1
00:00:03,120 --> 00:00:10,719
You know what? I'm done. Between forced updates, bloatware, AND AI FEATURES THAT

2
00:00:08,320 --> 00:00:17,840
NOBODY asked for. I mean, what in heaven's name are you doing, son?

3
00:00:14,719 --> 00:00:20,160
He's not dead. He just needs the right

4
00:00:17,840 --> 00:00:26,240
medicine. But I have an NVIDIA GPU. Isn't that a sin? According to Reddit,

5
00:00:23,199 --> 00:00:28,480
maybe it is true that the mere mention

6
00:00:26,240 --> 00:00:33,280
of NVIDIA in a Linux forum can start an argument. But when Lennis Torvoltz, the

7
00:00:31,359 --> 00:00:38,719
man himself, was here, he mentioned that Linux and NVIDIA have somewhat mended

8
00:00:36,399 --> 00:00:43,680
the relationship now that NVIDIA is balls deep in the AI bubble and [music]

9
00:00:40,719 --> 00:00:49,920
needs Linux developers on their side. Which raises the question, is AMD still

10
00:00:46,800 --> 00:00:52,079
the best GPU choice for Linux? And where

11
00:00:49,920 --> 00:00:56,559
does Intel [music] land? To find out, we spent weeks testing. And I don't just

12
00:00:54,320 --> 00:01:01,359
mean FPS numbers. I'm talking setup process and day-to-day usability, too.

13
00:00:58,879 --> 00:01:06,080
Now, as you might expect, AMD does remain a very strong choice. But you

14
00:01:03,680 --> 00:01:11,280
MIGHT BE SHOCKED, SHOCKED, I SAY, when you see how far the others, especially

15
00:01:08,640 --> 00:01:16,159
certain ones, have come. What never shocks anyone anymore, though, is my

16
00:01:13,439 --> 00:01:19,840
segways to our sponsor, Play Tracker. They keep all your gaming achievements

17
00:01:17,840 --> 00:01:24,320
in one place. It tracks all of your performance and stats across multiple

18
00:01:21,840 --> 00:01:28,479
games on multiple platforms and lets you compare and compete against friends. Try

19
00:01:26,640 --> 00:01:31,720
Play Tracker for free using our link down below.

20
00:01:34,321 --> 00:01:36,341
[music]

21
00:01:39,360 --> 00:01:46,079
Before we get into the results, let's talk about the setup process. We chose

22
00:01:43,119 --> 00:01:51,439
these five GPUs to represent both recent and slightly older gaming cards from the

23
00:01:48,399 --> 00:01:54,399
main brands and also Intel was there. As

24
00:01:51,439 --> 00:02:00,079
for our DRO, our original plan when this was an article on ltlabs.com was good

25
00:01:57,280 --> 00:02:05,119
old Ubuntu 2404LTS or long-term support. Why that one?

26
00:02:02,880 --> 00:02:10,160
Well, while there are numerous more gaming focused dros like Bazite or Steam

27
00:02:08,000 --> 00:02:14,560
OS, we decided to approach this experiment with the idea that we wanted

28
00:02:11,840 --> 00:02:20,080
a stable, userfriendly, wellsupported desktop experience that can also game.

29
00:02:18,319 --> 00:02:23,680
Maybe we should do a video going through Linux gaming dros, though. Comment which

30
00:02:22,000 --> 00:02:27,120
ones that you guys would want to see. Right out of the gate, we ran into a

31
00:02:25,280 --> 00:02:32,400
problem that will likely be familiar to any Linux early adopter. Driver [music]

32
00:02:29,520 --> 00:02:37,519
support. Out of our five GPUs, four of them failed to work out of the box. Some

33
00:02:34,560 --> 00:02:41,760
were easy fixes. For the 460Ti, Ubuntu has a driver on their additional drivers

34
00:02:39,519 --> 00:02:47,040
page, and after a short reboot, it worked flawlessly. But the 5060 Ti, that

35
00:02:45,280 --> 00:02:50,400
was a bit of a different story. See, back when we originally did this

36
00:02:48,400 --> 00:02:55,360
testing, Ubuntu didn't have a driver for it at all. Meaning, we had to settle for

