LLC levels in relation to durability

[Melodist]

Hello there,

 

Do you know how LLC levels translate to durability? Which LLC levels would generally be heavily degrading for the CPU? Is there like a common good value in terms of performance to durability because I don't think running at 100% constantly is good for the CPU.

 

Haven't messed around a lot with LLC in the past due to X99 and Server generation CPUs and would appreciate some practical insight on LLC

 

[SolarNova]

LLC doesnt degrade CPU's.

[Melodist]

10 minutes ago, SolarNova said:

LLC doesnt degrade CPU's.

Why does der b8uer then say lower level LLCs are better for durability?

[Melodist]

I mean, there must be a setting that creates less overshoots vs profiles which won't compensate enough?

[SolarNova]

26 minutes ago, Melodist said:

Why does der b8uer then say lower level LLCs are better for durability?

Because under high OC's ..and extreme OC's for which Roman is known for, LLC can take an already very high voltage, even higher, thus it may cause degradation over time.

 

Now im going to make an assumption here, something i hate doing, but .. considering ur asking about LLC, its most likely ur not going to be doing any high end OC'ing, so LLC really isnt going to cause any degradation at stock or mid level OC's.

 

LLC is their to help keep OC's stable, it adds voltage as needed, and becomes more useful the higher you push the OC. And if ur the kind of person pushing an OC to the limit, ur not likely to be concerned about degradation.

 

In the end u should be monitoring CPU voltage when u OC anyway, LLC is just another setting that affects voltage. Monitor and adjust. LLC itself doesnt cause degradation, heat and voltage does. Thats what Roman is referring to.

 

Example. IIRC i have my LLC set to 'extreme' on the 3930k. and i use an offset voltage OC, together my CPU usually runs at a sustained max vcore of 1.362v, BUT on occasion can peak at 1.376v, and i've been runnign it liek that for over 6 years. I OC'ed to a temp target ( I didnt want to go above 80c) whilst also ensuring i stayed under 1.4v. A 3930k can run hotter and can run at a higher voltage, so i still have room for a higher OC should i need it in the future.

[Melodist]

6 minutes ago, SolarNova said:

Because under high OC's ..and extreme OC's for which Roman is known for, LLC can take an already very high voltage, even higher, thus it may cause degradation over time.

 

Now im going to make an assumption here, something i hate doing, but .. considering ur asking about LLC, its most likely ur not going to be doing any high end OC'ing, so LLC really isnt going to cause any degradation at stock or mid level OC's.

 

LLC is their to help keep OC's stable, it adds voltage as needed, and becomes more useful the higher you push the OC. And if ur the kind of person pushing an OC to the limit, ur not likely to be concerned about degradation.

 

In the end u should be monitoring CPU voltage when u OC anyway, LLC is just another setting that affects voltage. Monitor and adjust. LLC itself doesnt cause degradation, heat and voltage does. Thats what Roman is referring to.

 

Example. IIRC i have my LLC set to 'extreme' on the 3930k. and i use an offset voltage OC, together my CPU usually runs at a sustained max vcore of 1.362v, BUT on occasion can peak at 1.376v, and i've been runnign it liek that for over 5 years. I OC'ed to a temp target ( I didnt want to go above 80c) whilst also ensuring i stayed under 1.4v. A 3930k can run hotter and can run at a higher voltage, so i still have room for a higher OC should i need it in the future.

So I guess undervolting and running a higher LLC is better than running a higher voltage and lower LLC? Because I always see people dialing in a high voltage with a low LLC which just makes the chip run hot and unstable from the getgo, making overshoots even worse, which would explain why people struggle so much with the 9900k because their conventional way of overclockling by dialing in a high voltage at low LLC level doesn't work?

[SolarNova]

Just now, Melodist said:

So I guess undervolting and running a higher LLC is better than running a higher voltage and lower LLC? Because I always see people dialing in a high voltage with a low LLC which just makes the chip run hot and unstable from the getgo, making overshoots even worse, which would explain why people struggle so much with the 9900k because their conventional way of overclockling by dialing in a high voltage at low LLC level doesn't work?

