1
00:00:00,240 --> 00:00:06,879
today's video is all about using the right hardware for the job so we

2
00:00:05,200 --> 00:00:10,480
actually learned a couple of years ago you can check that video up here if you

3
00:00:08,240 --> 00:00:16,560
want that just because you've got a workload that traditionally scales

4
00:00:12,719 --> 00:00:19,760
really well with lots of processor cores

5
00:00:16,560 --> 00:00:22,240
like video encoding doesn't necessarily

6
00:00:19,760 --> 00:00:26,880
mean that it will scale infinitely and we ran into a situation where adding

7
00:00:24,720 --> 00:00:32,160
more processing course to our video encoding server actually resulted in

8
00:00:29,519 --> 00:00:37,360
significantly less performance for spending significantly more money so we

9
00:00:35,600 --> 00:00:42,719
actually found that the sweet spot was somewhere in the 8 to 10 core range with

10
00:00:40,640 --> 00:00:47,039
frequency actually being far more important than core count beyond that

11
00:00:44,879 --> 00:00:53,280
point fast forward a couple of years though and the 6950x 10 core overclocked

12
00:00:51,120 --> 00:00:56,079
extreme edition processor that we equip this thing with

13
00:00:54,640 --> 00:01:01,600
is not only no longer the ruler of the roost

14
00:00:58,960 --> 00:01:07,119
it's actually significantly slower than even consumer-grade a-core chips that

15
00:01:04,640 --> 00:01:10,560
you can buy today for a fraction of the cost

16
00:01:08,159 --> 00:01:16,320
so we're going to pull this bad boy out get it dusted off swap out some hardware

17
00:01:13,439 --> 00:01:18,880
and see just how fast we can make her fly

18
00:01:18,000 --> 00:01:24,840
uh you guys ready for me to pull down the render server

19
00:01:22,640 --> 00:01:30,920
all good sweet thanks guys

20
00:01:28,960 --> 00:01:35,280
speaking of all good lttstore.com is full of all good cool

21
00:01:33,600 --> 00:01:38,360
stuff like this hoodie the house stealth hoodie is finally in stock go check it

22
00:01:37,280 --> 00:01:47,540
out that's lttstore.com we're going to have that linked in the video description

23
00:01:50,479 --> 00:01:55,439
before we go any further i feel like i should explain why we need a render

24
00:01:53,920 --> 00:02:00,079
server in the first place you might think well if you've got a powerful

25
00:01:57,439 --> 00:02:04,719
workstation surely you can cut together your video export it and upload it to

26
00:02:02,320 --> 00:02:08,640
youtube all without sending it out to a separate server for processing and we

27
00:02:07,119 --> 00:02:12,480
could do that but there are some workflow advantages to doing things the

28
00:02:10,479 --> 00:02:17,520
way that we do so whether we're shooting on our reds or on our canon c200s we

29
00:02:15,680 --> 00:02:21,760
take whatever file type that we originally shot we dump it into our

30
00:02:19,680 --> 00:02:26,800
premiere project and we set premiere to use a cineform timeline so cineform is a

31
00:02:24,800 --> 00:02:30,080
mezzanine codec that has great performance on the timeline and it

32
00:02:28,560 --> 00:02:34,959
retains a great deal of the image quality from the original file it also

33
00:02:32,640 --> 00:02:42,800
has some other advantages so compared to codecs that use iframes like h.264

34
00:02:37,920 --> 00:02:45,280
cineform is much easier to kind of uh

35
00:02:42,800 --> 00:02:50,560
chop pieces out of without having to completely re-render your file so

36
00:02:48,160 --> 00:02:54,879
something that we can do is if we set our timeline to cineform render all the

37
00:02:52,720 --> 00:02:59,760
files out in cineform and then make a change like pulling a sponsor spot out

38
00:02:57,200 --> 00:03:04,480
for flow plane we can very quickly export two different versions of that

39
00:03:01,840 --> 00:03:08,720
video without taking up a bunch of time on the editor's workstation allowing

40
00:03:06,400 --> 00:03:12,720
them to get back to work now we could take that finished cineform file and

41
00:03:10,560 --> 00:03:17,519
upload it directly to youtube only problem is it's anywhere from 30 to 60

