Voltage and temperature 14600k + msi b760 mortar max

enot92515ae02e7

New member
Joined
Jan 19, 2025
Messages
10
Hi, can you please tell me if this is a normal temperature and voltages for my processor. And are my bios settings ok or can I improve them?
Cooler - pentawave z06d
LLC- mode 7
IA CEP -off
CPU ratio P core - 53
E core - 40
Intel speed shift - off
OCCT-25119-124150.png


Без имени.jpg
 
The only question I would like to ask is whether I should set the CPU Ratio settings myself, which respectively disable turbo boost and EIST or leave the settings in automatic mode?

Leave all the intelligent boosting and downclocking mechanisms active nowadays, also Speed Shift. These mechanisms are among the main innovations that happened since the days of Pentium 4, when it was replaced by the first "Core" CPUs. If you set fixed CPU multipliers and don't let it boost or clock down, you make the CPU behave like a faster, but dumb Pentium 4.

Well, some people may not have lived through the Pentium 4 days, but it was regarded as a "brute-force" CPU in the end, with excessive power draw. Back then, it was the first CPU to ever draw more than 80W (which was very extreme at the time), later even >100W. Instead of intelligent solutions, it only relied on its (fixed) high clock speeds to gain any performance improvements, and higher and higher voltages were needed to gain those high frequencies (3+ GHz).

Reviewers back then were appalled by the higher power draw, calling it a "monster" already, and saying that 80-100W is more than what is bearable for a desktop PC. Well, maybe they were right, but with inventions like the tower cooler (up until ~20 years ago, basically all coolers were top-blow coolers similar to the Intel Boxed ones) and improved case design etc., we now only call CPUs "monsters" when they have 2x-3x that power draw.

Anyway, back to the features, those things like letting the CPU decide the best clock speeds (and thus voltages) for the situation according to the workload, you shouldn't take that away from it. Not even overclockers really use fixed multipliers anymore, most of them realized that an OC that cancels out the intelligent mechanisms belongs to the past. Heck, maybe even overclocking altogether, with how much Intel pushes the CPUs from factory nowadays. The best thing you can do is lower the voltages a little to where it's still stable, for example by lowering CPU Lite Load. But the rest, you don't have to intervene so much and set fixed speeds.
 
Оставьте все интеллектуальные механизмы повышения и снижения частоты активными в наши дни, включая Speed Shift. Эти механизмы являются одними из главных инноваций, появившихся во времена Pentium 4, когда его заменили первые процессоры «Core». Если вы установите фиксированные множители процессора и не позволите ему повышать или понижать скорости, вы заставите процессор вести себя как более быстрый, но тупой Pentium 4.

Ну, некоторые люди, возможно, не пережили дни Pentium 4, но в конечном итоге он считался процессором «грубой силы» с чрезмерным энергопотреблением. Тогда это был первый процессор, который потреблял более 80 Вт (что в то время было очень экстремально), позже даже >100 Вт. Вместо интеллектуальных решений он выбирает только свои (фиксированные) высокие тактовые частоты для получения каких-либо улучшений производительности, а для получения этих высоких частот (3+ ГГц) требуются все более высокие напряжения.

Тогда обозреватели были потрясены более высоким энергопотреблением, назвав его «монстром» и заявив, что 80-100 Вт — это больше, чем терпимо для настольного ПК. Что ж, возможно, они были правы, но с такими изобретениями, как башенный кулер (еще ~20 лет назад все кулеры были с верхним обдувом, похожими на коробочные Intel) и улучшенным корпусом и т. д. д., мы теперь называем процессоры «монстрами», только когда они потребляют в 2-3 раза больше энергии.

