Trouble overclocking 4790k with 1600mhz RAM- bottleneck?

[whiteboi420]

Hello

I'm getting my gaming pc back into action and want to overclock the cpu if possible. The issues I've had with this in the past were cooling and instability. Atm I have a corsair 120 rad which works fine at stock speeds, but the moment I try even a 100mhz oc and run a stress test it climbs to 100 degrees in a few seconds and the pc turns off. I can fix that with a bigger rad, but my other problem was that even in those few seconds I had to test the oc before the shut down I could see that it wasn't stable. Freezes galore. Increase the oc and windows won't boot. I read that slow ram can be a bottleneck when trying to oc the cpu. Atm a have 2x8gb of 1600mhz cl9 ddr3 ram. New ram is too big an investment for me atm, so I wanted to know if I can effectively oc my current ram to run at around 2100mhz which I read works well for oc'ing the cpu. I don't care about latency trade offs as based on my understanding I just need higher ram speeds. Applying xmp profile didn't work, it made the cpu oc even more unstable.

Thanks a lot!

[GrockleTD]

1 hour ago, whiteboi420 said:

Hello

I'm getting my gaming pc back into action and want to overclock the cpu if possible. The issues I've had with this in the past were cooling and instability. Atm I have a corsair 120 rad which works fine at stock speeds, but the moment I try even a 100mhz oc and run a stress test it climbs to 100 degrees in a few seconds and the pc turns off. I can fix that with a bigger rad, but my other problem was that even in those few seconds I had to test the oc before the shut down I could see that it wasn't stable. Freezes galore. Increase the oc and windows won't boot. I read that slow ram can be a bottleneck when trying to oc the cpu. Atm a have 2x8gb of 1600mhz cl9 ddr3 ram. New ram is too big an investment for me atm, so I wanted to know if I can effectively oc my current ram to run at around 2100mhz which I read works well for oc'ing the cpu. I don't care about latency trade offs as based on my understanding I just need higher ram speeds. Applying xmp profile didn't work, it made the cpu oc even more unstable.

Thanks a lot!

A:is your AIO pump working B: did you change the cooling curve at all C: when did you last change the thermal paste D: what voltage are you using?

[DocSwag]

Faster ram might alleviate bottlenecks, but it won't allow the cpu to OC higher.

[Psittac]

I would fix the cooling issue first, namely the things @Grockle88 mentioned followed by a better cooler if none of that works.

 

After that I would refresh your knowledge of overclocking proceedure.  My 3570k could overclock a bit on a 120mm rad so there is an issue with your clocking and/or cooling technique.

[whiteboi420]

9 minutes ago, Grockle88 said:

A:is your AIO pump working B: did you change the cooling curve at all C: when did you last change the thermal paste D: what voltage are you using?

A: It's working, I can hear it buzz slightly. B: I tried, but for some reason adjusting the fan curve in the corsair software doesn't work. I can only change the static speed. In any case I tried cranking them up to max and temp still reached 100c. I agree that's weird, it shouldn't be that bad. C: It's still got the pre-applied paste, haven't changed it since Jan 2015 when I built it. D: I don't remember exactly, but I started with normal and went all the way to the max safe rating for the 4790k and I still got freezes at 4.5ghz.

[whiteboi420]

5 minutes ago, Psittac said:

I would fix the cooling issue first, namely the things @Grockle88 mentioned followed by a better cooler if none of that works.

 

After that I would refresh your knowledge of overclocking proceedure.  My 3570k could overclock a bit on a 120mm rad so there is an issue with your clocking and/or cooling technique.

Right, the cooling is indeed a problem but only when oc'ing, which is why I wanted to clear things up with whether oc'ing the ram would help my cpu run more stable. At stock speeds it sits at max 65c under gaming load with no throttling so a new rad isn't worth it if I also have do dish out another 250 bucks for new ram as well. Can't afford it atm since I also need a new gpu : / thanks!

[Jurrunio]

1. Use thermal paste bought seperately. Preapplied paste like these dont perform well.

 

2. Make sure VCCIN is 0.5V higher than your target Vcore.

 

3. Max safe Vcore should be 1.3V, can raise to 1.35V if it still runs cool.

[Psittac]

faster memory will in no way affect the stability of a cpu clock

 

You really need to get a good application of Arctic Silver 5, Cryonaught or some other reputable interface material. 3yo preapplied tim isn't going to last.

 

That is more than likely your largest concern.

 

Then I would look into clocking procedure.  I haven't been up on my clocking, but I'll give some base advice.

First start with stock speeds and a manual vcore set to the default vcore setting (find this with a good monitoring tool like hwinfo64)

Run stress test's and monitor temps with real temp, make sure you are looking at per core temps and NOT package temps.  Most bundled monitor programs use package temp, this is a dangerous and useless number. I would actually recommend a baseline temp test using the correct software on optimized defaults so you know whats going on.  I recommend Aida64 trial for a stress test and run only the cpu.