42
00:03:15,760 --> 00:03:22,879
gigabytes for the average Linus tech tips video which even though our internet here is

43
00:03:20,159 --> 00:03:27,040
really fast still takes a long time to upload to youtube because there are

44
00:03:24,720 --> 00:03:30,959
transfer speed limits on their side so what we need to do then instead is take

45
00:03:29,120 --> 00:03:36,159
our cineform file here you can see we're remoted into that server right now and

46
00:03:33,120 --> 00:03:38,159
convert it to h.264 which will not only

47
00:03:36,159 --> 00:03:43,120
upload faster on youtube it will also process faster and more reliably

48
00:03:41,280 --> 00:03:48,239
here comes oh man that's heavy

49
00:03:45,840 --> 00:03:48,239
okay

50
00:03:49,040 --> 00:03:56,400
oh all right let's see how

51
00:03:53,680 --> 00:04:00,480
brutal this thing is inside ready

52
00:04:00,480 --> 00:04:08,000
wow that's actually not that bad it's almost like

53
00:04:05,360 --> 00:04:10,959
dust filters are a really really good idea

54
00:04:09,360 --> 00:04:15,599
so let's walk through how this thing was configured in the first place we've got

55
00:04:13,200 --> 00:04:20,400
a triple 120 millimeter fan with some of Noctua's industrial ppc fans on it for

56
00:04:18,160 --> 00:04:24,479
cooling it's actually you can't really tell but it's uh it's lifted about this

57
00:04:22,960 --> 00:04:29,440
much off the bottom of the server chassis so that as these fans

58
00:04:27,120 --> 00:04:33,440
draw air through the front of the case these ones suck that cool air up through

59
00:04:31,759 --> 00:04:38,639
the radiator and then the warm air kind of gets blown along this way for our

60
00:04:36,000 --> 00:04:42,639
reservoir we went with a like a metal one because i didn't want any risk of

61
00:04:40,880 --> 00:04:47,680
leaking whatsoever you can see it actually has very little evaporation too

62
00:04:45,280 --> 00:04:50,720
which is really nice a d5 pump because reliability

63
00:04:49,040 --> 00:04:55,600
is absolutely the most important thing in a build like this and raid one drives

64
00:04:53,840 --> 00:05:00,240
uh those aren't that reliable of drives those are 10 year old force 120. it

65
00:04:57,759 --> 00:05:04,000
doesn't matter we've got 32 gigs of RAM this was the stuff we painted actually

66
00:05:01,919 --> 00:05:08,320
for a project quite a long time ago and then i went with a gtx 980. now normally

67
00:05:06,639 --> 00:05:12,400
you wouldn't need like a gaming graphics card in something like this but some of

68
00:05:10,400 --> 00:05:16,240
the workloads that it handles are cuda accelerated and we did get a performance

69
00:05:14,400 --> 00:05:20,320
improvement from that so we chucked it in there so the only things we're really

70
00:05:18,639 --> 00:05:27,520
replacing today are going to be the motherboard the CPU and the memory and

71
00:05:23,039 --> 00:05:30,080
then i also grabbed a an rtx 2080 ti to

72
00:05:27,520 --> 00:05:35,039
see if maybe after we upgrade our CPU there could be some benefit from a

73
00:05:31,520 --> 00:05:36,639
faster GPU so let's get started

74
00:05:35,039 --> 00:05:41,680
so pretty much every aspect of this is getting an upgrade we are dropping down

75
00:05:38,639 --> 00:05:43,520
from 10 gore's to a course but because

76
00:05:41,680 --> 00:05:48,160
they're faster ones i'm still expecting a performance improvement we're going

77
00:05:45,520 --> 00:05:54,080
from 26 66 megahertz memory all the way up to 3200 megahertz memory and that is

78
00:05:50,960 --> 00:05:56,000
with lower timings to boot and we're of

79
00:05:54,080 --> 00:06:01,440
course going from this older x99 motherboard to a new z391

80
00:05:59,360 --> 00:06:04,720
so let's hope that since my last project using this thing i had to ultimately

81
00:06:03,039 --> 00:06:10,400
pull it out of my system because of a Thunderbolt snafu

82
00:06:07,280 --> 00:06:12,319
our inventory team didn't actually

83
00:06:10,400 --> 00:06:16,720
take the water block off the motherboard because that would be a bit of a hassle

