1
00:00:00,160 --> 00:00:07,759
This is the ROG RTX5090 Astral, one of,

2
00:00:04,560 --> 00:00:10,880
if not the fastest graphics cards on the

3
00:00:07,759 --> 00:00:13,360
planet. You could fit so much FPS in

4
00:00:10,880 --> 00:00:18,000
this bad girl. But even gaming on topofthe-line hardware like this, and

5
00:00:16,480 --> 00:00:22,800
ignoring that you basically need to be the Sultan of Brunai to get one, feels

6
00:00:20,560 --> 00:00:27,199
like every new game that comes out has terrible performance. So to get the

7
00:00:25,039 --> 00:00:32,800
frames we want, we had to find another way. Hi. Oh jeez, I didn't see you

8
00:00:29,679 --> 00:00:35,360
there. Just kidding. You're huge. For

9
00:00:32,800 --> 00:00:39,120
the uninitiated, Splave holds the number one all-time spot on the hardwarebot

10
00:00:37,600 --> 00:00:43,200
overclocking leaderboard, holds countless overclocking world records,

11
00:00:40,960 --> 00:00:47,520
and build some of the world's fastest computers over at Splave PC. And he's

12
00:00:45,600 --> 00:00:52,640
going to show us how to take this pile of parts and make it even faster than it

13
00:00:50,480 --> 00:00:57,120
would otherwise be pretty much step by step. So, if you happen to have an

14
00:00:54,640 --> 00:01:00,640
$8,000 computer at home, you can follow right

15
00:00:58,039 --> 00:01:04,400
along. But seriously, you might want to because the same principles we're using

16
00:01:02,480 --> 00:01:09,119
today should apply to pretty much any modern build. And when we're done, not

17
00:01:06,720 --> 00:01:14,320
only will our machine be faster, it will also probably draw less power. I mean,

18
00:01:13,040 --> 00:01:20,799
probably, right? Like, do we know that for sure? 890 W. Holy. In case it

19
00:01:19,040 --> 00:01:25,759
doesn't though, our friends at CSIC sponsored this video and sent over their

20
00:01:22,960 --> 00:01:32,320
very best power supply unit, the Prime Titanium 1600 watt, which literally can

21
00:01:28,880 --> 00:01:36,720
handle two RTX 5090s. Looks so small

22
00:01:32,320 --> 00:01:40,079
when he holds it. That's what she said.

23
00:01:36,720 --> 00:01:40,079
It's going to be that kind of video. All

24
00:01:46,360 --> 00:01:51,600
right, we have kind of an order of operations for our builds, but I'm going

25
00:01:50,320 --> 00:01:54,720
to play follow the leader today. So, what do you want to start with? Uh, I

26
00:01:53,439 --> 00:02:01,759
usually start with the motherboard. Okay, everyone has the same order of operations. The ASRock X870E Taichi was

27
00:01:59,840 --> 00:02:06,640
your choice. Can you walk us through why this is your preferred motherboard for

28
00:02:03,520 --> 00:02:08,000
tuning? The BIOS is very in-depth and

29
00:02:06,640 --> 00:02:11,039
there's lots of options that we can play with. I couldn't help noticing none of

30
00:02:09,759 --> 00:02:16,800
what you said really over overlapped with the main selling points that motherboard manufacturers focus on.

31
00:02:14,480 --> 00:02:22,560
Like, oh, we have lightning gaming USB and a fancy thing for releasing your

32
00:02:18,640 --> 00:02:24,239
PCIe cards. You're just more of a nuts

33
00:02:22,560 --> 00:02:28,959
and bolt kind of guy. I do like this thing. Oh, that's a really quick M.2

34
00:02:26,720 --> 00:02:32,480
heat sink release. Yeah. Other than that, no. I like the overclocking stuff.

35
00:02:30,959 --> 00:02:35,200
So really, you chose this motherboard the same way that other influencers

36
00:02:34,000 --> 00:02:40,239
choose their motherboards, the most expensive one. Man knows what he likes.

