1
00:00:00,560 --> 00:00:08,080
now if i'm Intel at this point i'm hoping beyond hope that ryzen

2
00:00:05,839 --> 00:00:13,280
has reached his full potential zen 2 basically spelled the end for the

3
00:00:11,040 --> 00:00:16,720
lakes in enthusiast's eyes so much so that it's now starting to leak into the

4
00:00:14,799 --> 00:00:21,439
mainstream even but unfortunately for Intel those enthusiasts have been busy

5
00:00:19,920 --> 00:00:27,519
and it ain't the end one usmas is that how i pronounce that

6
00:00:24,160 --> 00:00:29,920
of the ryzen drm calculator fame reached

7
00:00:27,519 --> 00:00:36,559
out to us with a special little project that he's calling workstation tool

8
00:00:32,960 --> 00:00:39,920
it's an automatic overclocking and

9
00:00:36,559 --> 00:00:43,120
simultaneous under volting tool that

10
00:00:39,920 --> 00:00:46,000
promises to both improve performance

11
00:00:43,120 --> 00:00:51,840
and decrease power consumption on your AMD ryzen processor is it over does AMD

12
00:00:49,680 --> 00:00:56,320
have the high ground let's fire it up

13
00:00:53,520 --> 00:01:00,719
and see for ourselves shall we and we're going to see if today's video

14
00:00:57,840 --> 00:01:04,000
is brought to you by glasswire yes it is

15
00:01:02,239 --> 00:01:07,840
instantly see your current and past network activity with glasswire you can

16
00:01:05,920 --> 00:01:12,479
detect malware and lock badly behaving apps on your pc or Android device and

17
00:01:09,760 --> 00:01:15,840
use offer code Linus to get 25 off at the link down below

18
00:01:23,200 --> 00:01:28,320
now i know what you guys are thinking come on

19
00:01:26,400 --> 00:01:33,439
this is just another one of those pc performance tune-up tools registry

20
00:01:31,520 --> 00:01:37,840
cleaners that claims it's going to improve performance but actually doesn't

21
00:01:35,119 --> 00:01:42,720
do anything and i understand your skepticism we even had someone reach out

22
00:01:40,400 --> 00:01:47,439
about sponsoring a video with their performance tuning tool

23
00:01:45,119 --> 00:01:51,680
and i installed this thing and i'm like okay so i've got all these numbers that

24
00:01:49,119 --> 00:01:55,280
say oh you are going 30 faster on networking but i was like hey what

25
00:01:53,360 --> 00:01:59,119
benchmarks can i run to demonstrate it's working they were like oh you want to

26
00:01:56,880 --> 00:02:03,439
run benchmarks whoa what no

27
00:02:00,159 --> 00:02:05,360
we have from one usmas screenshots

28
00:02:03,439 --> 00:02:08,800
showing in real benchmarks that people actually use like cinebench as much as a

29
00:02:07,680 --> 00:02:13,840
9 performance improvement after running

30
00:02:11,120 --> 00:02:19,760
this software and the craziest part he's providing it to the community for free

31
00:02:16,879 --> 00:02:26,080
now it is a little bit complicated to set up which is why i brought an Anthony

32
00:02:23,280 --> 00:02:29,760
hello it's not as complicated as zen underclocking has been in the past

33
00:02:28,640 --> 00:02:34,720
first you need to make sure that you're using a zen 2 CPU zen one and zen plus

34
00:02:32,560 --> 00:02:39,080
as well as renoir epus are not currently supported you'll also need to make sure

35
00:02:36,239 --> 00:02:44,080
your BIOS is up to date with the gsa 1.0.0.4 combo pi being the lowest one

36
00:02:41,840 --> 00:02:47,280
osmis has tested to work while you're in your BIOS there are still some settings

37
00:02:45,840 --> 00:02:52,800
you'll need to change as well you need to make sure your CPU load line calibration setting is set to level 4

38
00:02:50,560 --> 00:02:56,480
for ASUS motherboards or level 3 for other vendors and you'll want to set CPU

39
00:02:54,959 --> 00:03:00,000
phase control to standard if your motherboard has a setting for it ASUS

40
00:02:58,160 --> 00:03:03,920
motherboards also have a CPU current capability setting which you should set

41
00:03:01,519 --> 00:03:07,599
to 100 and once all that's done save your changes and boot into Windows where

42
00:03:05,760 --> 00:03:11,920
you'll want to make sure you've got net framework 4.8 or higher installed and

43
00:03:10,159 --> 00:03:15,599
then download ryzen master as this is how the app gets its power frequency and

44
00:03:13,760 --> 00:03:18,720
temperature readings with all that done you're almost ready to get started

