New PC with i7-13700K and Z790 Gaming Plus WiFi - undervolt experience

V

Vassil_V

Member
Joined
Jul 14, 2024
Messages
89
Two days ago I finished a build with an i7-13700K (with AK620 cooler) and a MSI Z790 Gaming Plus motherboard. During my research while deciding what components to go with for the PC, I naturally came upon the dreaded issues with instability and degradation of 13th and 14th gen i7 and i9 CPUs, so I had already done some reading and preparation before building the PC. Before installing Windows I went ahead and changed some settings in the BIOS to the Intel recommended ones, as follows:
PL1 = PL2 = 253W
ICCMax set to 307A
IA CEP = enabled
eTVB, TVB, TVB voltage optimization = enabled
C-States = enabled (I've left C1E disabled)
I also set the TJMax to 95C instead of 100C, for peace of mind.

Note that the latest stable BIOS for my MB doesn't actually have the Intel Default Settings preset, but PL1/PL2, ICCMax are set automatically as above by selecting "Boxed Cooler" mode. There are two beta BIOS versions with Intel Default Settings, but I'd rather avoid betas. If somebody has experience with those, I'd appreciate the feedback.

I don't really understand the AC and DC loadline setting - if I understand correctly, MSI have many (23) different presets for that, called Lite Load Control. Intel recommend that AC=DC. Here, I've selected "Intel Default" mode, which shows up as AC=DC=110 in "advanced mode", which I believe means 1.1 oHms. According to official Intel 13th and 14th gen datasheet, this is the maximum value but I'm not sure what to think of that. Nevertheless, I left it at "Intel Default".

After installing Windows, drivers and all updates, I installed HWInfo and Cinebench 23. CB score was around 29000 and thermal throttle was instantaneous, so I started looking at voltages. What really surprised me was that the maximum VID (a bit later I learned to look at the VCore instead) was hitting 1.5V under light load. During all-core CB load it was about 1.35-1.37
After some experimenting with Lite Load modes ranging from Mode 13 (default for "Normal" profile) to Mode 9, I saw the voltage decreasing and temps improving, but the CB score gets significantly decreased - I saw as low as 20K points with Lite Load = 9. I believe this is because of IA CEP. As a result, I went back to the "Intel Default" preset where AC=DC=110 (as I also would prefer to stick with Intel's recommended settings), and decided to try just setting a negative CPU voltage offset.

Note - Loadline Calibration Control is set to Auto, haven't tested any of the 7 different modes.

I switched the CPU voltage to "Adaptive + Offset", and after some testing, it seems that -0.125V is stable - no crashes in any benchmarks or tests (including Prime95), nor in any of the games I've tested so far. -0.130V was also stable for some tests, but I decided to dial it back to -0.125V to be on the safer side.

With -0.125V offset, the CB 23 score for a single run is around 30700, maitaining 5.3 for P cores and 4.1/2 for E cores, at 225W. The voltage during this load is around 1.25-1.26V which seems great to me. The 10-minute run sees a bit of thermal throttling and the score drops a little, with the clocks decreasing by 100-200Mhz for both the P and E cores.
Idle CPU package temp (currently while writing this) is 32C, ambient temp = 24C.
During gaming, the average ranges between 50C and 70C and overall I'm very impressed by the temps during gaming.

I am very curious about voltage under light load and idle - when the CPU is downclocked (P cores at 3.3 and E cores at 2.6), the voltage can drop to as low as 0.7V, the absolute lowest I've recorded is 0.66V. On the otherside, when the clocks get boosted (but under light load) the Vcore can spike up to around 1.35-1.37V.
Do those voltages sound okay and what are you seeing on your 13th/14th gen i7/i9s?

My goal is stability and longevity (as much as I can influence this), I'm not after overclocking. I've also turned off enchanced turbo boost and turbo boost 3.0, and manually set the ratio limits to my two "star" cores to 53, so there is no boosting to 5.4GHz.

Overall, I am quite happy with my results but I'm under the impression that most people approach undervolting by adjusting Lite Load and Load Line Calibration, so I feel like I'm missing something.

Thanks to everyone who has read everything, and I'll welcome any feedback, suggestions, comments and personal experiences!
 
