What are the missing variables that demonstrate the actual causation between cooling and results?

[TPCEA]

In my own testing, and in the testing of some other people, I've noticed anomalies where the data we're tracking does not reflect upon the CPU/GPU temperatures, nor does it correlate with the clock speeds.

This means that we're not tracking the right things in HWInfo or whatever. Yet, there are so many options and I do not know what else to consider. I track averages and maxes for (DTS and Enhanced) CPU core temps and CPU package; GPU hot spot, memory junction and temperature; average effective clocks, max CPU/thread usage, and total CPU usage; motherboard and VRM temps, and the fan speeds. Admittedly, I do this testing in a case, unlike many others who use a test bench (horiz or vert), and, also unlike many others, I don't lock everything down in BIOS. I see these data points don't explain why I get the clocks, CPU/thread usage and total CPU usage, and they also don't correlate with the scores I get in CB24 and Time Spy.

Example: If you look at Kleo Yan's results (sorry, is it ok to link to that video?) the same is evident, even though he also tracks memory temps and SSD temps. He locks everything down in BIOS AFAIK, and he uses a test bench.




Does anyone know what's missing that explains why, for example, A1 and E1 have different temperatures but identical clock speeds? Thank you!

Edit: TO CLARIFY: I need to know which sensors to monitor to discover the causal relationship between cooling and performance. None of the above are doing that, nor are the ones I've been watching, nor does watts.
Example of what I'm watching at this time:

 

[ir_cow]

360 AIO vs 240 AIO? The video explains the 240 is on the side while the 360 is mounted to the top (I think). Just so happens the same picture is used for A1 and E1. Honestly this results pages is a bit out of context without watching the whole video. Also this guy kinda just jumps around in explaining the setup.

 

[Caring1]

Different CPU block on the AIO?
The 240 has a fan that helps cool the VRMs and surrounds.
I'm surprised by how much cooler the NVME is with a vertical GPU.
Seems to be worth the change.

 

[TPCEA]

360 AIO vs 240 AIO? The video explains the 240 is on the side while the 360 is mounted to the top (I think). Just so happens the same picture is used for A1 and E1. Honestly this results pages is a bit out of context without watching the whole video. Also this guy kinda just jumps around in explaining the setup.

So how is it that a 280 "won" over a 360 of the same brand and product line (minus lights) i.e. the Arctic Liquid Freezer II 280 and 360, but only one once, really, cuz F1 lost to E1 and A1? And since we now use LEDs that are on their own independent power line, we can't say that the LEDs are leaching the power. Well, actually, I recently got one with LEDs that steal power from the pump. Yay.

Also, A1 has better CPU clocks than C1 and D3 despite a higher temperature. Then there's the GPU clocks in C1 and D3, which match A1, despite different temperatures across the board. This is what I'm talking about. The temps don't show why the performance is what it is. The same thing is evident in my own testing. I appreciate your response, but it doesn't respond to my questions. If you can answer them, that would be much appreciated.

Different CPU block on the AIO?
The 240 has a fan that helps cool the VRMs and surrounds.
I'm surprised by how much cooler the NVME is with a vertical GPU.
Seems to be worth the change.

A CPU block? Do you mean the block, the cold plate and chamber(s)?
The 280 and 360 in the picture are both the same product line - both with VRM fan, same block. Nah, it still wouldn't be the block if they were different brands, unless the engineers really screwed up. Oh, wait, Jimmy who just finished their engineering degree may not actually know enough... ;P How do I know it's not the blocks? Because all of my testing (50+ tests and counting) is on ONE AIO mounted at the 3 points on the 5000X, intake/exhaust, tubes up/down, fans inside/between the case and rad, and, finally, I did the tests with the AIO both on a vertical mount and in the PCIe slot. The anomalies just keep happening. How can higher temperatures produce better results? Yet, they do. It's because I don't know which are actually the sensors that provide the causal link.

As for the NVME, I'm not at all surprised. First, because my testing upholds that, although the differences in the picture and the top contenders in my testing are marginal differences in terms of performance. Second, when you move the GC from the PCIe slot to an independent mount, you move its heat away from the SSDs so, yeah, a vertical mount is worth it IF the mount location provides enough airflow space so that it won't cause the GC to heat up too much. I had to get an EZDIY-FAB (overpriced) vertical mount that can be angled in order to counteract the overheating issue that the stupid vertical mount on the 5000X causes. It sits on the PSU shroud. It helps, sometimes significantly.

Still, I really need help identifying the correct sensors. Do you know?

