PC Build: MSI Z790 Tomahawk - i5 14600KF

[dasgeist3152b02d5]

Good morning. Just wanted to ask for opinions regarding the results of my i5-14600k that i just built a couple of days earlier.

On the MSI Bios I have PL1-PL2 at 181W and Current Limit at 200A per Intel. At idle the CPU voltage hovers, in general, from 1.235V to 1.37V. Temperatures were around 39C to 42C. When i ran Cinebench R23 multi thread only, I got 22487 points at 150A, roughly 180W and 87C package. P-Core 5 was the only one that throttled.

Idle

Cinebench

Do you think these are good results broadly speaking? I have not touched anything else on the Bios apart from Memory Try it, because with XMP, my ram will not go at 6400Mhz that is rated at, and I have it at 6000MHz. Before the tests and the bios tweak, I had the cooling profile of MSI at tower cooler, without any instability, or high temps, at least in games. The reason I did the Bios was of the current that was on the profile at 400A and that scared me a bit.

On the other hand on 3dMark on the nomad and timespy tests I get bellow average results for the system that I have.

Here is my config:
Case: Kolink Unity Code X
CPU: i5-14600KF
M/B: MSI Z790 Tomahawk WIFI MAX
Ram: G-Skill Trident DDR5 32G @ 6400MHz
GPU: Palit RTX 4080 Super
PSU: BeQuiet! Pure Power 12M 1000W
Cooling: DarkFlash DX-360 V2.6 rated at 288W TDP on top of the case
Fans: 6x DeepCool RF120 (3x on the side and three on the bottom as intakes) and one 120mm that came with the case acting as exhaust

The system has not broken more than 65c on day to day tasks. 87c was the maximum on Cinebench. On Alan Wake2 with everything on max 4K including path tracing I get 27 to 32 fps.

I just need your opinions.

Thanks in advance for the help.

p.s. Sorry for the captures they are from 65" Sony TV at a distance of 9 feet /3m

 

[citay]

At idle the CPU voltage hovers, in general, from 1.235V to 1.37V. Temperatures were around 39C to 42C.

Elevated, especially for this cooling. Hopefully the AIO's radiator fans are configured as exhaust, pulling/pushing out of the top? Otherwise you'd have way too many intake fans compard to the one exhaust fan.

The voltage should be below 1V in idle too, if everything is set up right in the BIOS, and if there's no software running in the background that keeps the CPU busy. But i see that it actually goes well below 1V, check your screenshots, so maybe you're looking at the wrong column. Can you also show a screenshot of the BIOS page where you also set the power limits?

Your CB23 score is in the ballpark, running it with HWinfo in the background always costs a couple hundred points, but if you calculate it to percentages, you're within the margin of error for your CPU model.

As for the system, it looks pretty good, except maybe the case, it doesn't quite convince me. Yes, it's huge, but no dust filters (except on the top, where you don't really need it if the AIO is configured as exhaust). And i personally don't like the "around-the-corner-airflow" cases too much. Simple front-to-back-airflow with full dust filters everywhere for me, please. Other than that, looks good.

There's no word about your BIOS version, which one you got?

About the screenshots, you should redo them properly. I will post my usual instructions again: Open "Sensors", then expand all sensors by clicking on the little <--> arrows on the bottom, also expand the columns of the sensors a bit so everything can be read. Make it three big columns of sensors (or four, if the screen resolution is high enough). In the end, it should be a screenshot with all the sensors visible at once, like this:

Make sure the power plan in Windows is on "Balanced". Do nothing on the PC for a while (couple minutes), so the "minimum" baselines for the values are established. After that time in idle, then produce full CPU load with Cinebench, and after completing a 10 minute run, when the CPU temperatures have stabilized at the highest level, take a screenshot of the sensor window and tell me the Cinebench score.

Now, don't take a photo of the screen, take a screenshot in Windows. Press the PrtScrn key (Print Screen, two above DEL). Or if you press ALT + PrtScrn, it takes a screenshot of the active window (without the taskbar). This is ten times faster than taking a photo, and ten times better quality.

 

[dasgeist3152b02d5]

Thanks for the reply. Yes the radiator fans are configured as exhaust. So 3 fans on the radiator and one on the back acting as exhausts, and 6 as intakes.

