Solved

Overclock losing stability

Please see the rig in my sig

Sorry for the long post, I'm really trying to keep this as short as possible, please bear with me as it is a TALE.

So I have been learning to overclock and at first it was working great but the performance keeps degrading pretty quickly. All I know is it has something to do with the voltage.

At first I was using Asus' eztuning wizard in the BIOS, but that proved unstable as the voltage was far too low, (basically they just make sure it will boot) the problems at that point were that games would stutter badly once every dozen seconds or so (sound and audio stop for about .25-.5 seconds) but the system as a whole was pretty stable.

But then I got much helpful advice and started doing it the right way. I had the "CPU SVID" support on as well (even though it says it is recommended off for overclocking, but I never really saw or looked at it), but I just kept increasing the multiplier, doing tests for about an hour or so (LinX twice, Prime95 Small FFT's for about 10min, Asus RealBench for 30min) play a couple games for a few minutes each to make sure it wasn't stuttering, rinse repeat, until I got to 4.7GHz, and then I starting failing tests off the bat. I assumed it was temp since I was hitting like 87C during stress tests so I put it back to 4.6GHz and called it a day.
I also posted about that online, was told it was not temp it was voltage, and decided to give 4.7GHz another crack. I then set my voltage to 1.35V (based on an Asus overclocking guide for my board family)(found SVID support and turned it OFF, also noticed I could no longer see my CPU voltage in CPU-z) fired it up and viola, it worked, passed all tests, and was even stable for gaming.

However the next day (I also leave my computer on always) I noticed there was some stuttering, even though it was kind of minor, and my system was a little wonky (WoW kept crashing, g-sync/v-sync wasn't working, fps drops all over the place). During a WoW raid I also noticed that my CPU was performing EXACTLY THE SAME as without any OC at all. I then got so frustrated I removed the CPU overclock entirely and reset the entire BIOS back to default EXCEPT the RAM OC which I kept at 3200MHz.

Anyways I decided to give one more shot since I was low FPS again and apparently I had gotten used to the 15-20fps increase I was getting with the overclock. I set voltage back to 1.35v, set my multiplier to 4.7GHz, left SVID ON, fired everything up. boom. 100% stable, nothing crashing, everything working right, gaming performance fantastic. Saw voltage was rock solid at 1.376V in CPU-z.

Next day, computer once again on all night in idle mode. Voltage was all over the place and way lower than what it was at the day before. like 1.29-1.34v, not even close. Set the voltage to 1.375v in the BIOS, and that is now where I am at. Games are micro stuttering every few minutes, just video very minor, performance dips and fps drops at heavy CPU usage times. (moreso than usual). Im at the end of my rope here.

Do I just need to turn off my computer when I am not using it? That seems like the safest option. Or am I just missing a setting somewhere? Does that 0.001v make that much of a difference?? (1.375 vs. 1.376) I just really don't know what I should do here.

I've been mostly playing Fallout 4 and WoW but I play BF1 to test the system as well which usually just produces the results more frequently than FO4. WoW's optimization is just trash so it tells me if there is stuttering within the first minute of being on if the CPU is stable or not, since it is constantly using a single core at 80-90%.
Some stats. I really don't know what all I am doing. I have almost every settings set to auto which is part of my problem but I have no idea what I need to tweak. There are about 4 other voltage settings or so and I am afraid to mess with them.

Ai OC tuner: Manual
Asus Multicore enhancement- Manual
CPU BCLK- 100MHz
CPU core ratio limit- 47
CPU SVID Support- Auto
CPU core/cache current limit max.- Auto
Min. CPU cache ratio- Auto
Max. CPU cache ratio- Auto
CPU core/cache voltage- Manual
CPU core voltage override- 1.375
DRAM voltage- Auto
CPU VCCIO voltage- Auto
CPU system agent voltage- Auto
CPU Graphics voltage- Auto
PCH Core voltage- Auto
CPU Standby voltage- Auto
(I can get actual numbers when I get home, this is going off memory and a video of the BIOS options)
(I DID NOT set the adaptive/offset mode as Asus suggested in their guide, as I have no idea what the point is)

Prime 95 Small FFT test temps: 93C Max
LinX test temps: 89C Max
Asus RealBench temps: 81C Max
Gaming temps: Spikes up to 65C, between 45-55C normally, as low as 24C at idle.
Ambient temp: ~21C

Using these guides:
https://rog.asus.com/articles/guides/guide-overclocking-core-i7-6700k-on-the-maximus-viii-extreme/
http://www.tomshardware.com/faq/id-2103175/bios-overclocking-beginners.html

NOTE: I am not a fan of testing the system for 24hours and such. There's no way I can do that. I work 7 days a week (at least a couple hours a day, and I am absolutely terrified that something will go wrong while I am not there, and I will fry my $4,000 rig. I am also not a fan of cooking my CPU at UNGODLY temps for that long. Gaming temps get nowhere NEAR stress test temps and I am trying to preserve as much of the life of the CPU as possible, as I fully intend to upgrade to 8th Gen when it comes out and I want to sell the 6700k. If I need to do a certain stability test for 24 hours I may do it once but there is no way I am willing to do more than that!
Reply to EpIckFa1LJoN
35 answers Last reply Best Answer
More about overclock losing stability
  1. it would be helpful to post your specs.and always,always,always overclock thru the bios.
    Reply to aldan
  2. aldan said:
    it would be helpful to post your specs.and always,always,always overclock thru the bios.


