Basic BIOS Settings Suggestions

[B4K4]

Sorry if some of this is kinda noob, but I just finished my first build in almost 10 years

• i5 12600K
• Cooler Master ML360 AIO
• Asus TUF Z690-PLUS WIFI
• Corsair Vengeance 2x16GB DDR5-5600 CL36
• Kingston Fury Renegade 1TB NVME
• Asus TUF RX 6800 XT OC
• Corsair RM850x

This is my first time getting an unlocked processor, so I was wanting to do a light tune.

I enabled XMP II and Asus Multicore Enhancement and got a score of 15795 in Cinebench R23. Everything I was reading said that the 12600K should be getting a score of about 17000...

I did some digging and found that Windows was in Balanced mode. So I switched it to Performance, re-benchmarked, and got 17405.

 

Problem is I'm now seeing a lot about Multicore Enhancement drastically affecting temps, stability, and life of the CPU.

So far the system seems perfectly stable and my temps during benchmark hit a max of only 66°C.

I've looked into disabling Multicore Enhancement and just using the Asus AI OC. But then I'm reading that AI overclocks typically increase core voltages way too much, and again just lead to higher temps.

And lastly I'm finding little info on whether XMP I or II is the way to go. From what I understand, XMP I lets the motherboard handle some settings while XMP II sets everything but could be unstable.

 

So I guess I'm just asking here for suggestions...

Asus Multicore Enhancement - enabled or disabled?

XMP - I or II?

AI OC - yay or nay?

Are there any other settings that I should look into that I'm maybe not aware of? I've seen mention of Intel Turbo Boost Tech and to check and make sure that it's enabled.

 

Thanks in advance for any help you guys are willing to offer

[RONOTHAN##]

8 minutes ago, B4K4 said:

Asus Multicore Enhancement - enabled or disabled?

 

If you aren't gonna do manual overclocking, leaving this enabled is fine. It doesn't really affect stability, and it does increase temps a bit because it causes it to hold turbo longer, but not as much as the AI OC stuff. 

 

10 minutes ago, B4K4 said:

XMP - I or II?

It doesn't matter that much. I would usually do 1 since it has a small performance uplift and I've never seen instability come from it, but your mileage may vary. If you end up with a memory related BSOD, switch over to 2. 

 

12 minutes ago, B4K4 said:

AI OC - yay or nay?

No, it uses way more voltage than necessary and it's usually a bit ambitious with the settings it wants to run. Either manually overclock or don't overclock at all. 

 

13 minutes ago, B4K4 said:

Are there any other settings that I should look into that I'm maybe not aware of? I've seen mention of Intel Turbo Boost Tech and to check and make sure that it's enabled.

Not really. Most of the other settings you want to mess with are related to manual overclocking, not automatic performance uplifts, and the ones that should affect performance like Turbo should be enabled by default. 

[B4K4]

2 hours ago, RONOTHAN## said:

If you aren't gonna do manual overclocking, leaving this enabled is fine. It doesn't really affect stability, and it does increase temps a bit because it causes it to hold turbo longer, but not as much as the AI OC stuff. 

 

It doesn't matter that much. I would usually do 1 since it has a small performance uplift and I've never seen instability come from it, but your mileage may vary. If you end up with a memory related BSOD, switch over to 2. 

 

No, it uses way more voltage than necessary and it's usually a bit ambitious with the settings it wants to run. Either manually overclock or don't overclock at all. 

 

Not really. Most of the other settings you want to mess with are related to manual overclocking, not automatic performance uplifts, and the ones that should affect performance like Turbo should be enabled by default. 

Thanks for the info, I appreciate it.

I'd like to eventually learn how to do some mild manual oc'ing. I just need to put in some time researching how to do it properly.

I'm guessing that manual overclocking is how I'm seeing 12600K scores of 20k-22k in Cinebench.

I'll of course just Google it, but do you recommend any specific resources for learning how to manual oc?