37
00:02:37,599 --> 00:02:45,840
Let's explain our choice of CPU. We went with the 9800 X3D because they're all 3D

38
00:02:43,920 --> 00:02:49,680
cores and there's no games we have to play with clocking. The 3D core is

39
00:02:48,239 --> 00:02:55,440
different than the regular cores. Just to be clear, when he says there's no games we'll have to play to do that, he

40
00:02:52,560 --> 00:02:59,440
doesn't mean video games that don't require that thing. Yeah, we don't have

41
00:02:57,680 --> 00:03:02,480
to fool around with it. I forgot we're installing a contact frame. And I could

42
00:03:01,280 --> 00:03:08,959
explain why we're doing that, but I'll let Alan do it. So, I like the contact frame because we've seen better memory

43
00:03:06,879 --> 00:03:12,080
overclocking with using the contact frame over the stock one. Hey, has

44
00:03:10,800 --> 00:03:16,239
anyone actually set you up with one of these yet? No. Seriously, you came out

45
00:03:14,560 --> 00:03:20,400
and collabed with us before and we never got you a screwdriver? Look at the

46
00:03:17,840 --> 00:03:24,879
backpack on the floor, silly. You can't you can't give this man the commuter

47
00:03:22,400 --> 00:03:28,720
bag. Like, look at this. He's like a little uh school girl. Yeah. What the

48
00:03:27,360 --> 00:03:32,480
what the We got to get him the fulls size bag, Jake. Coming back to memory, I

49
00:03:30,959 --> 00:03:35,799
was a little surprised to see that you really don't with higher speed memory.

50
00:03:34,319 --> 00:03:42,239
This is still DDR56000. It just runs at extremely

51
00:03:39,120 --> 00:03:43,680
tight latencies. So, is this the ideal?

52
00:03:42,239 --> 00:03:48,879
Yeah, because you want to keep the memory controller in like a 1 one sync

53
00:03:46,080 --> 00:03:54,239
for latency on AMD currently. And that's around maybe 6,800 on a really, really

54
00:03:51,599 --> 00:03:58,000
good sample. And a highly binned kit like this, realistically you can play

55
00:03:56,319 --> 00:04:02,480
around with it a little bit. If you want a little bit higher latencies and higher

56
00:04:00,319 --> 00:04:07,280
clock speeds, you can do that because these are just golden dies essentially.

57
00:04:05,200 --> 00:04:13,200
Now, Techbot did a piece recently and they found that DDR5 8000 could actually

58
00:04:10,400 --> 00:04:16,880
outperform DDR5 6000. And that's in spite of the fact that the memory

59
00:04:14,720 --> 00:04:21,280
controller runs at half the speed once you go above 6,000. But with Allen's

60
00:04:19,120 --> 00:04:25,120
help, we should be able to take that memory controller and boost it above

61
00:04:23,360 --> 00:04:30,240
6,000, getting overall better performance. So, you're thinking sweet

62
00:04:26,479 --> 00:04:32,639
spot 6400 CL28. Yeah, around there.

63
00:04:30,240 --> 00:04:37,160
Let's put in an SSD. We chose this because it holds data. It's a Rocket 5.

64
00:04:34,800 --> 00:04:42,400
Those are fast. Yeah, it's fast. It's a rocket. Look at it go. Oh god. For our

65
00:04:40,080 --> 00:04:47,440
case, we went for the fractal torrent, which is built for torrents of air

66
00:04:45,120 --> 00:04:51,919
running through it. Okay, there's all your fresh air for the GPU right there.

67
00:04:49,600 --> 00:04:56,800
Honeycomb mesh at the back, top-mounted power supply, and nothing but airflow at

68
00:04:54,960 --> 00:05:02,160
the front. I'm just disappointed we didn't get to use one of the coolest

69
00:04:58,240 --> 00:05:03,440
features of this case, these 180x 38 mm

70
00:05:02,160 --> 00:05:08,479
fans that would normally be in the front, but we're going to have an AIO

71
00:05:05,520 --> 00:05:14,160
liquid cooler. So, they go over there. For our good enough AIO cooler, we've

72
00:05:11,280 --> 00:05:20,320
gone with the Lumen S36 RGB. And we chose it because Fractal's nice. Fractal

73
00:05:17,919 --> 00:05:24,240
is nice. All right, Jake knows why we chose it. It's because it is a

74
00:05:22,080 --> 00:05:28,560
nonasitech design, so you're not bound to their sort of patented how everyone

75
00:05:26,560 --> 00:05:32,320
else does it design. The pump is actually embedded in the radiator, which

76
00:05:30,479 --> 00:05:36,800
detracts from cooling less than you would probably think because that's

77
00:05:34,080 --> 00:05:40,560
right where the fan hub is, right? You weren't getting any air flow there

78
00:05:38,080 --> 00:05:44,639
anyway. It also looks nice, just like Fractal is. Oh, and we scratched it.