    Specs are in the sig, as noted in the first line of the post, OC is done through BIOS.
    Reply to EpIckFa1LJoN
  3. sorry,i see it now.dont know your level of experience in overclocking so dont be offended here.reset your bios settings and start over.start by increasing your cpu multiplier gradually until the system becomes unstable.back the multiplier off until it becomes stable.this is your maximum oc with stock voltage.now increase multiplier until unstable again.increase cpu voltage gradually until stable.test with a reliable tool ie prime 95 or occt.sorry if this seems basic,but its how ive always done it.
    Reply to aldan
  4. aldan said:
    sorry,i see it now.dont know your level of experience in overclocking so dont be offended here.reset your bios settings and start over.start by increasing your cpu multiplier gradually until the system becomes unstable.back the multiplier off until it becomes stable.this is your maximum oc with stock voltage.now increase multiplier until unstable again.increase cpu voltage gradually until stable.test with a reliable tool ie prime 95 or occt.sorry if this seems basic,but its how ive always done it.


    Yes, I've done that. Unfortunately there is external voltage regulation going on that pushes the voltage far past the 1.2v "base" of the 6700k. From what I have researched so far it seems most older CPU's voltage is constant, whereas mine is dynamic, I guess if no one has any better advice I will try to set it to adaptive mode and set the offset to 1.38V and see what happens. Also unfortunately... the voltage regulation either with Skylake CPUs or my board is very different than of previous generations. I never once crashed or got BSOD as stated in the beginners guide, even when the voltage set was FAR off the voltage required for stability (1.376v required for 4.7GHz, had it set to 1.35V). Furthermore, I have absolutely no explanation as to why my CPU was fairly stable at 4.7GHz at 1.35V but not at all at 1.345v or 1.355v, but 100% stable at 1.376v, also why 1.375V is giving me trouble when 1.376V is perfectly fine, I am not sure how big of a difference 0.001V makes. Going even further, I have no way of regulating/monitoring the voltage with CPU SVID disabled, as it disables the reading in CPU-Z (I can check the Asus AI Suite but that reading was always different than the one given by CPU-Z so idk).

    This is just getting out of my league a bit, and I am really just afraid of tampering too much and ruining my CPU/MB. I don't really have another $700 to replace them right now. But I really can't stand stock performance on my two favorite games... (FO4 is not meant to be run at 100Hz 3440x1440 so it acts up, on stock settings I will get as low as 45fps, with the OC I hardly ever drop below 75, so its a huge difference, same thing with WoW, single core usage on stock bottlenecks the system and I drop under 50 in raids, with the stable OC, it's more like 70.) So you can kinda see my dilemma.
    Reply to EpIckFa1LJoN
  5. well you certainly come across like you know what you are doing.so your pc regulates your cpu voltage even when set on manual in the bios?if so youve got me stumped.
    Reply to aldan
  6. aldan said:
    well you certainly come across like you know what you are doing.so your pc regulates your cpu voltage even when set on manual in the bios?if so youve got me stumped.


    Yes. For whatever reason there are multiple regulators. Enabling the CPU SVID Support (enabling external voltage regulation?), shows my voltage in CPU-z, but it is dynamic and it will range from 1.29ish to 1.34ish, but yesterday with it enabled, it was at 1.376v and not budging one bit. That is why I am stumped. I did not changed one setting between the two readings but they are very drastic. Disabling SVID support, I can no longer read my voltage on CPU-z, so I can only assume it doesn't change, but I have absolutely no idea, because I can't see it.

    I may come across like I know what I'm doing. I'm a quick learner but right now I am completely in over my head. I have no idea what I am doing. I kinda fell ass backwards into a good OC voltage there at 1.376v because that's what it was reading constantly on CPU-z and I had a stable OC at 4.7GHz, but then it went to hell and the voltage was too low and all over the place.

    Maybe I just need to follow the Asus Guide to the letter.
    Reply to EpIckFa1LJoN
  7. So when you go back to stock settings, what is your RAM clocked at? you mentioned "leaving the RAM OC" which is troubling to me. XMP is a pre-set OC for your ram, but you are heavily OC the CPU which will effect RAM Clock stability or vice-versa.

    I would reset and OC your core first as you have been doing and then try it with stock ram speeds.
    Reply to TheFluffyDog
  8. TheFluffyDog said:
    So when you go back to stock settings, what is your RAM clocked at? you mentioned "leaving the RAM OC" which is troubling to me. XMP is a pre-set OC for your ram, but you are heavily OC the CPU which will effect RAM Clock stability or vice-versa.

    I would reset and OC your core first as you have been doing and then try it with stock ram speeds.


    The only thing I have changed with the RAM is the target frequency, which is what my RAM is rated at, 3200MHz. The base is 2133MHz, which means my RAM is under clocked at default. The voltages there are still set to auto. XMP is also set to manual, so it shouldn't be interfering.
    Reply to EpIckFa1LJoN
  9. aldan said:
    well you certainly come across like you know what you are doing.so your pc regulates your cpu voltage even when set on manual in the bios?if so youve got me stumped.


    There are 2 types of voltage regulation on these platforms.