Last edited:
Sort by date Sort by votes
Unfortunately, I don't think a single guide can apply to 10th and 11th gen, and also 12-14th, but it doesn't hurt to try. I am open to whatever questions you might have, though I probably won't have an answer to all. :D
 
Upvote 0 Downvote
BTW, in regards to the high voltage spikes under light loads, I would add two points. [1] Many voltage spikes will be far too fast for HWInfo to capture, which makes these spikes even more of a concern. [2] The electrical spec for 10th gen appears to state 1.52v as the Max voltage, but this appears to have been upped to 1.72v for Raptor Lake. Is the new limit really safe? In the past, staying under 1.4v allowed for some headroom for momentary transient spikes; now that headroom is gone. I mean, Intel had years to find the electrical boundaries of the 14nm process. I doubt that Intel has accumulated the same level of test data for the 10nm. You and others are right to be concerned about high voltages, even under light loads. Perhaps we can find another undervolting approach that will work better. Of course, it would have to be as stable as your current config.
vg_vassile15a502e8 said:
Unfortunately, I don't think a single guide can apply to 10th and 11th gen, and also 12-14th, but it doesn't hurt to try. I am open to whatever questions you might have, though I probably won't have an answer to all. :D
Click to expand...
That's me. Always looking for that silver bullet! :rolleyes:
 
Last edited:
Upvote 0 Downvote
vg_vassile15a502e8 said:
Unfortunately, I don't think a single guide can apply to 10th and 11th gen, and also 12-14th, but it doesn't hurt to try. I am open to whatever questions you might have, though I probably won't have an answer to all. :D
Click to expand...
You and I have been talking about the observation you made that light demand workload appeared to be spiking voltages beyond what would be considered normal/acceptable. That observation was something I spotted in your very first post on this thread, and I couldn’t explain it to myself let alone anyone else. I had not seen that behavior before, but assumed that it had something to do with the higher core count than previous Intel generations. Then in your second post, you increased your negative voltage offset having elected to stay with the new Intel Defaults mode as your CPU Lite Load Mode. In response to your observation that LLC dropped to the most droopy (mode 8), I created my table. Right?

So here‘s my issue: Intel just announced that all the issues we have been talking about to do with 13th/14th gen instability, as well as the potential for degradation, were due to high voltages - presumably as a result of a microcode bug…the second bug after the eTVB bug. Are there any others?

https://community.intel.com/t5/Proc...ty-Reports-on-Intel-Core-13th-and/m-p/1617113

I have a problem with the above announcement that I cannot resolve. If high voltages were an issue, why would the recent BIOS changes in connection with the Intel Baseline/Performance/Extreme advisory (whatever they called it) introduce a mechanism (i.e. the “Intel Defaults” CPU LL Mode) that would set AC_LL at the max setting (i.e. 1.1 mOhm)? Would this not lead to a massive increase in voltage versus the previous defaults of CPU LL = 9, 12, etc.? …especially under light workloads… I mean, had you not undervolted like crazy, where would your voltages be? And how many users would know that they needed to manual undervolt (or even be proficient at it) to get light load voltages under control? Plus, with manual undervolting, the responsibility for a stable system falls entirely to the user.

This whole thing makes no sense to me. And being that Intel is going to be releasing yet another major code change that affects voltage, I would not be surprised if people have to start over again trying to stabilize their systems…because the rules keep changing. In this regard, it will be interesting to see what changes you will have to make to your manual undervolting, etc. after the new microcode and BIOS update. Please keep us posted. Your experiences may help others.

BTW, I think you will find some of the more technical posts in this Reddit thread of interest. I certainly did.
https://www.reddit.com/r/intel/comments/1e9mf04

In a related matter, Intel appears to have come up with a great way to "neuter" (get ahead of) Gamers Nexus oxidation testing being that they can only afford to test a couple of dead CPUs. In other words, you just have to make the statement that it was a limited number of early production CPUs. Now, while that might be completely accurate, it might also be a wonderful way to dismiss the whole oxidation issue while you run down the warranty clock.

If anyone is reading this post and needs some context, I suggest reading the entire thread. It’s not that long.

EDIT: Below is a user comment on the latest TechLevel1 video. This hammers the point home in a very concise fashion:
1721749619615.png
 
Last edited:
Upvote 0 Downvote
FlyingScot said:
You and I have been talking about the observation you made that light demand workload appeared to be spiking voltages beyond what would be considered normal/acceptable. That observation was something I spotted in your very first post on this thread, and I couldn’t explain it to myself let alone anyone else. I had not seen that behavior before, but assumed that it had something to do with the higher core count than previous Intel generations. Then in your second post, you increased your negative voltage offset having elected to stay with the new Intel Defaults mode as your CPU Lite Load Mode. In response to your observation that LLC dropped to the most droopy (mode 8), I created my table. Right?

