well, i cant tell you what voltage you need for that. 16-19-19 is the best you see on 3600 for your IC at 1.35V. the formula is (timing/(frequenz/2))*1000= ns. you need to test yourself. if scaling where perfect we would not need to bin but could calculate all timings.
2x16GB DDR4-3200 Kit HyperX Fury
- Ersteller olan
- Erstellt am
If we can assume that I got Hynix CJR, there is something better. But sure I agree with you that my case is not the best by far.
Karhu RAM test is good but not the end to all. you see, there are different types of tests. I would rank Aida 64 stresstest light. Karhu and mem test medium. And things like GSAT heavy. i think there was something called lynpack thats a mix of GSAT and Prime 95, that would crush almost everything. Its not important for you but you should know. I can run 16-18-18-34 4400mhz 1,53V hours on karhu but the first youtube video i play has sound errors like every 10s like the mic of the host is damaged. Ever wondert what the Skew controll in the dram timings is? its not performance. just stability. the more stable you like to go the deeper you dive into how dram actually works. down to the mesh the transistors sit on. stability is relative. you can have a karhu error but never any issuses or BSOD like hell.
My rules are: focus on one stresstest that you think fits your needs.
a setting is stable until you prove the opposit.
use real world scenarios like games that are notorious for crashing or the application(s) that you are using
a stresstest cant cover 100%
be patient and try over multiple sessions.
dont force settings and try different and new things
it must be repeatable. if you start your PC next morning and it wont boot or is instable, your settings are wrong. even if they where stable before.
have a goal
Overclocking RAM is not easy and straight forward like CPU or GPU. That can be good or bad. RAM OC is the hardest of them all. Even lower settings can perform better given the right timings. you might see Aida 64 memory benchmark all over but its not real. i would use aida 64 CPU photoWorxx benchmark. its one number. it does scale with CPU frequ. like memory benchmark. its very consistent. its somewhat real world. you may like to disable HT on your CPU and only activate 6 cores. thats where the benchmark performs best. its good to see progress and what timings do to performance.
My rules are: focus on one stresstest that you think fits your needs.
a setting is stable until you prove the opposit.
use real world scenarios like games that are notorious for crashing or the application(s) that you are using
a stresstest cant cover 100%
be patient and try over multiple sessions.
dont force settings and try different and new things
it must be repeatable. if you start your PC next morning and it wont boot or is instable, your settings are wrong. even if they where stable before.
have a goal
Overclocking RAM is not easy and straight forward like CPU or GPU. That can be good or bad. RAM OC is the hardest of them all. Even lower settings can perform better given the right timings. you might see Aida 64 memory benchmark all over but its not real. i would use aida 64 CPU photoWorxx benchmark. its one number. it does scale with CPU frequ. like memory benchmark. its very consistent. its somewhat real world. you may like to disable HT on your CPU and only activate 6 cores. thats where the benchmark performs best. its good to see progress and what timings do to performance.
Zuletzt bearbeitet:
CPU 4800 rpm, rly?)
After spending number of hours unfortunate looking for the good point with HX432C16FB3K4 and CMK32GX4M2B3200C16R i turned my attention to PVS416G400C9K you noticed before. They came to me several hours ago and after quick survey i stay at 4100c17 which doesn't seems record-breaking but suits me well for the moment. Gotta say, this upgrade definitely not worth it cost, but i can console myself being upon the concept of learning by doing
Maybe later there will be a chance for me to check all of them better. And again, thanks for you advices dude!Anhänge
i would use the 1302 Bios. the 1502 is not as shit as 1401 but in my experience 1302 is the best beside 0805. just put 1,5V on it and with that timings you should be able to 4266/4300. VCCSA nad VCCIO are far to low. if you set them auto they should go up to 1,4V. that save but you can lower it later to about 1.34V
Wonder why, but looks like 3D performance falls with 4100c17cr2, lol. I usually got ~202 fps with standart 1080p preset in Furmark 1.21.1.0, now it comes around 198-199 fps. It was scaled from 190 to 202 with previous memory optimisations on current system...
you did not improve your subtimings
tRDRD_sg and tWRWR_sg can be lowered to 6 no problem
tRDWR_sg and dg should be fine on 10
tWRRD_sg 28 and _dg 23
you cant boot higher easy because only 2 DIMMs my bad.
tWR can go to 10. you can set that better with tWRPRE at 31 instead of tWR. tWR=tWRPRE-21
on 4100 you should also be able to 16-17-17-32 with 1.5V
tRDRD_sg and tWRWR_sg can be lowered to 6 no problem
tRDWR_sg and dg should be fine on 10
tWRRD_sg 28 and _dg 23
you cant boot higher easy because only 2 DIMMs my bad.
tWR can go to 10. you can set that better with tWRPRE at 31 instead of tWR. tWR=tWRPRE-21
on 4100 you should also be able to 16-17-17-32 with 1.5V
Zuletzt bearbeitet:
Ok i'll give it a try.
Ye i know that ASUS behavior. I found VCCSA/VCCIO over 1.32v may become harmful for continuous use. Currently there was no issues for me to run 4100 and below with 1.05v, what the purpose of this raise can be?
its not really stable. if you would start prime 95 with RAM use it will crash. you should at least have 1.25V on IO and SA. what ever you have read that says 1.32 may become harmfull is far outdated. teh 14nm node from intel is so good by now that you can run IO and SA well over 1.4V and it wont be harmfull. i would say even 1.5V should be fine, but you dont need so much and additional heat means lower clock headroom. 1.3V is just fine.
you may look here for some ideas what is normal. 4x8Gb use a bit more by default https://www.hardwareluxx.de/communi...c-ergebnis-thread-kein-quatschthread.1235257/
in the start post you see under the spoilers by Ergebnisse that the Names at the end of the summary is red. click on it to get the screen shots
Beitrag automatisch zusammengeführt:
you may look here for some ideas what is normal. 4x8Gb use a bit more by default https://www.hardwareluxx.de/communi...c-ergebnis-thread-kein-quatschthread.1235257/
Beitrag automatisch zusammengeführt:
in the start post you see under the spoilers by Ergebnisse that the Names at the end of the summary is red. click on it to get the screen shots
Zuletzt bearbeitet:
Unfortunately it won't boot at with 1.45v VCCSA/VCCIO 1.3v. Not tested each separately yet.
That's truth. I noticed that boot on previous saved preset is not really stable, despite of good memtest, lol.
Yep, i taken my current setup exactly from this thread, by BF2-Gamer:
4133 MHz | 17-18-18-38-320-2T | 1.408V | 1.256vIO | 1.256vSA | 8700K | ASUS MX Hero Mode1 2004 | G.Skill F4-4400C19D-16GTZKK | 11/2017 | BF2-Gamer
His timings looks too lose to improve it that agressive way, you think?
thats a hero X, so could be he cant make it better on that board. also only 1.4V. i would reset bios and just let the IO/SA auto with main timings and 4100MHz. if that helps you can go back to the subs
or because of only 2 DIMMs you cant get it down... dont know
Beitrag automatisch zusammengeführt:
or because of only 2 DIMMs you cant get it down... dont know
Compared several presets close to 4100 and noticed that everyone keep tWRWR_dr and tWRWR_dd on 9, while i sitting down on 6 lol. Hope that's was my problem
After tWRWR fix fps in Furmark raised to 205. Aida performing the same. So funny
Beitrag automatisch zusammengeführt:
After tWRWR fix fps in Furmark raised to 205. Aida performing the same. So funny