    Adaptive Voltage (voltage scales with response to clock rate)
    C states (voltage drops during idle for power savings)

    I would leave voltage on manual and just leave C-States alone as these are unlikely the cause of the issue.
    Reply to TheFluffyDog
  10. EpIckFa1LJoN said:
    TheFluffyDog said:
    So when you go back to stock settings, what is your RAM clocked at? you mentioned "leaving the RAM OC" which is troubling to me. XMP is a pre-set OC for your ram, but you are heavily OC the CPU which will effect RAM Clock stability or vice-versa.

    I would reset and OC your core first as you have been doing and then try it with stock ram speeds.


    The only thing I have changed with the RAM is the target frequency, which is what my RAM is rated at, 3200MHz. The base is 2133MHz, which means my RAM is under clocked at default. The voltages there are still set to auto. XMP is also set to manual, so it shouldn't be interfering.


    you are manually change the ram clock to 3200?
    Reply to TheFluffyDog
  11. TheFluffyDog said:
    EpIckFa1LJoN said:
    TheFluffyDog said:
    So when you go back to stock settings, what is your RAM clocked at? you mentioned "leaving the RAM OC" which is troubling to me. XMP is a pre-set OC for your ram, but you are heavily OC the CPU which will effect RAM Clock stability or vice-versa.

    I would reset and OC your core first as you have been doing and then try it with stock ram speeds.


    The only thing I have changed with the RAM is the target frequency, which is what my RAM is rated at, 3200MHz. The base is 2133MHz, which means my RAM is under clocked at default. The voltages there are still set to auto. XMP is also set to manual, so it shouldn't be interfering.


    you are manually change the ram clock to 3200?


    Target DRAM Frequency - Auto -> 3200MHz

    I could change it but I notice a SERIOUS speed decrease with it at 2133 vs 3200. I mean what's the point of spending an extra $50 or so on RAM if i'm using 2/3rd of its rated speed.

    Stock on MB is 2133MHz at 1.2v, mine is 3200MHz at 1.5v, that bad?
    Reply to EpIckFa1LJoN
  12. EpIckFa1LJoN said:
    TheFluffyDog said:
    So when you go back to stock settings, what is your RAM clocked at? you mentioned "leaving the RAM OC" which is troubling to me. XMP is a pre-set OC for your ram, but you are heavily OC the CPU which will effect RAM Clock stability or vice-versa.

    I would reset and OC your core first as you have been doing and then try it with stock ram speeds.


    The only thing I have changed with the RAM is the target frequency, which is what my RAM is rated at, 3200MHz. The base is 2133MHz, which means my RAM is under clocked at default. The voltages there are still set to auto. XMP is also set to manual, so it shouldn't be interfering.


    If you raised the RAM Clock to 3200Mhz without adjusting the timings to the proper XMP profile timings, than that will definitely make it unstable. XMP adjust the frequency and the timings to theire rated numbers. AS you may have noted there is a dominator series with frequency rated from 2133-3200 with XMP and 1.2-1.35V. all of those kits come from the same assembly line and XMP just ups the speeds and at the same time increases latency to stabilize the speed. IF you are increasing the speed without increasing the latencies it will become unstable without additional voltage. Also XMP will up the voltage from the standard DDR$ voltage of 1.2V to the kits rateed voltage of 1.35V. Based on what you are saying you probably running 3200Mhz @ CL12 1.2V if XMP is not turned on.
    Reply to TheFluffyDog
  13. EpIckFa1LJoN said:
    TheFluffyDog said:
    EpIckFa1LJoN said:
    TheFluffyDog said:
    So when you go back to stock settings, what is your RAM clocked at? you mentioned "leaving the RAM OC" which is troubling to me. XMP is a pre-set OC for your ram, but you are heavily OC the CPU which will effect RAM Clock stability or vice-versa.

    I would reset and OC your core first as you have been doing and then try it with stock ram speeds.


    The only thing I have changed with the RAM is the target frequency, which is what my RAM is rated at, 3200MHz. The base is 2133MHz, which means my RAM is under clocked at default. The voltages there are still set to auto. XMP is also set to manual, so it shouldn't be interfering.


    you are manually change the ram clock to 3200?


    Target DRAM Frequency - Auto -> 3200MHz

    I could change it but I notice a SERIOUS speed decrease with it at 2133 vs 3200. I mean what's the point of spending an extra $50 or so on RAM if i'm using 2/3rd of its rated speed.

    Stock on MB is 2133MHz at 1.2v, mine is 3200MHz at 1.5v, that bad?


    There is NO WAY you can perceive the difference between 2133DDR4 and 3200. unless you are a professional overclocker running liquid nitrogen for RAM benchmarks only. Games will never benefit from that. Also, you havent adjsuted you timings so i would assume thats an issue as well. Try setting the ram to stock speeds and try your over clock again. IF its stable then turn on XMP, dont OC the ram yourself, its a whole other part of overclocking and the stability testing for RAM is much longer and needds atleast an overnight stability test.


    I really think that the ram is the issue so lets see if we can try this next
    Reply to TheFluffyDog
  14. TheFluffyDog said:
    EpIckFa1LJoN said:
    TheFluffyDog said:
    So when you go back to stock settings, what is your RAM clocked at? you mentioned "leaving the RAM OC" which is troubling to me. XMP is a pre-set OC for your ram, but you are heavily OC the CPU which will effect RAM Clock stability or vice-versa.

    I would reset and OC your core first as you have been doing and then try it with stock ram speeds.


    The only thing I have changed with the RAM is the target frequency, which is what my RAM is rated at, 3200MHz. The base is 2133MHz, which means my RAM is under clocked at default. The voltages there are still set to auto. XMP is also set to manual, so it shouldn't be interfering.