I honestly wouldnt know how the 9900k reacts to LLC in regards to OC'ing, havnt looked into it.

[Melodist]

6 minutes ago, SolarNova said:

I honestly wouldnt know how the 9900k reacts to LLC in regards to OC'ing, havnt looked into it.

But you get what I mean? Having a lower core voltage let's the Chip Run more stable and cooler, which also translates to better stability, running a high LLC to keep it tight at the low voltage?

 

People just always dial in high voltages at low LLC, which is why their overclocks with high core CPUs barely work. Because it is a very unstable high voltage where the high voltage by itself also destabilizes the chip.

 

[Falkentyne]

So much misinformation in this thread.

LLC *DOES NOT* overvolt and has NEVER added voltage.

LLC lowers resistance, which reduces vdroop.

The lowest resistance you can get to is 0 mOhms.

However due to the 'feedback loop', this causes wild transient swings when LLC is increased this much.  A high vcore with high transient spikes AND high current *Can* degrade CPU's even at low overclocks, not JUST because of transient spikes (this would happen very slowly, as spikes only last for microseconds), but also because of electrical specifications being exceeded as well (high RMS on-die vcore with high current (Amps)---example: at 193 amps current load on an 8 core CFL processor, OR 138 amps on a 6 core CFL, On-die sense voltage should NOT exceed 1.22v-1.23v on ambient cooling!

 

And yes it is better to use a higher vcore and a lower LLC.  Transients will be better and vdroop will be more.  This actually can increase stability slightly at a certain "on-die sense" measured voltage if you are borderline stable.  (Example: LLC Turbo + 1.27v load VR VOUT (differential on-die sense voltage via ADC controller value) needed at 5 ghz, might be 1.25v load VR VOUT needed with LLC High for the same stability, although the bios set vcore would probably be about 30-50mv higher here).

 

Proof:

(Note: Maximus XI boards allow proper vcore measuring via vcore sio on-die sense, other boards do NOT, unless they have VR VOUT access)

(this shows that a too high LLC with very high current can DECREASE stability)

 

https://rog.asus.com/forum/showthread.php?109211-Question-about-Transient-Response-(to-Shamino-and-Raja)

 

Quote

I fired up the my Maximus XI Gene + 9900K to see if I could replicate your behavior.

Core = 4.7G
Cache = 4.4G

P95 29.1 FMA3 Small FFTs 15K

LLC=6, Vcore set = 1.130V, Vcore read = 1.066V: 1 thread failed after 6 minutes
LLC=6, Vcore set = 1.140V, Vcore read = 1.074V: pass 20m+

LLC=8, Vcore set = 1.075V, Vcore read = 1.074V: 1 thread failed after 2 minutes
LLC=8, Vcore set = 1.085V, Vcore read = 1.083V: 1 thread failed after 4 minutes
LLC=8, Vcore set = 1.095V, Vcore read = 1.092V: 1 thread failed after 2 minutes
LLC=8, Vcore set = 1.105V, Vcore read = 1.101V: 1 thread failed after 9 minutes
LLC=8, Vcore set = 1.115V, Vcore read = 1.110V: 1 thread failed after 6 minutes
LLC=8, Vcore set = 1.125V, Vcore read = 1.119V: 1 thread failed after 2 minutes
LLC=8, Vcore set = 1.135V, Vcore read = 1.137V: pass 1h+

I repeated it again with LLC=6, Vcore set = 1.140V, Vcore read = 1.074V and 1 thread failed after 14 minutes. Probably 10-20mV extra would pass for 1h+.

This seems even worse than what you reported. One aspect to think of is with higher voltages the temperatures will increase and worsen stability, resulting in even higher voltage required. It would be interesting with a direct comparison with the same CPU/cooling but different boards. I could possibly try to get an Aorus Master next week. I've got a Maximus XI Apex, but I doubt that would perform much better for this specific test case. Primarily the additional VRM components should yield lower VRM temperature.
 

 

This shows why a too high LLC can decrease stability (look at the transient DIPS).

This also shows that LLC does NOT add voltage.  The bios voltage is the target voltage.