With a baseline I would go back and make sure speed step and all of that other auto bull crap is off then set the core multiplier to all core and bump the multiplier up one.  Test, monitor and evaluate.  Keep bumping the core multiplier until the stress test fails or other stability issues, then drop the multiplier back to stable and bump the vcore. Rinse and repeat until temps are not viable.

 

There are several other settings you can use and many other ways to test for stability but this is the most straight forward way to get a raw core overclock.

 

Also there are variable autoish ways to overclock but I started years before that was a thing so this is the method I use.

[Mick Naughty]

So youre saying the system is freezing while stress testing? As in when the cpu is being completely taxed that other stuff is laggy? Sounds about right to me.

 

Need to drop voltage if you are using auto. Even for stock and for a single step in the multiplier.

 

Idk about the memory, I tend to buy it once right and not have an issue.

[whiteboi420]

12 hours ago, Psittac said:

faster memory will in no way affect the stability of a cpu clock

 

You really need to get a good application of Arctic Silver 5, Cryonaught or some other reputable interface material. 3yo preapplied tim isn't going to last.

 

That is more than likely your largest concern.

 

Then I would look into clocking procedure.  I haven't been up on my clocking, but I'll give some base advice.

First start with stock speeds and a manual vcore set to the default vcore setting (find this with a good monitoring tool like hwinfo64)

Run stress test's and monitor temps with real temp, make sure you are looking at per core temps and NOT package temps.  Most bundled monitor programs use package temp, this is a dangerous and useless number. I would actually recommend a baseline temp test using the correct software on optimized defaults so you know whats going on.  I recommend Aida64 trial for a stress test and run only the cpu.

With a baseline I would go back and make sure speed step and all of that other auto bull crap is off then set the core multiplier to all core and bump the multiplier up one.  Test, monitor and evaluate.  Keep bumping the core multiplier until the stress test fails or other stability issues, then drop the multiplier back to stable and bump the vcore. Rinse and repeat until temps are not viable.

 

There are several other settings you can use and many other ways to test for stability but this is the most straight forward way to get a raw core overclock.

 

Also there are variable autoish ways to overclock but I started years before that was a thing so this is the method I use.

Thanks for the tips, I accept I'm a noob at this so it's really helpful yeah I'd like to do it manually, I have an auto oc option but that would take the satisfaction away and it's really generous with the voltage. Plus even that wasn't stable at 4.5.

[whiteboi420]

11 hours ago, Mick Naughty said:

So youre saying the system is freezing while stress testing? As in when the cpu is being completely taxed that other stuff is laggy? Sounds about right to me.

 

Need to drop voltage if you are using auto. Even for stock and for a single step in the multiplier.

 

Idk about the memory, I tend to buy it once right and not have an issue.

Yeah pretty much, I fire up Intel burn test and it freezes momentarily a few times before it overheats and shuts down. I'm gonna do something about the cooling, either try a new thermal paste and re-sitting the block on the cpu or get a bigger aio. Tbh I want to go balls to the wall with delidding liquid metal and a 280 aio (max I can fit in my matx case), but I find it very strange that even with the auto oc option and plenty of voltage it's not stable even at +100mhz. I thought the ram was the issue but if you guys are saying it doesn't matter then I guess not. Could just be a bad chip, but I haven't heard of an unlocked cpu not even being able to take +100 on the clock speed. Really bummed out about it. Could better cooling actually improve stability?

[Mick Naughty]

Could be a few things but if its really too hot, may be the issue overall. Auto oc is even worse when it comes to voltage. 

 

Had mine hit 98c even on a custom loop. Its much better now. Think I had it around 4.8 at 1.28v. Run at 4.6 at 1.26 now. Prolly gonna turn the voltage down or oc it back up. 

[whiteboi420]

1 minute ago, Mick Naughty said:

Could be a few things but if its really too hot, may be the issue overall. Auto oc is even worse when it comes to voltage. 

 

Had mine hit 98c even on a custom loop. Its much better now. Think I had it around 4.8 at 1.28v. Run at 4.6 at 1.26 now. Prolly gonna turn the voltage down or oc it back up. 

Nice, 4.8 on a 4770k is pretty sweet. I read that's pretty much where they top out. Yeah I'm not entirely sure what the connection is, but I know you can get some pretty crazy oc's with sub zero cooling so I guess lower temps can improve stability? I'll try that, hopefully it works out. Cuz idk what else could be the problem. Thanks!