45
00:03:16,959 --> 00:03:22,959
download and extract the archive to a folder you can access we use the desktop

46
00:03:21,040 --> 00:03:27,360
and then download and extract cinebench from maxon's website not the Windows

47
00:03:25,360 --> 00:03:30,720
store this version doesn't work the app actually uses cinebench as a useful

48
00:03:29,120 --> 00:03:34,720
before and after comparison between settings and maintains a database of

49
00:03:32,720 --> 00:03:40,000
your performance results when all that's done we just run wst in the main folder

50
00:03:37,519 --> 00:03:43,200
just like this here's the app and what you'll see here

51
00:03:42,159 --> 00:03:46,560
at the top is a bunch of information constantly

52
00:03:45,280 --> 00:03:50,239
updating this is coming from ryzen master on the left hand column we have

53
00:03:48,400 --> 00:03:56,480
the core numbers so we've got core 1 core 2 core 3 core 4 on ccx1 5678 on

54
00:03:54,159 --> 00:03:59,120
ccx2 etc in the middle we have the effective clock speed of each of these

55
00:03:58,159 --> 00:04:04,560
cores right now they're basically nothing because the system is more or less idle

56
00:04:02,400 --> 00:04:10,959
and on the right hand side we have a number that indicates exactly how good

57
00:04:07,840 --> 00:04:13,280
the core is for overclocking this is the

58
00:04:10,959 --> 00:04:17,919
number the BIOS indicates to Windows to tell it which cores it should more

59
00:04:15,280 --> 00:04:22,560
heavily load so for our system here looks like ccx1 has pretty good cores

60
00:04:21,040 --> 00:04:27,120
ccx2 is it's all right czx3 not so great four

61
00:04:25,680 --> 00:04:31,600
that seems to be the lowest out of them all five is pretty good actually that's

62
00:04:29,680 --> 00:04:36,160
the best of them all we also have CPU usage CPU temperature current voltage

63
00:04:34,240 --> 00:04:41,520
total package power as well as some other AMD related power metrics here let's

64
00:04:39,680 --> 00:04:44,880
take a look at the settings what the app does is it takes

65
00:04:43,280 --> 00:04:49,199
the reference frequency here which is the starting level and the maximum

66
00:04:46,639 --> 00:04:54,080
frequency that you enter in which is the end level and what it does is it steps

67
00:04:52,000 --> 00:04:57,600
through each of those frequencies testing it each time to see which cores

68
00:04:56,720 --> 00:05:03,440
are actually running at those frequencies which ones are erroring out which ones are getting too hot that kind

69
00:05:01,360 --> 00:05:08,479
of thing as it does that that's where cycle time comes in 240 seconds that is

70
00:05:06,400 --> 00:05:13,759
four minutes is the default for how long it uh tests each iteration you can

71
00:05:11,520 --> 00:05:16,800
imagine this takes quite a while to do but the outcome should be pretty great

72
00:05:15,919 --> 00:05:22,800
now a couple of other settings here are a little bit confusing ccx delta this is

73
00:05:20,720 --> 00:05:28,320
actually set on a per CPU basis and it should be set automatically based on

74
00:05:25,600 --> 00:05:30,880
what your CPU is that being said it's not something to really worry about

75
00:05:29,840 --> 00:05:37,120
ryzen 3 to 5 and 7 i think are good for 75

76
00:05:34,720 --> 00:05:41,440
megahertz ryzen 9 i think is and

77
00:05:37,759 --> 00:05:42,880
fifty and uh Threadripper is 75 to 100.

78
00:05:41,440 --> 00:05:45,680
we also have the testing mode here currently it's on avx light

79
00:05:44,800 --> 00:05:50,400
which one osmos actually says is

80
00:05:48,080 --> 00:05:56,400
similar in terms of power draw to what you might expect from a full avx

81
00:05:53,759 --> 00:06:00,080
workload except it's not quite as hot so it's useful for testing coolers that are

82
00:05:58,479 --> 00:06:04,960
a little bit less capable than what we have here the only other thing that's going to be relevant for most people is

83
00:06:03,280 --> 00:06:09,919
the reference voltage and target voltage again this is the start and end points

84
00:06:07,840 --> 00:06:14,080
much like the frequency here these are all pre-configured with values that the

85
00:06:12,639 --> 00:06:17,759
software kind of thinks you might want to use so i mean if you want you can

86
00:06:16,240 --> 00:06:21,919
just hit start and see what happens otherwise you can tweak these things like the

87
00:06:20,080 --> 00:06:25,680
maximum temperature or where you want to start and stop just to

88
00:06:23,360 --> 00:06:30,560
kind of narrow things down or yeah try your luck aside from that there's not a