Bios version is the latest from MSI. Installed immediately before installing windows. Version ME_16.1.30.2361.

I have relocated the dust filters that came with it. One is in the bottom intake and one on the side intake. Only thing that alarms me, and I have to verify, are the high idle voltages and the XMP that does not work. I have 2x16Gb GSkill Trident Z5 RGB DDR5-6400 CL32-39-39-102 1.40V and, with XMP, the system simply crashes. I have set them at 6000 with Memory Tryit! in the Bios, and is stable.

I will try your suggestions and return with my results.

 

[dasgeist3152b02d5]

Elevated, especially for this cooling. Hopefully the AIO's radiator fans are configured as exhaust, pulling/pushing out of the top? Otherwise you'd have way too many intake fans compard to the one exhaust fan.

The voltage should be below 1V in idle too, if everything is set up right in the BIOS, and if there's no software running in the background that keeps the CPU busy. But i see that it actually goes well below 1V, check your screenshots, so maybe you're looking at the wrong column. Can you also show a screenshot of the BIOS page where you also set the power limits?

Your CB23 score is in the ballpark, running it with HWinfo in the background always costs a couple hundred points, but if you calculate it to percentages, you're within the margin of error for your CPU model.

As for the system, it looks pretty good, except maybe the case, it doesn't quite convince me. Yes, it's huge, but no dust filters (except on the top, where you don't really need it if the AIO is configured as exhaust). And i personally don't like the "around-the-corner-airflow" cases too much. Simple front-to-back-airflow with full dust filters everywhere for me, please. Other than that, looks good.

There's no word about your BIOS version, which one you got?

About the screenshots, you should redo them properly. I will post my usual instructions again: Open "Sensors", then expand all sensors by clicking on the little <--> arrows on the bottom, also expand the columns of the sensors a bit so everything can be read. Make it three big columns of sensors (or four, if the screen resolution is high enough). In the end, it should be a screenshot with all the sensors visible at once, like this:

Make sure the power plan in Windows is on "Balanced". Do nothing on the PC for a while (couple minutes), so the "minimum" baselines for the values are established. After that time in idle, then produce full CPU load with Cinebench, and after completing a 10 minute run, when the CPU temperatures have stabilized at the highest level, take a screenshot of the sensor window and tell me the Cinebench score.

Now, don't take a photo of the screen, take a screenshot in Windows. Press the PrtScrn key (Print Screen, two above DEL). Or if you press ALT + PrtScrn, it takes a screenshot of the active window (without the taskbar). This is ten times faster than taking a photo, and ten times better quality.

just redid the tests. I am attaching pictures of the settings and tests. First are the Bios settings. Since then i have changed the ram settings to 6200MHz.

Next I am attaching the HWinfo results and Cinebench (the Cinebench results are with the new memory settings ie 6200MHz). The first idle photo is what I see most of the time and the second happens once in a while, think of it as momentary spikes. Your opinion is greatly appreciated.

 

[citay]

Bios version is the latest from MSI. Installed immediately before installing windows. Version ME_16.1.30.2361.

The ME version is seperate from the BIOS, and has to be updated from within Windows, but seeing as you did an update before installing Windows, it will have been the BIOS update. And indeed, your BIOS screenshots confirm this.

The BIOS settings seem more or less normal, meaning as expected with one of those latest beta versions, except one thing that strikes me as odd, "CPU Lite Load" on Mode 22, that is a very high default mode, one of the highest i've seen so far, and it means quite a lot of additional VCore (core voltage). This could be something that MSI thought necessary to set in the beta BIOS, to prevent any potential instability, but if they thought like that, it was the wrong way to go about it. If we look at this thread, we might have a bit of a pattern.

Coming to the HWinfo numbers. The CPU enters power-saving modes properly in idle, power draw goes to under 10W as it should be. Still, we have some intermittent load in the background, making it go to 55-65W, this has nothing to do with any BIOS settings, it's purely because something is running in the background. You'd have to check in the task manager (CTRL-SHIFT-ESC), in the advanced view under processes, sorted by CPU, there you should see what is causing CPU load.