79
00:05:43,199 --> 00:05:50,759
Let's have you talk about your thermal paste while I mount this stupid radiator. We're using Splave's own

80
00:05:48,479 --> 00:05:57,199
thermal paste. Splave one is my thermal paste and it's very metallic, so it can

81
00:05:54,240 --> 00:06:02,080
be difficult to spread. You can either heat it up a little bit with like a

82
00:05:58,800 --> 00:06:03,600
haird dryer or you can just He-Man it,

83
00:06:02,080 --> 00:06:07,759
push really hard with your thumb on the spatula and get a nice thin layer.

84
00:06:05,919 --> 00:06:11,840
Right. Yeah. See, for some people that's an option. I usually use that option. I

85
00:06:10,080 --> 00:06:16,880
think your forearms are probably about twice as big around as my biceps. For my

86
00:06:15,039 --> 00:06:20,800
effort, how much performance benefit do I get? You're going to see significant

87
00:06:19,039 --> 00:06:27,039
improvement in temperature, which will lead, if you use, say, PBO

88
00:06:23,919 --> 00:06:28,639
um on AMD, you'll gain performance just

89
00:06:27,039 --> 00:06:34,880
from using better thermal paste. But we're not using that today, are we?

90
00:06:31,360 --> 00:06:37,919
We're using this, a thermal paste that

91
00:06:34,880 --> 00:06:39,919
apparently costs tens of dollars per

92
00:06:37,919 --> 00:06:44,400
gram. Correct. And what the hell is this? So, this is

93
00:06:42,240 --> 00:06:52,319
our hybrid paste. It looks like normal thermal paste. Sure does, but we have a

94
00:06:47,600 --> 00:06:54,880
special blend of gallium and other

95
00:06:52,319 --> 00:06:58,080
ingredients that I will not disclose. Sounds like you're trying to sell me

96
00:06:55,759 --> 00:07:02,000
fried chicken. We have our special blend of spices. There's salt and some other

97
00:07:00,639 --> 00:07:06,000
we're not going to tell you. So, basically, it has liquid metal in it.

98
00:07:04,639 --> 00:07:11,440
And we're trying to give you the benefits of liquid metal without the

99
00:07:08,319 --> 00:07:12,960
negatives of say shooting liquid metal

100
00:07:11,440 --> 00:07:16,560
across your motherboard and killing it. It being impossible to clean up. Yes.

101
00:07:14,720 --> 00:07:20,240
Liquid metal drying out and leaving you with no cooling. We have a backup layer

102
00:07:18,560 --> 00:07:23,280
of regular thermal paste. So, it's kind of giving you the best of both worlds.

103
00:07:21,840 --> 00:07:27,039
Let's see how challenging this one is to spread. I'll try not to use too much.

104
00:07:24,880 --> 00:07:31,120
Okay. It's expensive. We're going to really target where the compute dies are

105
00:07:29,840 --> 00:07:36,080
here. Put a little bit up there for the IO die. There you go. Oh. Oh, wow. That

106
00:07:33,919 --> 00:07:40,319
spread super easy. This is the part where we break the traditional order a

107
00:07:37,840 --> 00:07:44,880
little bit. I want to put this GPU in. But why don't you explain why we chose

108
00:07:42,400 --> 00:07:48,720
the ROG Astral other than because you think it's funny that you can lift it

109
00:07:46,639 --> 00:07:51,440
and I can't. I got I got to put this down. Can you hold this? Thank you.

110
00:07:50,000 --> 00:07:57,599
Obviously, just from your what you were just saying is is beefcake of a cooler

111
00:07:54,479 --> 00:08:00,080
here. Asus is all about overclocking.