 

[cvaldes]

My guess is some of the data variances are due to the inherent differences from manual thermal paste application and cooler installation.

Not to mention sample sizes below statistical significance.

All of the cooling product performance reviews really need to be taken with a large grain of salt. They are just general observations and not gospel. In the end, you need to try out whatever device in your own specific build with your actual parts and see if it provides a satisfactory result. There are always manufacturing variances that will not allow me to duplicate exact performance that some reviewer from Tom's Hardware, AnandTech, FPS Review, etc. achieves.

So when I was shopping around for AIOs, I read some reviews and ended up with Product X. I installed it and got performance results that were in the ballpark for Site A's review of Product X. That's good enough for me, I can put it behind me and move on with my life.

If Product Y on Site B has performance that 0.5 °C better than Product X on Site A, it's pretty pointless for me to chase that difference. Even if Site A has hundreds of product reviews of AIO coolers, there's still the sample size limitation, the fact that a human being applies the thermal paste and attaches the coldplate to the CPU mount.

 

[TPCEA]

My guess is some of the data variances are due to the inherent differences from manual thermal paste application and cooler installation.

Not to mention sample sizes below statistical significance.

All of the cooling product performance reviews really need to be taken with a large grain of salt. They are just general observations and not gospel. In the end, you need to try out whatever device in your own specific build with your actual parts and see if it provides a satisfactory result. There are always manufacturing variances that will not allow me to duplicate exact performance that some reviewer from Tom's Hardware, AnandTech, FPS Review, etc. achieves.

So when I was shopping around for AIOs, I read some reviews and ended up with Product X. I installed it and got performance results that were in the ballpark for Site A's review of Product X. That's good enough for me, I can put it behind me and move on with my life.

If Product Y on Site B has performance that 0.5 °C better than Product X on Site A, it's pretty pointless for me to chase that difference. Even if Site A has hundreds of product reviews of AIO coolers, there's still the sample size limitation, the fact that a human being applies the thermal paste and attaches the coldplate to the CPU mount.

Thank you for your response.

I'm not able to comment on Mr. Yan's efforts, and what you wrote makes sense, but I'm not remounting my AIO between tests, and I'm seeing the same odd disconnect in over 50 tests. Lower temps with worse results than higher temps just don't make sense, but I keep seeing it. Where is this problem occurring?

Yes, everything needs to be taken with a grain of salt, and for many, many reasons, but this isn't going to help us to do a better job at testing. Variances in TIMs are, I would suspect but haven't delved into, likely well within the range of normalcy, unless there's an ulterior motive, or the person made a significant error in TIM application, or they kept switching TIMs. I'm gonna have to get a LGA 1700 sized slice of Kryosheet so that I can make sure to test not only with the included TIM, but have a baseline that shows how much of a difference the stock TIM provides (+ or -), and I'll also test with Super Flower Megacools instead of the stock fans, because some AIOs are so weak that those fans will make a significant difference, I suspect.

Yes, it's true that insignificant variances from one tester to another are...insignificant. It's also true that some (many?) are shilling for whoever sends them money - technowhores, really. Almost everything in my inventory was purchased with my own earnings, and the 3 fans SilverStone sent me to review, no strings attached, received the same treatment as every other fan I've done an initial review of - an impartial one, although I did feel slightly guilty but realized that products that are gifted are ultimately a marketing trick to try to purchase a favorable review - but no contract to support that hope.

I don't know why I'm not actually getting responses to the question I (I guess poorly) asked as far as which sensors to monitor. If you know which sensors should the causal relationship between cooling and performance, I'd love to know.

 

[cvaldes]

These days I don't fixate on individual sensor readings. If the primary components (CPU, GPU, motherboard, SSDs) are within an acceptable range of operating temperature, I think the cooling system is doing its job.

The most important sensors I pay attention to are my ears. I focus on acoustics. If the fans are quiet and the system is relatively cool, everything is set up correctly in my mind. I'm not going to whip out some sort of soundmeter to see if Fan X from Brand A is 0.2 dB/A quieter than Fan Y from Brand B.

So I paid a little more for my Noctuas and Arctic fans than some no-name Chinese fan or an old-school Sunon. That's fine, I can live with it. And Noctua and Arctic will provide better customer support and warranty service anyhow than some fly-by-night seller on AliExpress.

I'm not the guy who keeps a temperature gauge running in my menu bar or some performance overlay like MSI Afterburner/RTSS. When I'm on my computer I have more important things to do than fixate on component temperatures. About the only time I look at these numbers more carefully is when I'm configuring a new build. On the hottest day of the year (typically in August), I might review the fan curves in some of my builds to see if I can optimize some but often I make no adjustments.

