Cooling problems persist on LGA1700

[ShrimpBrime]

When I built this rig a year ago I was chasing ram speeds of 7800, and I got 7400 stable with a lot of testing and tweaking. Then the bios updates caused instability so I stopped trying and have been stable on 7200 with default SA ever since.


Yeah I am thinking that plus maybe I don’t have a pristine slice of wafer in my 14700k... plus mid grade GB is I got what I paid for. Oh well!

Maybe I will swap the contact frame back in and try under volting *shrug emoji* shoulda just gone with a high end MB and 14900k to at least get a somewhat better binned cpu. Oh well. With no major cpu advancements coming I do not see a reason to burn more cash for a few percent gaming fps.

You won't gain even a few percent cause your are hitting 100c on synthetic loads, not games. There is definitely nothing to worry about.

On the custom loop, can do 5.7P and 4.5E cores respectively at room temp (75-80f). 5.8ghz P cores will just heat soak faster. That is tuning with LLC only. V-core on auto. This is realistic clocking, not show boat. That's what you read next. But for the purpose of seeing where the wattage is at, and what you get from it.

With chilling, 8P cores only, I can achieve 6.1/6.2ghz in the 300w area. With all P and E cores, can get it to pull 400w. That is 1.5v effective v-core. The E cores package is roughly worth 100w.

P cores only, 180w. So at defaults, the E-cores are worth roughly 73w. Wait till you realize this thing pushes 85-90c anyways....

2P cores at 6.9ghz 1.66v effective, 1 core usage reported 61w. So if both cores used, closer to 125w.

Yes, I haveth screen shots of all this. I do not have a screen shot of default effective working v-core, but I want to say it droops to 1.288v with LLC lvl3. Or third to lowest setting. 253w.

I found you can easily cool 4ghz on air. I mean 4ghz Cache, 4ghz P-cores and 4ghz E-cores manually dialed in, I like to set LLC high to prevent v-droop though. It helps. 0.950v to 1.0v does the trick nicely.
__________________

Tune it the way you want to. But Intel designed them to hit 100c and throttle or droop frequency after a time period if throttle is never reached. Turbo boost time limits is why I manually clock. The cpu is probably drooping on people in game no doubt. Just doesn't seem people talk about it, perhaps it's just not realized. AMD on the other hand boosts and droops with thermals. Seems to have better control. Intel, nah Fk it. Just roast it and it'll droop it's way back to 99c eventually... haha.

 

[fusseli]

I have been pondering, does ram speed or 1.45v ram effect cpu temperatures due to the IMC? It occurs to me that my 7200 is rock solid stable but could be part of the 14700K heat.

My 13900KS was stable at 4x16gb at 6400 but I’ve since moved it to 2x48gb of 6600 Dominator Titanium that is also stable and runs cool at 6.1. After a few minutes of 8 cores at 5.6 (p core only) it is maybe 87C and sustained single thread at 6.1 is 80C. All cores it runs 5.0-5.6 at 90C (it appears to throttle but never goes over 90C). It’s an older H150i RGB Elite 360mm AIO and no contact frame / stock intel. All as reported by AI Suite:
VCore 1.225v
CPU L2 1.216v
CPU SA 1.297v
MC 1.368v
CPU input 1.825v
CPU-z ST score 967

 

[Carlyle2020hs]

If 1.45 is the default voltage, no.
Otherwise a fan is in order if a stick goes above 52c.

Thanks for the pics. Nice cable management.

Could you shoot a macro pic of your cpu die after a fresh AIO removal pls?
Use a scondary light source to let the surface reflect the pastes imprint (try more than one angle).

i´ll take a coolers imprint as well

I can´t judge the mount or your handywork without proper nudes please

ps
You will remeber the first time you´ve closed the OG cpu mount and how hard it was?
Now it isn´t anymore since the cpu gave in a little.

Stop switching around the LGA mounts please and only use the aftermarket one.

You´ve mucked it up in the first place to try the OG one.
Now it will take months and at least three digits of heat cycles for the cpu to relax (almost) into its original form

Don´t use Loctite.
That´s only for virgin cpus that get embedded directly into a frame.

 

[ir_cow]

I have been pondering, does ram speed or 1.45v ram effect cpu temperatures due to the IMC? It occurs to me that my 7200 is rock solid stable but could be part of the 14700K heat.

