turn on the computer, hey? Charles actually brought up an overclocking buddy of his.

142
00:09:50,880 --> 00:09:55,840
They're a member of the same team. This is Max, it's actually his memory that we're borrowing.

143
00:09:55,840 --> 00:10:00,240
And board. So do you wanna tell us what you're changing? We're going to change the performance preferences

144
00:10:00,240 --> 00:10:04,040
from the Intel default settings to the ASUS advanced OC profile.

145
00:10:04,040 --> 00:10:07,280
Well, multi-core enhancement, we are going to enable it and remove all the limits.

146
00:10:07,280 --> 00:10:10,400
We're going to set manual for AI overclock tuner.

147
00:10:10,400 --> 00:10:15,560
NPU boost level three, DRAM timing control. The Apex has a lot of nice, useful profiles.

148
00:10:15,560 --> 00:10:19,240
You have 9,600 gear four for the 24 gig sticks.

149
00:10:19,240 --> 00:10:23,240
Oh, cool. 9,000 for the 16 gig sticks and the 24 gig sticks.

150
00:10:23,240 --> 00:10:26,440
So we're gonna run 9,000 on the 24 gig sticks and just leave it at that.

151
00:10:26,440 --> 00:10:31,240
So you don't even have to tune any of the sub timings or any of that crap, they just did all the work for us.

152
00:10:31,240 --> 00:10:36,240
Yep, make it really easy. Now, is that something that you would find on anything other than like an Apex,

153
00:10:36,240 --> 00:10:39,880
like a top tier board? I think the lower end boards do have some features.

154
00:10:39,880 --> 00:10:43,240
The Strix does have that. The Strix has that as well, very good. We aren't as aggressive.

155
00:10:43,240 --> 00:10:47,600
I don't think you're going to be running 9,000 on a four dim stick. Well, the cache is 49.

156
00:10:47,600 --> 00:10:51,720
49? Seems high, but. If you want to go 48, that's fine.

157
00:10:51,720 --> 00:10:55,800
All right, we'll go 48 for the start. We'll take it to 48. Hacking the main frame.

158
00:10:55,800 --> 00:11:00,400
So the D2D ratio is the interconnect. There is a chiplet design.

159
00:11:00,400 --> 00:11:03,680
So you have your SOC on here, your memory controller

160
00:11:03,680 --> 00:11:07,960
is I think on there too. And if we cared about onboard GPU performance, for instance,

161
00:11:07,960 --> 00:11:11,720
maybe we'd care about that or? Yeah, it increases performance across the board.

162
00:11:11,720 --> 00:11:16,240
Memory band with really high speeds, with like no NPU boost on this.

163
00:11:16,240 --> 00:11:19,720
That's right, the IMC is on a separate chiplet, isn't it?

164
00:11:19,720 --> 00:11:24,000
Latency is kind of an issue. Intel, okay, you'll get them next time, boys.

165
00:11:24,000 --> 00:11:27,640
It's a very strange design. It's effectively a monolithic CPU

166
00:11:27,640 --> 00:11:32,320
that is a chiplet on a single die. So we're going to change our V-Core to 1.35 volts.

167
00:11:32,320 --> 00:11:36,160
That's all we need, huh? You tend to not need as much voltage as you would

168
00:11:36,160 --> 00:11:40,000
going to the colder you go, but it really is dependent on the chips. There's a leakage factor,

169
00:11:40,000 --> 00:11:43,520
which is how much power these chips draw. And typically a higher leakage chip

170
00:11:43,520 --> 00:11:49,520
will require less voltage and will scale better with temperature. Ironically, the leaky chips are not necessarily

171
00:11:49,520 --> 00:11:54,360
the ones that Intel wants for the chips they're going to ship to consumers.

172
00:11:54,360 --> 00:11:58,440
They tend to not be as efficient when you're operating them normally.

173
00:11:58,440 --> 00:12:03,440
And then we're going to change it to from regulation mode, which is using the DLVR on the CPU.

174
00:12:03,440 --> 00:12:06,440
So the VRM will output a higher voltage to the CPU

175
00:12:06,440 --> 00:12:10,000
and the CPU steps it down. We're going to bypass that with power gate mode.

176
00:12:10,000 --> 00:12:14,360
Got it. So we're just going to let our board do the work because we know that we have an outstanding board

177
00:12:14,360 --> 00:12:19,080
that's going to deliver nice stable power. Something we'd never recommend that you actually do

178
00:12:19,080 --> 00:12:23,400
for real world performance, but if you're just trying to chase the highest possible number.