So here‘s my issue: Intel just announced that all the issues we have been talking about to do with 13th/14th gen instability, as well as the potential for degradation, were due to high voltages - presumably as a result of a microcode bug…the second bug after the eTVB bug. Are there any others?

https://community.intel.com/t5/Proc...ty-Reports-on-Intel-Core-13th-and/m-p/1617113

I have a problem with the above announcement that I cannot resolve. If high voltages were an issue, why would the recent BIOS changes in connection with the Intel Baseline/Performance/Extreme advisory (whatever they called it) introduce a mechanism (i.e. the “Intel Defaults” CPU LL Mode) that would set AC_LL at the max setting (i.e. 1.1 mOhm)? Would this not lead to a massive increase in voltage versus the previous defaults of CPU LL = 9, 12, etc.? …especially under light workloads… I mean, had you not undervolted like crazy, where would your voltages be? And how many users would know that they needed to manual undervolt (or even be proficient at it) to get light load voltages under control? Plus, with manual undervolting, the responsibility for a stable system falls entirely to the user.

This whole thing makes no sense to me. And being that Intel is going to be releasing yet another major code change that affects voltage, I would not be surprised if people have to start over again trying to stabilize their systems…because the rules keep changing. In this regard, it will be interesting to see what changes you will have to make to your manual undervolting, etc. after the new microcode and BIOS update. Please keep us posted. Your experiences may help others.

BTW, I think you will find some of the more technical posts in this Reddit thread of interest. I certainly did.
https://www.reddit.com/r/intel/comments/1e9mf04

In a related matter, Intel appears to have come up with a great way to "neuter" (get ahead of) Gamers Nexus oxidation testing being that they can only afford to test a couple of dead CPUs. In other words, you just have to make the statement that it was a limited number of early production CPUs. Now, while that might be completely accurate, it might also be a wonderful way to dismiss the whole oxidation issue while you run down the warranty clock.

If anyone is reading this post and needs some context, I suggest reading the entire thread. It’s not that long.

EDIT: Below is a user comment on the latest TechLevel1 video. This hammers the point home in a very concise fashion:
View attachment 190688
Click to expand...
I think we're in uncharted territory, and the best we can do is make an educated guess, observe, and take corresponding action.
My initial action of undervolting was because of high temps and high voltage spikes, but also high voltage under load (so, across the board). This was with the Intel Default loadline profile (on my MB), so AC=DC=110=1.1mOhms. This is an important note because the default way my MB booted up was not with this Intel Default loadline, it was with a custom MSI mode with lower AC value, which effectively means an undervolt compared to AC=110. The question is, what is considered "standard" or "default" settings? Having a baseline of what's considered default, can then answer the question where does the undervolting begin. I think there are way too many motherboards and motherboards vendors, many different BIOS versions, and I am 100% sure there will be differences in the default AC values (which affect the voltage), therefore the CPU's observed voltage under various conditions, even if we take the same CPU and test is across a variety of 15-20 Z690 and Z790 motherboards. This in my opinion is definitely part of the problem. Let's say Intel's CPUs didn't have a problem with the voltage requests during specific conditions, they all worked as Intel intended them to work from the start. Then you get a varierty of motherboards and install the same CPUs on them, and observe different voltages - this would all be coming from the MB settings, for example, the AC loadline. Which of the observed readings would you say is the "default" one? If the CPU is stable on one MB with its default settings, but not stable on another one, is it a problem with the CPU or the default MB settings? Because if a CPU is working properly (and within safe limits), it is never a guarantee how much it can be undervolted and still remain stable.

As far as I've read online, most Intel Default loadline settings of different motherboard vendors put the AC=DC=1.1 mOhms, matching the maximum value found in Intel's official 13/14 gen CPU datasheet. However, Intel themselves say that AC can be set lower, indirectly, through saying that DC can be set lower according to the MB's capabilities, and that AC should match DC. Okay, but DC should be calibrated against the LLC, but the LLC only takes effect as load on the CPU increases, when Vdroop is introduced. Therefore at idle, it shouldn't really make a difference. Which means that by setting DC lower, thus AC lower, you are effectively undervolting your CPU compared to a higher AC/DC setting, which is again technically within Intel's specifications. This leads me to the question what even are the default voltages for a given CPU? Is it determined by the motherboard and default BIOS settings, because this is how it looks right now.