Finally, the situation with Cinebench load. Scores are ok, especially with HWinfo running, but this is unsatisfactory cooling performance. With 181W power limits, you are not losing any performance (the 14600K(F) naturally doesn't go much higher than that), but you're already getting thermal throttling, on an E-core of all things. So maybe something about the cooling is not dialed in, for example check the fan curves. Your screenshots use a pretty big font so i can't see the fan sensors. they are just cut off at the bottom right of the list.

In summary, the CPU performance is ok, the cooling performance is not as good as it should be, and lastly, the "CPU Lite Load" setting might need some tuning down. That would also help with the temperatures, however, your cooler shouldn't begin to struggle slightly below 200W already. If they claim 288W TDP for it, it should at least be good for 250W worth of heat until it gets into trouble.

 

[dasgeist3152b02d5]

The ME version is seperate from the BIOS, and has to be updated from within Windows, but seeing as you did an update before installing Windows, it will have been the BIOS update. And indeed, your BIOS screenshots confirm this.

The BIOS settings seem more or less normal, meaning as expected with one of those latest beta versions, except one thing that strikes me as odd, "CPU Lite Load" on Mode 22, that is a very high default mode, one of the highest i've seen so far, and it means quite a lot of additional VCore (core voltage). This could be something that MSI thought necessary to set in the beta BIOS, to prevent any potential instability, but if they thought like that, it was the wrong way to go about it. If we look at this thread, we might have a bit of a pattern.

Coming to the HWinfo numbers. The CPU enters power-saving modes properly in idle, power draw goes to under 10W as it should be. Still, we have some intermittent load in the background, making it go to 55-65W, this has nothing to do with any BIOS settings, it's purely because something is running in the background. You'd have to check in the task manager (CTRL-SHIFT-ESC), in the advanced view under processes, sorted by CPU, there you should see what is causing CPU load.

Finally, the situation with Cinebench load. Scores are ok, especially with HWinfo running, but this is unsatisfactory cooling performance. With 181W power limits, you are not losing any performance (the 14600K(F) naturally doesn't go much higher than that), but you're already getting thermal throttling, on an E-core of all things. So maybe something about the cooling is not dialed in, for example check the fan curves. Your screenshots use a pretty big font so i can't see the fan sensors. they are just cut off at the bottom right of the list.

In summary, the CPU performance is ok, the cooling performance is not as good as it should be, and lastly, the "CPU Lite Load" setting might need some tuning down. That would also help with the temperatures, however, your cooler shouldn't begin to struggle slightly below 200W already. If they claim 288W TDP for it, it should at least be good for 250W worth of heat until it gets into trouble.

Good morning. Thanks for the help! It has been invaluable.

Fan curves for all fans are as follows:
At or bellow 40c 30%
At or bellow 60c 50%
At or bellow 70c 70%
At or bellow 80c 80% (Have changed it since then and have put it at full tilt)
Pump is at full tilt at 2800 rpm
The 6x DeepCool RF120M Fans and the exhaust, at full tilt reach, roughly, 1500rpm
The DarkFlash fans at full tilt reach roughly 1800rpm

Rerun Cinebench and at, roughly, the 2min mark I have throttle at either P-Core 2 or P-Core 5 (Retried it with fans at full tilt same result). Also when I tried to install a game from FitGirl that has highly compressed files, I had the CPU package reach 95c and throttle at cores 2 and 5. Is it possible that the thermal paste is not correctly applied? I did one large dot at the center and four smaller ones in the corners of the lid.

Also I have the 3-pin pump connected to the M/B and the SATA connector to the power supply. However on their site (https://www.darkflash.com/product/twisterdxv2.6360) it seems that the SATA is used for RGB and not for the pump. It is confusing. Still, I don't think that would cause any problem.

Also do you think that the voltages are bit high based on my previous pics? On Cinebench lowest was 1.21V and highest was 1.48V

Sorry to keep bugging you, but the whole Intel thing makes me second guessing myself if is should have gone with AMD instead.

Thanks once again for all your help!

 

[citay]

Yeah, for the fan curves, at 85 or 90°C at the latest, you want to have them at 100% PWM signal (4-pin fans) or 12V (3-pin fans). Same for the pump at those temps.

Rerun Cinebench and at, roughly, the 2min mark I have throttle at either P-Core 2 or P-Core 5 (Retried it with fans at full tilt same result). Also when I tried to install a game from FitGirl that has highly compressed files, I had the CPU package reach 95c and throttle at cores 2 and 5.