 

[GerKNG]

Physics.
The RGB Fans of the Arctic AIO are different and perform significantly worse.

 

[TPCEA]

These days I don't fixate on individual sensor readings. If the primary components (CPU, GPU, motherboard, SSDs) are within an acceptable range of operating temperature, I think the cooling system is doing its job.

The most important sensors I pay attention to are my ears. I focus on acoustics. If the fans are quiet and the system is relatively cool, everything is set up correctly in my mind.

I'm not the guy who keeps a temperature gauge running in my menu bar or some performance overlay like MSI Afterburner/RTSS.

In other words, and I was really hoping otherwise, you don't know. That's ok. I appreciate your honesty.

Personally, I do not care about noise until it gets to the level that extended exposure to it will cause damage like, oh, working at Amazon KCVG. Totally random example, of course. XD

I do, however, appreciate that other people value quietness more than or equal to performance, so I will also do tests of noise from 10-100% speed. In an actual case, not a test bench, because almost no one (statistically speaking, this is true, but I know that there must be plenty of people using them for SOMETHING, otherwise companies wouldn't keep making them and the price for one would be ridiculous) runs their computer in a test bench all the time.

I don't keep a temperature gauge running, either, but when I'm doing testing I have two FCs that show me temperature sitting near my PC's base. Unsurprisingly, the one that's a few inches from it always shows a higher temperature than the one that's over 1 foot away and there's a piece of wood and stuff in the way. "Stuff" is a highly technical term that means a partial piece of cardboard. Sometimes, I even add a tall paperboard box to block the flow of heat from the side vent into that area so I get a proper reading. Mostly that's just needed when I've got the side vent set to exhaust.

Being a tester is so much fun.

Physics.
The RGB Fans of the Arctic AIO are different and perform significantly worse.

Good point.
280=2x140mm fans:
Airflow:
72.84 cfm | 123.76 m3/h

Static Pressure:
2.4 mmH2O

360=3x120mm fans:
Airflow:
48.8 cfm

Static Pressure:
1.85 mmH2O

However, there are two problems with this.

1. A1 (360) outperformed F1 (280), despite better temps overall.
2. My testing (again, over 50 tests) have been performed on just ONE AIO with the same fans, except once when I tried out the Megacools - and I'm not counting that test with these.

Finally, and again, do you know which sensors will demonstrate the causal relationship between cooling and performance?

 

[cvaldes]

In other words, and I was really hoping otherwise, you don't know. That's ok. I appreciate your honesty.

Personally, I do not care about noise until it gets to the level that extended exposure to it will cause damage like, oh, working at Amazon KCVG. Totally random example, of course. XD

I do, however, appreciate that other people value quietness more than or equal to performance, so I will also do tests of noise from 10-100% speed.

Ultimately I see these cooling components as part of a holistic solution.

The primary purpose of these parts are to move heat away from the components that generate it to outside of the chassis.

I only pay attention when I'm configuring a new build. Some of my computers I really don't have to think about it, like the Mac mini M2 Pro (in my System Specs).

One thing for sure, I'm not going to outwit some thermal designer who has a Ph.D from one of the world's best engineering programs in Mathematics, Physics, or Mechanical Engineering. Someone who has probably done this for 20+ years and has multiple patents with exciting names like "thermal zone monitoring in an electronic device" and "computer thermal system". Those guys picked the components and set up the fan curve in my Mac mini. They did a good job.

Like many I do care about fan acoustics. If I didn't I'd set all fans in my PC builds at 100%. And for me, noise is a distraction that comes before hearing damage. All the mechanical designers at these companies that design and produce fans have some sort of awareness about acoustics. No computer fan will get to the point where it will cause permanent hearing loss. Annoyance is a far greater issue for PC components in terms of sound.

As a tester, it makes zero sense to say "I don't care about noise" because many people who buy fans actually care A LOT about noise. Acoustic characteristics are a very important metric about fans especially for consumers since your ears will be about a meter away from them. Acoustics might be less important to people who run datacenters but that's not what TechPowerUp's audience is about.

If you are doing your fan tests with your primary audience being datacenter IT managers, yeah, you can put less emphasis on fan acoustics. Those guys will be focusing on performance per watt, performance per dollar, performance-TCO ratios, and maybe durability rather than noise-normalized performance.

 

[TPCEA]

No computer fan will get to the point where it will cause permanent hearing loss.

