RAM overclocking

[Hyomoto]

I have two test benches set up, one for Intel and one for AMD. However, between them the same RAM settings will not work. At first I thought it might be the CPU, so I set up another AMD build with a different mother board but everything else identical and ... again the RAM doesn't accept the same settings. Generally speaking, for the GPU and the CPU overclocking is pretty stable, I've never run into a situation where my bench settings failed on a different motherboard.
But RAM? It appears that not only do the brands, speeds and amount affect how RAM can be clocked, but also the motherboard.

I have to give the developers kudos, at first glance the game does not appear to have this kind of depth at all. As someone who apparently enjoys logging this type of data, the rabbit hole is quite deep and my hat is off to you.

[Soggyy#1]

I find that ASUS boards can handle a higher OC than other ones.
Maybe inspecing the specs of the board can say something about it.
my highest score is 22100



The RAM and BASE CPU overclock go hand in hand
The higher the base CPU teh higher the RAM.
If you are in Free build get teh corsaid 4700 16G
This can over clock to 5560 without failure at 1.75V

[🦊Λℚ𝓤ΛƑΛᗯҜᔕ🦊]

Originally posted by Hyomoto:

I have two test benches set up, one for Intel and one for AMD. However, between them the same RAM settings will not work. At first I thought it might be the CPU, so I set up another AMD build with a different mother board but everything else identical and ... again the RAM doesn't accept the same settings. Generally speaking, for the GPU and the CPU overclocking is pretty stable, I've never run into a situation where my bench settings failed on a different motherboard.
But RAM? It appears that not only do the brands, speeds and amount affect how RAM can be clocked, but also the motherboard.

I have to give the developers kudos, at first glance the game does not appear to have this kind of depth at all. As someone who apparently enjoys logging this type of data, the rabbit hole is quite deep and my hat is off to you.

You might be shocked to realize that this is pretty darn close to how it works in real life in regards to ram overclocking as well.

[Hyomoto]

I tried looking through the board descriptions but they don't seem to give any data on even what the max speeds they allow for is. The more I play, the more of my complaints about the game revolve around information that is incomplete or missing rather than missing features or parts. It turns out the game is actually pretty robust once you hit level 13, and I suppose I enjoy the discovery aspect that goes along with the OC, but when you find inconsistencies like this you can't consult the board specs for more info.

I also had an odd test yesterday, I believe with a Ryzen 5 1400, it gave the same bench score at 16 and 32 GB of RAM, but the 1700X gave a WILDLY different score at 16 and 32 GB. So it seems out of all the pieces in the game the 'RAM connection' is the most ... complex part of the overclock. The CPU and GPU are always stable at certain voltages and clocks as long as you have enough power and cooling. RAM on the other hand... well, it gets interesting doesn't it?

[🦊Λℚ𝓤ΛƑΛᗯҜᔕ🦊]

Originally posted by Hyomoto:

I tried looking through the board descriptions but they don't seem to give any data on even what the max speeds they allow for is. The more I play, the more of my complaints about the game revolve around information that is incomplete or missing rather than missing features or parts. It turns out the game is actually pretty robust once you hit level 13, and I suppose I enjoy the discovery aspect that goes along with the OC, but when you find inconsistencies like this you can't consult the board specs for more info.

The maximum speed is set by the processor, not the motherboard. The motherboard has almost nothing at all to do with it. With modern processors the memory controller has been intregrated into the CPU it's self for many generations now since 2011. And as such it's "Silicon Lotto" which cpu will allow higher ram than a different one. There's RNG in this game and every cpu you spawn has a different initial value in PCBS as to what it's maximum overclock and maximum ram seped will end up being.

[Brittimus]

IRT is right that the processor controls the speed, but the motherboards do have different limits to what you can set the RAM to without changing the base clock. Some will OC RAM higher than others but by changing the base clock (assuming you can) you can make them all do whatever speed you want. Not correcting anything he said because there's nothing to correct but just adding a little more to it.

And you shouldn't get different scores using different memory sizes. As long as you're using the same number of memory channels and the same speed it doesn't matter how many GB you're using. Unless they have patched that.

[Hyomoto]

In the tests I've run, different stick sizes give different results as well. Now it's possible some of that was motherboard dependent, but just as an example:
Shean 8GB @ 1.65 2496
Shean 4GB @ 1.65 2388
Shean 2GB @ 1.65 2448

Originally I was verifying these tests by running them again in the other test bench. As for the CPUs there might be a RNG seed, but I've yet to come across it. A Ryzen 5 1500X is always stable at 1.4v and 3875Mhz. Now, for my baselines I don't mess with the bus speed, so some of my personal numbers are definitely low by a few Mhz, but I've never run into a Ryzen 5 1500X that will not clock out at 1.4v and 3875Mhz. RAM on the other hand, same stick, different motherboards can give different results.

As for the RAM, different CPUs and the same motherboard give me the same results on the same stick. I'll do some more tests, but I haven't seen any RNG thus far. Perhaps it's only on the upper end.

[Brittimus]

I think I understand now what you're saying. You're saying those are the speeds you can achieve without crashing correct? In that case you've already encountered RNG, that's why they're different. My point is that it doesn't matter how many gigs of ram you have for 3DMark. Just how many memory channels you're using and the speed of your RAM. If I'm correct in what you're saying then in my experience the motherboard does not have an effect on how high you can clock RAM, that's all to do with the individual RAM sticks.

[🦊Λℚ𝓤ΛƑΛᗯҜᔕ🦊]

Originally posted by Hyomoto:

In the tests I've run, different stick sizes give different results as well. Now it's possible some of that was motherboard dependent, but just as an example:
Shean 8GB @ 1.65 2496
Shean 4GB @ 1.65 2388
Shean 2GB @ 1.65 2448

Originally I was verifying these tests by running them again in the other test bench. As for the CPUs there might be a RNG seed, but I've yet to come across it. A Ryzen 5 1500X is always stable at 1.4v and 3875Mhz. Now, for my baselines I don't mess with the bus speed, so some of my personal numbers are definitely low by a few Mhz, but I've never run into a Ryzen 5 1500X that will not clock out at 1.4v and 3875Mhz. RAM on the other hand, same stick, different motherboards can give different results.

As for the RAM, different CPUs and the same motherboard give me the same results on the same stick. I'll do some more tests, but I haven't seen any RNG thus far. Perhaps it's only on the upper end.

Are you sure all of those ram sticks are the same Mhz speeds?

Also the RNG for CPU's only comes in to play when you're trying to run ram at or above 4500+ Mhz in PCBS.

[Brittimus]

Also are you doing this testing in free build or career mode? In free build I don't think you're actually testing the same instance of a you're actually getting a new copy of the part every time you pull on from inventory. In career mode on the other hand you are using the same instance each time so you would be able to control for RNG in career mode.

[Hyomoto]

Yeah, these are all career mode tests. Free build lacks the whole "get really insane about the numbers to fill orders as cheaply as possible" that seems to drive me to do these tests :P

I would definitely believe the RNG is slated towards the high numbers, I do all my benches at 0.5v increments up to 1.5 for the CPU and 1.65 for the RAM. As for the sticks themselves though, I haven't ~yet~ noticed two same brand, same speed, same size sticks that perform differently. The memory channels thing is a good point, that would explain why 2 sticks and 4 sticks have different numbers. I have a very, very incomplete set of data on multiple sticks of RAM right now, but the full set I do have is for the 2GB Shean TITAN and 1 and 2 sticks have the same clock/voltage pairs, but 4 sticks is different.

Thanks for the tip!