Placeholder Image

字幕表 動画を再生する

  • What's the guy's just two cents here.

  • We're gonna talk about RAM once again, because if you haven't seen our first video leading to this one, I highly recommend that you pause this video and go and watch the 1st 1 which was specifically does ram matter for gaming computers.

  • What we did in that particular video as we tested single channel and dual channel with different capacity configurations.

  • So force for eight and 16 both single and dual channel on a few different titles that we figured would have different types of ram utilization.

  • And the results were not surprising to me.

  • They were surprising to fill end to a lot of you in the comments were kind of surprised as well.

  • So if you haven't seen part one technically of this little Ram Siri's please go and watch that first.

  • But today we're gonna talk specifically about RAM speed.

  • Not to be confused, Ramstein.

  • That's something different entirely.

  • Building upon their highly popular eight series cases, the new H to 10 I 5 10 I 5 10 elite and 7 10 I from NZX T offer a sleek yet functional chassis for PC and busiest ample interior space and air flow make.

  • Th Siri is the perfect choice for both air cooling as well as water cooling, while the intelligent features offer integrated temperature monitoring and lighting control seamlessly.

  • To learn more about the H.

  • Siri's chassis from NZX tea for your next build, head to nzx dot com.

  • So we didn't do any different configurations in terms of channel or capacity because we already talked about that, obviously.

  • So for testing today we had 16 gigabytes of G skill, tried, etc.

  • And this was specifically the part number.

  • Um, where is a partner on this thing?

  • F 43866 C 18 Q Dash 32 g e T z k w Basically, it was a 32 gigabyte kit where comes with four of these, but these are 3800 and 66 megahertz damage, which is kind of funny because we're talking 101 3100 versus 1 33 base block, right?

  • So Ram is just like a CPU in the sense that it's in a sick and what that means.

  • It has an operating speed and operating frequency and all that sort of stuff.

  • But what confuses a lot people regarding memory, because I think a lot of people just start discussing like, Well, how much ram do we need?

  • 24 8 16 32 But that's only a small piece of the puzzle when it comes to rent.

  • Now, when it comes to gaming computers, obviously the most important piece in your system is the graphics card that's responsible for handling all the post processing and all the graphics prettiness of your game.

  • But your CPU people often we'll talk about your gravis card could potentially bottleneck.

  • Your CPU are not considering how RAM is also playing into that whole equation.

  • So we did.

  • Today is we took 16 gigabytes of the G skill, and even though it's rated at 38 66 we ran it at 4000 at its member, its stock timings, which in this particular set, have another yes, seal 18 1919 35 which is not very tight.

  • This is the same ram that we actually used this exact set.

  • Two more room over there.

  • No, over there on the computer.

  • This is what we use when we're doing the world over clocking challenge with Steve, uh, gamers nexus and we actually ran these all the way down to, like, seal 11.

  • Extremely tight.

  • But what we did was we just took the factory timings, and then we changed the actual ram speed.

  • So what we did was we did what, Five different speeds.

  • I think it was four different speeds.

  • So we did 4000 that we did 3200 which we've kind of find to be the sweet spot.

  • A lot of Ram kids.

  • Now we're coming about 3200 megahertz, cause it's gotten a lot cheaper to manufacture rams.

  • You're getting more speed out of it.

  • We also then tested 2600 not 26 66 like we did in the previous video and then 2133 which is actually the DDR four base speed for memory.

  • I think technically it's 2400 now, but 2133 is where DDR four pretty much bottoms out in terms of its base clock.

  • So we did that particular test.

  • Um and then when we also did, was a comparison between the fastest memory speed at 14 40 versus the slowest memory speed at 14.

  • 40 and we tested across the board with 10.

  • 80 p at all those intervals Because again, we're looking for bottle necking here with the CPU and how Ram speed specifically equates to those particular titles.

  • Now, we had a slight title change on this one.

  • We went we got rid of doom because doom will run on a freaking raspberry pi Practically not really bled close four gigabytes at 2133.

  • Single channel was still 200 maker or 200 acres.