89
00:06:27,840 --> 00:06:34,800
whole lot else to the program itself so uh let me go ahead and set a short cycle

90
00:06:32,880 --> 00:06:37,840
time and uh hit start and see what happens so the first thing that's going

91
00:06:36,160 --> 00:06:42,160
to happen is cinebench r20 is going to pop up and it's going to do a baseline

92
00:06:39,759 --> 00:06:46,880
run this is your CPU as it is normally so let's see what we get here it's a

93
00:06:44,560 --> 00:06:51,360
3970x looks like there was a previous run on this that was 16 927 let's see

94
00:06:49,199 --> 00:06:55,120
what we get here 16 932 a little bit better now it's going to go through it

95
00:06:53,360 --> 00:07:00,160
set the parameters that we have here so it uh we only have 4 000 set for both

96
00:06:57,520 --> 00:07:03,280
the reference and max frequency and 1100 for the reference and target voltage one

97
00:07:01,759 --> 00:07:07,840
of the most interesting and really exciting things about this program is

98
00:07:05,520 --> 00:07:11,440
that not only does it do all of this it still works with all of AMD's power

99
00:07:09,599 --> 00:07:15,599
management features you're basically getting the best of both worlds between

100
00:07:13,360 --> 00:07:20,240
a straight underclock as well as overclocking per ccx it should make a

101
00:07:17,840 --> 00:07:26,000
huge difference so that's it run the utility and everything's dialed in i'm

102
00:07:23,120 --> 00:07:30,319
ready to get my performance gains basically yeah it does take a little bit

103
00:07:27,919 --> 00:07:33,680
more effort than that uh it does require a little bit of tweaking from the user's

104
00:07:31,919 --> 00:07:39,199
perspective especially for a chip that one of us hasn't actually tested yet

105
00:07:35,919 --> 00:07:41,520
which is the 3970x he's got a 3960x so

106
00:07:39,199 --> 00:07:46,560
it took me a little bit of doing and a few bsods later but here we are and

107
00:07:44,639 --> 00:07:50,880
should have everything all dialed in for a nice underbolt it's actually a little

108
00:07:48,560 --> 00:07:55,680
bit slightly overclocked as well so it's about 3.9 on all cores with a couple of

109
00:07:53,599 --> 00:07:59,599
ccx's dialed back one or two ticks that's really cool having that level of

110
00:07:57,520 --> 00:08:04,720
granularity where you can just like AMD would tune them at the

111
00:08:01,919 --> 00:08:09,440
factory go okay yeah this one's ever so slightly better than that one let's

112
00:08:07,280 --> 00:08:13,919
turn it up just a little bit more super cool yeah like you could do this

113
00:08:10,879 --> 00:08:16,800
manually with ryzen master but like from

114
00:08:13,919 --> 00:08:20,960
experience it takes a lot longer than this does and it is not as automated so

115
00:08:18,960 --> 00:08:24,400
like to do all of this would have taken so much longer because i had to set up

116
00:08:22,400 --> 00:08:27,440
all the testing myself i had to crank everything on my own and you'd have to

117
00:08:25,919 --> 00:08:32,800
be sitting there staring at it the whole time yeah okay so i can run cinebench then yeah

118
00:08:30,960 --> 00:08:37,760
theoretically there's not really a whole lot to go wrong here if it's overclocked

119
00:08:35,440 --> 00:08:42,240
then it's gonna be faster unless it crashes unless it crashes boy these

120
00:08:40,080 --> 00:08:46,240
things are fast ain't they oh yeah now one tip we got from one asmos via email

121
00:08:44,720 --> 00:08:52,560
as we were going back and forth about this was that the tool works best for

122
00:08:49,360 --> 00:08:53,360
cpus that have more ccx's

123
00:08:52,560 --> 00:09:01,360
so wait what was our before number the before number was uh 16 9 something that

124
00:08:58,880 --> 00:09:05,360
is a smaller performance gain than i expected it's small but at the same time

125
00:09:03,760 --> 00:09:08,560
the actual overclock was small it's within about 100 megahertz of what it

126
00:09:06,959 --> 00:09:12,320
would normally be doing so that could come down to us just having a chip that

127
00:09:10,959 --> 00:09:16,240
doesn't undervolt and overclock particularly well then yeah i tried

128
00:09:14,480 --> 00:09:19,519
doing a more aggressive overclock and while i could get stable results it

129
00:09:17,839 --> 00:09:24,640
wasn't consistent and i think that might be down to either our chip or our

130
00:09:22,560 --> 00:09:28,320
load line calibration or our phase control settings while dialing it in