84
00:06:13,600 --> 00:06:17,520
for me to reinstall and they did

85
00:06:16,720 --> 00:06:23,199
oh they took the CPU out of it too uh

86
00:06:21,440 --> 00:06:27,919
guys i guess i'm gonna have to put the whole

87
00:06:24,800 --> 00:06:29,919
thing back on again oh

88
00:06:27,919 --> 00:06:34,720
oh they took it apart even more than it comes apart to start with

89
00:06:32,960 --> 00:06:38,080
this is going to take me a bit all right so that's done let's go ahead

90
00:06:36,880 --> 00:06:44,319
and pop this bad boy in here

91
00:06:42,000 --> 00:06:49,039
and oh yeah

92
00:06:46,479 --> 00:06:54,400
this board is wider than standard atx so i'm gonna need to move my ssds

93
00:06:52,240 --> 00:06:57,199
ah well that's some good double-sided tape

94
00:06:59,280 --> 00:07:05,520
i think i bet this drive now let's try and put the board in

95
00:07:03,599 --> 00:07:09,759
all right just need to plug in all the fans

96
00:07:08,080 --> 00:07:14,639
plug in our power and now we need to swap out our original CPU

97
00:07:13,360 --> 00:07:21,520
block for this one that cools the entire board

98
00:07:18,319 --> 00:07:23,840
and CPU and vrm and all that

99
00:07:21,520 --> 00:07:27,280
now i have a crazy idea normally

100
00:07:25,039 --> 00:07:31,520
you would think well Linus the best way to do that is probably to drain the

101
00:07:28,880 --> 00:07:35,680
whole loop connect everything and then refill it but i think we can

102
00:07:33,440 --> 00:07:43,479
skip the draining step as long as we keep the part of the loop that we open

103
00:07:38,400 --> 00:07:43,479
at the highest point just a second

104
00:07:45,680 --> 00:07:52,560
i'm gonna prop it up on my graphics card

105
00:07:50,560 --> 00:07:57,199
i'm just gonna crack this open i can't find my vice grips so this is not really

106
00:07:55,759 --> 00:08:01,840
optimal but ah so everything that's under me is

107
00:07:59,919 --> 00:08:07,280
waterproof i think yeah tubing fittings reservoir

108
00:08:04,080 --> 00:08:08,319
industrial ppc fans

109
00:08:07,280 --> 00:08:12,800
there let's go ahead and undo the other side

110
00:08:11,120 --> 00:08:17,680
yeah not much in there now here's the risky part i need to keep these two

111
00:08:14,720 --> 00:08:22,319
level because if one goes too high the water will go down down down down

112
00:08:20,080 --> 00:08:26,560
down and splurge out the other side oh hi

113
00:08:24,240 --> 00:08:31,440
oh i think it's in now let's see if this one reaches

114
00:08:29,039 --> 00:08:36,240
i should just rename the channel to like low effort water cooling

115
00:08:33,839 --> 00:08:41,159
tips forget it i'm just gonna get a mac pro

116
00:08:43,279 --> 00:08:49,519
no i'm kidding that's not actually a better solution so these are 3 200

117
00:08:47,040 --> 00:08:52,399
megahertz c14 now if Gigabyte's not impressed with

118
00:08:51,200 --> 00:08:58,160
what i'm doing with their high-end gaming components i'm sure g-skill is on

119
00:08:55,519 --> 00:09:02,959
a whole other level here let's throw the GPU back in and then we're good to go

120
00:09:00,160 --> 00:09:06,240
this is really a very horrible mounting mechanism

121
00:09:04,959 --> 00:09:10,640
i'm not even going to try and screw it in this time i don't know why i bothered

122
00:09:08,080 --> 00:09:15,040
okay so all that's left is to fire it up and see if it still

123
00:09:12,720 --> 00:09:20,920
works there's definitely water in there

124
00:09:17,519 --> 00:09:20,920
come on baby

125
00:09:22,080 --> 00:09:30,000
try giving it ye old reseat ncmos clear

126
00:09:26,399 --> 00:09:30,000
and see if that resolves it

127
00:09:32,080 --> 00:09:38,720
hey whoa whoa hey hey hold on

128
00:09:36,080 --> 00:09:42,160
no idea what happened but it's up whoa

129
00:09:40,160 --> 00:09:46,320
BIOS has been reset nice so theoretically all we have to do is