Think carefully, you've written a blanket statement. As an ex-English teacher, there are few situations where a blanket statement is 100% true. I already have something in mind (server fans can exceed 80dBA). However, I agree that the likelihood of finding a PC fan that will hurt your ears is very low.

Ultimately I see these cooling components as part of a holistic solution.

That's fine - that's your need level.

One thing for sure, I'm not going to outwit some thermal designer who has a Ph.D from one of the world's best engineering programs in Mathematics, Physics, or Mechanical Engineering. Someone who has probably done this for 20+ years and has multiple patents with exciting names like "thermal zone monitoring in an electronic device" and "computer thermal system". Those guys picked the components and set up the fan curve in my Mac mini. They did a good job.

You and me both. I'll never hold a patent.

As a tester, it makes zero sense to say "I don't care about noise" because many people who buy fans actually care A LOT about noise.

1) It is my personal preference - so what? - you have no way to invalidate my personal preference. 2) Did you not read where I stated that others care about noise?

Acoustic characteristics are a very important metric about fans especially for consumers since your ears will be about a meter away from them.

To some PC users, not to all. I can sit here and listen to the harmonic resonance of my AIO's fans and tune it out. It's just not worth being bothered by inconsequential noises, at least not for me. If the dBA reaches a high enough level, then I will consider if there's a way to reduce it, or I will finish my work and let it go back down to normal (barely audible). Heck, my window fan is louder than my PC most of the time, and it's over 2 meters away, whereas my PC is a foot from my left knee.

Acoustics might be less important to people who run datacenters but that's not what TechPowerUp's audience is about.

If you are doing your fan tests with your primary audience being datacenter IT managers, yeah, you can put less emphasis on fan acoustics.

Irrelevant and probably inaccurate, since server rooms are usually separate from where people do their work, unless they're working on the servers and there's no remote terminal set up.

But none of this is germaine to my topic. Can you please stay on topic?

 

[ARF]

Does anyone know what's missing that explains why, for example, A1 and E1 have different temperatures but identical clock speeds? Thank you!

Because 5.03 GHz is the max clock in your case, while it depends on the CPU itself and its surrounding software whether that max clock will be reached at 100%, or at 99.2%.

I'm surprised by how much cooler the NVME is with a vertical GPU.
Seems to be worth the change.

I am surprised that those who designed the motherboards made the extremely illogical decision to put the M.2 slots directly under the hottest component in the PC case - the GPU.

 

[TPCEA]

Because 5.03 GHz is the max clock in your case, while it depends on the CPU itself and its surrounding software whether that max clock will be reached at 100%, or at 99.2%.

I realize that Kleo Yan's CPU maxes out at 5.03 GHz. Mine isn't OC'd, and the max I've seen is about 4,564.8 MHz, which was while running Cinebench 2024's CPU Multi-core test and my GPU was at its lowest temp for that test (21C+ambient), which doesn't stress my system as much as Prime 95 (for the CPU) and Time Spy (for the GPU). Is there a bell curve whereby results decline prior to TJMax? That doesn't seem to make sense, so...

Unfortunately, what the max clock speed is doesn't answer my question about the inconsistencies between runs, it just shows that the inconsistencies demonstrated from the data of the tracked sensors are not showing causality.

What I need to know is, in HWiNFO64, which sensors will show the causal relationship between cooling and performance. I do not see that in his results, mine, and other testers who actually show their data (raw or adjusted). Watts is a consequence of cooling and can indicate increased performance, but it is not the CAUSE of increased performance - it is the result of whatever was positively affected such that performance could increase because temperatures SOMEWHERE decreased in a helpful (to the CPU) way, resulting in a higher power draw. Where is/are the actual key point(s) that, when cooled, definitively cause better performance? We are tracking the wrong sensors!

Note, I do understand since I don't have climate control, that it's going to potentially impact my results but, as with Kleo and others, I'm seeing higher temps giving better results than lower temps sometimes. Also, they're using climate control and most use a test bench!

Additionally, on my PC, even though I don't OC, as the ambient temperature reaches a certain point (which is around 22C, without A/C) and the load is high, my computer becomes sluggish even though I haven't reached tjmax on my CPU and GPU. Since my CPU and GPU aren't throttling, something else is getting too hot.

