3600 C18 or 3600 C17 Ram

[mk3624]

Trying to figure out which one to buy for the Ryzen 3000 build
16gb, same brand, etc.
The C17 one cost $50 CAD more than the C18.
Wondering if the performance and overclock ability justify the price.
Lets say 90% game 10% work.

[Rhaemond]

3600 CL18 is equal to 10 nanoseconds while 3600 CL17 is equal to 9.4444444 seconds. Since this is only a difference of a point something nano seconds, it makes no difference in the real world tbh. Go with the cheaper one. 

[nick name]

Those are both high CL timings.  Can you afford lower?  That high I would recommend isntead something like a 3200CL14 b-die kit and overclocking that faster.  

[mk3624]

2 hours ago, nick name said:

Those are both high CL timings.  Can you afford lower?  That high I would recommend isntead something like a 3200CL14 b-die kit and overclocking that faster.  

It cost roughly $70 more than the 3600 CL18 (same brand, series,etc) 16gb, $20 more than the 3600CL17. So I guess I wont be going for the 3600CL17 kit.
But i feel like the $70 should go for a better CPU than ram? 

[mk3624]

5 minutes ago, starcoaster said:

are you using ryzen? if so, your speed and latency matters. for intel, it really doesn't matter very much as those CPU's are not heavily memory-bound.

Planning to upgrade from i5-3570k to a r7 3700x in a week or so (whenever reviews are out and there is stock in Canada).

[nick name]

12 minutes ago, starcoaster said:

okay yeah memory will be important to you then. To be honest, ryzen has diminishing returns on anything over about 3200mhz, so i'd personally start there and look for the kit with the tightest timings i could find.

 

3200 CL14

No, claiming diminishing returns after 3200MHz isn't a fair assessment without qualification.  There are definite gains past 3200MHz.  And my RAM kit at 3600MHz  14-15-14-14 timings out perform 3200MHz 12-13-12-12 timings.  There is plenty of performance to be had going past 3200MHz.  You're not just picking at bones.

[nick name]

24 minutes ago, mk3624 said:

It cost roughly $70 more than the 3600 CL18 (same brand, series,etc) 16gb, $20 more than the 3600CL17. So I guess I wont be going for the 3600CL17 kit.
But i feel like the $70 should go for a better CPU than ram? 

If the price difference is that much then spending more on a CPU or GPU makes sense to me too.  Wait, that's Canadian doll hairs.  Ok.  So if someone was in the market for 3200CL16 RAM that wasn't b-die then I would say 3600CL17 or maybe CL18 is better.  But if you can afford it then finding Samsung b-die is worth it.  I just don't know that $70 Canadian is gonna find you a better CPU or GPU.  

[mk3624]

12 minutes ago, nick name said:

I just don't know that $70 Canadian is gonna find you a better CPU or GPU.  

True. The best bet is probably I save a slightly more for a r7 3800x instead of r7 3700x but that need 3rd party reviews to see if that is justifiable.
I guess in the end I will just have to wait for full reviews in 2~3 days before I do anything.

[melete]

33 minutes ago, nick name said:

No, claiming diminishing returns after 3200MHz isn't a fair assessment without qualification.  There are definite gains past 3200MHz.  And my RAM kit at 3600MHz  14-15-14-14 timings out perform 3200MHz 12-13-12-12 timings.  There is plenty of performance to be had going past 3200MHz.  You're not just picking at bones.

There's diminishing returns on price anyway, a B die kit is 50% more expensive in many cases. But you're right, everything needs qualification. Performance with RAM (or any CPU hardware) is always relative to what application you're throwing at the RAM, differences in synthetic benchmarks like Geekbench may not actually reflect real-world performance gains depending on the use case.

[nick name]

9 hours ago, melete said:

There's diminishing returns on price anyway, a B die kit is 50% more expensive in many cases. But you're right, everything needs qualification. Performance with RAM (or any CPU hardware) is always relative to what application you're throwing at the RAM, differences in synthetic benchmarks like Geekbench may not actually reflect real-world performance gains depending on the use case.

I completely agree that there are imperceptible improvements when moving up in speed or going lower in timings and that you need to measure them to be noticed.  

[Mathieu9836]

Try to OC the cheaper kit to 3600 CL17, it will probably do it with without more voltage, if it does then try 1.36 volts up to 1.4