WEBVTT

00:00:00.160 --> 00:00:08.480
since our last video on this topic a lot has changed AMD took the CPU core wars

00:00:05.359 --> 00:00:11.440
to the next level NVIDIA turned rtx on

00:00:08.480 --> 00:00:17.520
and apple dropped the new mac pro but has RAM usage changed do you wanda

00:00:15.360 --> 00:00:23.359
really need one and a half terabytes of RAM or we'll just a dab do you to find

00:00:20.240 --> 00:00:25.359
out we tested a variety of workloads

00:00:23.359 --> 00:00:29.439
everything from opening browser rams and chrome

00:00:26.400 --> 00:00:32.559
excuse me browser tabs to high

00:00:29.439 --> 00:00:35.600
resolution gaming to 4k video editing to

00:00:32.559 --> 00:00:37.680
complex flow simulation on a single 4

00:00:35.600 --> 00:00:43.040
gig stick of RAM the smallest available for ddr4 all the way up to a monstrous

00:00:41.280 --> 00:00:47.760
256 gig kit and everything in between and

00:00:46.079 --> 00:00:52.079
how much do you need well as always the answer is

00:00:50.480 --> 00:00:56.960
it depends and as always here's a segue to our

00:00:54.160 --> 00:01:01.359
sponsor xsplit xsplit makes powerful streaming pc apps for streamers vloggers

00:00:59.280 --> 00:01:04.479
and more check out their vcam app to change or blur your background without a

00:01:03.039 --> 00:01:07.799
green screen we're gonna have that linked below

00:01:15.280 --> 00:01:21.840
just like last time we're going to track page file usage with performance monitor

00:01:19.520 --> 00:01:26.799
and open up tasks until we notice activity in the page file that threshold

00:01:24.400 --> 00:01:31.360
where the system starts to swap data out of RAM and onto the boot drive is

00:01:29.200 --> 00:01:36.640
important not because you would instantly see your system slow down to a

00:01:34.000 --> 00:01:41.680
crawl but because that's where you could see a dip in system responsiveness when

00:01:39.360 --> 00:01:46.000
switching between apps or even an increase in load times for programs that

00:01:43.680 --> 00:01:49.840
have been paged to your SSD or hard drive before we begin though i want to

00:01:47.920 --> 00:01:55.920
talk through one of the most important choices for a video like this

00:01:52.079 --> 00:01:58.320
for our CPU we went with the 24 core AMD

00:01:55.920 --> 00:02:02.960
threadripper 3960x because it allowed us to test every one

00:02:00.399 --> 00:02:07.840
of our RAM configurations on the same ASUS zenith 2 extreme motherboard

00:02:05.759 --> 00:02:10.720
eliminating that as a variable we also went with that because we wanted a

00:02:09.360 --> 00:02:15.200
system that was reasonably representative of what will become

00:02:13.120 --> 00:02:19.680
available to general consumers over the next few years we could have plonked a

00:02:17.520 --> 00:02:23.680
64-core processor onto that same motherboard but

00:02:21.440 --> 00:02:27.760
that ain't going mainstream while the way things are going 24 core actually

00:02:26.640 --> 00:02:32.319
might thanks AMD we went with an rtx 2080

00:02:30.239 --> 00:02:36.640
super graphics card of course our mp600 one terabyte boot drive and a variety of

00:02:34.080 --> 00:02:40.560
Corsair vengeance memory that was chosen simply because Corsair has an incredibly

00:02:38.640 --> 00:02:44.959
broad memory lineup that covers any config we could want and they sent it

00:02:42.319 --> 00:02:50.319
over to us at four gigabytes everything is about as you would expect we're

00:02:47.040 --> 00:02:52.560
sitting at around 50 usage after logging

00:02:50.319 --> 00:02:57.440
into Windows before we've even done anything simply opening chrome and

00:02:54.800 --> 00:03:01.280
watching a 4k youtube video immediately pings the page file crossing our

00:02:59.120 --> 00:03:05.040
threshold if you avoid video the good news is that you can open six to seven

00:03:02.959 --> 00:03:10.640
tabs of various text and image based sites before you see page values and we

00:03:07.680 --> 00:03:16.879
got to about 125 tabs before chrome decided it had enough and completely

00:03:13.120 --> 00:03:19.440
pooped itself eight gigs was much better

00:03:16.879 --> 00:03:25.680
we comfortably handled three 4k videos from youtube and 27 tabs at the same

00:03:22.560 --> 00:03:28.400
time before we saw any page file usage

00:03:25.680 --> 00:03:32.720
630 tabs later we saw the system become unresponsive and laggy anytime we tried

00:03:30.560 --> 00:03:36.879
to load or reload a page maybe on Linux four gigs would be okay

00:03:34.879 --> 00:03:42.720
but eight gigs looks like the lowest we'd recommend for a good Windows 10