112
00:07:57,599 --> 00:08:04,240
So, they support with GPU tweak software and stuff. They make it real easy for us

113
00:08:02,240 --> 00:08:07,199
to get all the performance out of these cards. We have a bit of an

114
00:08:05,360 --> 00:08:10,960
unconventional cooling setup for this build. Fresh intake right here, all for

115
00:08:09,280 --> 00:08:15,360
the GPU. And we're pulling fresh air into this radiator for the CPU. And if

116
00:08:13,120 --> 00:08:21,120
our RAM gets a little bit warmer, then the PSU suck all the hot air out.

117
00:08:18,280 --> 00:08:25,440
Sure. Hey, speaking of power supply, now is a great time to talk about our

118
00:08:22,800 --> 00:08:29,120
sponsor for this build, CIC. Naturally, a build of this caliber deserves the

119
00:08:27,199 --> 00:08:33,760
very best from CSI's lineup, and that's why we went with the

120
00:08:30,680 --> 00:08:36,479
TX600 watt prime unit. This thing is 80

121
00:08:33,760 --> 00:08:41,519
plus titanium efficient. It's ATX 3.1, meaning it has the newer 12volt 2x6

122
00:08:39,039 --> 00:08:46,720
power connector. In fact, it has two of them in case you're building an AI PC or

123
00:08:44,159 --> 00:08:52,800
something like that. And as always, it comes with CIC's acclaimed lengthy

124
00:08:49,680 --> 00:08:54,800
warranty. It reaches 94% peak efficiency

125
00:08:52,800 --> 00:08:59,200
according to Cybernetics and has very low ripple. It basically doesn't get any

126
00:08:57,440 --> 00:09:02,480
better than this except for the version of this that has like a better fan in it

127
00:09:00,959 --> 00:09:08,000
from Noctua, but we don't need that today because apparently this is not going to be a quiet build. No, we're not

128
00:09:05,920 --> 00:09:11,600
worried about that right now. Neat. All right, let's get a monitor. Something

129
00:09:09,680 --> 00:09:15,200
for you to note is we are going to be running at 1080p for all of our testing

130
00:09:13,360 --> 00:09:19,760
today because when you are extreme overclocking or whatever, you are going

131
00:09:17,440 --> 00:09:23,600
to want to run into situations where you could be CPU bottlenecked so that as you

132
00:09:22,160 --> 00:09:28,640
increase the speed of your CPU, you can really see the differences and measure them. Oh my god, I'm dead again. Oh, I

133
00:09:27,040 --> 00:09:33,680
missed. How did I miss that? He's right there. Come on, buddy. So, we're getting

134
00:09:30,640 --> 00:09:35,200
like anywhere from 450 to 600 FPS. Once

135
00:09:33,680 --> 00:09:40,720
you get into really high frame rates, things start to get a little variable.

136
00:09:38,480 --> 00:09:44,080
So, you can beat that, right? Yeah, let's do it. Okay. Why don't you walk us

137
00:09:42,320 --> 00:09:50,240
through what you're doing? Sure. Uh, right now, just going to raise the CPU

138
00:09:46,800 --> 00:09:52,800
clocks. Fixed clock. No boost. No boost.

139
00:09:50,240 --> 00:09:56,640
No boost. Oh. So, well, that's not the meta I was quite expecting, but sure.

140
00:09:54,480 --> 00:10:00,800
We'll just be conservative at first and then sure. A conservative 5.4 GHz. We

141
00:09:59,200 --> 00:10:06,399
also increased the voltage a little bit. That made more of a difference than I

142
00:10:03,120 --> 00:10:10,000
was necessarily expecting.

143
00:10:06,399 --> 00:10:12,399
Maybe 6 to 700 now. Damn, we're getting

144
00:10:10,000 --> 00:10:16,079
6 to 700 FPS. And you haven't really even pushed it yet. No, we could go

145
00:10:14,000 --> 00:10:19,920
further. Okay, well, we should do that. If all this seems too easy for us to

146
00:10:18,079 --> 00:10:22,800
bother flying a professional out here, that's because everything we've been

147
00:10:21,120 --> 00:10:29,279
doing so far is the easy stuff that you don't really have to know anything to do. Accurate. I'm trying to say that the

148
00:10:27,519 --> 00:10:32,800
stuff that he knows is stuff we're going to get to in a minute. Like, I could

149
00:10:30,720 --> 00:10:39,040
have done that. Can you give me the dumb dumb version of what exactly it is you

150
00:10:34,640 --> 00:10:41,079
just changed? So, memory timings.