37
00:02:52,959 --> 00:03:00,400
the official driver from NVIDIA, which if you ask any Linux head, they're going

38
00:02:57,360 --> 00:03:03,440
to tell you is a big no no. As for our

39
00:03:00,400 --> 00:03:05,920
Radeon 970, it too was lacking bakedin

40
00:03:03,440 --> 00:03:10,640
drivers, and AMD hadn't yet rolled out their official installer. Finally, the

41
00:03:08,080 --> 00:03:14,000
Arc B580 appeared to be working until the moment we fired up a game, and it

42
00:03:12,239 --> 00:03:18,400
showed its true colors. Turns out that the easiest fix was to use Kubuntu,

43
00:03:16,159 --> 00:03:22,480
which uses the KDE desktop environment rather than Gnome. Of course, all of

44
00:03:20,319 --> 00:03:27,280
these issues have since been resolved. Yay, open source. So, what's the point

45
00:03:25,920 --> 00:03:31,760
of bringing them up? Well, we said we'd talk about the experience, and it turns

46
00:03:29,120 --> 00:03:37,440
out that for us, the bottom line is that it comes down less to which brand of GPU

47
00:03:34,720 --> 00:03:42,799
you choose and more down to the age and the exact timing. The easiest fix across

48
00:03:40,480 --> 00:03:48,159
the board for all of them was simply to move off of LTS and use Ubuntu 25, which

49
00:03:46,400 --> 00:03:51,760
just uses a newer version of the Linux kernel that had bakedin support for all

50
00:03:50,080 --> 00:03:57,280
of our cards. With everything up and running, how's the day-to-day life?

51
00:03:53,680 --> 00:03:59,840
Well, this is where things get totally

52
00:03:57,280 --> 00:04:04,480
fine. Actually, most people these days, except for you, Dave, and by the way,

53
00:04:02,239 --> 00:04:08,560
you are so good at using computers, do a huge percent of their work in a browser.

54
00:04:06,560 --> 00:04:13,360
Google Docs, Teams, support tickets, even some design work can now be done

55
00:04:10,560 --> 00:04:17,840
fully within a browser. And even for the work that isn't, some combination of

56
00:04:15,680 --> 00:04:22,560
Linux alternative software, virtual machines, or remote desktop applications

57
00:04:20,400 --> 00:04:26,000
could probably allow most people to make the switch. Of course, the day-to-day

58
00:04:24,400 --> 00:04:29,520
usability of Linux is not the point of this video. The point of this video is

59
00:04:27,680 --> 00:04:35,360
the difference in that experience between our various cards. And I kind of

60
00:04:33,040 --> 00:04:38,479
hate to say it, but it's kind of a draw. Again, I know that's not very

61
00:04:36,960 --> 00:04:42,800
satisfying, and I'm sure that you would much rather that I praise the corporate

62
00:04:40,720 --> 00:04:47,600
brand that you love and bash the one that you really hate. But Lin Torvalds

63
00:04:45,280 --> 00:04:52,400
himself is doing his day-to-day work right now on an Intel arc. Not because

64
00:04:50,479 --> 00:04:56,880
of how special it is compared to the other options, but because for his work,

65
00:04:54,479 --> 00:05:00,639
it just plain didn't matter that much. He actually cared more about how quiet

66
00:04:58,479 --> 00:05:05,280
the cooler was. With that said, there will be some huge exceptions to this

67
00:05:03,440 --> 00:05:08,720
depending on the software you use. There could be major differences in

68
00:05:06,880 --> 00:05:14,320
performance, compatibility, and stability. But which card works best for

69
00:05:11,840 --> 00:05:18,880
training LLMs or whatever, that's beyond the scope of this video. Which brings us

70
00:05:15,840 --> 00:05:20,560
to the big question. What about the

71
00:05:18,880 --> 00:05:24,720
games? Now, before we show you the numbers, it's important to understand

72
00:05:22,080 --> 00:05:28,560
that our usual testing software doesn't work quite the same across Windows and

73
00:05:26,800 --> 00:05:33,520
Linux. That's why we stuck to simple in-game benchmarking tools. Also, for

74
00:05:30,960 --> 00:05:38,720
that reason, we can't draw a definitive conclusion about exactly how the Linux

75
00:05:36,720 --> 00:05:43,600
and Windows performance compares to each other. What we can do is get a broad

76
00:05:41,120 --> 00:05:48,320
impression, and we can say that across all of our three brands, if you were

77
00:05:45,759 --> 00:05:52,639
hoping for a big win for Linux, unfortunately, that still isn't the

78
00:05:50,240 --> 00:05:58,960
case. Across the board, we saw lower performance on Linux at 1080p, both with

79
00:05:55,680 --> 00:06:00,639
and without rate tracing, and at 1440p.