00:03:39.440 --> 00:03:44.799
experience with 16 and 32 gigs we stuck

00:03:42.720 --> 00:03:50.560
with the three 4k youtube videos can only watch so many videos at the same

00:03:46.319 --> 00:03:52.560
time and made it to 430 and 730 other

00:03:50.560 --> 00:03:56.799
tabs respectively before tapping that page file now chrome wasn't very happy

00:03:54.640 --> 00:04:00.959
when we hit our max tabs in either case but as long as we didn't touch anything

00:03:58.720 --> 00:04:05.360
else the computer was actually still usable it's safe to say that anything

00:04:03.040 --> 00:04:10.239
beyond this is pretty much colossal overkill on to the games shadow of the

00:04:08.000 --> 00:04:15.840
tomb raider calls for eight gigs of RAM for its minimum spec and 16 for its

00:04:13.439 --> 00:04:19.840
recommended spec rainbow six siege wants six gigs of RAM minimum and only eight

00:04:17.759 --> 00:04:24.560
gigs recommended and finally we've got cs go that needs just

00:04:22.960 --> 00:04:28.560
two gigs of RAM just two can basically run on an overclocked

00:04:27.120 --> 00:04:32.720
potato at this point all of our games were run at 1440p and

00:04:30.720 --> 00:04:37.120
the highest settings with no motion blur obviously

00:04:33.840 --> 00:04:40.160
starting at 4 gigs of RAM cs go well as

00:04:37.120 --> 00:04:42.639
advertised ran just fine at 278 frames

00:04:40.160 --> 00:04:49.600
per second with no noticeable issues but while both tomb raider and siege

00:04:45.840 --> 00:04:51.840
seemed to run well at 90 and 154 FPS

00:04:49.600 --> 00:04:57.120
respectively what's interesting here is that these great looking numbers did not

00:04:54.560 --> 00:05:02.479
translate to a great gameplay experience due to occasional frame rate dips

00:05:00.000 --> 00:05:07.600
both games reached about 60 page file usage so i think we know what to blame

00:05:04.800 --> 00:05:12.240
tomb raider jumped to 103 FPS when we switched to 8 gigs of RAM and smoothed

00:05:10.240 --> 00:05:16.160
out considerably with almost no stutters or hitching to be seen and rainbow six

00:05:14.720 --> 00:05:22.400
also improved to 155 FPS so okay not really but just

00:05:20.160 --> 00:05:26.240
like tomb raider all the stuttering was gone

00:05:23.360 --> 00:05:30.320
csgo added a whopping six frames but sarcasm aside it was a solid experience

00:05:28.320 --> 00:05:34.880
with no issues as we'd expect so eight gigs then is still plenty for

00:05:32.960 --> 00:05:39.120
gaming well here's the thing on a sanitized

00:05:36.880 --> 00:05:42.320
benchmarking Windows install nothing else is going to be running in the

00:05:40.240 --> 00:05:46.720
background that's not representative of the real world and when tomb raider and

00:05:44.479 --> 00:05:52.320
siege were running they were using seven to seven and a half gigs of RAM and the

00:05:49.280 --> 00:05:55.280
page file hit 20 usage on occasion so

00:05:52.320 --> 00:05:59.120
having chrome discord and or a music app open in the background would likely

00:05:57.280 --> 00:06:03.680
affect performance to some degree fortunately moving up to 16 gigs gives

00:06:01.600 --> 00:06:07.360
us plenty of room to work with just like there's plenty of room in your closet

00:06:04.880 --> 00:06:11.600
for our merch ltdstore.com average frame rates either stayed steady or improved

00:06:09.680 --> 00:06:15.840
and we had plenty of headroom to record our gameplay or stream in the background

00:06:13.680 --> 00:06:20.880
past this point though there was no real performance gain or loss adobe premiere

00:06:18.800 --> 00:06:24.880
cut four and eight gigs down pretty quick with the four gig setup crashing

00:06:22.880 --> 00:06:28.960
on a regular basis anytime we tried to render just a one minute test clip and

00:06:26.960 --> 00:06:33.919
as for our eight gig config it did manage to render the clip but it took a

00:06:31.600 --> 00:06:38.560
hot minute as the kids say the video that we're using is a mix of edited 4k

00:06:36.560 --> 00:06:43.360
and 8k footage from one of our past videos at 16 gigs premiere was usable

00:06:41.520 --> 00:06:47.600
with our render taking just 3 minutes and 23 seconds

00:06:44.960 --> 00:06:51.600
but scrubbing and playback was like watching an old movie picture reel

00:06:49.360 --> 00:06:57.759
remember guys this is on a Threadripper and at both one quarter and 1 8 scale

00:06:55.039 --> 00:07:02.080
so we'd say this is okay as long as you're willing to wait around to create

00:06:59.440 --> 00:07:07.280
proxies before you edit so those are lower resolution versions of your clips