130
00:09:45,519 --> 00:09:50,959
press 5.1 gigahertz oc

131
00:09:49,120 --> 00:09:56,560
and it'll work now this will probably take a few minutes changing chipsets

132
00:09:53,519 --> 00:09:58,720
and check this shiz out my friends we

133
00:09:56,560 --> 00:10:02,320
are booted not bad

134
00:10:00,480 --> 00:10:05,839
now neither of our network controllers have drivers here so i'm gonna have to

135
00:10:03,600 --> 00:10:09,519
download those but then we should be uh sitting pretty pretty quick here all

136
00:10:07,920 --> 00:10:12,560
right so while i reboot for a driver install we're just gonna top up the

137
00:10:10,959 --> 00:10:18,000
water here we did lose a little bit of water and our new block

138
00:10:14,720 --> 00:10:20,480
takes up a lot more of it so let's just

139
00:10:18,000 --> 00:10:24,160
do some precision pouring with the Linus tech chips water bottle

140
00:10:31,440 --> 00:10:37,440
not a supported use case for the Linus tech tips water bottle that's about as

141
00:10:35,200 --> 00:10:40,880
full as it was before i guess sick

142
00:10:38,880 --> 00:10:45,200
so the purpose of what we're doing here is to ensure that we're actually getting

143
00:10:42,959 --> 00:10:48,000
the 5.1 gigahertz that were advertised when the system's idle it's not

144
00:10:46,640 --> 00:10:52,079
necessarily going to be running at that speed but let's go ahead and hit it with

145
00:10:50,160 --> 00:10:56,320
a heavy CPU load and see what happens hey not bad

146
00:10:54,800 --> 00:11:00,560
okay then we survived our cinebench test so that

147
00:10:58,320 --> 00:11:05,040
means the last thing for us to do is

148
00:11:01,680 --> 00:11:06,399
try to enable xmp

149
00:11:05,040 --> 00:11:12,399
which will adjust our memory automatically

150
00:11:08,560 --> 00:11:14,880
to its faster speed and tighter timings

151
00:11:12,399 --> 00:11:18,240
i don't have a lot of hope for this it's a good sign

152
00:11:16,399 --> 00:11:22,320
that's a good sign too now that we've got that dialed in let's make sure it's

153
00:11:19,680 --> 00:11:27,120
not completely unstable do a quick cinebench extreme run

154
00:11:24,640 --> 00:11:30,320
eight threads at over five gigahertz ed oh yeah and your server has RGB now

155
00:11:29,600 --> 00:11:37,120
yeah so we seem to be stable we actually got 10 more points went from 568 to 578 so

156
00:11:35,920 --> 00:11:39,680
we're gonna go ahead and shut down cinebench here no i don't want to save

157
00:11:38,959 --> 00:11:46,640
it and uh ed show me where our sample project is why don't we do the tech link

158
00:11:45,120 --> 00:11:52,000
now TechLinked even though it doesn't take that long to render is the kind of

159
00:11:49,920 --> 00:11:57,360
project where the faster it goes the nicer it is because it's only about 10

160
00:11:54,399 --> 00:12:02,240
minutes but because we write shoot and edit it all within the same day

161
00:11:59,839 --> 00:12:07,040
the faster it exports renders and uploads the faster everybody gets to go

162
00:12:04,160 --> 00:12:10,720
home so let's go ahead and process is it just this one right here

163
00:12:09,040 --> 00:12:13,440
uh so i'm going to remove this from the queue

164
00:12:12,240 --> 00:12:17,040
and we're going to find out

165
00:12:16,000 --> 00:12:19,440
what do you think think it's going to be faster

166
00:12:22,800 --> 00:12:26,079
did it crash stopped

167
00:12:26,959 --> 00:12:36,320
did even the reset switch not work that's pretty crash let's turn xmp off

168
00:12:33,120 --> 00:12:39,760
and people wonder why i don't like xmp

169
00:12:36,320 --> 00:12:42,000
never works all right we've got one more

170
00:12:39,760 --> 00:12:46,560
trick up our sleeves i pulled two of the sticks i think 16 gigs of RAM should be

