• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Overclocking Help: What Am I Doing Wrong?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Jems

New Member
Joined
Feb 15, 2019
Hopefully I can get some answers here. I've been banging my head against a wall trying to further my OC and still be stable.

Let's start with my system (Should be in my sig, but might as well put it here as well). These settings ran a blended test of Prime95 version 29 build 8 for 15 hours without any errors.

CPU: I9 9900k @ 5024 MHz
MOBO: MSI MEG ACE
Ram: Corsair Dominator @3350 MHz
GPU: MSI GeForce GTX 970 (from prior build)
C: NVMe Samsung SSD 970 1 TB
D: 6TB VSD. (Very slow disk!)
Cooling: EKWB Custom Water Loop (560mm rad)
Case: Thermaltake Tower 900
MediaBay: Kingwin Media and Fan Controller
OS: Win10
PSU: Corsair Hx850i

More info:
Currently OCed to 5024
vCore: set at 1.325; 1.344 under load (Prime95 with AVX-Blended Test)
Max Temps: 78C (15 hours Prime95 blended).
DRAM: 1.34-1.36
UEFI updated yesterday

CPU OC Settings:
BCLK: Set to 100 (AIDA reads 101)
Multiplier: x50
EIST, C-State, SpeedShift disabled
vCore set to 1.325 with override mode.
Ratio mode set to fixed
Ring Frequency set to run @ 300 MHz less than effective clock.
All other settings at auto​

So, I've tried pushing up the bus, tried the multiplier, neither worked. Any setting higher than what I currently have ends with a freeze or BSOD. I've increased the vCore to no avail (except for heating up the system, of course). I've played with increasing and lowering ram speeds, increasing and lowering ring speeds, and nothing. Changing the BCLK Amplitude has not affect (tried 800mV and 1100mV). I haven't played with the slew rate as it's been a long time since I've touched it, and I do not know what BCLK ORT is so I haven't adjusted it at all.

As for the crashes, I've worked through several of the errors causing BSODs but still have two that keep showing up: IRQL_NOT_LESS_OR_EQUAL (parameters showed it was a fault trying to write to an address in virtual memory (IRQL was DISPATCH_LEVEL but that's too specific for my knowledge). I've tested memory, but everything's okay (increased paging files as well). The second one is a NTOSKRNL.exe and the third one is a Hall.dll. All three of these make me think that I have to adjust RAM settings before I push higher, but any adjustment to RAM seems to either make it less stable or have no affect (although, I admit I haven't gone about this part in a logical fashion. Just make a couple adjustments out of desperation).

So, does anyone have any ideas? I find it really difficult believing I can be stable for 15 hours at 5025 and completely unstable with any setting above that no matter the vCore setting. Perhaps I'm wrong, but I find it more likely that I'm missing something or am ignorant of something since I've only overclocked three systems and don't spend a lot of time doing it.
 
You just hit the limit. I had the same thing happen could overclock i7 9700k to 4.9GHz and would not go to 5.0GHz.
 
This^ 5025 MHz isn't a bad OC for that chip, it seems to be in the range most are getting from what I've seen. As long as temps are acceptable, I'd agree that that's all there is.
 
These chips do not have a hard wall per say.

Is the board on the latest bios? If not, flash it.
Have you tired you ram at xmp settings?
 
According to silicon lottery and all the post I seen they have a hard limit.

They are not. There is thermal limit, not hard limit in any way. Once you drop temps you will see that without raising the voltage your chip will make 100-200MHz more. Now the main problem is to keep it at low enough temps ... at least on ambient temps it's not possible to OC them much above stock. Even the best water cooling (not saying about chilled water etc) will keep these chips at 90°C+ under full load.
 
They are not. There is thermal limit, not hard limit in any way. Once you drop temps you will see that without raising the voltage your chip will make 100-200MHz more. Now the main problem is to keep it at low enough temps ... at least on ambient temps it's not possible to OC them much above stock. Even the best water cooling (not saying about chilled water etc) will keep these chips at 90°C+ under full load.

Yes they have a hard limit there is one in this thread now. I can count the number of posts I've seen like this one. Temperature is always a factor with overclocking anything. A hard limit means it takes a lot more voltage for the next step. This is old overclocking news.
 
Hard limit in my eyes is, like Sandy Bridge, you were multiplier limited regardless of temp.
 
Yes they have a hard limit there is one in this thread now. I can count the number of posts I've seen like this one. Temperature is always a factor with overclocking anything. A hard limit means it takes a lot more voltage for the next step. This is old overclocking news.

That is not 'hard limit'.
That is 'wall'.
You cannot break a hard limit, but you CAN break a wall down with voltage or temp.
 
They are not. There is thermal limit, not hard limit in any way. Once you drop temps you will see that without raising the voltage your chip will make 100-200MHz more. Now the main problem is to keep it at low enough temps ... at least on ambient temps it's not possible to OC them much above stock. Even the best water cooling (not saying about chilled water etc) will keep these chips at 90°C+ under full load.
Hard limit in my eyes is, like Sandy Bridge, you were multiplier limited regardless of temp.
That is not 'hard limit'.
That is 'wall'.
You cannot break a hard limit, but you CAN break a wall down with voltage or temp.
These... :)

A hard limit means it takes a lot more voltage for the next step. This is old overclocking news.
That is old overclocking news... however, its called a voltage wall. There isn't a "HARD" limit on these CPUs. Like Mandrake said... a "hard" wall is something more of the Sandybridge and its CPU multiplier). THAT was a hard limit (though you could go more with overclocking BCLK... a hard limit is on the multiplier. One works, the next doesn't, period, no matter what voltage or cooling.