00:07:04.240 --> 00:07:09.039
or if you stick with 1080p 32gb and

00:07:07.280 --> 00:07:13.599
beyond is gonna help bring your render times even lower and work with much

00:07:11.599 --> 00:07:17.680
larger and more complicated project files a lot of the usage here though is

00:07:15.759 --> 00:07:22.160
going to come down to your individual workflow needs our editors mark and a

00:07:20.319 --> 00:07:26.960
prime have slightly different machines with marks having 64 gigs of RAM and

00:07:24.639 --> 00:07:31.280
prime's having 128 both of them have premiere after effects photoshop chrome

00:07:29.120 --> 00:07:36.800
word and excel open while editing a video and depending on the video mark

00:07:33.360 --> 00:07:39.520
will see 85 to 95 usage while he edits

00:07:36.800 --> 00:07:45.840
and a prime will see 60 to 70 percent sometimes spiking to 90 utilization or

00:07:43.840 --> 00:07:51.199
gigs now up until this point in the video i would forgive you for thinking

00:07:47.520 --> 00:07:53.599
that 256 gigs is overkill for anything

00:07:51.199 --> 00:07:58.800
but while nothing that average folks are likely to do will ever touch it

00:07:55.759 --> 00:08:01.120
engineers and scientists will often have

00:07:58.800 --> 00:08:06.240
huge amounts of data to deal with to the point where such a system loadout might

00:08:03.919 --> 00:08:11.360
not even be a luxury but rather a necessity check out our RAM punisher

00:08:08.960 --> 00:08:15.759
flow simulation benchmark the more memory you have the more variables you

00:08:13.680 --> 00:08:18.800
can account for improving the airflow calculations around this cybertruck

00:08:17.919 --> 00:08:24.879
model this thing can easily suck up

00:08:21.720 --> 00:08:27.440
256 gigs of RAM and actually prefers to

00:08:24.879 --> 00:08:31.520
have more shout out to flo joe by the way for this benchmark another example

00:08:29.599 --> 00:08:36.959
is a professional composer and mixer like neil parfit who regularly has

00:08:33.839 --> 00:08:38.800
projects reach upwards of 215 gigs

00:08:36.959 --> 00:08:42.880
thanks to all of the tracks and instrument sets that he needs loaded

00:08:40.640 --> 00:08:46.959
having them in memory means instant access now one thing our test was not

00:08:45.279 --> 00:08:51.760
able to account for today was the increase in memory bandwidth as we

00:08:49.279 --> 00:08:56.800
scaled from single channel to dual and quad channel configurations that's why

00:08:54.240 --> 00:09:02.240
we focused on page file usage as our threshold rather than on the performance

00:08:59.519 --> 00:09:07.040
differences but stay subscribed because that is something that we would like to

00:09:04.000 --> 00:09:09.440
explore going forward so then in the end

00:09:07.040 --> 00:09:14.640
conclusion you're probably fine with just a solid 8 to 16 gigs of RAM for

00:09:12.080 --> 00:09:20.000
your everyday use and gaming if you do photo and video editing bump that up to

00:09:16.720 --> 00:09:22.720
32 or 64 or 128 if you do it

00:09:20.000 --> 00:09:26.240
professionally and beyond that well it's probably just bragging rights unless you

00:09:24.399 --> 00:09:31.040
know for a fact that you're doing something that benefits from it

00:09:29.040 --> 00:09:34.720
which you know we're not judging bragging rights is okay too

00:09:33.120 --> 00:09:40.399
just like you'll have bragging rights by switching to our sponsor xsplit xsplit's

00:09:38.080 --> 00:09:44.560
vcam allows you to remove replace and blur your background without a green

00:09:42.399 --> 00:09:49.040
screen it works with any webcam and there's no need for complicated lighting

00:09:46.399 --> 00:09:52.320
setups it uses ai to automatically know the difference between you and your

00:09:50.560 --> 00:09:56.240
background and allows you to customize your background to anything you want

00:09:54.160 --> 00:10:02.240
vcam works with other xsplit broadcast tools but also obs slack hangouts skype

00:10:00.000 --> 00:10:07.279
and more and they've got 24 7 support with lifetime updates it costs as little

00:10:04.160 --> 00:10:09.440
as 9.95 us for three months and lifetime

00:10:07.279 --> 00:10:13.120
licenses available for just forty dollars use offer code Linus tech tips

00:10:11.600 --> 00:10:17.120
we're gonna have that down below to get ten percent off all right if you guys

00:10:15.360 --> 00:10:22.160
like crazy RAM videos maybe check out our recent one opening up chrome tabs on

00:10:19.279 --> 00:10:26.000
a system with two terabytes of ride it's a wild

00:10:23.360 --> 00:10:30.000
and sometimes very tedious ride but don't worry we cut out the tdm it's just

00:10:28.160 --> 00:10:35.600
it's a fun video stop what do you people want everybody's

00:10:32.800 --> 00:10:35.600
messaging me