151
00:10:39,040 --> 00:10:46,320
Okay. Less dumb. Basically, lower number good for the

152
00:10:44,240 --> 00:10:50,560
most part. Mhm. Except for when it's not. But some of them we tune a little

153
00:10:48,160 --> 00:10:54,959
bit. Yeah. And some of them we're literally setting to less than half.

154
00:10:53,040 --> 00:11:00,399
Biggest one probably is refresh interval. Almost always the max value

155
00:10:58,000 --> 00:11:04,079
here is is going to be stable and okay. it will raise the temperature of your

156
00:11:02,079 --> 00:11:06,959
memory which can kind of show you too that it's working harder. And for the

157
00:11:05,440 --> 00:11:11,240
people out there, you know, who wanted to kind of learn this, how do they learn

158
00:11:10,079 --> 00:11:16,720
that 403 is an optimal

159
00:11:14,120 --> 00:11:25,600
TRFCSB refresh cycle time? Some literature will tell you your TRA should

160
00:11:20,720 --> 00:11:29,680
be like twice your TRCD divided by the

161
00:11:25,600 --> 00:11:32,560
square root of pi * 4, right? But if you

162
00:11:29,680 --> 00:11:38,279
just test it yourself and and lower it and see a gain in a benchmark or a game,

163
00:11:35,200 --> 00:11:42,480
then that is a good value for you,

164
00:11:38,279 --> 00:11:44,880
right? So, literature be damned. Try it

165
00:11:42,480 --> 00:11:50,000
and if it it works for you, then then go with that. What are the odds if someone

166
00:11:47,120 --> 00:11:54,720
had a 9800 X3D and they bought a really nice kit of memory that they could just

167
00:11:52,480 --> 00:11:59,839
plop these numbers in? These numbers in uh pretty high. Okay. Yeah. If you had

168
00:11:57,200 --> 00:12:05,279
the good memory, decent IMC. Oh, here we go. We're

169
00:12:01,000 --> 00:12:07,760
in. We got 100 more megahertz. We're

170
00:12:05,279 --> 00:12:11,519
out. Hopefully I don't get backbanned for just like popping into matches and

171
00:12:09,600 --> 00:12:16,279
exiting them after I get killed a bunch. Oh, see, here's the thing. Sometimes you

172
00:12:13,920 --> 00:12:20,639
can have too much of a good thing. That was too much clock speeds. Feel like not

173
00:12:19,200 --> 00:12:26,800
all of this is going to make it into the final cut. So, if you want the extras,

174
00:12:23,639 --> 00:12:29,200
LMG.gg/flowplane. This is a rocket.

175
00:12:26,800 --> 00:12:33,440
Look at it go. Oh god. Oh no. Katy Perry, you told your wife you'd be

176
00:12:30,800 --> 00:12:36,480
leaving by now? If she wanted someone who's punctual, she could have married

177
00:12:35,120 --> 00:12:43,200
someone punctual. There are a lot of religious types in the overclocking community. No. What the kind of question

178
00:12:40,480 --> 00:12:46,680
is that? Are you okay in there, Lauren Sanchez? Ow.

179
00:12:46,959 --> 00:12:52,240
Realistically, I'm sure we could squeeze a few more mehz out of it, but the idea

180
00:12:50,399 --> 00:12:56,240
was that you legitimately would be able to follow along at home. And these are

181
00:12:54,320 --> 00:13:00,240
pretty attainable settings if you have decent hardware. So, let's go ahead and

182
00:12:58,240 --> 00:13:07,200
do our after benchmark. It's a little bit better. How we doing? 780 average

183
00:13:03,680 --> 00:13:09,680
and 300 on the 1% lows. Okay, that's not

184
00:13:07,200 --> 00:13:14,560
that much. I mean, it's better. It's better. Yes. I mean, we're getting into

185
00:13:11,839 --> 00:13:18,600
very high frame rate territory. We might be getting limited by more than just the

186
00:13:15,920 --> 00:13:24,160
clock speed here. Should we do like a overclocking dude benchmark for a second

187
00:13:20,959 --> 00:13:25,680
here? like like a Port Royal. Man, I

188
00:13:24,160 --> 00:13:30,399
remember when 3D Marks benchmarks were like cutting edge. Like you couldn't get a game that had graphics that good. And

189
00:13:28,720 --> 00:13:33,399
now it seems like they uh pretty good graphics.