    If you raised the RAM Clock to 3200Mhz without adjusting the timings to the proper XMP profile timings, than that will definitely make it unstable. XMP adjust the frequency and the timings to theire rated numbers. AS you may have noted there is a dominator series with frequency rated from 2133-3200 with XMP and 1.2-1.35V. all of those kits come from the same assembly line and XMP just ups the speeds and at the same time increases latency to stabilize the speed. IF you are increasing the speed without increasing the latencies it will become unstable without additional voltage. Also XMP will up the voltage from the standard DDR$ voltage of 1.2V to the kits rateed voltage of 1.35V. Based on what you are saying you probably running 3200Mhz @ CL12 1.2V if XMP is not turned on.


    Ok I'll try it. I found the tech specs for my RAM. XMP 2.0, it has SPD ratings and Tested ratings... I should use the Tested stats?

    I've been running the RAM like this for quite some time is it possible this is causing my issues and not the CPU?
    Also is it possible that I have damaged my RAM by doing what I have been? It seems like it is .15V over its tested rating.

    Told you I have no idea what I'm doing :D

    So what you're saying is to set everything back to default and start over with just the CPU, and then the RAM?

    I know for a fact RAM does play a pretty heavy role in FO4. (pretty stupid) But faster and bigger RAM makes a much more significant impact in that game, so the RAM speed is important to me.
    Reply to EpIckFa1LJoN
  15. also as a side note, when you increase RAM frequencies you increase how fast is cycles. When you increase latency you increase the number of cycles required for a command to take place. By increasing both you effectively cancel some of the benefits of the faster speed. THis is why XMP works out of the box, because they speed up one thing and slow down the other. High ram speeds do benefit for things like 3D modeling, where after a command has been executed by the ram and the read cycle begins, the file size is very large and read cycle takes up the majority of the operating time. In games, there are more commands with smaller file sizes and the benefits on higher speeds are canceled by the increase in higher latencies.

    ^FYI
    Reply to TheFluffyDog
  16. EpIckFa1LJoN said:
    TheFluffyDog said:
    EpIckFa1LJoN said:
    TheFluffyDog said:
    So when you go back to stock settings, what is your RAM clocked at? you mentioned "leaving the RAM OC" which is troubling to me. XMP is a pre-set OC for your ram, but you are heavily OC the CPU which will effect RAM Clock stability or vice-versa.

    I would reset and OC your core first as you have been doing and then try it with stock ram speeds.


    The only thing I have changed with the RAM is the target frequency, which is what my RAM is rated at, 3200MHz. The base is 2133MHz, which means my RAM is under clocked at default. The voltages there are still set to auto. XMP is also set to manual, so it shouldn't be interfering.


    If you raised the RAM Clock to 3200Mhz without adjusting the timings to the proper XMP profile timings, than that will definitely make it unstable. XMP adjust the frequency and the timings to theire rated numbers. AS you may have noted there is a dominator series with frequency rated from 2133-3200 with XMP and 1.2-1.35V. all of those kits come from the same assembly line and XMP just ups the speeds and at the same time increases latency to stabilize the speed. IF you are increasing the speed without increasing the latencies it will become unstable without additional voltage. Also XMP will up the voltage from the standard DDR$ voltage of 1.2V to the kits rateed voltage of 1.35V. Based on what you are saying you probably running 3200Mhz @ CL12 1.2V if XMP is not turned on.


    Ok I'll try it. I found the tech specs for my RAM. XMP 2.0, it has SPD ratings and Tested ratings... I should use the Tested stats?

    I've been running the RAM like this for quite some time is it possible this is causing my issues and not the CPU?
    Also is it possible that I have damaged my RAM by doing what I have been? It seems like it is .15V over its tested rating.

    Told you I have no idea what I'm doing :D



    yeah, jsut set all the ram settings to stock. speed voltage everything. DO your core overclock again (which you are doing a good job on) and then after its stable, just hit the XMP button. that will raise everything accordingly to the rated specs. run your stress test again to ensure. Sometimes a high OC on the core and a high ram clock can interfere a little.
    Reply to TheFluffyDog
  17. TheFluffyDog said:
    also as a side note, when you increase RAM frequencies you increase how fast is cycles. When you increase latency you increase the number of cycles required for a command to take place. By increasing both you effectively cancel some of the benefits of the faster speed. THis is why XMP works out of the box, because they speed up one thing and slow down the other. High ram speeds do benefit for things like 3D modeling, where after a command has been executed by the ram and the read cycle begins, the file size is very large and read cycle takes up the majority of the operating time. In games, there are more commands with smaller file sizes and the benefits on higher speeds are canceled by the increase in higher latencies.