You say I've undervolted like crazy, but note that there are people who run AC values of 20 or lower, which would bring the voltage at more or less the same level as with my offset undervolt. On another note, in my first post I mentioned that I'm coming from a laptop with an i9-12900HX, and because a few months ago I did a lot of research but strictly for Intel HX mobile CPUs, I noticed even back then that people with 13 gen HXs were able to undervolt them massively, some running a -200mV offsets and were happy. On laptops you don't have the options to change AC/DC loadlines, LLC, along with many others, as the BIOSes are a lot more limited, so you have no choice but to stick with offsets, which are most often applied via Throttlestop or XTU. My point is that 13 gens apparently have a lot of extra voltage pumped into them, which judging by my research about undervolting HX mobile CPUs, also applies to the mobile line-up. It surely didn't apply to my 12 gen i9-12900HX, which didn't take too kindly do any larger undervolt offsets. Why that is, I really don't know, but there is clear evidence.

We'll see what will happen with the new August microcode which aims to fix the wrong voltage requests, I'm super curious to read reports about how voltage changes and whether/how much undervolting is affected.

I'm now enjoying a vacation in Greece but when I get back home I'll definitely catch up on the news!
 
Upvote 0 Downvote
Ive read this thread all the way through and made many adjustments with your tips.

For information, this is my build:
MSI MAG Z790 Tomahawk Wifi MAX
i9 14900k
Deepcool LT720
G-Skill Trident Z5 DDR5 6400

My Cinebench 24 score was about 2100, with the TDP reaching nearly 400 watts and my CPU absolutely cooking itself. I lowered the PL1 and PL2 to 253-300 and reached about the same score.
With the new MSI BIOS update, the scores plummeted. It was due to AI-CEP being enabled with a Load Lite of 10. Increasing that to 16 made it stable, but again cooking my CPU.
I decreased the voltages with -0,110, dropped the Load Lite to a stable 12, PL1-PL2 at 200-253 and 310 Amps.

My current Cinebench 24 score is 1951, so its about 10% slower with the intel recommended settings, but the temperatures average around 66 degrees, with peaks to 85. For some reason, the Cinebench R23 ran for 10 minutes straight, without throttling, and when I took a screenshot, it started to throttle at 85 degrees.
Intel recommended me to RMA the processor due to thermal throttling.

I will wait until after the microcode patch before I do that, because this just seems to be that weird voltage spike it has during low effort tasks.
 

Attachments

  • CinebenchHWinfo.jpg
    CinebenchHWinfo.jpg
    1.4 MB · Views: 188
Upvote 1 Downvote
migrainefilm153402d2 said:
Intel recommended me to RMA the processor due to thermal throttling.
Click to expand...

What? Thermal throttling is what happens with every CPU if the power draw is higher than what the cooling can deal with. How is that grounds for an RMA?

edit: Oh, you're getting it at mid-80°C temperatures. But it might be a case of very mild, brief throttling. You mustn't forget, HWinfo is sampling way too slow, there might've been a momentary peak for the briefest split-second that it didn't catch, and the of course the "Thermal throttling" bit is set. You'd basically have to check if the frequencies take a additional hit over what they already do (due to the power limits).
 
Upvote 0 Downvote
citay said:
What? Thermal throttling is what happens with every CPU if the power draw is higher than what the cooling can deal with. How is that grounds for an RMA?

edit: Oh, you're getting it at mid-80°C temperatures. But it might be a case of very mild, brief throttling. You mustn't forget, HWinfo is sampling way too slow, there might've been a momentary peak for the briefest split-second that it didn't catch, and the of course the "Thermal throttling" bit is set. You'd basically have to check if the frequencies take a additional hit over what they already do (due to the power limits).
Click to expand...
That was my thought exactly, I do not want to RMA because of thermal throttling, the P cores keep boosting to 5,686Mhz and it never overdraws any power, I really think that those throttle spikes are due to the buggy power draw. It does not throttle during benchmarking with sustained performance.