Intel Core i7-12700K Specifications

Lithography	Intel 7
Max Turbo Frequency	5.0GHz
Performance-core Max Turbo Frequency	4.9GHz
Efficient-core Max Turbo Frequency	3.8GHz
Performance-core Base Frequency	3.6GHz

Here's a test I'm currently running in Prime 95, Small FFTs (the most demanding):

What sensors am I missing? I know people want to see watts but for which sensor? And which sensors show critical bottleneck areas on a mainboard?

 

[Waldorf]

@TPCEA
always go with the sensors that show the hottest temps

you will always have variations, even doing the same thing on the same setup multiple times, and it includes temps.
just a tenth amount of difference in certain clocks will have an impact (e.g. bus at 100.0 vs 100.x), and like others said repasting, room temp not steady..

one reason some cooler reviews are done with a fake-cpu "heater" with constant output, not affected by anything (much).

setting up the cooler comes down to what type of aio it is (refillable or CLC) and what you prefer to have priority for temps.
blowing hot air in might drop your cpu temp a little, but will make the case/mb/gpu/drives hotter than they need to be, impacting boost clocks of the gpu, dropping fps.
so for gaming it should be mounted to exhaust heat outside the case, mounted to the case side if there is an option, so the top fans can exhaust the hot air from the gpu.

with arctic you want the fans on inside of rad, pushing thru, as the P-series fans prefer that.

no computer fan will damage your ear from noise, as all 1st world countries have (work) noise limits, and those incl pc's (and their fans).
no company will waste money to produce two different types, one unit within noise limits to sell in countries with noise regulation,
and another version (loud) for other countries, so you only get stuff that is below a harmful threshold.

 

[ShrimpBrime]

Watts is not consequence of cooling.

The cooling is the consequence of watts.

5w chips don't need a heat sink.

15w-50w need a heat sink, but maybe not a fan.

Usually higher than 50w needs active cooling with fans on a heatsink.

You simply convert electrical energy to a thermal energy with a cpu. It moves a little data in the process.

_______

Sluggish when warm ambient.

The PC doesn't have an external ambient temp sensor. So the temps are not accurate in the sense it's compared to something. This causes the deviation between 2 identical systems. And why multimeter with temp probes are accurate.

Most power supplies dont like over 40c operating temps. Something to consider paying attention to. Memory, hard drives and voltage regulators.

Another thing to consider is a warm processor leaks voltage. It is possible to have lower performance just from this. Voltage keeps transistors operational. As voltage leaks, some of the transistors will stop working. When too many stop, this can be witnessed as a hang. See it all the time, I'm more into over-volt than under however.

 

[TPCEA]

@TPCEA
always go with the sensors that show the hottest temps

you will always have variations, even doing the same thing on the same setup multiple times, and it includes temps.
just a tenth amount of difference in certain clocks will have an impact (e.g. bus at 100.0 vs 100.x), and like others said repasting, room temp not steady..

one reason some cooler reviews are done with a fake-cpu "heater" with constant output, not affected by anything (much).

setting up the cooler comes down to what type of aio it is (refillable or CLC) and what you prefer to have priority for temps.
blowing hot air in might drop your cpu temp a little, but will make the case/mb/gpu/drives hotter than they need to be, impacting boost clocks of the gpu, dropping fps.
so for gaming it should be mounted to exhaust heat outside the case, mounted to the case side if there is an option, so the top fans can exhaust the hot air from the gpu.

with arctic you want the fans on inside of rad, pushing thru, as the P-series fans prefer that.

no computer fan will damage your ear from noise, as all 1st world countries have (work) noise limits, and those incl pc's (and their fans).
no company will waste money to produce two different types, one unit within noise limits to sell in countries with noise regulation,
and another version (loud) for other countries, so you only get stuff that is below a harmful threshold.

I know that but the sensors I watch, the ones that Kleo Yan watches, and the others who show their data are not revealing the cause. That is what I'm asking for. I appreciate your response, but do you know the answer to my actual question?

As to no computer fans being able to damage your ears, I say to you the same thing as the last person: this is not true unless you ignore server fans. I have the bgears b-blaster 140x38R, and that thing is pretty noisy, but I have fans that are even louder than that. I assure you, Waldorf, there are fans that are loud enough to do damage.

Watts is not consequence of cooling.

The cooling is the consequence of watts.

I appreciate your response. Totally disagree on this. Watts produce heat while fueling performance. Enough heat can slow down the computer. If you cool off the computer enough, not only will it recover but, with a low enough temperature, it will actually exceed the previous speeds and use more watts because it is processing more data. This is amply demonstrated in so many test videos and articles I've seen and, further, sub-ambient efforts are irrefutable proof of this.

Sluggish when warm ambient.