    ^FYI


    Okay, so does that mean my RAM is crap for gaming? I'm replacing it when I upgrade to 8th Gen anyways... what should I get?
    Reply to EpIckFa1LJoN
  18. Ok so at stock on my RAM, the latency timings are 15-15-15-36, at 2133MHz, at 1.2v, so will I see any difference at XMP 2 with 16-18-18-36 at 3200MHz, at 1.35v? or is it moot?

    http://www.techspot.com/review/1089-fallout-4-benchmarks/page6.html
    http://wccftech.com/fallout-4-performance-heavily-influenced-by-ram-speed-according-to-report/
    Reply to EpIckFa1LJoN
  19. EpIckFa1LJoN said:
    Ok so at stock on my RAM, the latency timings are 15-15-15-36, at 2133MHz, at 1.2v, so will I see any difference at XMP 2 with 16-18-18-36 at 3200MHz, at 1.35v? or is it moot?

    http://www.techspot.com/review/1089-fallout-4-benchmarks/page6.html
    http://wccftech.com/fallout-4-performance-heavily-influenced-by-ram-speed-according-to-report/


    it is moot. Also the chip uses an integrated memory controller, so by keeping the ram speed low you may even get some extra clocks out of the core. 4.5 vs 4.7 on the core will yield much more than anyhting ram could offer. Especially DDR4 (there are difference in DDR4 operation and DDR3 both are moot in game performance, but DDR4 is already faster than DDR3 so it becomes even more moot)
    Reply to TheFluffyDog
  20. TheFluffyDog said:
    EpIckFa1LJoN said:
    Ok so at stock on my RAM, the latency timings are 15-15-15-36, at 2133MHz, at 1.2v, so will I see any difference at XMP 2 with 16-18-18-36 at 3200MHz, at 1.35v? or is it moot?

    http://www.techspot.com/review/1089-fallout-4-benchmarks/page6.html
    http://wccftech.com/fallout-4-performance-heavily-influenced-by-ram-speed-according-to-report/


    it is moot. Also the chip uses an integrated memory controller, so by keeping the ram speed low you may even get some extra clocks out of the core. 4.5 vs 4.7 on the core will yield much more than anyhting ram could offer. Especially DDR4 (there are difference in DDR4 operation and DDR3 both are moot in game performance, but DDR4 is already faster than DDR3 so it becomes even more moot)



    So should I even bother trying to OC the RAM to its tested ratings? Because as is in some parts of the game I struggle to stay much over 70. I'll set all to default and see what happens but from the benchmarks it looks like a couple hundred MHz of RAM on DDR3 have a very significant impact in FO4.
    Reply to EpIckFa1LJoN
  21. EpIckFa1LJoN said:
    TheFluffyDog said:
    also as a side note, when you increase RAM frequencies you increase how fast is cycles. When you increase latency you increase the number of cycles required for a command to take place. By increasing both you effectively cancel some of the benefits of the faster speed. THis is why XMP works out of the box, because they speed up one thing and slow down the other. High ram speeds do benefit for things like 3D modeling, where after a command has been executed by the ram and the read cycle begins, the file size is very large and read cycle takes up the majority of the operating time. In games, there are more commands with smaller file sizes and the benefits on higher speeds are canceled by the increase in higher latencies.

    ^FYI


    Okay, so does that mean my RAM is crap for gaming? I'm replacing it when I upgrade to 8th Gen anyways... what should I get?


    Your ram is fine, and is as "fast" as they say it is, its just that games will not benefit. IF you did video rendering of videos in 4k you might actually see significant scaling with an increase in ram speed (ie. 2133-3200 could yield upwards of 10% performance) and if you are a professional rendering and testing rendered scenes 10-15 times a day, this is a significant investment. Also, if you are building a PC and really getting into it, it will eventually become bragging rights. :p for me, i collect limited edition GPU's i have a 980 ti MSI Lightning and a 980 ti EVGA Classy annnnd a Power Color Devil 13 295x2. And only the classy is in my build, the rest just sit there because they are "best of the best". Dominator Ram is kinda like that. It is definitely a solid, well build, nice looking ram kit, but its really not worth any extra money in a gaming benchmark. But if you like you like it, and if you are really into it, your allowed to like it :p
    Reply to TheFluffyDog
  22. Okay then. We'll i'll try resetting that and starting from scratch. Will probably be until Friday that I know if it is even stable or not (I have raid tonight probably run it at stock and deal with the low fps for that, and I work all day tomorrow, so Thursday is the only day I will have to really work on it much..) I'll let you know how it goes. Thanks for the help.


    So for future reference... RAM speed over 2133MHz is more about bragging rights, and I should just get lower latency RAM? C15 or lower, maybe?
    Reply to EpIckFa1LJoN
  23. EpIckFa1LJoN said:
    TheFluffyDog said:
    EpIckFa1LJoN said:
    Ok so at stock on my RAM, the latency timings are 15-15-15-36, at 2133MHz, at 1.2v, so will I see any difference at XMP 2 with 16-18-18-36 at 3200MHz, at 1.35v? or is it moot?

    http://www.techspot.com/review/1089-fallout-4-benchmarks/page6.html
    http://wccftech.com/fallout-4-performance-heavily-influenced-by-ram-speed-according-to-report/


    it is moot. Also the chip uses an integrated memory controller, so by keeping the ram speed low you may even get some extra clocks out of the core. 4.5 vs 4.7 on the core will yield much more than anyhting ram could offer. Especially DDR4 (there are difference in DDR4 operation and DDR3 both are moot in game performance, but DDR4 is already faster than DDR3 so it becomes even more moot)



    So should I even bother trying to OC the RAM to its tested ratings? Because as is in some parts of the game I struggle to stay much over 70. I'll set all to default and see what happens but from the benchmarks it looks like a couple hundred MHz of RAM on DDR3 have a very significant impact in FO4.