My CPU does not crash, hang or BSOD at any given time, I just have a little stutter every once in a while in the audio when gaming, not during rendering or light tasks (12 hours of youtube/snes9x, no stutter), so I think I'm still "safe" with my CPU

Here is the intel support response:
Thank you for sharing the test results. As we can see you are facing thermal throttling. Since the issue persists ever after performing the troubleshooting steps, we have good news, your product is under warranty, and we'll proceed with a Standard Warranty Replacement. The average turnaround time for this process is 5-7 working days and there won't be charges at your end.
 
Upvote 0 Downvote
migrainefilm153402d2 said:
Ive read this thread all the way through and made many adjustments with your tips.

For information, this is my build:
MSI MAG Z790 Tomahawk Wifi MAX
i9 14900k
Deepcool LT720
G-Skill Trident Z5 DDR5 6400

My Cinebench 24 score was about 2100, with the TDP reaching nearly 400 watts and my CPU absolutely cooking itself. I lowered the PL1 and PL2 to 253-300 and reached about the same score.
With the new MSI BIOS update, the scores plummeted. It was due to AI-CEP being enabled with a Load Lite of 10. Increasing that to 16 made it stable, but again cooking my CPU.
I decreased the voltages with -0,110, dropped the Load Lite to a stable 12, PL1-PL2 at 200-253 and 310 Amps.

My current Cinebench 24 score is 1951, so its about 10% slower with the intel recommended settings, but the temperatures average around 66 degrees, with peaks to 85. For some reason, the Cinebench R23 ran for 10 minutes straight, without throttling, and when I took a screenshot, it started to throttle at 85 degrees.
Intel recommended me to RMA the processor due to thermal throttling.

I will wait until after the microcode patch before I do that, because this just seems to be that weird voltage spike it has during low effort tasks.
Click to expand...

Hey, I just saw your comment. How have those settings been treating you, is everything stable? I see that the R23 score is a bit lower compared to normal for 14900K, which I believe is due to IA CEP interfering with the lowered Lite Load mode. Have you tried setting Lite Load to mode 16 (a.k.a Intel Default, a.k.a 110/110), and increasing the negative offset to -0.15V? Looking at the Vcore values on your screenshot I think it should be possible to reduce it further, while at the same time increasing performance due to no?/less interference from IA CEP.

Also, I have a hypothesis regarding why IA CEP doesn't seem to interfere with setting negative offsets, while it does when lowering Lite Load. Undervolting through lowering the AC load line is basically done on the motherboard's side, decreasing the voltage supplied by the motherboard to the CPU, without affecting the requested voltage by the CPU. Therefore, IA CEP "sees" this lower supplied voltage, notes that it's much lower compared to what the CPU asked for, and says "I'm downclocking this b@!ch to keep it stable". However, setting a negative offset is applied to the V/F curve, so the CPU requests lower voltage for any given frequency, so IA CEP is not bothered by it.
The above is highly likely to not be completely true, this is just how I see it.

Also, I can confirm that I have successfully found my point of instability using a negative offset, even with IA CEP = enabled - Cinebench 2024 crashes 100% of the time when I set the negative offset to -175mV.
However, -150mV has been rock solid so far.
 
Upvote 0 Downvote
vg_vassile15a502e8 said:
The question is, what is considered "standard" or "default" settings? Having a baseline of what's considered default, can then answer the question where does the undervolting begin. I think there are way too many motherboards and motherboards vendors, many different BIOS versions,
Click to expand...
You had made the above observation in your last post, which I now see as a major part of the problem. Hey, I actually felt better about not being the only one thoroughly confused/frustrated by Intel’s recommended defaults chart they put out…and then changed… In the recent 30min video by Steve at Gamers Nexus he railed on what a mess it is and how it stinks of being full of loopholes where Intel can wiggle out from being responsible for any CPU damage caused by high temps/volts, etc.

The comment you made above is right on point. We don’t even have a common definition of “Intel Defaults” when we are talking with each other. It’s a mess. Without a universal starting point, it’s exceedingly difficult for us to help each other. Everybody is offering one, two or three settings to change at a time but the results the users are seeing appear to be very inconsistent. On top of that, the rules will likely keep changing every time a new microcode update gets pushed out.

This whole Raptor Lake thing has given me a headache. My only recourse is to fall back to a very methodical and measured approach where I try and arrive at my own “baseline” - and then try to improve upon those results one step at a time.