Yes, this is the first problem, the unsatisfactory cooling performance. Why is this cooler not performing up to its capabilities. The thermal paste application sounded ok. Maybe it's one of those coolers where the cold plate (what contacts the CPU's IHS) is not shaped ideally for LGA1700 CPUs without a contact frame. Now, bear with me here.

As you may know, on LGA1700 there's the issue of the ILM (Independent Loading Mechanism, what clamps the CPU into the socket) practically bending the CPU in the middle, something that has been known for over two years. First, the remedy was to put washers under the socket mechanism, alleviating some ILM pressure from the CPU, counting on the cooler to potentially do more of that work (i also use the washer mod). Then later it gave rise to the professionally made contact frames, which replace the top part of the socket mechanism altogether, also see here. And also here.

Contact frames are good for lowering CPU temperatures (by preventing bending in the middle of the IHS and thus improving contact between the IHS and the cold plate / cooler base). Ideal would be the latest Thermal Grizzly one, it is machined to the highest standards. der8auer (TG CEO) actually came up with the idea in the first place, the others are just copycats, but of course the TG frame also costs more.

When the base plate / cold plate of the cooler doesn't take that bending into account, the contact won't be ideal, and the temperatures will be higher. This prompted Noctua ro release their newest air cooler in three versions, with different shapes:

Which version of the NH-D15 G2 should I buy (LBC, standard or HBC)?

Our standard recommendation is the standard NH-D15 G2 because it’s the best allrounder and most versatile of the three versions: It provides the best results on LGA1851, AMD AM5 using the included offset mounting, and on LGA1700 CPUs when utilisi...

faqs.noctua.at

So this could explain it to some extent, if this was the case. But maybe not fully. The performance is off by a bit too much. Don't get me wrong, 200W effective is not nothing, that's quite a lot already. But why already thermal throttling with this big cooler. So the simple fact is, with 181W power limits, they are still not low enough for this cooler, this is independent of what is going on with the voltages, this is purely talking about the thermal load which overpowers the cooler. It doesn't matter if you put a 14900KS in there and limit it to 181W, you would have the same result.

Having said that, we should work on lowering the voltages to your individual CPU's actual requirements. I explain that here in step 2). As i said, your CPU Lite Load setting seems way too high. If you can lower it a bunch, that means lower voltages, and as a consequence, lower power draw and heat. So the thermal throttling problem might be taken care of that way. Note that this still doesn't explain why the cooler gets into trouble at 181W power limits already. But if the full load power draw can be lowered enough, maybe you don't have to worry about that so much anymore.

Also I have the 3-pin pump connected to the M/B and the SATA connector to the power supply. However on their site (https://www.darkflash.com/product/twisterdxv2.6360) it seems that the SATA is used for RGB and not for the pump. It is confusing. Still, I don't think that would cause any problem.

Your board has ARGB support, so connect it like on the right side of the first picture, ARGB cable onto the board's header, SATA not connected, Reset Switch front panel cable not turned into RGB control button, but kept on JFP1 header of the board. The pump gets powered via a fan header, this should be on PUMP_FAN on the board. The other three fans, daisy-chained, could then go on CPU_FAN.

 

[dasgeist3152b02d5]

Yeah, for the fan curves, at 85 or 90°C at the latest, you want to have them at 100% PWM signal (4-pin fans) or 12V (3-pin fans). Same for the pump at those temps.



Yes, this is the first problem, the unsatisfactory cooling performance. Why is this cooler not performing up to its capabilities. The thermal paste application sounded ok. Maybe it's one of those coolers where the cold plate (what contacts the CPU's IHS) is not shaped ideally for LGA1700 CPUs without a contact frame. Now, bear with me here.

As you may know, on LGA1700 there's the issue of the ILM (Independent Loading Mechanism, what clamps the CPU into the socket) practically bending the CPU in the middle, something that has been known for over two years. First, the remedy was to put washers under the socket mechanism, alleviating some ILM pressure from the CPU, counting on the cooler to potentially do more of that work (i also use the washer mod). Then later it gave rise to the professionally made contact frames, which replace the top part of the socket mechanism altogether, also see here. And also here.