[RONOTHAN##]

26 minutes ago, B4K4 said:

I'm guessing that manual overclocking is how I'm seeing 12600K scores of 20k-22k in Cinebench.

Yeah, that's how you get it up that high. 

 

27 minutes ago, B4K4 said:

I'll of course just Google it, but do you recommend any specific resources for learning how to manual oc?

There are some YouTube channels that do focus on it, Der8auer, SkatterBench, and Actually Hardcore Overclocking to name a few, though they go into the more advanced methods of overclocking and other nuances of the platform. For just a basic overclock, it's not super difficult, the way I like to think of it is a balancing act between voltage, clock speed, and temperature. Changing one affects the other two, and your goal is to maximize the clock speed while keeping the temperature and voltage below safe limits (100C and 1.45V respectively, though with 12th gen you will hit temp limits before you get anywhere near voltage limits). My general process is as such:

1. First, decide if you want to go for multicore score or game speed. This only applies to 12th gen, but you can clock the ring (the interconnect between the cores and the memory controller) way higher when the E cores are disabled, which can impact some workloads like games a lot. Disabling the e cores doesn't reduce the multi core score by all that much, especially if you're overclocking the P cores as well (average R23 score after overclocked with the P cores off is ~16K), but it's lower enough that you have to decide which is more important to you for whether you'd leave the E cores on or off. 
2. Run a couple benchmarks initially to see where your performance started to make sure you're actually improving. You want to do multiple, as there can be regressions in one benchmark but good performance in another with certain setups. If you want to disable E cores, I'd recommend doing that before benchmarking fully. 
3. Once that's done, go into the BIOS and set some initial settings.
   1. First, go through and disable all the power management settings (C-States, SpeedStep, etc.), though leave Intel Turbo Boost enabled still (you can't overclock if this is disabled for some reason). You can turn those back on later.
   2. Next, set the ratios. You want 48 for the P cores, (if they're enabled) 40 for the E cores, and 40 for the Ring (might be called Cache or Uncore, I'll refer to it as ring though since that's what I'm used to calling it). Those ratios should be stock (or at least close enough to stock, I'm going off memory for them) and therefore should just work. 
   3. Configure the LLC settings. They should be in the Digi+ VRM menu of the ASUS BIOS, and you want to configure that to Level 4 LLC. This sets the amount of VDroop you have, and without going into the specifics about it Level 4 should give a good balance between voltage regulation and not dropping down a ton. 
   4. Finally, set the initial voltage. Set the VCore mode to manual/override (I forget what ASUS calls it) and set the voltage to 1.35V. 
4. Boot into Windows, double check to make sure the frequency applied in either HWInfo64 (great for monitoring temps and other sensor data) or CPU-Z (much quicker for just checking frequency) and run a stress test of your choice for about 5 minutes to make sure it's not immediately unstable. Some of the bigger ones are Prime95 Small FFTs, OCCT, and Linpack Xtreme, they all have their pros and cons.