80
00:05:58,960 --> 00:06:05,440
Of course, that's not the question that's most on our minds right now.

81
00:06:02,800 --> 00:06:09,440
Rather, we want to know which GPU maker loses the least performance switching

82
00:06:08,000 --> 00:06:15,120
over to Linux. Starting with [music] Intel, the B580 was about 30% worse on

83
00:06:13,120 --> 00:06:19,759
average. That means that most of your games are going to run a full class of

84
00:06:17,280 --> 00:06:24,479
card worse. And you might even find outliers like what we saw with Dota 2

85
00:06:22,080 --> 00:06:29,600
that are frankly unacceptable. I mean, look at this. This game has a native

86
00:06:27,360 --> 00:06:33,600
Linux build. That means no Proton compatibility layer and yet it's getting

87
00:06:31,600 --> 00:06:38,000
barely over half of the performance compared to Windows. As for NVIDIA,

88
00:06:35,680 --> 00:06:41,440
well, things got a little spicy here to say the least. Do you remember I

89
00:06:39,680 --> 00:06:46,560
mentioned that our original article tested on Ubuntu 24? Well, in that test,

90
00:06:44,160 --> 00:06:50,479
the older 4060 Ti was still within striking distance of our Windows tests

91
00:06:48,400 --> 00:06:56,800
in a fair number of our games. Some were further off, but was pretty close. As

92
00:06:53,280 --> 00:06:59,520
for our 5060Ti, I mean, it is a faster,

93
00:06:56,800 --> 00:07:05,360
better card, but the difference between Windows and Linux was about 30% just

94
00:07:02,479 --> 00:07:09,599
like we saw with Intel. Yikes. Though not as yikes as the internet might have

95
00:07:07,360 --> 00:07:14,880
you believe. What is as yikes is the fact that upgrading to Ubuntu 25 for the

96
00:07:12,639 --> 00:07:19,919
retesting that we did for this video saw an overall regression in Linux

97
00:07:17,199 --> 00:07:24,160
optimization across both of our NVIDIA cards. or I don't know, maybe they tuned

98
00:07:22,479 --> 00:07:29,280
the Windows drivers to be even better. All I can say for sure is that NVIDIA

99
00:07:26,560 --> 00:07:33,840
could use just a few of their trillions of dollars to do a little bit better

100
00:07:31,280 --> 00:07:38,319
here. Bringing us finally to AMD end. Well, it won't be a huge surprise to

101
00:07:36,479 --> 00:07:42,000
many of you, but it turns out that if you let open- source folks do open

102
00:07:40,160 --> 00:07:47,039
source things with your drivers, you might get better open-source performance

103
00:07:44,319 --> 00:07:52,160
with some games like Cyberpunk 2077 seeing almost no performance loss and

104
00:07:49,840 --> 00:07:57,919
even some outliers like Strange Brigade on our RX970 that actually performed

105
00:07:55,199 --> 00:08:02,639
better on Linux. For gaming, then AMD emerges once again as a clear winner.

106
00:07:59,919 --> 00:08:09,520
So, if we factor that in, let's see here. We've got AMD wins gaming.

107
00:08:05,360 --> 00:08:12,160
Everything else is a draw. AMD wins.