  • Yeah, 200 megahertz for the game, I guess.

  • 200 mega frames a second.

  • So we replaced that title instead with Ghost re Kon wildlands, which is a much larger type of game with a much more intensive graphics engine as well as more CPU included in that particular title.

  • Whether that we still use will the war craft floating over over Mar which, as you guys know, the major cities condemn finally become a bit of a bottleneck.

  • We found what was the lowest F.

  • P.

  • S portion as we were kind of flying around and then we used that as a static shot, no movement just to see what happens with the F.

  • P s Wow It's kind of hard to benchmark because there's no track is never the same players online.

  • They're never in the same place.

  • So we had to kind of let it run for a while on the average out, the F P s there.

  • But other than that, everything else was at maximum settings.

  • And I guess we'll just go and show the slides just like last time, and we'll talk about it.

  • We found some interesting results in this one, so there's definitely since I talked about so.

  • Sentiments are 20 doesn't really care about memory speeds, and the only reason we threw that woman in there is we needed some sort of a control that was just seep you.

  • Now Blender might have showed a bit of a difference, but this is still more or less how it pertains to gaming ATT least in this discussion.

  • And our 20 was just put in there sort of a sanity check now battle Frazar offer while buffet.

  • The latest expansion 4000 megahertz gave us 140 f.

  • P s average, but as we drop the memory down to 3200 we saw a three f ps increase and from the minimum to the maximum and the average, that increase was across the board.

  • So it wasn't just like a margin of error.

  • But once we drop down to 2600 we lost quite a bit.

  • F p s down to 130 average frames per second.

  • Again, this is at 14.

  • 40.

  • We don't do 10 80 in war the workout, because literally it made zero difference to the F.

  • B s.

  • But then at 2133 we dropped all the way down to 119.

  • Those results aren't really that surprising.

  • If you watched our previous video where we did four gigabytes at 133 megahertz, we saw a significant drop in performance with RAM speed as well, because we know that this is a very CPU intensive title.

  • I think a good title to test that there was a way to actually measure it would be Minecraft, cause that's all compute.

  • And then GPU is just being used for, like, some shade er's and stuff.

  • But it's very basic.

  • It's all CPU intensive, so that would have been interesting to talk about far cry five.

  • That was just a linear scaling as speed was faster.

  • So we're the F.

  • B s 4150 average attended P at 1 41 all the way down to 1 27 and then 1 17 That's like that's exactly what you would kind of expect.

  • But that's not always the case, because if we look at wild lands that gave us the same weird result as BF a where 4000 we got 129 at 3200 we got the same score at 1 29 were being baking up a little bit.

  • But then, once we drop a low 3200 then we saw a linear drop off of 1 17 at 2600 down to 109 at 2133.

  • It looks like the honest, sweet spot, which a lot of people have said, and I've sort of agreed with, although I'm usually willing to go down to like 262,800 on memory because you can easily over clock the memory a little bit.

  • If you don't mess with timings and save a little bit of money, money, save a lot of money, but with how inexpensive 3000 and 3200 dims are becoming right now, it looks like 3200 is the sweet spot.

  • Now, if you take a look at the previous video eight gigabyte dual channel versus 16 gigabyte dual channel, I would still say those on a budget should definitely consider taking a look at eight gigabytes with 3000 ish megahertz sticks.

  • But what about if you're not running 10.

  • 80 p?

  • Because that was the other question of pumping.

  • Her head was like, J, you've got a 28 e t I Not over clocked, by the way, just locked in terms of fan speed and power limiting all that, so wouldn't we wanted to alleviate the graphics card fluctuating, affecting our scores.

  • We only wanted the variable to be our ram speed, but we found with 14 40 p 4000 megahertz and far cry if I was 135 average f p s.

  • But at 2133 we dropped about what 19 f b s down to 116.

  • So that showed us that that particular title was definitely wanting the capacity like we showed our previous video, but also wants the speed to go along with it.

  • But if we go ahead and look at that on the flip side and we take a look at our 14 40 p test with wild lands, there were the exact same number.

