You guys mentioning CAS latency (CL) -- good on you -- because that has traditionally been a lot more important than 400 mhz difference in RAM clocking, in terms of sim performance.   Has that changed in the last 10-15 years?  I don't know because the last time I tested it was probably in 2005.  :)

 

17 hours ago, Mace said:

You guys mentioning CAS latency (CL) -- good on you -- because that has traditionally been a lot more important than 400 mhz difference in RAM clocking, in terms of sim performance.   Has that changed in the last 10-15 years?  I don't know because the last time I tested it was probably in 2005.  🙂

 

Nope still important, that's something you need to look out for when you see fast ram at a price to good to be true, Check the latency. There are memory calculators out there that will compare different ram speeds with different timings, great for comparisons. But in all honesty you are chasing for the last few FPS you can get when you are increasing your ram speed. There is usually a better bang for your buck, think CPU and GPU first. But if your already rocking a fast CPU and GPU but with older ram then it might be worth a look.

So which one would be faster between these 2

DDR4 3200MHz with CL16

DDR4 3600MHz with CL18

 

8 hours ago, captain420 said:

So which one would be faster between these 2

DDR4 3200MHz with CL16

DDR4 3600MHz with CL18

 

In terms of primary latency (CAS) they are equal, so the RAM with higher bandwidth will be faster in some workloads.  

But their primary latency is different, how are they equal?

 

7 hours ago, Noel said:

But their primary latency is different, how are they equal?

 

You have to calculate the time to derive latency.  CAS rating is measured in clock cycles, not nanoseconds.  

You can apply the simple equation CL/MTR = primary latency 

The actual equation is more complex, but this simplified equation allows us to compare the latency of different memory products.

In this case:

16/3200 = .005 and

18/3600 = .005

so the primary latency of these two RAM modules would be the same, but because the 3600 modules run at a higher clock speed, they provide more bandwidth.  

If you were to compare say 3200 CL14 RAM with the same 3600 CL18 RAM it might be a more difficult decision, as the 3200 CL14 RAM will have lower latency, while the 3600 CL18 RAM will have higher bandwidth.  Some workloads will perform better with the lower latency RAM while others will perform better with the higher bandwidth.  

 

I see, thank you.  And vis a vis realtime applicability, in what types of conditions would 12.5% greater bandwidth manifest?  I.e., in what circumstances does the buss reach full capacity?

Can you also comment on Command Rate?  What happens to effective bandwidth when we go from TN1 to TN2?  I remember back when I was trying to squeeze every bit of bandwidth I was able to set a command rate of 1 on that memory and it remained stable.

Edited May 12, 20206 yr by Noel

So hardly much difference between the 2 types of RAM I posted, then which would you go with? Which one would be better for flight simming, if there's any significant speed performance between the 2 that is. If not, then I'll go with the 3200MHz CL16 one.

In a memory benchmark it should be around 12.5% faster, equating to 12.5% greater bandwidth.  So this begs the question with regard to flight sims, in what circumstances does is realtime bandwidth need exceed the bandwidth limit associated with 3200mHz memory?  This is the only answer that matters really because f it never hits the ceiling then 3600mHz memory will not matter one iota.   And, what tool would one use to prove this?

On 5/11/2020 at 10:08 PM, captain420 said:

So hardly much difference between the 2 types of RAM I posted, then which would you go with? Which one would be better for flight simming, if there's any significant speed performance between the 2 that is. If not, then I'll go with the 3200MHz CL16 one.

I would check out buildzoids videos on ram, he has some really good recommendations on which sticks offer the best bang for the buck, and very good explanations on ram overclocking and bandwidth.  From what Ive been watching, 3200 cl16 isnt that great of a bin, I would look for 3600 cl16.  There are still some very affordable b die kits out there, as well as micron rev e.  Based on buildzoids videos I have ordered 3 different ram kits for my 10900k build and i will test them out to see which overclocks best.  It's tough because I want 32 gigs but all of the z490 boards are daisy chain, but supposedly with better ram overclocking. So im going to test 4 dimms of rev e, 4 dimms of b die, and 2 16gb dimms of b die and see what works best. 

 

option 1 Samsung B die 32gb (2x16) kit:

https://www.newegg.com/g-skill-32gb-288-pin-ddr4-sdram/p/N82E16820232378?Item=N82E16820232378

 

Option 2 cheap fast Samsung b die 32gb, I ordered 2 2x8 kits so 4 dimms total:

https://www.newegg.com/patriot-16gb-288-pin-ddr4-sdram/p/N82E16820225142?Item=N82E16820225142

 

Option 3 cheaper fast micron rev. e 32gb,  2x 2x8 kits, 4 dimms:

https://www.newegg.com/ballistix-16gb-288-pin-ddr4-sdram/p/N82E16820164185?Item=N82E16820164185

 

Edited May 16, 20206 yr by Pilot53