179
00:12:23,400 --> 00:12:28,560
It is very fun. Once we're done running PCMark 7, it'll have the screen up with the result.

180
00:12:28,560 --> 00:12:32,760
We're going to open up four copies of the CPU ID. We're going to set up four boxes.

181
00:12:32,760 --> 00:12:36,960
We're going to set up the CPU, the amount of the board, the memory tabs.

182
00:12:36,960 --> 00:12:41,440
And then we're going to do a screenshot. And then we're going to have a world record for you.

183
00:12:41,440 --> 00:12:44,520
That easy? That easy. You got to be kidding me.

184
00:12:44,520 --> 00:12:48,680
Well, we got to save the result. All right, here we go.

185
00:12:48,680 --> 00:12:52,520
PCMark 7, let's go boys. Wait, before we actually do the run,

186
00:12:52,520 --> 00:12:57,080
I mean, can we just put it under load and see how stable the temperature is?

187
00:12:57,080 --> 00:13:00,120
Okay, okay. W prime? Yep.

188
00:13:00,120 --> 00:13:03,880
Okay. Are we at 64.4? Yeah, okay. It's a very popular older benchmark.

189
00:13:03,880 --> 00:13:07,720
Wait. We got down. Did that run? Yeah.

190
00:13:07,720 --> 00:13:11,400
It didn't even move. That is crazy.

191
00:13:11,400 --> 00:13:14,960
It's gone up a degree and a half. And so my expansion valve will kick in

192
00:13:14,960 --> 00:13:19,800
and it'll compensate for the load. Balance out now. It is so cool.

193
00:13:19,800 --> 00:13:22,820
And we're going the other way now in temperature. Look at this.

194
00:13:22,820 --> 00:13:27,320
Yeah, he's right. That's the wrong way. He's right. We're going back down.

195
00:13:27,320 --> 00:13:32,320
We went from negative 104.4 to like 0.5, 0.6,

196
00:13:33,440 --> 00:13:38,560
back to 0.5 still. It'll fluctuate a little bit. We're within like 0.3.

197
00:13:38,560 --> 00:13:45,400
0.5, yeah. But with the e-cores on and running around 5.8 or 5.7,

198
00:13:45,400 --> 00:13:49,840
we might have like a two degree fluctuation. And that's with a 285K.

199
00:13:49,840 --> 00:13:53,160
With the Raptor Lake, we might have like a four degree. Right.

200
00:13:53,160 --> 00:13:57,840
He's on the Intel team, but you know, if we're being honest, she was a hungry boy.

201
00:13:57,840 --> 00:14:02,280
So we're just finishing up and we're at our coldest temperature yet. So we're pretty damn stable

202
00:14:02,280 --> 00:14:06,280
if we're able to run like a stability test right now. So this is a thermal expansion valve right here.

203
00:14:06,280 --> 00:14:11,120
And this is a expansion valve that responds under load. So now it'll go back down in temperature again

204
00:14:11,120 --> 00:14:15,600
as we have no load. Right, okay. But then it'll go, oh, hold on a second.

205
00:14:15,600 --> 00:14:18,840
We're good. I don't need all of this extra coolant

206
00:14:18,840 --> 00:14:22,480
and it'll back it off. Yeah. Cause there's nothing you can do

207
00:14:22,480 --> 00:14:28,000
about the temperature of the refrigerant, right? It's within reason, right?

208
00:14:28,000 --> 00:14:31,040
You're not going to run your compressors at like a variable load.

209
00:14:31,040 --> 00:14:34,920
This is not a sophisticated, you know, modern heat pump or air conditioning system.

210
00:14:34,920 --> 00:14:38,080
These are running at a fixed speed.

211
00:14:38,080 --> 00:14:43,800
So what you would do then to achieve this kind of temperature control is you're adjusting the flow rate

212
00:14:43,800 --> 00:14:47,160
through the evaporator head. So it's the, my understanding is the temperature

213
00:14:47,160 --> 00:14:50,840
of the refrigerant should be basically the same ish. Correct.

214
00:14:50,840 --> 00:14:54,800
But we just juice it way more when we need more cooling.

215
00:14:54,800 --> 00:14:58,680
Right, it opens up the valve more. Cool. Are we ready to do the run

216
00:14:58,680 --> 00:15:01,720
or did you guys need to go double check something? No, no, we're all good to go.

217
00:15:01,720 --> 00:15:06,440
We just got to press the go button. Yeah, you can really see how much he runs benchmarks.