Give it enough voltage and keep it cool.. it isn't a hard (unbreakable) wall.
 
Last edited:
These... :)

That is old overclocking news... however, its called a voltage wall. There isn't a "HARD" limit on these CPUs. Give it enough voltage and keep it cool.. it isn't a hard (unbreakable) wall.

I have not seen a hard limit with enough cooling on any CPU it's all about voltage and cooling that is the ending point for all silicon chips. Name a processor that everyone reached the same exact hard limit, you can't because every processor is different in quality. Your definition of hard limit does not exist with silicon chips.
 
I have not seen a hard limit with enough cooling on any CPU it's all about voltage and cooling that is the ending point for all silicon chips. Name a processor that everyone reached the same exact hard limit, you can't because every processor is different in quality. Your definition of hard limit does not exist with silicon chips.
Pretty sure I just did (Mandrake did first) with Sandybridge. Yes, every CPU in that generation is different, however it has more of a "hard" limit to it because the multiplier would not budge no matter the cooling or voltage. BOOM. Done once you hit that multi. In fact, that is how we binned them back in the day.... that flashing cursor... 'member that gents?? THAT is a "hard" limit.

THIS, what we are seeing here is NOT a hard limit as the CPU/OP is limited by cooling and or voltage. Sandybridge CPUs, even though they would vary max multiplier, each one had a HARD limit where the multiplier wouldn't work NO MATTER what cooling or how much voltage. That is just how those worked. So, it does exist... and each CPU had its own 'hard' limit on the CPU multiplier which would limit its total speed as BCLK was extremely limited on that platform. These chips are quite different.

I mean look at the words you used... how is it a 'hard' limit when all it takes is voltage or cooling? We aren't talking subambient to make a point here either (though it proves the point pretty well). That isn't hard, that is make the adjustment and go (assuming cooling can support it). In the end, you are using the term hard limit in a manner many are not used to... at all. We don't take it to mean as you say it does...the way we are using/defining it you will find is more commonly held. I mean 4 veteran overclockers are telling you otherwise... maybe keep an open mind here?

Back to the OP.. he is limited by his voltage and cooling (or maybe the board). The point is, the chip does not have a hard wall but an artificially imposed limit due to another restriction (which isn't the CPU itself, but other things). You can split hairs all you want, but, that CPU isn't at a hard limit.
 
Last edited:
9th gen chips are a bit different than the previous gens. They need much better cooling. There are not so big voltage steps as long as CPU has low enough temp.
Voltage/temps always work both ways ... you set higher voltage to stabilize cpu but that causes higher temps which are destabilizing it. Sometimes lower voltage is helping in overclocking. Sometimes you have to play with additional voltages/settings which are not always available on cheaper motherboards. There has to be some balance and 9900K is, as I already said in other thread, reaching its limits on ambient temps. I have exactly the same. On AIO it runs at 5GHz stable ... on better cooling is passing 6GHz in benchmarks.
You can talk about "hard" wall regarding IMC or motherboard memory support. There you have clocks which you won't pass, no matter what you do.
 
Pretty sure I just did (Mandrake did first) with Sandybridge. Yes, every CPU in that generation is different, however it has more of a "hard" limit to it because the multiplier would not budge no matter the cooling or voltage. BOOM. Done once you hit that multi. In fact, that is how we binned them back in the day.... that flashing cursor... 'member that gents?? THAT is a "hard" limit.

THIS, what we are seeing here is NOT a hard limit as the CPU/OP is limited by cooling and or voltage. Sandybridge CPUs, even though they would vary max multiplier, each one had a HARD limit where the multiplier wouldn't work NO MATTER what cooling or how much voltage. That is just how those worked. So, it does exist... and each CPU had its own 'hard' limit on the CPU multiplier which would limit its total speed as BCLK was extremely limited on that platform. These chips are quite different.

I mean look at the words you used... how is it a 'hard' limit when all it takes is voltage or cooling? We aren't talking subambient to make a point here either (though it proves the point pretty well). That isn't hard, that is make the adjustment and go (assuming cooling can support it). In the end, you are using the term hard limit in a manner many are not used to... at all. We don't take it to mean as you say it does...the way we are using/defining it you will find is more commonly held. I mean 4 veteran overclockers are telling you otherwise... maybe keep an open mind here?

Back to the OP.. he is limited by his voltage and cooling (or maybe the board). The point is, the chip does not have a hard wall but an artificially imposed limit due to another restriction (which isn't the CPU itself, but other things). You can split hairs all you want, but, that CPU isn't at a hard limit.

Well I did not hit a hard limit with Sandybridge. I did hit a hard limit with i7 9700k at 4.9GHz no voltage increase would let me run at 5.0GHz.
 
Well I did not hit a hard limit with Sandybridge. I did hit a hard limit with i7 9700k at 4.9GHz no voltage increase would let me run at 5.0GHz.
Just because it wasn't reached, doesn't mean it wasn't there. SB had a multi hard wall unlike any chip I can recall. You with SB is where this guy is at... a soft limit. Call it what you will... ;)

Let's focus on the OP, please. This has been hashed out in perpetuity.. we'll agree to disagree with what a 'hard limit' is for CPUs. :)
 
Last edited:
Back