В любом случае, когда речь идет о функциях, таких вещах, как определение ЦП выбора наилучшей таковой частоты (и, следовательно, напряжения) для ситуации в соответствии с рабочей программой, вы не должны отнимать это у него. Даже оверклокеры больше не используют фиксированные множители, большинство из них поняли, что разгон, который отменяет интеллектуальные механизмы, принадлежит прошлому. Черт, может быть, даже разгон вообще, раньше, как сильно Intel сейчас выжимает ЦП с завода. Лучшее, что вы можете сделать, это немного снизить напряжение, чтобы оно все еще было стабильным, например, снизив загрузку CPU Lite. Но в остальном вам не нужно так сильно вмешиваться и сохранять фиксированные скорости.
I keep experimenting, but what I don't understand is that I lowered the voltages quite well, now 1.2 - 1.210 is the maximum I can see from the sensors, but why does the processor keep consuming quite a lot? 170w, with a max temperature of 85 degrees Celsius. I was expecting a lower temperature and consumption reading after all
 
Я продолжаю экспериментировать, но я не понимаю, что я снизил напряжение довольно хорошо, теперь 1,2 - 1,210 - это максимум, который я вижу по датчикам, но почему процессор продолжает потреблять довольно много? 170 Вт, при максимальной температуре 85 градусов Цельсия. Я все-таки ожидал более низких показаний температуры и потребления
 
I keep experimenting, but what I don't understand is that I lowered the voltages quite well, now 1.2 - 1.210 is the maximum I can see from the sensors, but why does the processor keep consuming quite a lot? 170w, with a max temperature of 85 degrees Celsius. I was expecting a lower temperature and consumption reading after all

Are you testing in OCCT? That is a theoretical stress test, designed to drive the CPU to the absolute limit with dirty tricks. If you want realistic full load, use Cinebench R23 Multi. Also, instead of HWmonitor, use HWinfo for monitoring, it's superior. I link it in my guide, and there i also mention how to set it up so you see all the sensors at once (or at least most of them).

So, if you leave HWinfo sensors running, do a 10-minute Cinebench run as mentioned in my guide, and then show me a screenshot of the HWinfo sensors and tell me the points from Cinebench, i can check what is up.

In general, 14th gen (the true models, meaning the -K ones) are all more power-hungry than 13th gen, and considerably more than 12th gen. The 14th gen -K models were the peak of high power draw and inefficiency. But we'll see what happens during Cinebench.
 
Are you testing in OCCT? That is a theoretical stress test, designed to drive the CPU to the absolute limit with dirty tricks. If you want realistic full load, use Cinebench R23 Multi. Also, instead of HWmonitor, use HWinfo for monitoring, it's superior. I link it in my guide, and there i also mention how to set it up so you see all the sensors at once (or at least most of them).

So, if you leave HWinfo sensors running, do a 10-minute Cinebench run as mentioned in my guide, and then show me a screenshot of the HWinfo sensors and tell me the points from Cinebench, i can check what is up.

In general, 14th gen (the true models, meaning the -K ones) are all more power-hungry than 13th gen, and considerably more than 12th gen. The 14th gen -K models were the peak of high power draw and inefficiency. But we'll see what happens during Cinebench.
HWinfo.jpg

I hope I did everything right, sorry that the program is not in English
 
Sorry, can you set the BIOS to English and post the same screenshots? For HWinfo it's also better in English, those terms are all used in English internationally anyway. But for HWinfo i see enough on your screenshot. In the BIOS i see a few settings that might explain some oddities, but more when i can see it all in English.
 
Sorry, can you set the BIOS to English and post the same screenshots? For HWinfo it's also better in English, those terms are all used in English internationally anyway. But for HWinfo i see enough on your screenshot. In the BIOS i see a few settings that might explain some oddities, but more when i can see it all in English.
INFOOO.jpg

Bios1.jpg

bios2.jpg

Bios3.jpg

Bios4.jpg

Bios5.jpg

Well, I hope I did everything right this time. I'll wait for your answer.
 
Yes, this is better, thank you. So, i see a few things that i would set differently.

1) Directly under OC, set "CPU Cooler Tuning" to [MSI Performance]. This usually includes the highest power limits that make sense to use (253W) for any CPU, and it should set the CPU Current Limit to 307A, which also makes sense. Not that your CPU reaches these, but with a 14th gen -K series it's nice to have some kind of limits set.

2) Set "CPU Base Clock" to [100 MHz]. A lot of other clocks are derived from this, and it's best to keep it on 100 MHz. You will never ever feel for example the RAM clocks being 2.5% higher (because the RAM performance doesn't rise as much from this, and the overall performance not at all). It's just another variable that should be removed by sticking to the spec of 100 MHz. You can then enable "LOCK BCLK 100.0" if it becomes available, which will disable so-called Spread Spectrum.

