Best gaming settings for 4133mhz ram?

[Kitsan]

I game at 1080p and im trying to push my system more to being 240 fps stable in Apex and OW with less lows so I can buy a 240hz monitor.

 

I have this kit https://www.viper.patriotmemory.com/viper-steel-dram it arrived today and im really looking forward to installing it when I get home tonight but I have been told not to use XMP here: 

 

So what settings should I use to get the best FPS boost out of it? 

 

I have a relatively weak Z370P D3 motherboard right now (I am looking at replacing it next month with one of the Z390 Aorus Master boards) and a 8700K which clocks to 5ghz with daily use stability.

 

 

[HoneyBadger84]

5 minutes ago, remus243 said:

I game at 1080p and im trying to push my system more to being 240 fps stable with less lows so I can buy a 240hz monitor.

 

I have this kit https://www.viper.patriotmemory.com/viper-steel-dram it arrived today and im really looking forward to installing it when I get home tonight but I have been told not to use XMP here: 

 

So what settings should I use to get the best FPS boost out of it? 

 

I have a relatively weak Z370P D3 motherboard right now (I am looking at replacing it next month with one of the Z390 Aorus Master boards) and a 8700K which clocks to 5ghz with daily use stability.

 

 

Wouldn't worry about it too much. The RAM should be the last thing you tweak on a new setup as it'll be quite time consuming to get it as tight as possible whilst also being stable. 

 

Something in the neighborhood of 4000MHz CL17 like I'm running would serve you well.

 

Note: the Z390 AORUS Master prefers all DIMMs be populated for optimal memory overclocking. That's coming from several extreme overclockers. I like to think myself a good example of that since I'm running CL17-17-17 at 4000MHz 1.4V and I literally had no clue what I was doing when I started off overclocking this ram. Watched a Buildzoid video and a few others and I went with it. 

[Kitsan]

Hmm... I don't really want to spend another £170 to populate the other dimms tbh, is there any other board I could consider that isn't any more expensive than the master?

 

Also what CL17 timings like 17-22-22-40 or? 

[xg32]

23 minutes ago, remus243 said:

Also what CL17 timings like 17-22-22-40 or? 

for 4133, assuming you are using b-die, it's either cl17 or 18, i'd just go with 17-17-17-37 or 18-18-18-38, voltage should be around 1.4 or 1.45 and vccio/sa should be about 1.25v.

 

The aorus ultra, taichi would both work, not many choices about 4000, i'd just go with the master.

[HoneyBadger84]

46 minutes ago, remus243 said:

Hmm... I don't really want to spend another £170 to populate the other dimms tbh, is there any other board I could consider that isn't any more expensive than the master?

You don't HAVE to populate all DIMMs on the board. It just overclocks better with all DIMMs populated, due to the design of how the slots and cpu are "wired" to each other. 

 

The Master is a superior board for everything else. 

 

And I did mean 17-17-17-38 2T with a tRFC of 400. I was hoping to get CL16 stable at 4000MHz but so far haven't had any luck. 

 

There's no real reason to want speeds above 3200MHz CL13 or CL14 unless you're running things that actually need the extra raw bandwidth. You actually have better RAM performance at 3200MHz CL13 than at most higher speeds because they require looser timings unless you're willing to push well beyond the maximum recommended daily use voltage of 1.5V. 

[Kitsan]

Oh okay, your earlier comment made me feel like that Master suffered if it wasnt running 4 dimms but if it's still better than the rest then im okay with that

 

It's already been firmly established in the other thread that at 1080p gaming faster/more bandwidth...etc ram helps sorry and politely it's also not what I was asking

[Kitsan]

On 3/5/2019 at 9:30 AM, xg32 said:

for 4133, assuming you are using b-die, it's either cl17 or 18, i'd just go with 17-17-17-37 or 18-18-18-38, voltage should be around 1.4 or 1.45 and vccio/sa should be about 1.25v.

 

The aorus ultra, taichi would both work, not many choices about 4000, i'd just go with the master.

I'm a bit confused, I used the 18 18 18 38 timing and 5ghz oc and my 3d mark time spy score went down. It also doesn't fully recognise my memory at 4133mhz?