   1. Prime95 Small FFTs is about as hot a stress test as you can find, if you're pushing your chip it will end up at 100C in this while being like 80C in Cinebench. It's usually the most effective at finding instability though, it's just generally considered too unrealistic and will reduce your max overclock somewhat significantly compared to the other two. 
   2. Linpack Xtreme is a more balanced stress test, it's much more in line with a workload you can actually see in the real world, though it is dependent on having relatively fast memory speed for it to run at full speed (E.G. you'd have to do a memory overclock first, which is a very tall order for someone who's never done a CPU overclock), and it doesn't work properly with the E cores on. 
   3. OCCT is in the middle of those two, it has a lot of different settings and it does have some cool features to it, though it's paid software so you'd have to use the free trial if you don't want to spend the ~$5 a month for it, and the free trial has some 
5. Once the stress test completed:
   1. If it was stable and temps were fine, increase the P core multiplier. 
   2. If it was stable and temps weren't fine, decrease voltage by 0.01V.
   3. If it wasn't stable and temps were fine, increase the voltage by 0.01V.
   4. If it wasn't stable and temps weren't fine, decrease the voltage by 0.01V and P core multiplier by 1.
6. Repeat from Step 4 until you found the upper limit of the P cores
   1. You can also mess with the CPU PLL voltage as this can help with P core stability a bit, but it's little enough that I wouldn't blame you for not trying. It sweet spots, so the way you want to find the optimal value is to lower the voltage to the point where you're just barely crashing, then increase the CPU PLL by 0.015V from the stock 0.9V until it stops crashing until you get to ~1.1V
7. Once you found the P core limit, Repeat Step 4 and 5 with the E cores (assuming they're enabled). 
   1. There is also the E core PLL voltage and E core L2 Cache voltage, these can help, though not by enough to make a noticeable difference. IIRC E core L2 voltage you want to just leave at 1.25V on 12th gen, though in my experience (13th gen admittedly, not 12th) it just doesn't really do anything.
8. Once you found the limits for the P and E cores, move onto the ring. Repeat Steps 4 and 5 with the ring multiplier. 
   1. There are even more voltages this time. There's the ring PLL, SA PLL, and System Agent voltage that affect ring stability. The most important of these is System Agent, and usually higher is better for this with ring overclocking, but higher can also cause weird memory issues so it might not be a good idea to mess with this unless you're going to be stress testing your RAM at the same time. 
9. After you've found the initial balance for the chip, run a long final stress test. I'd want at least an hour, more likely 2, for my main system, though run it for as long as you like (I'd say anything more than 24 hours is pretty excessive though). If you run into instability, do the following:
   1. If the temps get too high and/or you start throttling, decrease voltage by 0.01V until it's below 100C. 
   2. If you get a crash and you aren't yet at the temp limit, increase the voltage by 0.01V until it works again. 
   3. If you get a crash and you're at the temp limit, decrease the multipliers one by one (I go in the order ring, E core, P core, but it's up to you) until it becomes stable, then start adding back the ones you removed earlier to figure out which was bad (if not all of them). 
10. Finally, re-run the initial benchmarks to see how much performance you gained. 