I said that in my last message.

The PC doesn't have an external ambient temp sensor. So the temps are not accurate in the sense it's compared to something. This causes the deviation between 2 identical systems. And why multimeter with temp probes are accurate.

I have two separate FCs with thermocouples in different positions to help me know the ambient temperature. I'm pretty confident that I'm getting a pretty good idea of my ambient temperature. My multimeter doesn't have probes and I'm not sure where to find ones that would work with it.

Most power supplies dont like over 40c operating temps. Something to consider paying attention to. Memory, hard drives and voltage regulators.

Good to know. Kleo and others paid attention to memory, VRMs and storage, but there was nothing conclusive, and I paid attention to VRMs, too, with nothing conclusive. One reviewer, I think it was HW Canucks, tested VRM fans and, while temps dropped, there was no change in performance.

Another thing to consider is a warm processor leaks voltage. It is possible to have lower performance just from this. Voltage keeps transistors operational. As voltage leaks, some of the transistors will stop working. When too many stop, this can be witnessed as a hang. See it all the time, I'm more into over-volt than under however.

I'm not sure what you mean by "leaks", but I think it's really about resistance caused by increased heat.

Please, do you have the answer to my question about which specific sensors to monitor? That is all I'm asking for but not one person has actually read what I've written and asked and given me an answer.

 

[ShrimpBrime]

I know that but the sensors I watch, the ones that Kleo Yan watches, and the others who show their data are not revealing the cause. That is what I'm asking for. I appreciate your response, but do you know the answer to my actual question?

As to no computer fans being able to damage your ears, I say to you the same thing as the last person: this is not true unless you ignore server fans. I have the bgears b-blaster 140x38R, and that thing is pretty noisy, but I have fans that are even louder than that. I assure you, Waldorf, there are fans that are loud enough to do damage.


I appreciate your response. Totally disagree on this. Watts produce heat while fueling performance. Enough heat can slow down the computer. If you cool off the computer enough, not only will it recover but, with a low enough temperature, it will actually exceed the previous speeds and use more watts because it is processing more data. This is amply demonstrated in so many test videos and articles I've seen and, further, sub-ambient efforts are irrefutable proof of this.

I said that in my last message.

I have two separate FCs with thermocouples in different positions to help me know the ambient temperature. I'm pretty confident that I'm getting a pretty good idea of my ambient temperature. My multimeter doesn't have probes and I'm not sure where to find ones that would work with it.


Good to know. Kleo and others paid attention to memory, VRMs and storage, but there was nothing conclusive, and I paid attention to VRMs, too, with nothing conclusive. One reviewer, I think it was HW Canucks, tested VRM fans and, while temps dropped, there was no change in performance.


I'm not sure what you mean by "leaks", but I think it's really about resistance caused by increased heat.

Please, do you have the answer to my question about which specific sensors to monitor? That is all I'm asking for but not one person has actually read what I've written and asked and given me an answer.

That wasn't meant to be picked apart.

Just because you know the ambient temp, your cpu does not. It's a temp algorithm, not a direct data source.

To your answer, the 2 main sensors would be cpu core and cpu package temps. Core is die temp, package is off-die temps.

For wattage, total package power, for the cpu.

And performance, by which method of measurements? Cinebench?

Is the measurements of Cinebench done over an average or pin point single case runs that determined an outcome to the curiosity of these questions?

 

[Waldorf]

@TPCEA
because its not relevant.
you can have mouse movement waking up a core, spiking power/temps, so until you start monitoring everything the os+sw does,
parallel to cpu usage and temps, you might get closer, but many things the cpu does (internally) and for whatever reason,
will make it probably impossible, short of being a chip/board manufacturer, and able to do things without using things like os for testing.


damage occurs because of (noise) level AND time, not just one.
sure there are fans loud enough, but not those that can be used in office/end user computers, and why i stated so.
it needs to be above 85db, to be harmful, and then only if its hours every day, which 1st world countries have laws regulating it that it wont happen.
e.g. in germany its 25db for any computer at work, so no, wont happen unless you are "misusing" a fan, not designed to be in a end user pc.


it gets slower when it gets hot, because it requires more voltage at higher temps.
ppl oc stuff using LN, because it reduces voltage requirement at every level, because its colder, not because there is more power going thru.
if it was different, there wouldnt be much of a limit to how much chips can take..

leakage has nothing to do with resistance, but how much power gets "away".

 

[TPCEA]

That wasn't meant to be picked apart.

Just because you know the ambient temp, your cpu does not. It's a temp algorithm, not a direct data source.