108
00:08:09,520 --> 00:08:16,240
Or do they? Our final test was to throw a couple of different dros into the mix

109
00:08:14,160 --> 00:08:20,800
because, let's face it, our entire comments section would be complaining if

110
00:08:17,759 --> 00:08:23,680
we didn't. Since AMD came out on top, we

111
00:08:20,800 --> 00:08:30,479
wanted to see how our RX7700 XT would handle Basite, Mangaro, Ubuntu 25, and

112
00:08:27,520 --> 00:08:34,159
Ubuntu 24 across three games. And what we found was that there was functionally

113
00:08:32,399 --> 00:08:38,080
no difference between the three, which I got to say is honestly a really good

114
00:08:36,080 --> 00:08:43,519
thing. I think the last thing that the Linux community wants is a war over

115
00:08:40,800 --> 00:08:47,519
which DRO performs the best for gaming. There's already enough of a war over the

116
00:08:45,760 --> 00:08:51,760
ease of use of gaming across various distros or the unique value propositions

117
00:08:49,920 --> 00:08:56,160
that they offer from pre-installed applications and features or slick UIs

118
00:08:54,640 --> 00:09:03,200
for organizing your favorite retro titles. So, in conclusion, then AMD

119
00:08:59,839 --> 00:09:06,320
wins, but it's also totally valid to go

120
00:09:03,200 --> 00:09:07,920
Linux on any GPU hardware. In fact, I

121
00:09:06,320 --> 00:09:12,080
think it's fair to say that for most people out there, it really doesn't

122
00:09:10,000 --> 00:09:16,640
matter that much. Certainly, it doesn't matter enough that you should wait until

123
00:09:14,640 --> 00:09:19,920
you can afford a new GPU before you take the Linux plunge. If you're interested

124
00:09:18,320 --> 00:09:24,080
in jumping ship from Windows, you're not alone right now. And I think that your

125
00:09:22,080 --> 00:09:27,680
bigger roadblocks are going to be around settling on a DRO that suits you and

126
00:09:26,320 --> 00:09:31,040
getting all of your devices and workflows running. But the good news is

127
00:09:29,360 --> 00:09:34,480
that there's an ever growing number of resources out there for new users. And

128
00:09:33,200 --> 00:09:37,600
we're going to link some of those below for you guys to take a look at. One way

129
00:09:36,240 --> 00:09:43,120
or another, the year of the Linux desktop is getting ever closer. And

130
00:09:40,080 --> 00:09:45,200
arguably, I would say is even today for

131
00:09:43,120 --> 00:09:50,399
people that mostly work inside a browser or would rather give up a little bit of

132
00:09:48,000 --> 00:09:54,160
performance than put up with Windows's Just like you're going to put

133
00:09:52,320 --> 00:09:58,000
up with this segue [music] to our sponsor, Play Tracker. You know

134
00:09:56,080 --> 00:10:01,839
what time of year is always a hoot? Spotify rap season. There's just

135
00:09:59,920 --> 00:10:05,440
something satisfying about seeing what music and moods you experience

136
00:10:03,600 --> 00:10:09,440
throughout the year. Well, Play Tracker is kind of like that, but for gaming. It

137
00:10:07,760 --> 00:10:13,040
works across multiple consoles and platforms. As you play games throughout

138
00:10:11,120 --> 00:10:16,880
the year, you earn experience, unlock rewards, and show it all off in a

139
00:10:14,959 --> 00:10:20,560
customizable user profile. It's more than just gamifying gaming, though.

140
00:10:18,720 --> 00:10:24,320
There's a fun social element to it as you compete against your friends,

141
00:10:22,240 --> 00:10:28,640
regardless if they prefer console or computer. Or if you prefer friendship

142
00:10:26,320 --> 00:10:31,760
over rivalry, you can create parties to share the credit for overcoming

143
00:10:30,160 --> 00:10:36,800
obstacles together. Plus, there are giveaways for when you complete challenges or you clear up your gaming

144
00:10:34,640 --> 00:10:40,720
backlog. Try Play Tracker for free today by using our link down below. If you

145
00:10:39,120 --> 00:10:45,279
guys enjoyed this video, why not check out the collab that we did with the

146
00:10:42,959 --> 00:10:49,760
Linus Torvtz, creator of the Linux kernel. He actually ended up going with

147
00:10:47,600 --> 00:10:54,000
an Intel ARC for his new computer and even talked a bit about how NVIDIA and

148
00:10:52,000 --> 00:10:56,480
Linux are getting along quite a bit better these