 

 

Top 2 are with the tighter timings and bottom 2 are old xmp timings

[xg32]

4 hours ago, remus243 said:

.

cpu-z is detecting it as 3733, if it won't go above 4000, u could also try cl16.

[HoneyBadger84]

On 3/5/2019 at 8:58 AM, remus243 said:

Oh okay, your earlier comment made me feel like that Master suffered if it wasnt running 4 dimms but if it's still better than the rest then im okay with that

 

It's already been firmly established in the other thread that at 1080p gaming faster/more bandwidth...etc ram helps sorry and politely it's also not what I was asking

Raw speed doesn't necessarily mean better performance. Latency is the big key, the lower the timings, the better your latency. 3200MHz RAM with very tight timings like CL12 or 1N/1T will actually perform the same or better than RAM running at 4000MHz CL18 in cpu intensive applications. 

 

And what I said was, for maximum RAM overclocking potential, you would want all 4 DIMMs populated on the AORUS Master. It is still very good at RAM OCing with only 2 DIMMs in. 

 

People telling you that "higher speed plainly equals more bandwidth" either haven't tested extensively enough to realize that statement is false, or they're just believing what they randomly read on the internet somewhere. It's just more complicated than that. 

 

Linus even has videos on this along with other YouTube tech channels. Raw speed isn't everything if you're sacrificing low timings as a result. 

 

Here's an article that talks about it further: https://www.crucial.com/usa/en/memory-performance-speed-latency

[Kitsan]

4 hours ago, xg32 said:

cpu-z is detecting it as 3733, if it won't go above 4000, u could also try cl16.

I think it might have something to do with my crappy motherboard, which settings would you recommend for CL16 please XG32?

[Kitsan]

4 hours ago, HoneyBadger84 said:

SNIP

Did you even read the thread I posted in the OP?

 

There are several benchmarks videos...etc which plainly show that at 1080p, that faster/higher bandwidth RAM helps on quite a few titles (not all), now please stop trying to push this rhetoric here. It's not what im asking, you are posting about something which isn't relative to this post but is relative to the post I linked in my OP.

[Kitsan]

CL16 has made it worse again... what am I doing wrong, it's still at 3733mhz?

[xg32]

4 hours ago, remus243 said:

.

i mean there's no way cl16 should be worse than cl18, keep in mind these are synthetic tests that does not represent actual performance, there's still a couple of minor tweaks u can do if it's stable (save current settings), the bus speed from 800mhz to 1ghz, and the nb multi from 44 to 46 or 48 (don't try 50). Keep it mind that ALL of this is for 2% extra performance when you are cpu bottlenecked on a 8700k, so it's just optimization, whatever you are running now is fine.

CL16 3733 is very good.

 

On a side note, i haven't done any synthetics at all on my build for benchmark reasons, what i get from cl14 3200 to cl16 4000 is just 2-3% fps bump in shadow of war, farcry5/nd and division 2 on the most intensive areas. That's it.

[Kitsan]

Is there another benchmark I can run to test the performance other than 3dmark?

[HoneyBadger84]

12 hours ago, remus243 said:

Did you even read the thread I posted in the OP?

 

There are several benchmarks videos...etc which plainly show that at 1080p, that faster/higher bandwidth RAM helps on quite a few titles (not all), now please stop trying to push this rhetoric here. It's not what im asking, you are posting about something which isn't relative to this post but is relative to the post I linked in my OP.

K dude not gonna argue whether or not it's valid to this conversation if you're going call it "rhetoric" lol

 

If your performance is going down that indicates instability. If you're testing for RAM settings you should be hitting it with AIDA64's Memory stress test for at least an hour, MemTest, and Prime95 large FFTs for at least an hour to check for stability. Chances are if the settings are not stable it will fail AIDA64 or prime95 in under 10 minutes. If it's near stability but still unstable it could go upwards of 30 minutes before failing. 

 

When you get a better board that will probably help a lot. Are you increasing your VCCSA and VCCIO voltages? That can help with instability as well. Setting them both at 1.2V is a pretty standard starting point. On SA don't go above 1.35-1.4, and IO about the same. 

[Kitsan]

Ahh okay thanks, i'll give memtest and prime95 a go