[B4K4]

19 hours ago, RONOTHAN## said:

Yeah, that's how you get it up that high. 

 

There are some YouTube channels that do focus on it, Der8auer, SkatterBench, and Actually Hardcore Overclocking to name a few, though they go into the more advanced methods of overclocking and other nuances of the platform. For just a basic overclock, it's not super difficult, the way I like to think of it is a balancing act between voltage, clock speed, and temperature. Changing one affects the other two, and your goal is to maximize the clock speed while keeping the temperature and voltage below safe limits (100C and 1.45V respectively, though with 12th gen you will hit temp limits before you get anywhere near voltage limits). My general process is as such:

1. First, decide if you want to go for multicore score or game speed. This only applies to 12th gen, but you can clock the ring (the interconnect between the cores and the memory controller) way higher when the E cores are disabled, which can impact some workloads like games a lot. Disabling the e cores doesn't reduce the multi core score by all that much, especially if you're overclocking the P cores as well (average R23 score after overclocked with the P cores off is ~16K), but it's lower enough that you have to decide which is more important to you for whether you'd leave the E cores on or off. 
2. Run a couple benchmarks initially to see where your performance started to make sure you're actually improving. You want to do multiple, as there can be regressions in one benchmark but good performance in another with certain setups. If you want to disable E cores, I'd recommend doing that before benchmarking fully. 
3. Once that's done, go into the BIOS and set some initial settings.
   1. First, go through and disable all the power management settings (C-States, SpeedStep, etc.), though leave Intel Turbo Boost enabled still (you can't overclock if this is disabled for some reason). You can turn those back on later.
   2. Next, set the ratios. You want 48 for the P cores, (if they're enabled) 40 for the E cores, and 40 for the Ring (might be called Cache or Uncore, I'll refer to it as ring though since that's what I'm used to calling it). Those ratios should be stock (or at least close enough to stock, I'm going off memory for them) and therefore should just work. 
   3. Configure the LLC settings. They should be in the Digi+ VRM menu of the ASUS BIOS, and you want to configure that to Level 4 LLC. This sets the amount of VDroop you have, and without going into the specifics about it Level 4 should give a good balance between voltage regulation and not dropping down a ton. 
   4. Finally, set the initial voltage. Set the VCore mode to manual/override (I forget what ASUS calls it) and set the voltage to 1.35V. 
4. Boot into Windows, double check to make sure the frequency applied in either HWInfo64 (great for monitoring temps and other sensor data) or CPU-Z (much quicker for just checking frequency) and run a stress test of your choice for about 5 minutes to make sure it's not immediately unstable. Some of the bigger ones are Prime95 Small FFTs, OCCT, and Linpack Xtreme, they all have their pros and cons.
   1. Prime95 Small FFTs is about as hot a stress test as you can find, if you're pushing your chip it will end up at 100C in this while being like 80C in Cinebench. It's usually the most effective at finding instability though, it's just generally considered too unrealistic and will reduce your max overclock somewhat significantly compared to the other two. 
   2. Linpack Xtreme is a more balanced stress test, it's much more in line with a workload you can actually see in the real world, though it is dependent on having relatively fast memory speed for it to run at full speed (E.G. you'd have to do a memory overclock first, which is a very tall order for someone who's never done a CPU overclock), and it doesn't work properly with the E cores on. 
   3. OCCT is in the middle of those two, it has a lot of different settings and it does have some cool features to it, though it's paid software so you'd have to use the free trial if you don't want to spend the ~$5 a month for it, and the free trial has some 
5. Once the stress test completed:
   1. If it was stable and temps were fine, increase the P core multiplier. 
   2. If it was stable and temps weren't fine, decrease voltage by 0.01V.
   3. If it wasn't stable and temps were fine, increase the voltage by 0.01V.
   4. If it wasn't stable and temps weren't fine, decrease the voltage by 0.01V and P core multiplier by 1.
6. Repeat from Step 4 until you found the upper limit of the P cores
   1. You can also mess with the CPU PLL voltage as this can help with P core stability a bit, but it's little enough that I wouldn't blame you for not trying. It sweet spots, so the way you want to find the optimal value is to lower the voltage to the point where you're just barely crashing, then increase the CPU PLL by 0.015V from the stock 0.9V until it stops crashing until you get to ~1.1V
7. Once you found the P core limit, Repeat Step 4 and 5 with the E cores (assuming they're enabled). 
   1. There is also the E core PLL voltage and E core L2 Cache voltage, these can help, though not by enough to make a noticeable difference. IIRC E core L2 voltage you want to just leave at 1.25V on 12th gen, though in my experience (13th gen admittedly, not 12th) it just doesn't really do anything.
8. Once you found the limits for the P and E cores, move onto the ring. Repeat Steps 4 and 5 with the ring multiplier. 
   1. There are even more voltages this time. There's the ring PLL, SA PLL, and System Agent voltage that affect ring stability. The most important of these is System Agent, and usually higher is better for this with ring overclocking, but higher can also cause weird memory issues so it might not be a good idea to mess with this unless you're going to be stress testing your RAM at the same time. 
9. After you've found the initial balance for the chip, run a long final stress test. I'd want at least an hour, more likely 2, for my main system, though run it for as long as you like (I'd say anything more than 24 hours is pretty excessive though). If you run into instability, do the following:
   1. If the temps get too high and/or you start throttling, decrease voltage by 0.01V until it's below 100C. 
   2. If you get a crash and you aren't yet at the temp limit, increase the voltage by 0.01V until it works again. 
   3. If you get a crash and you're at the temp limit, decrease the multipliers one by one (I go in the order ring, E core, P core, but it's up to you) until it becomes stable, then start adding back the ones you removed earlier to figure out which was bad (if not all of them). 
10. Finally, re-run the initial benchmarks to see how much performance you gained. 

This is awesome.

Thanks a lot for all your help.