Contact frames are good for lowering CPU temperatures (by preventing bending in the middle of the IHS and thus improving contact between the IHS and the cold plate / cooler base). Ideal would be the latest Thermal Grizzly one, it is machined to the highest standards. der8auer (TG CEO) actually came up with the idea in the first place, the others are just copycats, but of course the TG frame also costs more.

When the base plate / cold plate of the cooler doesn't take that bending into account, the contact won't be ideal, and the temperatures will be higher. This prompted Noctua ro release their newest air cooler in three versions, with different shapes:

Which version of the NH-D15 G2 should I buy (LBC, standard or HBC)?

Our standard recommendation is the standard NH-D15 G2 because it’s the best allrounder and most versatile of the three versions: It provides the best results on LGA1851, AMD AM5 using the included offset mounting, and on LGA1700 CPUs when utilisi...

faqs.noctua.at

So this could explain it to some extent, if this was the case. But maybe not fully. The performance is off by a bit too much. Don't get me wrong, 200W effective is not nothing, that's quite a lot already. But why already thermal throttling with this big cooler. So the simple fact is, with 181W power limits, they are still not low enough for this cooler, this is independent of what is going on with the voltages, this is purely talking about the thermal load which overpowers the cooler. It doesn't matter if you put a 14900KS in there and limit it to 181W, you would have the same result.

Having said that, we should work on lowering the voltages to your individual CPU's actual requirements. I explain that here in step 2). As i said, your CPU Lite Load setting seems way too high. If you can lower it a bunch, that means lower voltages, and as a consequence, lower power draw and heat. So the thermal throttling problem might be taken care of that way. Note that this still doesn't explain why the cooler gets into trouble at 181W power limits already. But if the full load power draw can be lowered enough, maybe you don't have to worry about that so much anymore.



Your board has ARGB support, so connect it like on the right side of the first picture, ARGB cable onto the board's header, SATA not connected, Reset Switch front panel cable not turned into RGB control button, but kept on JFP1 header of the board. The pump gets powered via a fan header, this should be on PUMP_FAN on the board. The other three fans, daisy-chained, could then go on CPU_FAN.

Good morning! My problems keep getting worse so I decided to contact you again. Since last time we spoke I changed my configuration a bit. I was gifted a Lian Li O11 vision and a Lian Li Galahad II trinity performance. Just to preface I had to put enormous pressure to mount the AIO on the socket, because the way it mounts is with springs below the screws on the four corners of the pump instead of just screws. I have also changed my CPU Lite Load to 10 and 12 and have disabled IA CEP, because it tanked my results in CineBench after reading your post.

Since day one I could not use XMP on my memory which should run at 6400MHz with 32-39-39-102.
With memory try it! I had them run at 6200 cl36 and for a time, until last time we spoke, it was running fine. I logged about 3 Hours on Cyberpunk, about 2 hours on Hellblade II and about 15 hours on Alan Wake. After that, due to a family emergency the pc went into storage until last Monday.

When I switched it on everything was fine until on Tuesday when I tried to play CyberPunk. I had two crashes, since then nothing. Hellblade II, however, has been nothing but crashes. Always a "Fatal Error" crash and twice the PC was locked. I even downloaded the Tekken II demo without issues. On event viewer for HellBalde i get the error "RADAR_PRE_LEAK_64".

Since then I have run Cinebench about four times with temps at max. 73 degrees and OCCT for 10 minutes. If I run OCCT with CPU and memory it, immediately, starts producing errors and after a while it hangs and I get a blue screen. Cinebench was very close to reference i5-14600K results. Also I have run the Intel Diagnostic tool with a Pass mark. However when I run Memtest I get Fail.

Now what I have done is to try the memory on all ram banks. XMP works only on Bank B2 for both sticks of ram. On every other bank XMP is not working.

My question is, do you think it is a faulty M/B or a CPU issue? I am on the latest BIOS. In total I have used the computer for 10 days. 5 days in the summer when I first built it and now when i had some time. It shouldn't be the CPU.