I don't think it causes CPU temp to go up, but I did observe memory erroring up if the CPU temp is too high. Like for example I can run 8600 MT/s without a CPU OC but with it on I'm stuck at 8200 long term and 8400 for a few hours.

 

[InVasMani]

You should be able to drop the L2, SA, CPU input a bit. Bumping up Vcore and to smaller extent MC can help if you're pushing higher peak frequency. The other stuff I didn't find mattered as much and in cases you can benefit from reducing them a bit though in the case L2 it impacts the cache performance and expectations some around that though you can reduce heat due to running higher speed and get more CPU frequency within power limits and ICCMAX instead conversely slowly which is a bit of a trade off. If you're running a bit lower MT/s and cache frequency with lower ring it's kind of a trade off. What works best is bound to vary by workload some extent. If you're trying to get like peak memory bandwidth MT/s and cache performance you probably don't want to do that, but for higher CPU frequency and ST/MT performance within heat and power budget which often outweighs the other benefits you might consider it.

Depends on software too if you're fine with more shorter burst performance as opposed to like Cinebench roasting your CPU in longer term benchmark what's more ideal with differ quite a bit on tuning. That said focusing more on short turbo burst performance and if you need more sustained performance dropping multiplier by like x1 isn't too difficult to improved sustained performance.

 

[fusseli]

Here’s the H115i when I took it off of the 14700K

And for my 14700K as reported by GBCC getting the high temps mentioned during all-core stress test:
Vcore 1.296v (Auto) and 1.332v (Normal) -- (bios setting)
CPU VDD2 1.4v
CPD DRAM Vdd/VddQ 1.125v
CPU input 1.8v
CPU-z ST score 912 (5.6 / all-core performance profile, but 99-100C all core load)

 

[kiddagoat]

You've mentioned instability and persistent nuisances since owning the board. I would remove any and all OC except for the XMP on the RAM. When I had my 13700K, I never really had many issues keeping it cool with a custom loop. Unless of course I ran something CPU intensive, and it would get into the low 90s. I was also using a thermalright contact frame.

Those chips are tuned pretty well out of the box and will boost accordingly.

You are trying to reign in the temps, so again turn everything back to stock for now to get a base line. I would highly recommend enforcing the intel default of 253W for the time being.

 

[fusseli]

You've mentioned instability and persistent nuisances since owning the board. I would remove any and all OC except for the XMP on the RAM. When I had my 13700K, I never really had many issues keeping it cool with a custom loop. Unless of course I ran something CPU intensive, and it would get into the low 90s. I was also using a thermalright contact frame.

Those chips are tuned pretty well out of the box and will boost accordingly.

You are trying to reign in the temps, so again turn everything back to stock for now to get a base line. I would highly recommend enforcing the intel default of 253W for the time being.

At all stock/auto, 14700K is getting:
Vcore 1.3v actually spikes up to 1.31v-1.36v under all core load
93C all-core stress test
67C single core
CPU-Z 881/14122

Gaming ~50% cpu load, 67C

 

[kiddagoat]

That seems more in line with the proper power limits. Before the power limits were enforced, my 13700k would regularly go 1.4-1.5V.

I also ended up having a defective chip that died after about a year.

Until it died, it had similar temps to what you are getting now.

 

[InVasMani]

Those temps seem reasonably normal. It should idle about 30c to 35c if you're on water or even less depending on your ambient temps.

 

[fusseli]

Those temps seem reasonably normal. It should idle about 30c to 35c if you're on water or even less depending on your ambient temps.

Yeah that is my idle range.

Maybe I just have a bad cpu/mobo combo and isn't ever going to OC well

 

[kiddagoat]

You have to factor in too that these newer CPUs are pushed to their limits out of the box. Any sort of gains with OC will come from undervolting to allow the chip to maintain those boost speeds while keeping thermal/power limits in check.

I am in the minority here as I don't OC and tinker like I used to, but I would just leave it for now and enjoy your system. The maybe 2-3% gains aren't worth it unless you are shooting for some leaderboard on a benchmark chart. The days of getting massive OC gains are over, be it CPU or GPU.

 

[iameatingjam]

I've never found the 14700k to be too hot at the 253w limit, and I only have a 240mm aio. I was expecting it to be hotter when I ordered it so I bought these overkill p12 max fans that go up to 3300rpm. Yet they never get any use, unless I'm doing a heavy all thread benchmark which I don't do very often...

Anyway though if you go pushing it beyond spec then yeah I imagine it can get pretty hot... keep it on its leash and it'll be fine heat wise. Thats my experience anyway.

