my ram OC's to 4000 CAS 18, should I just leave it at 3200 CAS 14 instead?

[Space Lynx]

I paid $187 for team dark pro cas 14-14-14-31 3200 ram 2x8gb... my motherboard MSI has a Memory Try It feature... well... it is stable at 4000 cas 18-18-18, so which is better? should i stick with lower latency?

 

[Deleted member 67555]

I paid $187 for team dark pro cas 14-14-14-31 3200 ram 2x8gb... my motherboard MSI has a Memory Try It feature... well... it is stable at 4000 cas 18-18-18, so which is better? should i stick with lower latency?

Test it.
I think you will find that your CPU has a preference to what speeds the memory runs at.
That's impressive either way you go.

 

[Space Lynx]

Test it.
I think you will find that your CPU has a preference to what speeds the memory runs at.
That's impressive either way you go.

Cool. Cinebench was fine, will try AIDA64 and prime95 next if stable at 4000 cas 18 im just gonna leave at that.

 

[dj-electric]

I'd love some numbers on it

 

[Space Lynx]

I need some help guys, Prime95 shows all workers working fine except number 2. what do I do? AIDA64 was fine for 5 minutes (will test it overnight later) but I wanted to try Prime... what does this fatal error mean? everything else seems to be working fine, AIDA uses 100% usage all 6 cores... but Prime95 is only using 4 cores for some reason...

 

[Vario]

I would run it at 3200 CL14 .
https://www.techpowerup.com/reviews...Memory_Performance_Benchmark_Analysis/10.html
Performance at this is high and it is the XMP rated speed so it should ideally be stable at that level without any future unstable hijinx.

I need some help guys, Prime95 shows all workers working fine except number 2. what do I do? AIDA64 was fine for 5 minutes (will test it overnight later) but I wanted to try Prime... what does this fatal error mean? everything else seems to be working fine, AIDA uses 100% usage all 6 cores... but Prime95 is only using 4 cores for some reason...

Means you are unstable. Prime 95 is more difficult to pass than AIDA.

 

[Space Lynx]

I would run it at 3200 CL14 .
https://www.techpowerup.com/reviews...Memory_Performance_Benchmark_Analysis/10.html
Performance at this is high and it is the XMP rated speed so it should ideally be stable at that level without any future unstable hijinx.


Means you are unstable. Prime 95 is more difficult to pass than AIDA.

aye good point. i will do XMP, then run porime95 again maybe that will fix fatal error on worker 2 as well

 

[Vario]

aye good point. i will do XMP, then run porime95 again maybe that will fix fatal error on worker 2 as well

I hope so. If that doesn't work try it with stock ram to rule out the ram, and then if still failing, it is probably your vcore that needs a bump.

 

[Space Lynx]

I hope so. If that doesn't work try it with stock ram to rule out the ram, and then if still failing, it is probably your vcore that needs a bump.

XMP fixed it, no more fatal error in prime95. ill just stick with XMP 3200 14-14-14-31

temps with Kryonaut are surprisingly good, I was expecting worse than this since I didn't go liquid metal

 

[cucker tarlson]

3200 CL14 is pretty good, 4000 CL18 might give you a few more fps in some scenarios, like 5% maybe, and only in heavily CPU intensive scenes. It's your choice whether it's worth your time and effort to tweak your memory for a minor change. I'd tweak that to achieve 4000 CL18 stable, might require bumping IMC/ram voltage, your sticks are samsung b-die ones, best for achieving fastest clocks at lowest cl.
However, running xmp is perfectly fine and you're not going to see any difference like 99.9% of the time. Like I said, 3200 cl14 is plenty fast.

 

[Vario]

3200 CL14 is pretty good, 4000 CL18 might give you a few more fps in some scenarios, like 5% maybe, and only in heavily CPU intensive scenes. It's your choice whether it's worth your time and effort to tweak your memory for a minor change. I'd tweak that to achieve 4000 CL18 stable, might require bumping IMC/ram voltage, but running xmp is perfectly fine and you're not going to see any difference like 99.9% of the time. Like I said, 3200 cl14 is plenty fast.

Do you have a source for that 5%?

 

[cucker tarlson]

No I haven't seen any worthy reviews for 4GHz ram in comparison with 3200. I'm guessing it's not gonna be bigger than 5%.

 

[Space Lynx]

well I can't figure out why, but 5.2 ghz all 6 cores is stable in AIDA64, Cinebench, TimeSpy, but for some reason Prime95 gives me fatal error on worker 2, it never crashes or freezes and other works keep going just fine... 0% cpu usage on number 2 core... but the other programs its 100% usage all 6 cores.

i tried cranking voltage real high as well, 1.425v even though it only needs 1.415 same issue right away it says worker fatal error on number 2 cpu....


so do i just say screw it and run at 5 ghz 0 avx offset loadline calibration on mode 5 and 1.415v? or is this prime95 fatal error an actual dangerous thing? nothing is freezing or crashing... so im inclined to say screw prime95...