171
00:12:44,240 --> 00:12:50,160
plenty for our workload here and we're gonna see if it's a little

172
00:12:47,839 --> 00:12:54,079
more stable without all the memory slots loaded up

173
00:12:51,440 --> 00:12:58,160
yep there it is good good good overclocked

174
00:12:55,920 --> 00:13:02,880
running at the right frequency all right

175
00:12:59,440 --> 00:13:04,720
let's try that sample project again

176
00:13:02,880 --> 00:13:09,440
now what's interesting is our CPU is actually only running at

177
00:13:07,040 --> 00:13:14,639
4.8 gigahertz right now i wonder if xmp is kind of screwing us

178
00:13:12,240 --> 00:13:20,240
here all right guys we are really close to results time here so for TechLinked

179
00:13:17,200 --> 00:13:21,680
our time to beat is nine minutes 48

180
00:13:20,240 --> 00:13:28,160
seconds and it looks like we're gonna come in

181
00:13:24,240 --> 00:13:28,160
well under that let's have a look here

182
00:13:28,240 --> 00:13:34,560
come on baby you have got to be kidding me

183
00:13:31,519 --> 00:13:36,480
nine minutes and 19 seconds

184
00:13:34,560 --> 00:13:39,920
maybe we really did need those other two cores

185
00:13:37,680 --> 00:13:44,000
so here's the situation we only beat our score by about 30

186
00:13:42,720 --> 00:13:47,680
seconds uh which is not the kind of improvement

187
00:13:46,079 --> 00:13:52,720
that we were looking for so there's a couple factors at play here

188
00:13:50,240 --> 00:13:57,440
one is that even though i don't think this is an avx workload it's kind

189
00:13:55,200 --> 00:14:04,880
of behaving like one so our CPU turbos to 5.05 or 5.1 in cinebench but only

190
00:14:01,360 --> 00:14:06,880
hits 4.8 here which looks like an avx

191
00:14:04,880 --> 00:14:10,880
offset so we're not getting quite the clock

192
00:14:08,399 --> 00:14:15,440
speed we were expecting and two it appears that my recollection may have

193
00:14:13,120 --> 00:14:21,440
been incorrect or adobe may have updated their encoder because our CPU is getting

194
00:14:18,480 --> 00:14:26,320
absolutely pinned to a hundred percent in many cases here

195
00:14:23,120 --> 00:14:27,839
it may be that we actually do need

196
00:14:26,320 --> 00:14:33,040
more cores so there's no way around it we need to

197
00:14:29,920 --> 00:14:33,040
reevaluate here

198
00:14:33,760 --> 00:14:39,680
this project has really challenged everything i thought i knew about our

199
00:14:37,519 --> 00:14:42,880
video export workload so what we're going to have to do is take things

200
00:14:41,600 --> 00:14:47,040
entirely back to the drawing board and to do that

201
00:14:45,440 --> 00:14:52,800
i've set up a test bench here so it's much quicker for me to swap things in

202
00:14:48,720 --> 00:14:55,920
and out with Intel's 18 core 7980xe with

203
00:14:52,800 --> 00:14:58,399
the goal being to determine well gee

204
00:14:55,920 --> 00:15:02,320
exactly how many cores do we need for this how well does it

205
00:15:01,120 --> 00:15:07,600
scale so the idea here is once we know that we

206
00:15:05,040 --> 00:15:10,000
can then find the fastest chip with that many cores

207
00:15:08,959 --> 00:15:14,399
and try this whole thing again so based on

208
00:15:12,720 --> 00:15:20,480
what we're looking at here with loads anywhere from about the 75 to 85 range

209
00:15:17,839 --> 00:15:24,079
on this thing we can guess and this will be a

210
00:15:21,600 --> 00:15:29,199
far from perfect guess but we can guess that maybe something in the neighborhood

211
00:15:25,440 --> 00:15:32,079
of 14 cores could be the right answer

212
00:15:29,199 --> 00:15:37,040
fortunately we have a very unique ace up our sleeve this courtesy of our friends

213
00:15:34,880 --> 00:15:41,680
over at puget because the only way to get this chip is through Intel's like

214
00:15:39,440 --> 00:15:50,880
secret special auction is a core i9 99 90 xe it actually has