    yeh i would just get your core clock in check first, then turn on the XMP settings. If i were you i wouldn't shoot for the moon if you dont plan on overnight stability tests. Maybe lock your voltage at 1.33 and just bump up the multiplier until it crashes. then go down one. IF it crashes again go down until it stops, and just bring the multiplier down until it stops crashing. you may not crash in the stability test but crash the next day or something, but again jsut lower the multiplier by one until it stops. for games, core is king, so dont worry about all the other overclocking stuff.
    Reply to TheFluffyDog
  24. TheFluffyDog said:
    EpIckFa1LJoN said:
    TheFluffyDog said:
    EpIckFa1LJoN said:
    Ok so at stock on my RAM, the latency timings are 15-15-15-36, at 2133MHz, at 1.2v, so will I see any difference at XMP 2 with 16-18-18-36 at 3200MHz, at 1.35v? or is it moot?

    http://www.techspot.com/review/1089-fallout-4-benchmarks/page6.html
    http://wccftech.com/fallout-4-performance-heavily-influenced-by-ram-speed-according-to-report/


    it is moot. Also the chip uses an integrated memory controller, so by keeping the ram speed low you may even get some extra clocks out of the core. 4.5 vs 4.7 on the core will yield much more than anyhting ram could offer. Especially DDR4 (there are difference in DDR4 operation and DDR3 both are moot in game performance, but DDR4 is already faster than DDR3 so it becomes even more moot)



    So should I even bother trying to OC the RAM to its tested ratings? Because as is in some parts of the game I struggle to stay much over 70. I'll set all to default and see what happens but from the benchmarks it looks like a couple hundred MHz of RAM on DDR3 have a very significant impact in FO4.


    yeh i would just get your core clock in check first, then turn on the XMP settings. If i were you i wouldn't shoot for the moon if you dont plan on overnight stability tests. Maybe lock your voltage at 1.33 and just bump up the multiplier until it crashes. then go down one. IF it crashes again go down until it stops, and just bring the multiplier down until it stops crashing. you may not crash in the stability test but crash the next day or something, but again jsut lower the multiplier by one until it stops. for games, core is king, so dont worry about all the other overclocking stuff.


    on the CPU or the RAM? Because it seems like I was able to get a great OC at 4.7GHz and I'm pretty happy there if it works. Temps are good and i'm very happy with the performance there. Even with the RAM situation how it is, I can only assume performance will increase once I reset the RAM to stock.
    Reply to EpIckFa1LJoN
  25. EpIckFa1LJoN said:
    TheFluffyDog said:
    EpIckFa1LJoN said:
    TheFluffyDog said:
    EpIckFa1LJoN said:
    Ok so at stock on my RAM, the latency timings are 15-15-15-36, at 2133MHz, at 1.2v, so will I see any difference at XMP 2 with 16-18-18-36 at 3200MHz, at 1.35v? or is it moot?

    http://www.techspot.com/review/1089-fallout-4-benchmarks/page6.html
    http://wccftech.com/fallout-4-performance-heavily-influenced-by-ram-speed-according-to-report/


    it is moot. Also the chip uses an integrated memory controller, so by keeping the ram speed low you may even get some extra clocks out of the core. 4.5 vs 4.7 on the core will yield much more than anyhting ram could offer. Especially DDR4 (there are difference in DDR4 operation and DDR3 both are moot in game performance, but DDR4 is already faster than DDR3 so it becomes even more moot)



    So should I even bother trying to OC the RAM to its tested ratings? Because as is in some parts of the game I struggle to stay much over 70. I'll set all to default and see what happens but from the benchmarks it looks like a couple hundred MHz of RAM on DDR3 have a very significant impact in FO4.


    yeh i would just get your core clock in check first, then turn on the XMP settings. If i were you i wouldn't shoot for the moon if you dont plan on overnight stability tests. Maybe lock your voltage at 1.33 and just bump up the multiplier until it crashes. then go down one. IF it crashes again go down until it stops, and just bring the multiplier down until it stops crashing. you may not crash in the stability test but crash the next day or something, but again jsut lower the multiplier by one until it stops. for games, core is king, so dont worry about all the other overclocking stuff.


    on the CPU or the RAM? Because it seems like I was able to get a great OC at 4.7GHz and I'm pretty happy there if it works. Temps are good and i'm very happy with the performance there. Even with the RAM situation how it is, I can only assume performance will increase once I reset the RAM to stock.


    this is what i would do step wise:

    Reset to stock in BIOS

    VCCIN 1.8-1.9V (keep this 0.5V above Vcore) <this is to improve voltage delivery from the MOBO to the Chip
    V-core to 1.3-1.4V (not adaptive)

    Look for your ring bus multiplier and set it to stock manually dont leave it on auto. for that chip its 40x or 4000Mhz. If you want you could even lower this to like 3.8 which may help you get a more stable over clock.(but at the very least set it to stock)

    then bring up core multiplier until it crashes. then do the step down. if you already have a good starting point with the voltage and clock rate you mentioned earlier feel free to start there, but make sure adjust VCCIN and the ring bus as i described. both of those should help with stability.

    use intel burn test to test (not the best program but definitely quick and will work for you because you dont want to over night it.) http://www.techspot.com/downloads/4965-intelburntest.html set the level to very high and run 10 cycles if it passes i would say you can move on to next ratio. if it fails set it down one and move to ram.

    When your happy, go back to bios and just turn on XMP. It will raise all the necessary numbers. then run your stress test again. At this point, give the FO4 situation, feel free to leave on XMP and bring down your multiplier so you can get a nice balanced performance profile for that title.