 

[rusty caterpillar]

I have a Gigbyte Aorus Z790 Elite X motherboard (wouldn't buy again, I miss Asus) and 14700k hits 100C and throttles.

I have tried with and without a Thermalright cpu pressure plate for LGA1700, no change.

I have re-installed the AIO cpu block several times, and upgraded from a 280mm to a 360mm AIO, and no change.

Any ideas?

Clogged accelerator plate inside the CPU AIO block maybe.

8:33

 

[fusseli]

Under-volting seems to make a difference. I setup an adaptive offset ranging -0.03v to -0.05v from 8 to 56 multipliers, and now on all-core stress it runs at 5.5ghz and only hits 90C after a few minutes.
CPU-Z 871/14716

I'll see if it still stable on 5.6 all-core:
CPU-Z 914/14852

Intel Core i7 14700K @ 5600 MHz - CPU-Z VALIDATOR

[32wsdg] Validated Dump by WH1 (2025-03-13 00:55:25) - MB: Gigabyte Z790 AORUS ELITE X WIFI7 - RAM: 32768 MB

valid.x86.fr

 

[fusseli]

Well, this is the last straw for me. Some sequence of events the last few days has made my system go from cool and quiet to completely unstable, the worst yet, from simply testing some basic undervolt settings and running stress tests in cpuz and Intel XTU. Even with a fresh bios reset the system freezes and crashes are thin about five minutes of booting. I have been able to restore stability by uninstalling Gigabyte GCC, Corsair iCue, and XTU, on yet another fresh bios reset.

I think I either have a dud cpu (shoulda binned it an i5, lol) or a dud MB, or both. I am picking up a refurb Asus Strix Z790-A and a new 14900k (because 9800x3d is unobtainable). Fuck this shit.

 

[kiddagoat]

Have you thought about just putting the system back to default setting and use the 253W default, what does it do?

 

[Bobaganoosh]

Well, this is the last straw for me. Some sequence of events the last few days has made my system go from cool and quiet to completely unstable, the worst yet, from simply testing some basic undervolt settings and running stress tests in cpuz and Intel XTU. Even with a fresh bios reset the system freezes and crashes are thin about five minutes of booting. I have been able to restore stability by uninstalling Gigabyte GCC, Corsair iCue, and XTU, on yet another fresh bios reset.

I think I either have a dud cpu (shoulda binned it an i5, lol) or a dud MB, or both. I am picking up a refurb Asus Strix Z790-A and a new 14900k (because 9800x3d is unobtainable). Fuck this shit.

You have to be a little careful with XTU (in my experience). it likes to fight with the BIOS settings sometimes, especially with things like V/F curves. If I had anything set there in the bios then tried to tweak it in XTU it would have issues...then they removed V/F curves on later version of XTU (not sure where it is currently I don't use it anymore). Anyway, you get too many pieces of software trying to fight over the same settings and it is a recipe for crashing. I was using XTU (or EVGA E-LEET) to tweak my settings within windows to test things, then I disabled the software and set things permanently in BIOS (not doing both at the same time).

Also, from your waterblock picture, I'd say it doesn't look awful, but you can definitely see a pressure difference from the middle to the outsides. That is pretty typical of a warped CPU IHS from the stock ILM. I also have seen stability issues from the motherboard bending when putting a waterblock onto the motherboard with the aftermarket contact frame, so I had to get an aftermarket back-plate to counteract that warping lol. It took a while, but I'm finally stable and getting great contact.

I had made this diagram of a side view of the stackup showing the issue. WIthout the cooler backplate (not shown in this picture), the space between the mounting bolt and the CPU socket's backplate was flexing. This was also partially because my water block just secures with nuts and bolts. There's no specific "bottom-out" retention method, so you have to be really careful to tighten things down evenly as well. I'm not sure any of this effects you, but it's just food for thought on how much bending and warping can all happen even you add the thermal grizzly/thermaltake/whatever aftermarket contact frame.

Edit: I forgot to mention, my issues with warping and stability were with 12900k. It took a while to solve because it involved testing different contact frames, buying a backplate, and trips to the hardware store to find the right nuts and bolts (and plastic washers) to secure the backplate to the motherboard, then put the top nuts on that mount the water block. I actually really struggled getting the 12900k stable with an aftermarket contact frame. The 13900k has worked perfect since my first install (with contact frame), but I used everything I learned with my 12900k struggles.