131
00:09:26,399 --> 00:09:32,160
it's best to have it on standard like i said but he came back and said hey

132
00:09:30,800 --> 00:09:34,959
if you're having issues with it with stability try sending it to a stream

133
00:09:33,839 --> 00:09:40,399
anyway and that should give you a little bit more overhead as far as stability

134
00:09:37,920 --> 00:09:43,839
goes it might increase your thermals but that's yeah i mean when you're dialing

135
00:09:42,080 --> 00:09:49,200
it in that's less of a problem so then should we have a look at what our thermals ended up like do you have the

136
00:09:46,880 --> 00:09:53,839
before thermals well i have if you go to the benchmark tab here this is our

137
00:09:51,120 --> 00:09:58,720
before run and this is our after run and our power usage is over there as

138
00:09:55,680 --> 00:10:00,000
well before and after wow so 30 watts

139
00:09:58,720 --> 00:10:04,160
less and ever so slightly greater performance

140
00:10:02,320 --> 00:10:07,120
yeah so if i were to keep the same level of power i could increase the

141
00:10:05,760 --> 00:10:10,880
performance more the only problem again being that

142
00:10:09,279 --> 00:10:15,600
this is a pre-production piece of software and not tested on every chip

143
00:10:13,760 --> 00:10:21,040
he's going to have it all massaged out by the time it's ready for release so

144
00:10:18,480 --> 00:10:26,160
the price is definitely right in a nutshell wst does automatically

145
00:10:24,160 --> 00:10:29,839
what we used to have to do manually when overclocking our processors and what i

146
00:10:28,079 --> 00:10:33,519
think is really cool is the fact that it uses the same

147
00:10:31,519 --> 00:10:37,839
standard tools for both measuring improvements in performance as well as

148
00:10:35,680 --> 00:10:43,279
validating stability that we as enthusiasts have been using for years

149
00:10:40,399 --> 00:10:47,600
and years now as you've seen here your mileage may vary you might not get

150
00:10:45,440 --> 00:10:51,200
a ton more performance or you might not even get a power consumption improvement

151
00:10:49,519 --> 00:10:54,640
there's nothing that one usmas can do about how good your chip came out of the

152
00:10:52,880 --> 00:10:59,839
factory but at least you've got it now and this

153
00:10:56,399 --> 00:11:01,760
really puts like custom levels of tuning

154
00:10:59,839 --> 00:11:05,839
overclocking in the hands of people who only really know how to press

155
00:11:03,839 --> 00:11:09,279
one button pretty neat

156
00:11:07,440 --> 00:11:14,880
you know what else is neat manscaped manscape's perfect package 3.0

157
00:11:12,720 --> 00:11:19,600
kit includes everything you need to take your junk grooming to the next level it

158
00:11:17,440 --> 00:11:23,040
includes their lawnmower 3.0 waterproof body trimmer with its ergonomic design

159
00:11:21,600 --> 00:11:27,440
and quality ceramic blades built with their advanced skin safe technology the

160
00:11:25,279 --> 00:11:33,120
blades are replaceable and they're 100 compatible with the lawnmower 2.0 and

161
00:11:29,600 --> 00:11:35,120
the 3.0 and it's got a powerful 7000 RPM

162
00:11:33,120 --> 00:11:39,360
motor and 600 milliamp hour lithium ion battery it's even cordless and

163
00:11:36,959 --> 00:11:44,000
waterproof so you can trim your ball hair wherever you happen to be even in

164
00:11:41,600 --> 00:11:47,279
the rain or more likely in the shower that actually makes a lot more sense

165
00:11:45,440 --> 00:11:51,519
they've got a new compact anti-tug adjustable trimmer guard and a built-in

166
00:11:48,959 --> 00:11:56,320
led light and you can get 20 off and free shipping on your perfect package

167
00:11:53,040 --> 00:11:57,839
3.0 when you check out our link down

168
00:11:56,320 --> 00:12:01,040
below totally intentional by the way

169
00:11:59,519 --> 00:12:05,279
thanks for watching guys if you enjoyed this video and you enjoy this kind of

170
00:12:03,120 --> 00:12:08,880
like tuning software wow have we done much in the way of like tuning software

171
00:12:06,959 --> 00:12:11,920
it hasn't really been much we never even we never even checked out his last piece

172
00:12:10,720 --> 00:12:15,839
of software so we're not going to throw to our video find a really good video on

173
00:12:14,240 --> 00:12:19,680
um dram calculator vram calculator yeah

174
00:12:17,839 --> 00:12:25,839
yeah find a good one on dram calculator and we'll just kind of throw to that

175
00:12:21,440 --> 00:12:25,839
it's also from one sms and also free