    Then going forward:

    if you notice a crash liek 2 days later ( i had this happen on a friends build the other day) just bring the multiplier down one. The only way to be sure about your overclock is to stress it using a program designed to stress it as soon as you do it otherwise your cant be sure if the crash was caused by something else (well you could always read the dump)
    Reply to TheFluffyDog
  26. If the ring bus is referring to the cache frequency. Mine is set to 4100 MHz, it is on auto as far as I'm aware. And right now I am using a combination of LinX, Prime95 Small FFT, and Asus RealBench to do my stress testing. Probably stop doing Prime95 since it heats up my CPU way more than the other two.
    Reply to EpIckFa1LJoN
  27. EpIckFa1LJoN said:
    If the ring bus is referring to the cache frequency. Mine is set to 4100 MHz, it is on auto as far as I'm aware. And right now I am using a combination of LinX, Prime95 Small FFT, and Asus RealBench to do my stress testing. Probably stop doing Prime95 since it heats up my CPU way more than the other two.


    yeh the cache is it. Just leave it at 4100 or 4000 but set it manually, other wise it will rise with core frequency and it wont help performance noticeably.
    Reply to TheFluffyDog
  28. TheFluffyDog said:
    EpIckFa1LJoN said:
    If the ring bus is referring to the cache frequency. Mine is set to 4100 MHz, it is on auto as far as I'm aware. And right now I am using a combination of LinX, Prime95 Small FFT, and Asus RealBench to do my stress testing. Probably stop doing Prime95 since it heats up my CPU way more than the other two.


    yeh the cache is it. Just leave it at 4100 or 4000 but set it manually, other wise it will rise with core frequency and it wont help performance noticeably.


    Okay so assuming I make it to get the core speed stable at 4.8GHz (I feel it is totally capable of it) and get the Cache to about 4500, everything is stable... Then I set the XMP?
    What do I do if that makes the system unstable? Increase the voltage?
    Reply to EpIckFa1LJoN
  29. EpIckFa1LJoN said:
    TheFluffyDog said:
    EpIckFa1LJoN said:
    If the ring bus is referring to the cache frequency. Mine is set to 4100 MHz, it is on auto as far as I'm aware. And right now I am using a combination of LinX, Prime95 Small FFT, and Asus RealBench to do my stress testing. Probably stop doing Prime95 since it heats up my CPU way more than the other two.


    yeh the cache is it. Just leave it at 4100 or 4000 but set it manually, other wise it will rise with core frequency and it wont help performance noticeably.


    Okay so assuming I make it to get the core speed stable at 4.8GHz (I feel it is totally capable of it) and get the Cache to about 4500, everything is stable... Then I set the XMP?
    What do I do if that makes the system unstable? Increase the voltage?


    If the RAM OC makes it unstable my friend, then your opening up the fight for 0.1Ghz. When your talking about holding everythign the same beside core voltage and core multiplier, usually a little more voltage can get you there, but when you start mixing all these different frequencies sometimes voltage only gets you so far. Also your already at 1.375V and i wouldn't want my chip much higher than that for a 24/7 OC. It should be safe up to 1.4V but if you need to go from 1.375-1.4V for 0.1Ghz i wouldnt do it. better off trying to get the RAM and core to agree at a lower voltage than turn down the ram and shoot for an extra 0.1.

    Now me, i would spend probabnly a good 20 hours over a few weeks. There are always places to squeeze one more 0.1Ghz out. and at the end of 2 weeks i might get an extra 0.2-0.3 from tweeking everything. But that means over night stress, and hours of settings and small benchmarks inbetween.

    I tried to ensure my advice was more for the "gamer/overclocker." now if you enjoying all this stress testing and tweaking i suggest you get a full Skylake OC guide from someone with a lot more experience than me and read about all the different settings as well as a MOBO and DDR4 guide and start really digging.
    Reply to TheFluffyDog
  30. I do not want to be a crazy overclocker. I definitely just want to get the most gaming performance out of my rig. So I'll just aim to make it stable at 4.6-4.7 rather than push it higher. I'm about to start tweaking and testing so I'll give it a few days to a week and see how it goes.
    Reply to EpIckFa1LJoN
  31. OK, so i've already run into a problem... I downloaded HWmonitor to check out voltages a little better and I see that there are two values for what I am trying to accomplish. VCORE and VID...

    When I set the voltage to override I am not sure what it is doing. And turning off SVID takes aways reading for VCORE.

    In any case I couldn't even get anything close to stable yesterday. Like I said I think I fell ass backwards into a decent overclock that only became unstable a day later....

    When I set the core voltage in my BIOS (1.38 at 4.7GHz), I saw the VCORE go up to 1.46v (aka WAAAY to dang high).

    So once again, I have absolutely no clue what I am doing.
    On the flip side... I tested it around with default values. 4.2GHz on Turbo boost, ram at default, and then I enabled XMP for my RAM, it underclocked my CPU to 4.0GHz and increased my RAM, but I noticed I got slightly worse performance in WoW but I actually got like 10 fps more in Fallout... So if I can get all this figured out that would be great... It doesn't help that I have no idea how to navigate my BIOS and set up the RAM manually so it doesn't interfere with the CPU by enabling the XMP.
    Reply to EpIckFa1LJoN
  32. EpIckFa1LJoN said:
    OK, so i've already run into a problem... I downloaded HWmonitor to check out voltages a little better and I see that there are two values for what I am trying to accomplish. VCORE and VID...