  • So we actually thought that those were the same engine.

  • But based on these results, I would say they're probably not the same engine.

  • Apparently, I need to breathe my phone.

  • My watches Tell me J you need to breathe.

  • I just get really passionate about this stuff.

  • So because raM's speed or the base clock is only a part of the story when it comes to Ram, there's also Ram timings, something I've honestly never done a video about on this channel.

  • But I played with a tongue when it comes to trying to get the most three D mark performance out of our CPU.

  • Because way also found that we were getting more F.

  • P s when we become hitting hard wall that we couldn't get past tightening up.

  • The timing's not only improved improved CPU, but GPU score as well.

  • So what we decided to do was take the sweet spot, which was 3200 megahertz on our dims.

  • And then we took the 10.

  • 80 p 3200 megahertz setting and took the timings from 18 1919 39 to seal 15 15 15 32.

  • So actually think it was actually steal 15 15 1932.

  • And so what we found here was that in far cry five being the title that had the most linear scaling with speed uh, a 10 80 pieces again, we want to keep that seep.

  • You being stressed, we found that we actually increased our score from 141 average F p s across four runs to 147.

  • So we gained six f p s by simply changing.

  • Our timing's a little bit, and those are not the fastest timings we could have gotten.

  • Like I said in our over clocking stuff, we've actually run these all the way down to seal 11 with that cast timing of it, I believe 28.

  • So this waas definitely not being pushed us far as it could.

  • So here's my recommendation at the end of the day, like our ballistics memory because it's a 24 100 stick actually has timings of 16 16 16 39.

  • So just, you know, one setting or one finding one click faster than that.

  • But again, like I said, what you often find us, the slower memory sometimes has tighter timings.

  • So what I need to ask you guys now is, What do you want us to do in the future?

  • Do you want to do a video about timing where we kind of go in there and see exactly how timings affect various titles?

  • I think this one's harder to test.

  • I think there will be a bigger margin of error range when you start dealing with timings.

  • One of the reasons why I think at 4000 we saw a weird increase in performance at 3200 versus 4000 is because we were pushing this room beyond its speck at that point.

  • So there could have been some air correction happening, no way to really quantify that.

  • But we did see a couple of those tests where 7200 either yielded a slight improvement or no change at all, and then under 32.

  • It just started to go bonkers, but yes.

  • Oh, my recommendation regarding memory is right around that 6200 megahertz, even with loose timings in dual channel is where you should be shopping.

  • And that would be the same thing for eight gigabytes in dual channel.

  • I think you would see the exact same type of linear scaling.

  • So if you guys have any good ideas for titles that we should test, um, I I showed we kind of did a slight title change by getting doom out of here because there was 00 impact whatsoever with timing and doom.

  • But I'd like to continue investigating Ram timings because I think everyone puts all the emphasis on GPU and CPU, but they forget that to get every ounce of performance out of your system.

  • If you pair it with terrible Ram or you even have that RAM slot, you don't have that gap between you're in.

  • A lot of people last video were like, Oh, I don't know if my rams and the right slot well, one Artie FM or read the freaking manual, and then the other thing you could look at is if you have four slots There should be a gap between your Ram sticks.

  • They're next to each other, and you got to empty Ram sticks.

  • Welcome to single channel, my friend, and gives a performance.

  • So you guys are enjoying these videos?

  • Give us some other topics you think we should delve into.

  • I'm enjoying doing this type of topic again.

  • We haven't done a lot of this exploratory explorer tive Explorer.

  • Welcome back.

  • No, I have, like a eyebrow hair, like in from my eye.

  • So I was, like, secret.

  • A slide past that flies past the dog, exerted chases, flies okay, anyway.

What's the guy's just two cents here.

字幕と単語

ワンタップで英和辞典検索 単語をクリックすると、意味が表示されます

B1 中級

メモリ速度は重要なのか? (Does Memory speed matter?)

  • 1 0
    林宜悉 に公開 2021 年 01 月 14 日
動画の中の単語