Again My Configuration:
Case: Lian Li O11 Vision
CPU: i5-14600KF
M/B: MSI Z790 Tomahawk WIFI MAX
Ram: G-Skill Trident DDR5 32G @ 6400MHz
GPU: Palit RTX 4080 Super
PSU: BeQuiet! Pure Power 12M 1000W
Cooling: Lian Li Galahad II trinity Performance. Fans set as intakes.
Case Fans: 3x Jungle Leopard S30 Magnetic inverted fans on the bottom as intakes and 2x normal as exhaust.

Bios Settings:
MSI Performance Profile (middle selection)
Long Duration Power Limit: 253W
Short Duration Power Limit: 253W
Current Limit: 400A
CPU Lite Load: 10 and 12
IA CEP Protection: Disabled
CPU has never exceeded 1.35V

I have also tried with Intel Default Power Profile with the same results.

Any help is much appreciated!

p.s. MemTest on Bank B2 passed. I am now running on B4 where I couldn't enable 6400MHz. I am testing at 6200MHz.

 

[FlyingScot]

I read through your whole backstory as your old post is new to me. I’m not sure what CiTay is going to suggest, but I’m of the opinion, given your major hardware changes, that you need to start at the very beginning. At a minimum, I would remove that AIO, take a picture of what the thermal paste looks like, pull the CPU and then very carefully inspect for bent CPU pins in the motherboard socket. Then I would do my homework on how to properly install the new AIO without having to use excessive force. My cooler has springs too, but I do not remember encountering excessive force. Something may not have been aligned correctly. The trick is always to tighten each corner a little at a time. Work on cars for any length of time and you find out how important that is. EDIT: Also, double-check that you are using the correct AIO mounting hardware for your particular motherboard CPU socket.

Then go over your whole system looking for any sign of a loose connection. Make sure all the standoffs (and screws) on your motherboard are in the right location. And make sure you have no extra standoffs, although this would require removing the motherboard. That’s up to you. But since this is a secondhand case, who knows what the previous owner did.

After that, comes double-checking the correct way to connect the AIO cables, and configure a fan curve on the aggressive side. I like to start out with quickly ramping up fan speeds above 65C to try to stay ahead of heat-soke. You can always adjust them down later. I would also reset all BIOS settings using the F6? - Optimized Defaults option and select the “Intel Defaults” CPU Cooler Tuning preset to begin your testing with. Basically, keep your changes to the minimum until you can establish a stable baseline. Probably also a good idea to not enable XMP until later on.

 

[dasgeist3152b02d5]

I read through your whole backstory as your old post is new to me. I’m not sure what CiTay is going to suggest, but I’m of the opinion, given your major hardware changes, that you need to start at the very beginning. At a minimum, I would remove that AIO, take a picture of what the thermal paste looks like, pull the CPU and then very carefully inspect for bent CPU pins in the motherboard socket. Then I would do my homework on how to properly install the new AIO without having to use excessive force. My cooler has springs too, but I do not remember encountering excessive force. Something may not have been aligned correctly. The trick is always to tighten each corner a little at a time. Work on cars for any length of time and you find out how important that is. EDIT: Also, double-check that you are using the correct AIO mounting hardware for your particular motherboard CPU socket.

Then go over your whole system looking for any sign of a loose connection. Make sure all the standoffs (and screws) on your motherboard are in the right location. And make sure you have no extra standoffs, although this would require removing the motherboard. That’s up to you. But since this is a secondhand case, who knows what the previous owner did.

After that, comes double-checking the correct way to connect the AIO cables, and configure a fan curve on the aggressive side. I like to start out with quickly ramping up fan speeds above 65C to try to stay ahead of heat-soke. You can always adjust them down later. I would also reset all BIOS settings using the F6? - Optimized Defaults option and select the “Intel Defaults” CPU Cooler Tuning preset to begin your testing with. Basically, keep your changes to the minimum until you can establish a stable baseline. Probably also a good idea to not enable XMP until later on.