    When I set the voltage to override I am not sure what it is doing. And turning off SVID takes aways reading for VCORE.

    In any case I couldn't even get anything close to stable yesterday. Like I said I think I fell ass backwards into a decent overclock that only became unstable a day later....

    When I set the core voltage in my BIOS (1.38 at 4.7GHz), I saw the VCORE go up to 1.46v (aka WAAAY to dang high).

    So once again, I have absolutely no clue what I am doing.
    On the flip side... I tested it around with default values. 4.2GHz on Turbo boost, ram at default, and then I enabled XMP for my RAM, it underclocked my CPU to 4.0GHz and increased my RAM, but I noticed I got slightly worse performance in WoW but I actually got like 10 fps more in Fallout... So if I can get all this figured out that would be great... It doesn't help that I have no idea how to navigate my BIOS and set up the RAM manually so it doesn't interfere with the CPU by enabling the XMP.


    VID is the set voltage from BIOS. VCORE is the actual supplied voltage. Under heavy load with intense OC, VCORE will usually dip. You can use the LLC function in the power settings to enable a boost in supply voltage under heavy load. its another one of those 4.6 vs 4.7 type tunes.

    I would say keave SVID off as it is likely to cause changes in voltages based on load. How about we try something more modest like a 4.5Ghz OC.

    try setting VCORE in BIOS to 1.35V and leave it in overide or manual. bump up the multiplier to 45. and fix cache @ 4.1 (stock). Then use a x45 multiplier and give it a stress test.

    I wouldn't worry too much about voltage monitoring with a setting like 1.35V. sure its a nice bump in voltage, but its not 1.45 or somehting, so your not on the boarder of killing your CPU. Also try to find where ASUS is hiding LLC (Load Line Calibration) in the BIOS and set it to max. That should help.

    The other thing is its very possible that your chip is just a shitty overclocker. you may not be able to hit 4.6-4.7 just because the chip is in the bottom 50% of all samples. When i got my 4670k i couldnt keep 4.5-4.6 (had to settle for 4.3) and when i swapped with a friend (who only wanted a 4.0Ghz OC) his chip went right to 4.6 with 5C cooler temps on the same cooler. So he agreed to switch since i wanted that OC.
    Reply to TheFluffyDog
  33. I actually just got fed up trying to interpret wether my system needed more or less voltage, (because under voltage is the same freaking crap problems as too much voltage), I gave up and looked for a guide online. Found a video that gives generic numbers, but changes everything you said. Cache, LLC, XMP, everything. 4.6GHz core, 4.1 cache, 1.31v VIN (something like 1.34v max VCORE) RAM is solid 16-18-18-36 3200MHz. Tried tweaking those settings to no avail. Increasing the voltage screwed it up, increasing the multiplier screwed it up... lowering it back screwed it up. So I reset to default, booted, raised it, tried again, seemed stable, started stress testing. (Skylake seems infinitely more touchy than previous generations.....)

    Everything is stable and fps is about 15-20 across the board. But most importantly stable.
    Passed all initial stress tests, while staying under 70C, passed 4 hours of RealBench Last night while I was at work, let it idle overnight, running it another 4 hours this morning to be sure. If it's stable I think we have a winner. All thanks to some random asian lady tech guru.
    Reply to EpIckFa1LJoN
  34. EpIckFa1LJoN said:
    I actually just got fed up trying to interpret wether my system needed more or less voltage, (because under voltage is the same freaking crap problems as too much voltage), I gave up and looked for a guide online. Found a video that gives generic numbers, but changes everything you said. Cache, LLC, XMP, everything. 4.6GHz core, 4.1 cache, 1.31v VIN (something like 1.34v max VCORE) RAM is solid 16-18-18-36 3200MHz. Tried tweaking those settings to no avail. Increasing the voltage screwed it up, increasing the multiplier screwed it up... lowering it back screwed it up. So I reset to default, booted, raised it, tried again, seemed stable, started stress testing. (Skylake seems infinitely more touchy than previous generations.....)

    Everything is stable and fps is about 15-20 across the board. But most importantly stable.
    Passed all initial stress tests, while staying under 70C, passed 4 hours of RealBench Last night while I was at work, let it idle overnight, running it another 4 hours this morning to be sure. If it's stable I think we have a winner. All thanks to some random asian lady tech guru.


    happy to see you got the XMP and your overclock. Its always a shame when your putting one thing down to gain another. Everything always feel better when you get it balanced
    Reply to TheFluffyDog
  35. Best answer
    Yeah, but I'm just happy its balanced.. I could have even lived with 4.5GHz and no Memory OC, as long as it was stable and running better than stock. Stock speeds are just too low on these poorly optimized games. (FO4 is optimized but not for over 60Hz and not for 21:9, having to mess with a lot of code to get it to run that way, not exactly ideal. Running it stock I would never drop under 60, but I need that 21:9 and its locked at 25 on my monitor lol, so I HAVE to unlock it which introduces a whole host of physics issues, which I am sure is straining my CPU. BF1 runs perfectly fine and dandy at stock lol, GPU usage almost never drops under 100% if I have the frames uncapped)

    Anyways here are my numbers. I apologize for the crappy quality, I have yet to figure out how to program my buttons on my keyboard... its pretty annoying... lol.
    Anyways, pic:
    Reply to EpIckFa1LJoN
Ask a new question Answer

Read More

Overclocking Intel RAM CPUs Memory Asus Performance