In regards to today’s post, [EDIT: You’re hypothesis is a good one, but] I think I remember CiTay saying that he had seen instances where CEP did mess with Undervolting the traditional way (I.e. offsets). I’ve seen other users really struggling with CEP, also. It’s why I have abandoned my ”little table“ for now.

Bottom line: We all have to be very careful that in our efforts to help each other that we are not advising people to make changes that are not compatible with their other settings. I bet if we polled everyone right now, we would see happy i9/i7 owners with three or more very different sets of settings. I’ve been working on something that I hope will highlight that observation.
 
Last edited:
Upvote 0 Downvote
Oh, you're definitely not the only one who's confused.. :X It is definitely a mess. At this point, I've read so many topics and have seen all kinds of different settings combinations, which people claim to work best for them. Some are using AC DC load lines at 1-1, some (as me) at 110-110, some 10/110, some 50/80, some 80/80, I mean.. there truly doesn't seem to be a general consensus on this. Some are combining reduced load lines with negative offsets, others are using only one or the other, third are using V/F point offsets (such as ScatterBencher, whose explanations are above my level to be honest). I genuinely don't know what should be considered default, the best I can go by is Intel's recommendations, but they only specify the max AC/DC, so anything below that is fair game.

Earlier today I tried setting Lite Load at Mode 9 (80=80), which seems to be the default for many people on MSI Z790 motherboards, with IA CEP=enabled, and without negative offset. CB R23 score dropped by 2800 points at 188W, while the Vcore under load was just about 10-20 mV higher compared to what I see with my settings. I assume this is because of IA CEP, but you see, what's confusing is that even though the VCore is a bit higher, IA CEP is interfering, while it's not when using my settings.
I would have tried turning IA CEP off, but the other problem is that even though the VCore under load is comparable, I saw spikes up to 1.43V shortly after booting, which is a no-no for me. When using Lite Load at Intel Default and the negative offset is set to -150mV, those spikes are up to 1.31V.
 
Upvote 0 Downvote
vg_vassile15a502e8 said:
Hey, I just saw your comment. How have those settings been treating you, is everything stable? I see that the R23 score is a bit lower compared to normal for 14900K, which I believe is due to IA CEP interfering with the lowered Lite Load mode. Have you tried setting Lite Load to mode 16 (a.k.a Intel Default, a.k.a 110/110), and increasing the negative offset to -0.15V? Looking at the Vcore values on your screenshot I think it should be possible to reduce it further, while at the same time increasing performance due to no?/less interference from IA CEP.
Click to expand...
It's very solid, except for Cinebench 24, that just stops working after a while. Cinebench 23, Prime95, you name it, works fine.

IA CEP is disabled, and dont notice anything weird about that at all. Granted, I dont know what 95% of the BIOS does
 
Upvote 0 Downvote
vg_vassile15a502e8 said:
Oh, you're definitely not the only one who's confused.. :X It is definitely a mess. At this point, I've read so many topics and have seen all kinds of different settings combinations, which people claim to work best for them. Some are using AC DC load lines at 1-1, some (as me) at 110-110, some 10/110, some 50/80, some 80/80, I mean.. there truly doesn't seem to be a general consensus on this. Some are combining reduced load lines with negative offsets, others are using only one or the other, third are using V/F point offsets (such as ScatterBencher, whose explanations are above my level to be honest). I genuinely don't know what should be considered default, the best I can go by is Intel's recommendations, but they only specify the max AC/DC, so anything below that is fair game.

Earlier today I tried setting Lite Load at Mode 9 (80=80), which seems to be the default for many people on MSI Z790 motherboards, with IA CEP=enabled, and without negative offset. CB R23 score dropped by 2800 points at 188W, while the Vcore under load was just about 10-20 mV higher compared to what I see with my settings. I assume this is because of IA CEP, but you see, what's confusing is that even though the VCore is a bit higher, IA CEP is interfering, while it's not when using my settings.
I would have tried turning IA CEP off, but the other problem is that even though the VCore under load is comparable, I saw spikes up to 1.43V shortly after booting, which is a no-no for me. When using Lite Load at Intel Default and the negative offset is set to -150mV, those spikes are up to 1.31V.
Click to expand...
Well, I'm very glad you updated us on your experiences. It makes me feel like I'm not just too dumb to understand this thing. There's so much background noise that it's exceedingly difficult to know which direction to go in, especially if you have an i9, with all its extra little microcode goodies.