218
00:15:06,440 --> 00:15:11,000
Is it the current world? Yep, that is. Do you ever run actual video games?

219
00:15:11,000 --> 00:15:14,840
Yeah. Oh, okay. I mean, some people get, they get so into the benchmarking

220
00:15:14,840 --> 00:15:19,440
they don't actually game. What's your game right now? Well, I just finished up battlefield four is Riker.

221
00:15:19,440 --> 00:15:23,240
Riker, sorry. He's still working his way through his backlog. Yeah.

222
00:15:23,240 --> 00:15:27,120
Okay, so we're ready. Okay, so we're gonna go. Click the go button. All right, I click go.

223
00:15:27,120 --> 00:15:31,360
I mean, we're all on the same page that just because I click go doesn't mean I set the record.

224
00:15:31,360 --> 00:15:35,000
Like this is- No, no, no, you click go, you set the world record. Yeah, I really don't think that's how it works.

225
00:15:35,000 --> 00:15:39,040
It's how it works. I really- I set the rules, that's how it works.

226
00:15:39,040 --> 00:15:44,040
How long does this take to run? So this is 15, 20 minutes. But actually what's hard about this for the Ellen too guys

227
00:15:44,040 --> 00:15:48,160
is the duration of this benchmark and pouring and keeping this thing under a consistent load.

228
00:15:48,160 --> 00:15:52,240
As to where the cascade can keep this under a very consistent load for a very long duration.

229
00:15:52,240 --> 00:15:55,400
In case you have to rerun it and rerun it again and rerun it again.

230
00:15:55,400 --> 00:16:00,620
Right. The cascade, you can just rerun this thing all day long and have it consistent runs every single time.

231
00:16:00,620 --> 00:16:04,360
So it's gonna be extremely hard to catch your score. Right, got it.

232
00:16:04,360 --> 00:16:07,760
It's gonna be extremely hard. My score. You're gonna be on the world record rankings

233
00:16:07,760 --> 00:16:11,080
for a long time and let the other guys chase you down.

234
00:16:11,080 --> 00:16:16,800
Good luck. Oh, we didn't get it on the first run.

235
00:16:16,800 --> 00:16:19,840
That's his fault. Can you stop assigning blame to everyone around here?

236
00:16:19,840 --> 00:16:23,640
Hey, he told me what settings to put in. He did do that. But you had the RAM set up though.

237
00:16:23,640 --> 00:16:27,080
Yeah. That's weird, that's not memory. Yeah, because it meant that that could be memory.

238
00:16:27,080 --> 00:16:30,280
That is CPU, I know for a fact it is. That wasn't CPU, we'll need a minute.

239
00:16:30,280 --> 00:16:33,840
Now you guys tuned one more thing. This is our third run.

240
00:16:33,840 --> 00:16:37,000
What did you guys change? The NPU boost, I dropped it down.

241
00:16:37,000 --> 00:16:40,560
But we're still at 9,000 mega transfers on the memory. I think about there.

242
00:16:40,560 --> 00:16:45,040
It's gotta be 15,017. Oh.

243
00:16:45,040 --> 00:16:48,680
So it'll take two runs. So close.

244
00:16:48,680 --> 00:16:51,800
We're so close. We got this.

245
00:16:51,800 --> 00:16:56,240
Okay, we got this, we got this. Graphics 315, we just need to bring up our GPU

246
00:16:56,240 --> 00:17:00,320
just a little bit. Oh yeah, we could do that. Just bring up our GPU a little bit and we're there.

247
00:17:00,320 --> 00:17:05,160
And now we play the waiting game. Here we are, run number two.

248
00:17:05,160 --> 00:17:09,520
Well, three, but who's counting? Oh no. Oh, it was data file corruption.

249
00:17:09,520 --> 00:17:13,520
Yeah. So we made it this far and then it's upset

250
00:17:13,520 --> 00:17:17,280
that it doesn't know the CPU name. Third time's a charm.

251
00:17:17,280 --> 00:17:21,240
This is a third run, right? Sure. I mean, it's movie magic. It can be whatever run you want it to be.

252
00:17:21,240 --> 00:17:25,520
If you edit it, it can be the first. All right. Oh, it went down.

253
00:17:25,520 --> 00:17:29,480
How'd we go down? You're on frame for a second to go up. You didn't put any GPU clock on it, man.

254
00:17:29,480 --> 00:17:32,280
Oh, I rebooted. One moment, please.

255
00:17:33,240 --> 00:17:36,480
Time to get the results from our second attempt.