3) A lowered CPU Lite Load is good. But then you don't have to mess with "CPU Load Line Calibration Control" (LLC), because that will essentially raise the voltages a bit again, and in my testing, it was completely counterproductive. Read the very last paragraph of the guide here (last paragraph of first post). Somehow a changed LLC massively undid everything i tried with a lower CPU Lite Load. So set LLC to Auto again and try only with a lowered CPU Lite Load. You may have to redo the stability testing to some degree, because now the "crutch" of the LLC is gone, but the overall end result should be better. Then you can do another Cinebench run and post a screenshot.

4) Minor detail, you can manually set "C1E Support" and "Intel Speed Shift" to [Enabled], i always do that.
 
Yes, this is better, thank you. So, i see a few things that i would set differently.

1) Directly under OC, set "CPU Cooler Tuning" to [MSI Performance]. This usually includes the highest power limits that make sense to use (253W) for any CPU, and it should set the CPU Current Limit to 307A, which also makes sense. Not that your CPU reaches these, but with a 14th gen -K series it's nice to have some kind of limits set.

2) Set "CPU Base Clock" to [100 MHz]. A lot of other clocks are derived from this, and it's best to keep it on 100 MHz. You will never ever feel for example the RAM clocks being 2.5% higher (because the RAM performance doesn't rise as much from this, and the overall performance not at all). It's just another variable that should be removed by sticking to the spec of 100 MHz. You can then enable "LOCK BCLK 100.0" if it becomes available, which will disable so-called Spread Spectrum.

3) A lowered CPU Lite Load is good. But then you don't have to mess with "CPU Load Line Calibration Control" (LLC), because that will essentially raise the voltages a bit again, and in my testing, it was completely counterproductive. Read the very last paragraph of the guide here (last paragraph of first post). Somehow a changed LLC massively undid everything i tried with a lower CPU Lite Load. So set LLC to Auto again and try only with a lowered CPU Lite Load. You may have to redo the stability testing to some degree, because now the "crutch" of the LLC is gone, but the overall end result should be better. Then you can do another Cinebench run and post a screenshot.

4) Minor detail, you can manually set "C1E Support" and "Intel Speed Shift" to [Enabled], i always do that.
Thanks for the answer, I'll try to do it today
 
Yes, this is better, thank you. So, i see a few things that i would set differently.

1) Directly under OC, set "CPU Cooler Tuning" to [MSI Performance]. This usually includes the highest power limits that make sense to use (253W) for any CPU, and it should set the CPU Current Limit to 307A, which also makes sense. Not that your CPU reaches these, but with a 14th gen -K series it's nice to have some kind of limits set.

2) Set "CPU Base Clock" to [100 MHz]. A lot of other clocks are derived from this, and it's best to keep it on 100 MHz. You will never ever feel for example the RAM clocks being 2.5% higher (because the RAM performance doesn't rise as much from this, and the overall performance not at all). It's just another variable that should be removed by sticking to the spec of 100 MHz. You can then enable "LOCK BCLK 100.0" if it becomes available, which will disable so-called Spread Spectrum.

3) A lowered CPU Lite Load is good. But then you don't have to mess with "CPU Load Line Calibration Control" (LLC), because that will essentially raise the voltages a bit again, and in my testing, it was completely counterproductive. Read the very last paragraph of the guide here (last paragraph of first post). Somehow a changed LLC massively undid everything i tried with a lower CPU Lite Load. So set LLC to Auto again and try only with a lowered CPU Lite Load. You may have to redo the stability testing to some degree, because now the "crutch" of the LLC is gone, but the overall end result should be better. Then you can do another Cinebench run and post a screenshot.

4) Minor detail, you can manually set "C1E Support" and "Intel Speed Shift" to [Enabled], i always do that.
Yes, these are probably the best settings for me, the temperature, although a little, but dropped. I pass the R23 test without problems, I turned it on for 10 minutes and I get very good points. Thank you again for all the explanations
Bench.jpg
 
Nice, and you're welcome. But Cinebench R23 isn't a very good stability test, it's only a good CPU benchmark. For stability testing you can use OCCT (CPU only test and Linpack test) for example, Prime95, things like that. Since they use some "dirty tricks" to stress the CPU, they will stress it more than CB23 can.
 
Back
Top