I think in Tom's most recent video (site= https://www.youtube.com/@MooresLawIsDead ) on the Raptor Lake situation, someone within Intel alluded to there being the possibility of many bugs in the Raptor Lake microcode - presumably due to the speed at which it was pushed out the door. At the time, Intel had bragged about getting Raptor Lake done in only like 9 -11 months. Now we see the consequence of that effort. A bit like something else I can think of but won't say on this forum.

The reason I bring this up is that we cannot rule out the possibility that the reason many of us are struggling to get the results we would expect from our changes is because the microcode is not reacting consistently. If you're trying to test something with logic bugs, you are going to have an impossible time figuring out what the heck is going on. We can only hope that this month's microcode fixes some of these bugs, assuming they exist. I think they do.
 
Last edited:
Upvote 0 Downvote
migrainefilm153402d2 said:
It's very solid, except for Cinebench 24, that just stops working after a while. Cinebench 23, Prime95, you name it, works fine.

IA CEP is disabled, and dont notice anything weird about that at all. Granted, I dont know what 95% of the BIOS does
Click to expand...
I'm sorry, I've misunderstood you regarding IA CEP. The lower score is explained by the PL1 and potentially ICCMax, which is understandable. Intel's recommendation for ICCMax for is 400A, as per the "Extreme" profile.

However, Cinebench 2024 crashing means it's not fully stable. Interestingly, during my most recent experiments, exactly Cinebench 2024 pointed towards instability issues, while R23 was running fine.
 
Upvote 0 Downvote
I have more stability with manual undervolt, and this is a very low liteload, stable for me is usually 9, so Its no surprise its crashing C24. Nothing else crashes, blender, premiere pro, after effects, games, nothing but C24 so its fine for me

Edit: Lite Load 8 works fine
 
Last edited:
Upvote 0 Downvote
FlyingScot said:
Well, I'm very glad you updated us on your experiences. It makes me feel like I'm not just too dumb to understand this thing. There's so much background noise that it's exceedingly difficult to know which direction to go in, especially if you have an i9, with all its extra little microcode goodies.

I think in Tom's most recent video (site= https://www.youtube.com/@MooresLawIsDead ) on the Raptor Lake situation, someone within Intel alluded to there being the possibility of many bugs in the Raptor Lake microcode - presumably due to the speed at which it was pushed out the door. At the time, Intel had bragged about getting Raptor Lake done in only like 9 months. Now we see the consequence of that effort. A bit like something else I can think of but won't say on this forum.

The reason I bring this up is that we cannot rule out the possibility that the reason many of us are struggling to get the results we would expect from our changes is because the microcode is not reacting consistently. If you're trying to test something with logic bugs, you are going to have an impossible time figuring out what the heck is going on. We can only hope that this month's microcode fixes some of these bugs, assuming they exist. I think they do.
Click to expand...
It's my pleasure to share this, I like tinkering with those stuff and hopefully somebody might find answers they're looking for in all those posts.
Will check out this video, thanks!
I'll update to the new BIOS with the *fixed* (allegedly) micro code when it comes out, it will be very interesting to see how it will affect the CPU's behaviour.
 
Upvote 0 Downvote
I wanted to thank OP (and others on the forum like FlyingScot and citay) for this thread. I recently put together a build with a 13700K and these posts have been helpful in guiding my tweaks. Thought I'd share the setup I settled on:

CPU: i7-13700K (mounted using Thermalright Contact Frame)
Cooler: Thermalright Phantom Spirit 120 Air Cooler
Motherboard: Z790 Tomahawk WiFi Max (on the latest BIOS that includes Intel Default settings, 7E25vA62, but it does not have the most recent 0x125 microcode update)

PL1 = 125W
PL2 = 125W
ICCMax = 307A
TJMax = 80C
Lite Load = Mode 16 (AC/DC = 1.1mOhms/1.1mOhms)
C-States = Enabled
Enhanced Turbo = Disabled
Intel Turbo Boost Max 3.0 = Disabled (the other Intel Boost and TVB options I left enabled)
IA CEP = Enabled (one note here: with the rest of these settings as they are, I tried disabling CEP and my all-core score in Cinebench R23 went down ~1000 points)
Max Core Ratios = All P-cores limited to 52x and all E-cores limited to 40x All P-cores limited to 53x and all E-cores limited to 41x
Core Voltage = Adaptive + Offset set to -100mv