 

[cucker tarlson]

Maybe 5.2GHz with -2 avx offset ? Tested that ?

 

[Space Lynx]

Maybe 5.2GHz with -2 avx offset ? Tested that ?

yep. anything above 4.5ghz it says fata error 2... prime95 only everything else works perfect....

maybe prime95 doesnt play well with the new Redstone 4 1803 Win 10?

 

[Deleted member 67555]

I have had systems stable for over a year in everything except Prime 95.
I have had systems crash and yet be Prime 95 stable...
I don't use Prime 95 to test stability anymore...my games do that.

 

[Space Lynx]

I have had systems stable for over a year in everything except Prime 95.
I have had systems crash and yet be Prime 95 stable...
I don't use Prime 95 to test stability anymore...my games do that.

yep I said screw it. went back to 4000 ram at 18 CAS wiuth memory Try it (those benches a person earlier in this topic linked were with CAS 23 at 4000... so yeah can't really use that as an argument.

and 5ghz even at 1.365v no downclocking, temps are not breaking 77 celsius in stress tests. -Prime95 = deleted screw that crap. games are running fine, temps are fine. gtx 1080 ti is boosting to 2050 core. i'd say this is a pretty epicly fast system and prime95 can eat my shorts. liek I said prime95 was never crashing, just one of the cores would say fatal and not run but the rest ran fine, so meh all 6 cores 100% working in everything else with 0 issues.

topic can be closed. im happy with my final settings - system specs updated if anyone wants to see it all, cheers mates

 

[John Naylor]

RAM performance is very application specific with respect to whether it benefits from lower CAS or more speed. One generic yard stick is CAS x 1000 / Speed (lower is better)

14 x 1000 / 3200 = 4.375 ns
18 x 1000 / 4000 = 4.500 ns

So outta the chute, I'd lean towards the 3200, only way to know what is serving you better is to test it with the apps in question

 

[Space Lynx]

THE ULTIMATE TANK OF 2005, MOST LEGENDARY PALADIN OF 2004 AND 2005 HAS ARRIVED!!!! LET THE TANKING BEGIN!

RAM performance is very application specific with respect to whether it benefits from lower CAS or more speed. One generic yard stick is CAS x 1000 / Speed (lower is better)

14 x 1000 / 3200 = 4.375 ns
18 x 1000 / 4000 = 4.500 ns

So outta the chute, I'd lean towards the 3200, only way to know what is serving you better is to test it with the apps in question

oh sweet mama... if only I were not a tight butt with my money:

 

[DarthBaggins]

I spent that on my 4x8GB 2400 CL10 kit (Corsair Dom's)
But I would aim for a good speed with a low CL

 

[Sasqui]

RAM performance is very application specific with respect to whether it benefits from lower CAS or more speed. One generic yard stick is CAS x 1000 / Speed (lower is better)

14 x 1000 / 3200 = 4.375 ns
18 x 1000 / 4000 = 4.500 ns

So outta the chute, I'd lean towards the 3200, only way to know what is serving you better is to test it with the apps in question

I've never seen such simple math before I was just going to point out this... 3200 @ CAS 14 is better than 4000 @ CAS 19. Too bad they don't show CAS 18 but the difference wouldn't be much

 

[Space Lynx]

I've never seen such simple math before I was just going to point out this... 3200 @ CAS 14 is better than 4000 @ CAS 19. Too bad they don't show CAS 18 but the difference wouldn't be much

View attachment 99482

thats CAS-19-23-23 thats much much different than CAS 18-18-18 :/ not sure that reference can be used to be honest

 

[Sasqui]

thats CAS-19-23-23 thats much much different than CAS 18-18-18 :/ not sure that reference can be used to be honest

If you compare the other numbers, the first one is the most significant. Yea, there will be a difference but tiny. What's interesting is the difference between 1T and 2T

 

[Deleted member 67555]

thats CAS-19-23-23 thats much much different than CAS 18-18-18 :/ not sure that reference can be used to be honest

Sure it can..
Start with the MSI try it settings then go into manual settings and try setting the first at 18 (or whatever)and restart...only takes your Mobo a few minutes stuck in boot loop before it recovers

 

[Space Lynx]

If you compare the other numbers, the first one is the most significant. Yea, there will be a difference but tiny. What's interesting is the difference between 1T and 2T

my 4000 is 18-18-18-21 1T so mine is one not two

Sure it can..
Start with the MSI try it settings then go into manual settings and try setting the first at 18 (or whatever)and restart...only takes your Mobo a few minutes stuck in boot loop before it recovers

why would I do this? I am very happy/stable with current cas 18 1T 4000