Thanks for the reply. The system is brand new. I only changed the case and the AIO which were also brand new. It is true that I had to use a lot of force to mount it, something that I haven't encountered before and I have built more than five systems with AIOs. Just for reference I had installed an Exos from Koolance external watercooling back in 2005! I also used Intel's default settings without any luck. I will check the CPU though as I am fearing that there is too much pressure. Temps are fine. If it is bend pins then i am in for a world of pain...;-)

 

[FlyingScot]

Thanks for the reply. The system is brand new. I only changed the case and the AIO which were also brand new. It is true that I had to use a lot of force to mount it, something that I haven't encountered before and I have built more than five systems with AIOs. Just for reference I had installed an Exos from Koolance external watercooling back in 2005! I also used Intel's default settings without any luck. I will check the CPU though as I am fearing that there is too much pressure. Temps are fine. If it is bend pins then i am in for a world of pain...;-)

Sorry about that. When you said “gifted”, I immediately assumed secondhand.

Given your knowledge level and experience, it does seen like focusing your investigation on the AIO would be the best place to start. But hopefully bent pins is not the issue. Keep us all posted.

 

[citay]

Just to preface I had to put enormous pressure to mount the AIO on the socket, because the way it mounts is with springs below the screws on the four corners of the pump instead of just screws.

Enormous pressure doesn't sound right. If you mean, you had to press hard on the screwdriver for the screws to "catch" in the thread, this is something i have encountered before as well. Usually it helps if you screw in the opposite direction first, as if to loosen them. Once you hear/feel a click from the screw, it means it caught the thread properly, and then you can tighten her down. Then you just follow the manual of the AIO, usually the screw would have some sort of defined end point, like the thread ending for example, so you can't really overtighten it. And yes, always tighten in a cross-pattern, by a turn or two each, to have equal pressure throughout the tightening process.

I have also changed my CPU Lite Load to 10 and 12

10 and 12? Or first 10, then 12?

Since then I have run Cinebench about four times with temps at max. 73 degrees and OCCT for 10 minutes. If I run OCCT with CPU and memory it, immediately, starts producing errors and after a while it hangs and I get a blue screen. Cinebench was very close to reference i5-14600K results. Also I have run the Intel Diagnostic tool with a Pass mark. However when I run Memtest I get Fail.

Yeah, there is severe instability, we have to find the source, seems like something RAM-related from what you said. The Intel Diagnostic Tool was never good for anything, really. Cinebench is just a benchmark and not a stability test, so the most it could do it freeze or crash, but then the instability is already huge. OCCT and Prime95 for example, they will check and validate the calculations, so they can immediately find miscalculations and report them to you. Also, due to them being true stress-tests and not a representation of real-world full load like Cinebench, they can use dirty tricks to make the workload more demanding and to provoke errors more easily.

If you have RAM errors in Memtest86, and looking at this,

Since day one I could not use XMP on my memory which should run at 6400MHz with 32-39-39-102.
With memory try it! I had them run at 6200 cl36 and for a time, until last time we spoke, it was running fine.

then the first order of business should be to back off the RAM speed even further and try again. DDR5-6400 with pretty tight timings, this is getting up there, meaning it might not always be a plug&play-type of setup. What you can do is, disable MemoryTryIt!, and instead, enable XMP, but then, before pressing F10 to save&exit, also set "DRAM Frequency" to DDR5-6000 by hand. See how it runs like that.

Oh, and update your BIOS to the vA8 final from https://www.msi.com/Motherboard/MAG-Z790-TOMAHAWK-MAX-WIFI/support

p.s. MemTest on Bank B2 passed. I am now running on B4 where I couldn't enable 6400MHz. I am testing at 6200MHz.

The slots are called A1, A2 (first channel) and B1, B2 (second channel) from left to right. With one module, test in slot A2, with two, in A2 and B2. Testing the modules individually in slot A2 makes sure that one of them isn't defective (which it would point to if you only had errors with one module on its own, even if you disable XMP altogether, but not with the other module on its own).

 

[dasgeist3152b02d5]

Enormous pressure doesn't sound right. If you mean, you had to press hard on the screwdriver for the screws to "catch" in the thread, this is something i have encountered before as well. Usually it helps if you screw in the opposite direction first, as if to loosen them. Once you hear/feel a click from the screw, it means it caught the thread properly, and then you can tighten her down. Then you just follow the manual of the AIO, usually the screw would have some sort of defined end point, like the thread ending for example, so you can't really overtighten it. And yes, always tighten in a cross-pattern, by a turn or two each, to have equal pressure throughout the tightening process.



10 and 12? Or first 10, then 12?