190
00:13:34,079 --> 00:13:41,680
Did good. We're a little above average for this hardware. We've got a good score. Okay, why don't we see what we

191
00:13:39,279 --> 00:13:45,440
can do with a faster GPU? Just an idea. Some fans going. Okay,

192
00:13:44,160 --> 00:13:50,279
this better go faster otherwise I think you might be out of a job.

193
00:13:48,160 --> 00:13:55,440
Whoa. Okay, hold on a second. Now we're Ted and Billy. Ted and

194
00:13:53,200 --> 00:14:01,760
Bill. We're excellent. I'm calling it here. 900 FPS. Really? I'm going to

195
00:13:59,199 --> 00:14:06,800
guess 778. 795. Okay, that is less more than I was

196
00:14:04,880 --> 00:14:10,480
expecting. What was it like? And our minimums are actually worse. Well, that

197
00:14:08,639 --> 00:14:14,000
might just be margin of error. So, now we're running Stellaris, which is like a

198
00:14:12,240 --> 00:14:18,240
top down kind of real-time strategy game. This benchmark simulates one year

199
00:14:16,480 --> 00:14:24,160
of in-game time. And we have kind of our pre-made map here. Um, it's very much a

200
00:14:20,560 --> 00:14:25,440
CPU benchmark. So, we'll see what it

201
00:14:24,160 --> 00:14:29,880
does. But we should We're getting only 51 FPS. It does. This is a very

202
00:14:27,839 --> 00:14:35,760
intensive game. Hey, that's like 10% improvement. Okay.

203
00:14:33,519 --> 00:14:38,720
Oh, lower number better. Yeah. On this one, this is the time it took to

204
00:14:36,959 --> 00:14:42,800
process. Heck yeah. Okay. So, then they're Well, that's awesome. All you

205
00:14:40,480 --> 00:14:48,240
have to do is get a master's degree in around with RAM timings or buy a PC for

206
00:14:44,959 --> 00:14:50,800
me. And the this guy the sales pitch

207
00:14:48,240 --> 00:14:54,560
always be selling. All right, let's put monkeying around and flash the BIOS.

208
00:14:52,720 --> 00:15:00,279
Okay, enough monkeying. Let's flash this thing. Okay. Woo. Sorry.

209
00:15:03,120 --> 00:15:10,160
This is on the big boy power BIOS. Okay.

210
00:15:07,839 --> 00:15:18,399
So, do Oh, we've got a we've got more slider now. But 120% of what? That is

211
00:15:14,160 --> 00:15:20,959
not as relevant anymore. I see. So the

212
00:15:18,399 --> 00:15:26,160
So is 100% a thousand? Technically, I don't know that. Okay. So I should turn

213
00:15:22,800 --> 00:15:29,279
it all the way up. We can try, but we

214
00:15:26,160 --> 00:15:30,720
can also here. Why don't you draw? Just

215
00:15:29,279 --> 00:15:37,279
give him the give him the remote, please. I can show you kind of in a non-

216
00:15:34,399 --> 00:15:45,920
3D benchmark would be cool. So we can see our amp amperage on the Uh-huh.

217
00:15:41,920 --> 00:15:48,320
Oh, yeah. Bad. Red is bad. Okay. And we

218
00:15:45,920 --> 00:15:54,240
we're all red. Oh, 12 volts times 10 amp. So, we're at like 120 watts per per

219
00:15:52,240 --> 00:15:57,680
pen. Do you want me to go higher or no? If we can go a little higher, I think.

220
00:15:55,920 --> 00:16:03,360
Yeah, we should. We should try. We're doing up to 11 amps now per pin. Should

221
00:16:00,800 --> 00:16:07,040
we try Port Royal? Sure. So, I think this should put us

222
00:16:05,279 --> 00:16:12,320
around a little better than where we were, but we'll go higher after this.