215
00:15:47,279 --> 00:15:52,320
fewer cores than a 9980xc at only 14

216
00:15:50,880 --> 00:15:59,440
versus 18 but it turbos to five gigahertz and its base

217
00:15:55,920 --> 00:16:01,440
clock is four gigahertz

218
00:15:59,440 --> 00:16:06,240
so why don't we give it a shot then charlie got nothing else to lose

219
00:16:04,000 --> 00:16:10,560
just time naturally

220
00:16:07,360 --> 00:16:13,040
changing out the CPU necessitates

221
00:16:10,560 --> 00:16:17,040
another motherboard swap so once again this poor

222
00:16:14,399 --> 00:16:21,199
aorus water cooled board is getting pulled from a project

223
00:16:18,959 --> 00:16:24,160
not because it did a bad job but just because

224
00:16:22,720 --> 00:16:27,240
it wasn't the right fit

225
00:16:31,759 --> 00:16:37,839
after all this going back to the original block

226
00:16:35,199 --> 00:16:37,839
so stupid

227
00:16:38,800 --> 00:16:44,880
how bad is it oh it's definitely dripping a bit

228
00:16:45,040 --> 00:16:52,079
now something to be aware of guys is that while electronics and

229
00:16:49,519 --> 00:16:55,360
water definitely don't mix as long as they're not powered the

230
00:16:53,680 --> 00:17:01,720
likelihood of doing permanent damage is actually very low

231
00:16:57,279 --> 00:17:01,720
not very low it's lower

232
00:17:04,880 --> 00:17:11,919
oh yeah i'm just looking pretty bad this time okay here we go

233
00:17:09,600 --> 00:17:11,919
and

234
00:17:12,880 --> 00:17:17,959
you're outta there beautiful

235
00:17:18,000 --> 00:17:24,480
where's that block yep it's on the floor

236
00:17:22,400 --> 00:17:28,160
so that goes to the inlet and that goes to the outlet

237
00:17:26,959 --> 00:17:32,880
go ahead and cut for the board i'm going with a primex 299 deluxe

238
00:17:31,280 --> 00:17:38,640
not for any particular reason other than that that's what i found on the shelf

239
00:17:36,160 --> 00:17:43,840
all right so here we go even though it's 14 compared to 18 cores

240
00:17:41,520 --> 00:17:51,440
because of its much higher frequencies this is a 200 205 watt TDP CPU

241
00:17:48,880 --> 00:17:57,360
and that's assuming that's accounting for Intel's optimism with respect to

242
00:17:54,559 --> 00:17:57,360
their tdps

243
00:17:59,200 --> 00:18:04,000
all right now we're ready blocks on

244
00:18:02,400 --> 00:18:07,919
cables are routed strap it in

245
00:18:05,760 --> 00:18:12,400
all right it's reassembled please work like just work

246
00:18:10,640 --> 00:18:16,960
it occurs to me this is a relatively new CPU and i have no idea when this

247
00:18:14,400 --> 00:18:22,640
board's BIOS was last updated heck yeah

248
00:18:18,880 --> 00:18:22,640
heck yeah all right it's a good sign

249
00:18:22,880 --> 00:18:26,919
techlinked transcode.mov

250
00:18:29,440 --> 00:18:36,000
and it doesn't look like it's going to be

251
00:18:32,960 --> 00:18:37,120
any faster at all curiously our CPU

252
00:18:36,000 --> 00:18:42,480
usage is significantly lower than it was on

253
00:18:40,080 --> 00:18:46,080
the 18 core okay

254
00:18:44,240 --> 00:18:50,880
and our time is looking like it's going to

255
00:18:47,760 --> 00:18:53,760
come in just over seven minutes for the

256
00:18:50,880 --> 00:18:58,240
raw encode which should put us right around the same time so we're at 9

257
00:18:56,400 --> 00:19:02,480
38 which is 10 seconds faster than our original

258
00:19:00,000 --> 00:19:07,600
configuration but actually 20 seconds slower than our water cooled overclocked

259
00:19:05,440 --> 00:19:10,960
9900k but i think we can do a little better

260
00:19:09,039 --> 00:19:14,400
here it occurs to me that just because xmp didn't work on that other board