Yeah, there is severe instability, we have to find the source, seems like something RAM-related from what you said. The Intel Diagnostic Tool was never good for anything, really. Cinebench is just a benchmark and not a stability test, so the most it could do it freeze or crash, but then the instability is already huge. OCCT and Prime95 for example, they will check and validate the calculations, so they can immediately find miscalculations and report them to you. Also, due to them being true stress-tests and not a representation of real-world full load like Cinebench, they can use dirty tricks to make the workload more demanding and to provoke errors more easily.

If you have RAM errors in Memtest86, and looking at this,



then the first order of business should be to back off the RAM speed even further and try again. DDR5-6400 with pretty tight timings, this is getting up there, meaning it might not always be a plug&play-type of setup. What you can do is, disable MemoryTryIt!, and instead, enable XMP, but then, before pressing F10 to save&exit, also set "DRAM Frequency" to DDR5-6000 by hand. See how it runs like that.

Oh, and update your BIOS to the vA8 final from https://www.msi.com/Motherboard/MAG-Z790-TOMAHAWK-MAX-WIFI/support



The slots are called A1, A2 (first channel) and B1, B2 (second channel) from left to right. With one module, test in slot A2, with two, in A2 and B2. Testing the modules individually in slot A2 makes sure that one of them isn't defective (which it would point to if you only had errors with one module on its own, even if you disable XMP altogether, but not with the other module on its own).

1.I had to press on the pump for the screws to catch and when one side did, the other needed a lot of downward pressure to catch. Same for the rest. Orientation of the tubes coming from the pump, are towards the RAM slots as per the manual.

2. CPU Lite Load at first was at 10 and, then, I used 12.

3. Ram at A2 at 6400Mhz passes. Ram at B2 passes only at 6200MHz. Both sticks at A2 work at 6400Mhz. I am already running MemTest86 at A2 and B2 at 6000MHz as you suggested (beat you to it . Will post results once it is done. This is with XMP disabled and MemoryTryIt! at 6000MHz. I have also reverted to default settings in Bios before running it.

4. Will update today. I am severely overworked right now, having to upload and test new retail software for 600 stores at work and now this.

Edit: I took out the retention clip for AMD that is similar to Intel and the springs on the screws are crazy tight. I have to put a lot of pressure on my thumb to compress them for what is worth.

 

[citay]

1. Yeah, that can be so, but it doesn't necessarily translate into excessive mounting pressure onto the CPU, it's just that the screws are difficult to put into the threads. I had the same happen on Noctua air coolers before. One side was already screwed in (just one screw on each side), but the other one, boy was that difficult to get into the thread. Even scratched the bottom of the cooler a bit in the process, from rubbing against the edge of the CPU's heatspreader. But once it's mounted, it's doing the job.

2. Ok. 10 was unstable?

3. You can also try with XMP, but DRAM Frequency at DDR5-6000. MemoryTryIt! will likely use different timings than XMP.

4. Your previous update took three months, and we picked up where we left off. So all in your own time.

 

[dasgeist3152b02d5]

1. I will see if there is too much pressure. I hope not

2. No. I just followed your advice that once i set it to a value that is stable it is better to go one up for stability headroom.

3. Will do once MemTest finishes.

4. My mistake. I should have clarified, that once i switched it on, the first thing i did was to update it to the September beta update 0x12b. I need to install the recent one.

5. Do you think the settings i had were good for the CPU. Did it need anything else?

 

[citay]

2. That's when Mode 9 would be unstable, and 10 would be the first stable one. Then you should set to Mode 11 for extra headroom. If Mode 10 was stable, but you not having tested any lower mode than that, then you don't actually know where it comes unstable, it could be at Mode 4, Mode 6, or even Mode 9 too. But then you don't need to set Mode 12.

3. BTW you can skip the Hammer test at the end, it takes a long time and it's not really about RAM stability, just about some theoretical vulnerability.

5. I would need to see the results in HWinfo after a Cinebench run again. Power limits of 253W are ok as long as your cooling can deal with it and keep it from the high 90°C range, and 253W is also the maximum i would allow for any CPU. The mode for CPU Lite Load, i would just try to lower it until you see the first instability in a stress test, don't know if you actually did that or just stopped at Mode 10. Because ever step lower is an improvement in the way the CPU is running, provided it stays stable.