256
00:17:36,480 --> 00:17:39,960
Movie magic, right? The same error again?

257
00:17:39,960 --> 00:17:43,000
Get CPU name. Big moment.

258
00:17:43,000 --> 00:17:43,960
First run.

259
00:17:46,760 --> 00:17:50,320
Next plan. We're gonna try and adjust the cache speed a little bit.

260
00:17:50,320 --> 00:17:54,560
Maybe try to tune the CPU speed up a little bit. We got second place again, but not quite.

261
00:17:54,560 --> 00:17:59,920
First place. What do you think's up? Well, if we look at the GPU Z tab here,

262
00:17:59,920 --> 00:18:04,680
we have our 59, is it 6X? It says it's 16X, right, gen five.

263
00:18:04,680 --> 00:18:08,600
Running it. But it's actually 8X. The SSD is in the wrong slot.

264
00:18:09,120 --> 00:18:13,880
Had it in the correct slot earlier, I don't know if somebody moved it or not. It got pulled and put back in at one point.

265
00:18:13,880 --> 00:18:19,120
Linus isn't here, so we'll blame him. Yes, we'll blame him. I'm pretty sure we have it on video that it was him.

266
00:18:19,120 --> 00:18:22,920
Okay, then let's move that over and let's try again.

267
00:18:22,920 --> 00:18:28,640
Whoa, whoa, whoa, whoa, whoa. Whoa, whoa, you guys blamed me for putting in the SSD.

268
00:18:28,640 --> 00:18:33,880
I didn't put in the SSD. Yeah, because you moved the 5090. I actually know who put it in.

269
00:18:33,880 --> 00:18:37,840
Oh, who? Unlike you guys.

270
00:18:37,840 --> 00:18:42,400
It was not me. I'm not a filthy rat, so I would never rat.

271
00:18:42,400 --> 00:18:45,960
Wait, do you have footage of who put it in? No, I don't, maybe.

272
00:18:45,960 --> 00:18:49,800
Editor, roll the clip if there is one. Oh, oh my God.

273
00:18:49,800 --> 00:18:54,880
Seriously? Well, fortunately it's only our first run.

274
00:18:54,880 --> 00:18:58,600
In case you're curious about the finer details, we just the CPU clocks a little bit,

275
00:18:58,600 --> 00:19:02,160
just the cache, and this is gonna be it.

276
00:19:02,160 --> 00:19:05,920
Ninth run. First try. Yeah, first try.

277
00:19:05,920 --> 00:19:09,600
Man, it's gonna be cool when we nail this. We got this. On the first try.

278
00:19:09,600 --> 00:19:13,120
Oh my God, it's so close. It's so close. Thousand watt BIOS?

279
00:19:14,080 --> 00:19:18,680
Ah? I think so. I think that would do it. On the next first try, we're gonna nail it.

280
00:19:18,680 --> 00:19:22,000
On the next first try. Yay, there it is.

281
00:19:22,000 --> 00:19:25,240
Nice, first place. Good job. Taiwan number one.

282
00:19:25,240 --> 00:19:29,560
It's only first try. You got a world record. The bad news is, this took all day,

283
00:19:29,560 --> 00:19:34,480
so I guess we're not gonna get to 3DMark today, but hey, this is staying here,

284
00:19:34,480 --> 00:19:40,120
so you can expect to see plenty of shenanigans make sure you absolutely pound that subscribe button.

285
00:19:40,120 --> 00:19:43,880
Like I'm talking a hole in your TV or monitor

286
00:19:43,880 --> 00:19:46,960
so you don't miss any of the future sub-zero updates

287
00:19:46,960 --> 00:19:50,240
with Charles coming up here in the future to build a single phase,

288
00:19:50,240 --> 00:19:54,360
and also just, I don't know, man, Elijah, we're gonna have to play with this thing like lots.

289
00:19:54,360 --> 00:19:58,320
Sounds good to me. Just like I'm gonna play with our sponsor.

290
00:19:59,320 --> 00:20:04,760
It's weird, right? Yeah. If you guys enjoyed this video, maybe go check out part one

291
00:20:04,800 --> 00:20:08,480
where we unboxed what I am calling the Beast, which it turns out,

292
00:20:08,480 --> 00:20:11,760
isn't that exactly what you call it? Yes. Cool.

293
00:20:11,760 --> 00:20:15,320
We both named it the same thing. Ironically, yeah.

294
00:20:15,320 --> 00:20:16,720
Good job on the world record.