 

https://www.overclock.net/forum/6-intel-motherboards/1638955-z370-z390-vrm-discussion-thread-412.html#post28022104

https://www.overclock.net/forum/6-intel-motherboards/1638955-z370-z390-vrm-discussion-thread-413.html#post28022572

[Melodist]

7 minutes ago, Falkentyne said:

So much misinformation in this thread.

LLC *DOES NOT* overvolt and has NEVER added voltage.

LLC lowers resistance, which reduces vdroop.

The lowest resistance you can get to is 0 mOhms.

However due to the 'feedback loop', this causes wild transient swings when LLC is increased this much.  A high vcore with high transient spikes AND high current *Can* degrade CPU's even at low overclocks.

 

And yes it is better to use a higher vcore and a lower LLC.  Transients will be better and vdroop will be more.  This actually can increase stability slightly at a certain "on-die sense" measured voltage if you are borderline stable.  (Example: LLC Turbo + 1.27v load VR VOUT (differential on-die sense voltage via ADC controller value) needed at 5 ghz, might be 1.25v load VR VOUT needed with LLC High for the same stability).

 

Proof:

(Note: Maximus XI boards allow proper vcore measuring via vcore sio on-die sense, other boards do NOT, unless they have VR VOUT access)

(this shows that a too high LLC with very high current can DECREASE stability)

 

https://rog.asus.com/forum/showthread.php?109211-Question-about-Transient-Response-(to-Shamino-and-Raja)

 

 

 

 

This shows why a too high LLC can decrease stability (look at the transient DIPS).

This also shows that LLC does NOT add voltage.  The bios voltage is the target voltage.

 

https://www.overclock.net/forum/6-intel-motherboards/1638955-z370-z390-vrm-discussion-thread-412.html#post28022104

https://www.overclock.net/forum/6-intel-motherboards/1638955-z370-z390-vrm-discussion-thread-413.html#post28022572

Yeah but lower resistance means more voltage is passing through, it is all relative. Have I written overvolting anywhere above?

 

Anyways that's what I wanted to see because apparently, with less LLC, he got away with lower voltage.

 

Anything written above wasn't wrong, it was just a false assumption of the LLC at a higher level being more beneficial, thanks for the info. Now I know I have to try running it at the lowest LLC possible for the OC to be stable without loosing too much voltage.

[Falkentyne]

8 minutes ago, Melodist said:

Yeah but lower resistance means more voltage is passing through, it is all relative. Have I written overvolting anywhere above?

 

Anyways that's what I wanted to see because apparently, with less LLC, he got away with lower voltage.

 

Anything written above wasn't wrong, it was just a false assumption of the LLC at a higher level being more beneficial, thanks for the info. Now I know I have to try running it at the lowest LLC possible for the OC to be stable without loosing too much voltage.

It was solarnova that mentioned overvolting (taking a high voltage even higher).  He got fooled by the SIO sensor problem.

https://www.overclock.net/forum/27686004-post2664.html

I was trying to show you (the op) that it is not always wise to lower bios voltage and raise LLC.  You can hurt stability at heavy loads.  But may be ok for gaming.

(Just don't ever use LLC8).

 

It's very important to distinguish between transient spikes/dips and RMS voltage.  No onboard sensors (not even VR VOUT) can read transient voltage swings.  They just happen far too fast (the oscilloscope shots pick them up easily).

 

The issue is what vcore sensors used to read (and still read on boards like the Asus Strix).

For example, set a Bios vcore of 1.30v and LLC8 (if 7 is the max in the Strix, then 7), the vcore sensor in the Strix will read about 1.36v sustained load vcore.

But that is because of power plane impedance affecting the SIO voltage sensor.  People who don't know this will assume that LLC7 is madly overvolting.

This issue has caused problems for years because no one knew what was going on.  Some people on older boards took probes directly to the VRM caps and read a voltage significantly lower at load than what the sensors were showing.

 

Maximus XI boards finally addressed this problem.

Gigabyte boards have done this since some of their Z370 boards, as they started using a VRM controller with hardware current monitoring (VR VOUT access), etc.