261
00:19:12,880 --> 00:19:19,520
doesn't mean it won't work on this one because it might have been optimized for quad channel operation

262
00:19:17,520 --> 00:19:23,440
so we're going to go ahead and

263
00:19:20,720 --> 00:19:26,400
try again with xmp good luck everyone

264
00:19:26,480 --> 00:19:34,160
well this puppy's definitely hauling now we're seeing turbo speeds floating right

265
00:19:31,039 --> 00:19:36,480
around the 4 gigahertz range

266
00:19:34,160 --> 00:19:41,120
and that estimated time remaining looks like it might be as much as 30 45

267
00:19:39,039 --> 00:19:45,280
seconds faster than our last run what remains to be seen is whether or not

268
00:19:43,520 --> 00:19:48,880
it's stable surprisingly was stable enough to finish

269
00:19:47,440 --> 00:19:52,880
the render all right good job xmp

270
00:19:50,960 --> 00:19:55,880
hey 850.

271
00:19:55,919 --> 00:20:01,679
wow so we actually crushed our original time

272
00:20:00,240 --> 00:20:05,679
that time goes to show you what more cores more

273
00:20:03,520 --> 00:20:08,640
clock speed and faster memory will do for you wow that's almost by a full

274
00:20:07,760 --> 00:20:13,919
minute so now that we've at least managed to make the machine faster

275
00:20:12,480 --> 00:20:17,520
i think i'm going to stop fooling around because this is two full days that the

276
00:20:15,840 --> 00:20:21,760
editors haven't had their transcode server but

277
00:20:19,280 --> 00:20:26,880
i'm also feeling pretty inspired right now because we did manage to get better

278
00:20:24,320 --> 00:20:32,159
results than this with that weird 1u Gigabyte overclockable server that we

279
00:20:29,679 --> 00:20:38,080
checked out a little while ago so what i think is that with a better designed

280
00:20:34,720 --> 00:20:40,159
water cooling system maybe a board with

281
00:20:38,080 --> 00:20:44,960
beefier vrms or something along those lines that we could push this further i

282
00:20:42,880 --> 00:20:48,880
think our 500 watt power supply might actually be holding us back as well so

283
00:20:47,679 --> 00:20:53,280
we're going to leave this one here for now but next time i'm going to be coming

284
00:20:51,120 --> 00:20:57,840
back at you guys with a custom designed enclosure that's really going to let

285
00:20:56,000 --> 00:21:01,280
this thing rip skillshare is an online learning

286
00:20:59,600 --> 00:21:05,600
community with thousands of classes in design business technology and more and

287
00:21:04,159 --> 00:21:10,480
their premium membership gives you unlimited access to high quality classes

288
00:21:08,240 --> 00:21:15,600
on must-know topics so you can improve your skills unlock new opportunities in

289
00:21:12,799 --> 00:21:18,559
life and do the work that you love it's more affordable than most learning

290
00:21:17,120 --> 00:21:22,559
platforms out there with an annual subscription costing less than ten

291
00:21:20,240 --> 00:21:25,760
dollars a month and it's got tons of great stuff like you can check out

292
00:21:23,840 --> 00:21:30,159
christopher dodd who is a top teacher at skillshare he's a self-taught web

293
00:21:27,760 --> 00:21:33,600
developer who inspires and educates students his classes go over the

294
00:21:32,000 --> 00:21:40,480
fundamentals of web development and he has over 6500 students signed up for his courses

295
00:21:38,240 --> 00:21:43,600
the first 500 of you to use the promo link in the description are going to get

296
00:21:41,760 --> 00:21:47,440
your first two months for free so don't wait guys go check out skillshare at the

297
00:21:45,919 --> 00:21:51,679
link below so thanks for watching guys if you disliked this video you can hit that

298
00:21:50,240 --> 00:21:56,159
button but if you liked it hit like get subscribed or maybe consider checking out where to buy the stuff we featured

299
00:21:54,400 --> 00:22:00,080
at the link in the video description also down there is our merch store which

300
00:21:57,919 --> 00:22:02,559
has cool shirts like this one that james is wearing

301
00:22:03,520 --> 00:22:10,000
it's like you're seeing double i'm also wearing it yeah lttstore.com uh also our

302
00:22:08,720 --> 00:22:13,200
community forum which you should definitely join