To your answer, the 2 main sensors would be cpu core and cpu package temps. Core is die temp, package is off-die temps.

For wattage, total package power, for the cpu.

And performance, by which method of measurements? Cinebench?

Is the measurements of Cinebench done over an average or pin point single case runs that determined an outcome to the curiosity of these questions?

Well, if you're going to make something as separate paragraphs, then it is pickapartable.

Thank you for getting to my question. It's much appreciated!

I see. How can an algorithm calculate the temp?? Please explain the equation.

I already track CPU core and package - both DTS and Enhanced. It does not show causality. It's very frustrating. Package, as I understand it, is above the core sensor. Is that correct?

TPP. Got it.

I was running P95 Small FFTs for 1 hour (now I'm on phase 3 to optimize the addition of more fans since I only used the 360 AIO fans and an A12x25 for exhaust, so I am only doing 30 minutes as 1 hr isn't really necessary or helpful), then I run CB24 CPU Multi-core, and then I run Time Spy. For the first two tasks, I've also got a video of some sort running, and I often do something else, like data entry. Each test run was done once for most configurations that I tested (50+), but sometimes I did additional tests when something seemed off, or I had the chance to do it over a range of temperatures. When I get to phase 4, I will be running multiple iterations (at least 3) for each AIO I test, and I will also test the AIO with the stock TIM and then a Kryosheet - separately and in combination with Super Flower Megacools on the AIO.

 

[ShrimpBrime]

Well, if you're going to make something as separate paragraphs, then it is pickapartable.

Thank you for getting to my question. It's much appreciated!

I see. How can an algorithm calculate the temp?? Please explain the equation.

I already track CPU core and package - both DTS and Enhanced. It does not show causality. It's very frustrating. Package, as I understand it, is above the core sensor. Is that correct?

TPP. Got it.

I was running P95 Small FFTs for 1 hour (now I'm on phase 3 to optimize the addition of more fans since I only used the 360 AIO fans and an A12x25 for exhaust, so I am only doing 30 minutes as 1 hr isn't really necessary or helpful), then I run CB24 CPU Multi-core, and then I run Time Spy. For the first two tasks, I've also got a video of some sort running, and I often do something else, like data entry. Each test run was done once for most configurations that I tested (50+), but sometimes I did additional tests when something seemed off, or I had the chance to do it over a range of temperatures. When I get to phase 4, I will be running multiple iterations (at least 3) for each AIO I test, and I will also test the AIO with the stock TIM and then a Kryosheet - separately and in combination with Super Flower Megacools on the AIO.

Temperature is assumed from voltage across on die thermistor. 1 for each core. I do not know the exact equation or values being used. My guess would be variations across whatever the thermistor is assigned, perhaps 5v.

But this isn't 4 cpu core die sensors each. How would we know if the center of a die is cooler than an outside edge, visa versa.

Then we have the temp program. Which one is most accurate?

Why not use HWMonitor opposed to HWInfo? Which one is closest to the bios reading I suppose would be the answer. Again, this all running through windows.

What about sensors that have a crazy reading instead of 0.0 or blank? Because the software is written with such a feature, it's just going to pin at a maximum assumed value.

Measurements. The best average value will be done with larger amount of tests. Is 3 enough to conclude average performance to average wattage used?

We just want consistent results mainly.
I would gather interest with v-core stock and LLC adjustments to accommodate possible drooping when undesired.

 

[TPCEA]

Temperature is assumed from voltage across on die thermistor. 1 for each core. I do not know the exact equation or values being used. My guess would be variations across whatever the thermistor is assigned, perhaps 5v.

But this isn't 4 cpu core die sensors each. How would we know if the center of a die is cooler than an outside edge, visa versa.

Then we have the temp program. Which one is most accurate?

Why not use HWMonitor opposed to HWInfo? Which one is closest to the bios reading I suppose would be the answer. Again, this all running through windows.

What about sensors that have a crazy reading instead of 0.0 or blank? Because the software is written with such a feature, it's just going to pin at a maximum assumed value.

Measurements. The best average value will be done with larger amount of tests. Is 3 enough to conclude average performance to average wattage used?

We just want consistent results mainly.
I would gather interest with v-core stock and LLC adjustments to accommodate possible drooping when undesired.

Thanks for the clarifications.

Which monitoring program do you think is best?

Of course 3 is not enough, but I'm not going to do 100 tests, either. If I can live forever (unlikely given my heart), I'll do each test 1,000 times, ok?