[Mick Naughty]

6 minutes ago, whiteboi420 said:

Nice, 4.8 on a 4770k is pretty sweet. I read that's pretty much where they top out. Yeah I'm not entirely sure what the connection is, but I know you can get some pretty crazy oc's with sub zero cooling so I guess lower temps can improve stability? I'll try that, hopefully it works out. Cuz idk what else could be the problem. Thanks!

That's my 4790k, I haven't really pushed the 4770k yet. They are both a little low for my needs so I just try to get what I can out of them.

[Psittac]

A cheap and easy thing to do is replace the Thermal Interface Material.  An overheating CPU will cause a ton of issues, it will throttle and make the system sluggish and act funny and even shut off to protect it's self.

 

I would save a delid until you are a little more experienced, there are things that can go wrong if you don't do it right.

 

Again start with something simple like TIM and go from there.  Then maybe upgrade cooling and once you can get any kind of stable clock and get some time tweaking the bios under your belt then think about delid.  You need to understand how your system act's instead of throwing hardware and mods at it.  Tweek test monitor and learn.

 

edit: FYI: I'm starting a clocking run on my main machine right now.  I threw numbers at it to hit 5ghz which I know isn't stress stable but figured meh, good enough.  And it was fine for the most part however gsync started glitching out and giving me black screens.  So I loaded optimized defaults for a day to make sure it wasn't something else and it proved to be the clock.  Now I'm at stock speeds with the random settings how I like it and the basic ones to manual, no speed step turbo boost or any of that, just a plain ole static multiplier, bclk and soon to be voltage. I let that run games for a couple hours and when I set a manual voltage I'll game for a couple hours with that before stress testing and monitoring.

 

remember that proper software is important.  I suggest Aida64; real temp and hwinfo64.  You should do some testing on stock to get a feel then set it all to manual with the same settings and after that change only one thing at a time.  Find the temps on manual voltage then bump the voltage and no speed increase, then bump the multiplier and test between everything.

 

edit update: Well I just tried stressing my machine and it crashes even on optimized default, so either there is something wrong with aida64 or my machine.  All that has changed since my previous clocking streak is an m.2, fresh OS and new monitor.  I don't feel like dealing with it tonight though.

[whiteboi420]

On 5/26/2018 at 2:44 AM, Psittac said:

A cheap and easy thing to do is replace the Thermal Interface Material.  An overheating CPU will cause a ton of issues, it will throttle and make the system sluggish and act funny and even shut off to protect it's self.

 

I would save a delid until you are a little more experienced, there are things that can go wrong if you don't do it right.

 

Again start with something simple like TIM and go from there.  Then maybe upgrade cooling and once you can get any kind of stable clock and get some time tweaking the bios under your belt then think about delid.  You need to understand how your system act's instead of throwing hardware and mods at it.  Tweek test monitor and learn.

 

edit: FYI: I'm starting a clocking run on my main machine right now.  I threw numbers at it to hit 5ghz which I know isn't stress stable but figured meh, good enough.  And it was fine for the most part however gsync started glitching out and giving me black screens.  So I loaded optimized defaults for a day to make sure it wasn't something else and it proved to be the clock.  Now I'm at stock speeds with the random settings how I like it and the basic ones to manual, no speed step turbo boost or any of that, just a plain ole static multiplier, bclk and soon to be voltage. I let that run games for a couple hours and when I set a manual voltage I'll game for a couple hours with that before stress testing and monitoring.

 

remember that proper software is important.  I suggest Aida64; real temp and hwinfo64.  You should do some testing on stock to get a feel then set it all to manual with the same settings and after that change only one thing at a time.  Find the temps on manual voltage then bump the voltage and no speed increase, then bump the multiplier and test between everything.

 

edit update: Well I just tried stressing my machine and it crashes even on optimized default, so either there is something wrong with aida64 or my machine.  All that has changed since my previous clocking streak is an m.2, fresh OS and new monitor.  I don't feel like dealing with it tonight though.

Did you manage to track down the problem?

 

I've been doing more research on it and it looks like cooling can actually influence stability at higher clocks. I don't get exactly how they are related, but this may be my problem as it overheats in a few seconds when oc'd even a bit. I'll try better thermal paste and reseating the block as it's currently rotated 90 degrees from how it should be based on the orientation of the logo. Could be that it's not covering the heat spreader properly and that's why it's not cooling as it should.

[Psittac]

17 hours ago, whiteboi420 said:

Did you manage to track down the problem?

 

I've been doing more research on it and it looks like cooling can actually influence stability at higher clocks. I don't get exactly how they are related, but this may be my problem as it overheats in a few seconds when oc'd even a bit. I'll try better thermal paste and reseating the block as it's currently rotated 90 degrees from how it should be based on the orientation of the logo. Could be that it's not covering the heat spreader properly and that's why it's not cooling as it should.

high temps certainly cause instability