At first, I tried fiddling with the V/F curve option, but I could not get the effect I desired. I gave up and went for a simple Adaptive + Offset like OP. I started with a -125mv offset, but after using it for a few hours my system locked up so I reduced it to -100mv just to be safe. I've been running that for over 24hrs with no issues so far. I've put the system through a suite of benchmarks and stress tests, and my minimum recorded Vcore is 0.654V and my maximum recorded Vcore is 1.244V 1.29V. The highest recorded VID request is 1.304V 1.39V. When hitting 125W under an all-core load, the CPU temps top out at about 59-60C. With a 6-8 thread load where the cores boost to max ratios the temps reach around 62-64C. Idle and very light load temps are 28-31C.

My two primary goals were to 1) maintain cool temps and reasonable power draw and 2) prevent the possibility of any degradation. I won't frequently be doing intense all-core loads with this system and even when I do it's not essential for me to squeeze out every last percentage point of performance. I get a decent enough 26400-26500 Cinebench R23 all-core score, and in lightly threaded applications the P-cores are able to boost to max ratio. I'm probably going to stick with this while I wait to see what the effects of Intel's incoming microcode update are.
 
Last edited:
Upvote 0 Downvote
garillon151802d3 said:
I wanted to thank OP (and others on the forum like FlyingScot and citay) for this thread. I recently put together a build with a 13700K and these posts have been helpful in guiding my tweaks. Thought I'd share the setup I settled on:

CPU: i7-13700K
Cooler: Thermalright Phantom Spirit 120 Air Cooler
Motherboard: Z790 Tomahawk WiFi Max (I'm on the latest BIOS that includes Intel Default settings, but it does not have the most recent 0x125 microcode update)

PL1 = 125W
PL2 = 125W
ICCMax = 307A
TJMax = 80C
Lite Load = Mode 16 (AC/DC = 1.1mOhms/1.1mOhms)
C-States = Enabled
Enhanced Turbo = Disabled (the other Intel Boost and TVB options I just left enabled)
IA CEP = Enabled (one note here: with the rest of these settings as they are, I tried disabling CEP and my all-core score in Cinebench R23 went down ~1000 points)
Max Core Ratios = All P-cores limited to 52x and all E-cores limited to 40x
Core Voltage = Adaptive + Offset set to -100mv

At first, I tried fiddling with the V/F curve option, but I could not get the effect I desired. I gave up and went for a simple Adaptive + Offset like OP. I started with a -125mv offset, but after using it for a few hours my system locked up so I reduced it to -100mv just to be safe. I've been running that for over 24hrs with no issues so far. I've put the system through a suite of benchmarks and stress tests, and my minimum recorded Vcore is 0.654V and my maximum recorded Vcore is 1.244V. The highest recorded VID request is 1.304V. When hitting 125W under an all-core load, the CPU temp sits at around 59-60C. With an 8 thread load, the P-cores boost to max ratio and the temps are around 62-64C. Idle and very light load temps are 28-31C.

My two primary goals were to 1) maintain cool temps and reasonable power draw and 2) prevent the possibility of any degradation. I won't frequently be doing intense all-core loads with this system and even when I do it's not essential for me to squeeze out every last percentage point of performance. I get a decent enough 26400-26500 Cinebench R23 all-core score, and in lightly threaded applications the P-cores are still able to boost to max ratio. I'm probably going to stick with this while I wait to see what the effects of Intel's incoming microcode update are.
Click to expand...
I'm happy that this thread has helped you!
I understand where you're coming from, but my personal advice is to not worry that much and give your CPU a chance to show more of what it's capable of. The settings you have currently set are, in my opinion, overly cautious, there isn't really a reason to set the TJMax at 80C, PL2 at 125W, and decrease the P and E cores clock speeds below stock. 1.25V max VCore is quite low and you can safely go up to 1.35V at least (but for me that's the ideal max), so I'd suggest that you at least set the turbo ratios to default, bump up PL2 to at least 150W (your cooler is very good and should be able to handle even more), and if you still prefer to be on the safe side with temperatures, set TJMax to 90C. Those settings are still plenty safe (even more than necessary), but will give you back some of the performance, especially in light loads due to the higher clocks.

At the end of the day, if you're happy with it that's what matters, but you have a good CPU, why limit it so much for questionable (if any) benefit.
 
Upvote 0 Downvote
Back
Top