Please explain that last line - it is a bit confusing.

 

[Noci]

Maybe not exactly what you're asking, but the sensors do have a 0.1% (DTS) up to 5% (thermal diode) deviation margin. So the temps in your example are not exceeding these ranges actually they are pretty close.

HWiNFO is just a monitoring utillity reporting the values that are reported by these sensors; https://www.hwinfo.com/forum/threads/cpu-temp-sensors-explanation.5597/ and as far as I know, it does a good job.

Causation means one thing causes another, in other words, action A causes outcome B, but in this case it is not that simple. Because of the fact that it is complex we have to look at correlation and correlation in it's simplest form is a relationship where action A relates to action B, but one event doesn't necessarily cause the other event to happen. But with chip performance vs cooling it is waaay more complex where action A has relations with multiple other actions of which some will happen and some not.
Because English is not my native language I did not fully understand your question, but I believe that core temps are the most important ones to keep an eye on as they trigger the built in safety mechanisms.

Keep in mind that the whole thermal dynamic interplay consists out of numerous variables of which we can only adjust a few. It is almost an art to find the sweet spot between all these settings and the monitoring is only a tool, not a solution.

 

[TPCEA]

Maybe not exactly what you're asking, but the sensors do have a 0.1% (DTS) up to 5% (thermal diode) deviation margin. So the temps in your example are not exceeding these ranges actually they are pretty close.

HWiNFO is just a monitoring utillity reporting the values that are reported by these sensors; https://www.hwinfo.com/forum/threads/cpu-temp-sensors-explanation.5597/ and as far as I know, it does a good job.

Causation means one thing causes another, in other words, action A causes outcome B, but in this case it is not that simple. Because of the fact that it is complex we have to look at correlation and correlation in it's simplest form is a relationship where action A relates to action B, but one event doesn't necessarily cause the other event to happen. But with chip performance vs cooling it is waaay more complex where action A has relations with multiple other actions of which some will happen and some not.
Because English is not my native language I did not fully understand your question, but I believe that core temps are the most important ones to keep an eye on as they trigger the built in safety mechanisms.

Keep in mind that the whole thermal dynamic interplay consists out of numerous variables of which we can only adjust a few. It is almost an art to find the sweet spot between all these settings and the monitoring is only a tool, not a solution.

I appreciate your response and knowing the margin of error is that large (sigh) is helpful.

Yes, I agree, we're talking about something very complex. Maybe not as complex as DNA, but there are lots of variables that contribute. I'm just hoping that someone can help me find the OTHER key ones since just looking at core, package, mobo, VRMs, RAM and storage temps is too simple. Of course, if there are 100 sensors that need to be monitored, that's going to be a *gasp* moment.

Thanks!

 

[ShrimpBrime]

Thanks for the clarifications.

Which monitoring program do you think is best?

Of course 3 is not enough, but I'm not going to do 100 tests, either. If I can live forever (unlikely given my heart), I'll do each test 1,000 times, ok?

Please explain that last line - it is a bit confusing.

Best?
Well smaller the better. HWInfo64 seems to use quite a bit of resources and can influence benchmark scores.
Gpu-Z is nice, simply reports package temps for cpu.

Maybe 10 or 20 times. Depends on consistency imo.

The LLC thing? What's the board default. Does changing anything make the pc boot LLC at a different level. Is it user controlled? Does the LLC get set low if bios determines high idle temps? Are you planning to controll LLC manually during your testing to determine if this could be a missing variable that isn't measured by the monitoring software....

 

[TPCEA]

Best?
Well smaller the better. HWInfo64 seems to use quite a bit of resources and can influence benchmark scores.
Gpu-Z is nice, simply reports package temps for cpu.

Maybe 10 or 20 times. Depends on consistency imo.

The LLC thing? What's the board default. Does changing anything make the pc boot LLC at a different level. Is it user controlled? Does the LLC get set low if bios determines high idle temps? Are you planning to controll LLC manually during your testing to determine if this could be a missing variable that isn't measured by the monitoring software....

I have the Asus TG H670-Pro Wifi D4. No clue about the default. I haven't messed with that setting at all.

I'll stick with HWiNFO64 if GPU-Z only gives me the package temp. Seriously, dude, I'm looking for causality, not one sensor. That I've already seen again and again isn't tied to causality by itself - not by a long shot.

Here's another question for you: when it comes to my mainboard, are the enhanced CPU package and core accurate or not in HWiNFO64? There's some info in the help file that says it tends to be inaccurate...Which is why I use DTS.