1
00:00:05,799 --> 00:00:15,240
Corsair RM series power supplies are optimized for silent operation click now

2
00:00:10,519 --> 00:00:17,240
to learn more I get asked so often lonus

3
00:00:15,240 --> 00:00:21,600
what RAM should I get what speed RAM should I get what capacity RAM should I

4
00:00:19,000 --> 00:00:24,560
get so capacity is pretty simple that's how many things you can have going at

5
00:00:23,000 --> 00:00:28,240
the same time so you look at your current system how much RAM you have how

6
00:00:26,720 --> 00:00:32,800
much your utilization is how much more would you like to be able to utilize there you go if you're not using

7
00:00:31,160 --> 00:00:36,440
professional applications like you know Da Vinci resolve and after effects and

8
00:00:34,840 --> 00:00:41,680
premere all at the same time you're probably good with anywhere from 8 to 16

9
00:00:38,640 --> 00:00:43,440
gigs of RAM for quite a while now when

10
00:00:41,680 --> 00:00:47,120
it gets into speed things get quite a bit more complicated because it's

11
00:00:45,000 --> 00:00:52,079
phenomenally confusing for people who don't know much about RAM speed there's

12
00:00:49,360 --> 00:00:59,320
frequency which will go anywhere on DDR3 from around 1333 all the way up to

13
00:00:55,480 --> 00:01:01,359
3,000 mahz and then there's latencies

14
00:00:59,320 --> 00:01:05,519
which are different okay so frequency predominantly affects the maximum

15
00:01:03,160 --> 00:01:09,360
bandwidth so that's how much data can be transferred to and from it at a time

16
00:01:07,360 --> 00:01:14,360
whereas latency predominantly affects how quickly it can respond to a request

17
00:01:12,159 --> 00:01:18,840
so these two things can affect performance in different applications in

18
00:01:16,080 --> 00:01:23,079
different ways however how much they affect things you might be surprised

19
00:01:21,040 --> 00:01:28,400
that it really isn't as much as you might expect so what we decided to do is

20
00:01:25,079 --> 00:01:30,960
we took some high performance DDR3 2.4

21
00:01:28,400 --> 00:01:38,360
GHz memory so that's 24 400 MHz memory and we ran it at 800 MHz okay we ran it

22
00:01:34,920 --> 00:01:40,840
at 1333 MHz 1600 MHz at what I'd

23
00:01:38,360 --> 00:01:47,320
consider to be a typical uh cast latency 9 so cl9 1,600 at cl8 so you tend to pay

24
00:01:45,360 --> 00:01:55,439
a little bit of a premium for a cl8 kit versus a 1600 kit 1866 and then for good

25
00:01:51,159 --> 00:01:57,960
measure at 2400 MHz then what we did is

26
00:01:55,439 --> 00:02:02,520
so we ran two games as well as cinebench just to give you some idea how it runs

27
00:01:59,320 --> 00:02:04,520
in games versus synthetics so here you

28
00:02:02,520 --> 00:02:07,600
can see the graph in front of me which basically goes okay well if you're

29
00:02:05,920 --> 00:02:11,800
looking at the same capacity RAM running all these different applications and

30
00:02:09,160 --> 00:02:15,959
different speeds you're looking at not a whole lot of difference in performance

31
00:02:13,400 --> 00:02:20,720
because what often happens with higher bandwidth so that it's higher frequency

32
00:02:17,840 --> 00:02:24,920
RAM is that it tends to sacrifice latency and then as you tighten up

33
00:02:23,040 --> 00:02:29,480
latencies you tend to sacrifice the maximum frequency to the point where

34
00:02:27,080 --> 00:02:33,360
they actually cancel each other out a lot and you end up spending more money

35
00:02:31,680 --> 00:02:38,120
for higher frequency RAM that might look better on paper but doesn't really

36
00:02:34,840 --> 00:02:40,760
deliver much in terms of a better

37
00:02:38,120 --> 00:02:45,720
performance you know number in the real world even our DDR 3800 which you

38
00:02:43,519 --> 00:02:50,760
actually can't really buy um we couldn't even find that um even our DDR 3800

39
00:02:48,920 --> 00:02:55,120
isn't really so much slower that it's going to make a difference so um with

40
00:02:53,360 --> 00:02:59,560
that said let's bring up the next graph which is the amount the number of

41
00:02:57,360 --> 00:03:04,920
dollars you're paying per FPS we took our gaming benchmarks and we graphed how

42
00:03:02,200 --> 00:03:08,239
much it's costing you per FPS because the frames per second didn't really

43
00:03:06,319 --> 00:03:13,560
change the performance didn't change much so spending more was basically just

44
00:03:10,720 --> 00:03:17,720
spending more and there are reasons why you might want better memory so uh for

45
00:03:16,040 --> 00:03:21,360
example if you like the aesthetic of a particular RAM kit versus another one

46
00:03:20,159 --> 00:03:26,200
that would be a reason to spend more money on it or if it has like cool lighting effects on it for example that

47
00:03:24,920 --> 00:03:30,480
might be another reason to spend more money but beyond that the way that I

48
00:03:28,000 --> 00:03:34,959
personally shop for money shop for money I shop for money the way I personally

49
00:03:32,159 --> 00:03:39,799
shop for RAM is I look for a brand that I trust so it should come with a

50
00:03:36,560 --> 00:03:41,599
lifetime warranty if it's worth its salt

51
00:03:39,799 --> 00:03:45,159
um it should be something that you know you can't find a whole lot of oh yeah

52
00:03:43,400 --> 00:03:48,879
this Sal all failed on me or whatever else because most of the good brands out

53
00:03:47,239 --> 00:03:51,959
there have pretty good reputations these days there's been a lot of consolidation

54
00:03:50,360 --> 00:03:56,000
and a lot of the bad RAM companies are just gone um and and that's really what

55
00:03:54,599 --> 00:03:59,879
I look for and you know who has a special edition RAM that like matches my

56
00:03:57,920 --> 00:04:06,120
motherboard really well stuff like that so there you go guys that is my take on

57
00:04:02,799 --> 00:04:09,200
memory yes there are situations where

58
00:04:06,120 --> 00:04:11,799
more memory bandwidth is needed and you

59
00:04:09,200 --> 00:04:17,359
can't possibly have enough most of these exist more like in servers not really in

60
00:04:15,200 --> 00:04:21,560
desktop environments leave a like on the video if you liked it leave a dislike if

61
00:04:19,000 --> 00:04:25,800
you disliked it and leave a comment if the results were not something you were

62
00:04:23,360 --> 00:04:39,579
expecting to see and as always don't forget to subscribe to lus Tech tips