223
00:16:09,279 --> 00:16:13,839
We'll go higher. No fail.

224
00:16:12,320 --> 00:16:19,120
Rats, is your card worse than mine? Realistically, if all you want is power

225
00:16:17,360 --> 00:16:22,480
consumption be damned, the highest clock speed you can get, a lot of this is

226
00:16:20,959 --> 00:16:26,560
going to come less down to your particular skill at entering the numbers

227
00:16:24,399 --> 00:16:30,240
and more down to how lucky you get in the silicon lottery, but only for

228
00:16:28,639 --> 00:16:33,759
certain things. You know, for the RAM, for instance, okay, you're actually

229
00:16:32,480 --> 00:16:40,079
going to need to educate yourself somewhat in terms of what all the timings mean. But that netted us a

230
00:16:37,759 --> 00:16:43,680
couple percent of the performance increase we saw

231
00:16:41,399 --> 00:16:49,360
today. But then a couple percent of a few percent is I guess a

232
00:16:45,720 --> 00:16:51,839
lot. Oh, hamburgers. Hamburgers. Our

233
00:16:49,360 --> 00:16:55,920
power in this building sucks. We've had all kinds of issues overclocking in this

234
00:16:53,680 --> 00:17:00,160
corner in the past. So part of his stability woes may just be that we don't

235
00:16:58,240 --> 00:17:03,360
have a particularly clean power source. And you can have a great PSU in your

236
00:17:01,839 --> 00:17:06,720
computer all you want, but if it has dirty input power, that is going to work

237
00:17:04,959 --> 00:17:11,600
against it. It feels more stable already. Does. Does it actually?

238
00:17:09,360 --> 00:17:16,799
Actually, a little. Yeah. Glitchy before. Yeah. Are you joking? No. No. It

239
00:17:14,880 --> 00:17:22,079
was being a little We're so close. Legendary. We're legendary. Let's go. We

240
00:17:19,439 --> 00:17:28,000
found what's after. Excellent. Let's go, dude. We are at like 3150.

241
00:17:25,760 --> 00:17:33,039
Dude, we're at the point where the performance curve, the bell curve is

242
00:17:30,640 --> 00:17:36,160
basically flat. What's after legendary? I don't There can't be something after

243
00:17:34,559 --> 00:17:41,840
legendary. Okay, you must know, right? Like, is there something after legendary? I really don't. I I never

244
00:17:39,280 --> 00:17:45,360
have even read that it's at legendary. Literally, this man only sees the

245
00:17:43,600 --> 00:17:51,760
number. Dude, we got this. Here we go. Here we go. Here we go. Here we go. Here we go. Here we go. Here we go. Here we

246
00:17:48,400 --> 00:17:54,400
go. Here we go. Here we go. Let's go. 3

247
00:17:51,760 --> 00:17:58,240
2 1. There's like four systems that are above this number. Can I see the

248
00:17:55,840 --> 00:18:03,440
hardware GPU Z thing? What does it say? It was drawing. 890 watts. Holy. You're

249
00:18:01,919 --> 00:18:09,679
telling me these 5090s ain't got no headroom? I feel like you can do more.

250
00:18:06,320 --> 00:18:11,919
Not today. Just need more temps. And we

251
00:18:09,679 --> 00:18:18,720
need to tell you about our sponsor. No, it's Cic. Oh, yeah. CIC, right? CIC.

252
00:18:15,919 --> 00:18:23,600
Good power. I mean, legendary power. Legendary even, you might say. We were

253
00:18:21,120 --> 00:18:26,799
able to drive a freaking 900 watt GPU off of this thing. Like, do you really

254
00:18:25,440 --> 00:18:30,160
need to know more than that? We're going to have CIC linked down below. We're

255
00:18:28,880 --> 00:18:34,080
also going to have Slav's thermal compound down there, as well as his uh

256
00:18:32,080 --> 00:18:37,120
PC system integration business. Thank you. If you guys enjoy these kinds of

257
00:18:35,760 --> 00:18:44,080
shenanigans, though, maybe you'll enjoy the last time that Alex and I tried using a modified BIOS to push a 1000

258
00:18:40,640 --> 00:18:44,080
watts through a GPU.