 

[fusseli]

Have you thought about just putting the system back to default setting and use the 253W default, what does it do?

Full bios reset to default was the freeze within five minutes of booting. This persisted after uninstalling XTU. I noticed that even task manager was freezing, or anything else querying the MB, as soon as being opened. It only went away after removing GCC and iCue also. Something got hosed that a MB reset doesn’t fix, even as simple as the monitoring programs that query the MB and don’t even change anything like XTU or tuning utilities do. At this point on this setup it’s a lost cause not worth a fresh windows install unless I change other hardware too, IMO.

You have to be a little careful with XTU (in my experience). it likes to fight with the BIOS settings sometimes, especially with things like V/F curves. If I had anything set there in the bios then tried to tweak it in XTU it would have issues...then they removed V/F curves on later version of XTU (not sure where it is currently I don't use it anymore). Anyway, you get too many pieces of software trying to fight over the same settings and it is a recipe for crashing. I was using XTU (or EVGA E-LEET) to tweak my settings within windows to test things, then I disabled the software and set things permanently in BIOS (not doing both at the same time).

Also, from your waterblock picture, I'd say it doesn't look awful, but you can definitely see a pressure difference from the middle to the outsides. That is pretty typical of a warped CPU IHS from the stock ILM. I also have seen stability issues from the motherboard bending when putting a waterblock onto the motherboard with the aftermarket contact frame, so I had to get an aftermarket back-plate to counteract that warping lol. It took a while, but I'm finally stable and getting great contact.

I had made this diagram of a side view of the stackup showing the issue. WIthout the cooler backplate (not shown in this picture), the space between the mounting bolt and the CPU socket's backplate was flexing. This was also partially because my water block just secures with nuts and bolts. There's no specific "bottom-out" retention method, so you have to be really careful to tighten things down evenly as well. I'm not sure any of this effects you, but it's just food for thought on how much bending and warping can all happen even you add the thermal grizzly/thermaltake/whatever aftermarket contact frame.
View attachment 389550

Yeah that’s a great point and it also is a reminder of the good reason why the water lock install instructions all say only to go finger tight on the thumb nuts with or without a contact frame.

There are people out there reporting those same issues with XTU. It could be root cause on this one the last few days, but the creepy part is that it persisted. Something is definitely wrong in this setup and everything I have tried in the last year finds a new instability. Good riddance!

 

[kiddagoat]

Yeah those gremlins do sound quite persistent. Especially after a full MB reset. Failsafe default usually offers stability to allow further troubleshooting.

 

[Bobaganoosh]

Yeah that’s a great point and it also is a reminder of the good reason why the water lock install instructions all say only to go finger tight on the thumb nuts with or without a contact frame.

lol I know, but I found out that my "finger tight" was tighter than they actually meant. you leave that sort of thing up to interpretation and I'm cranking it down. You know the old adage: "tighten until the bolt breaks, then back it off a quarter-turn"

 

[InVasMani]

Undervolting too much can also create instability so keep that in mind too.

 

[fusseli]

Yeah those gremlins do sound quite persistent. Especially after a full MB reset. Failsafe default usually offers stability to allow further troubleshooting.

Yeah, it's out of hand. And this whole system is hosed at this point. Since the crashing started even after a bios reset to default, I tried a system restore thinking maybe it was a registry or service issue from XTU/GCC/icue, and guess what the restore failed. Now random windows components like screen snip don't work anymore it's a freaking train wreck over here.

Undervolting too much can also create instability so keep that in mind too.

Yep didn't try over -0.03v for long before backing it off and then finding that it would freeze and reset no matter what. It's hosed. I'm selling this 14700k and MB separately and have a NIB 14900k on the way and a refurb strix Z790-A on the way for a fresh install on win11.

 

[ir_cow]

Why not just leave the Intel Defaults in place?

Also XTU is bad news. It holds your PC even when it's "off" ive found. Only solution I had was to load default and apply. If you uninstall, it should remove whatever is applied. I suggest not using any software to OC. They all suck and can easily kill the CPU. Not the first time this has happened to someone.

The other option is to put a contact frame on. It will drop your temps 5-10,c. But no matter what you do, the CPU will try to go to 100c or hit the wattage limit (PL1/PL2). Whatever comes first.

If you want to get extra mad, set it all to defaults and run prime95. Watch it go to 100c.

For troubleshooting, I would first see if one core is pretty far off while running Cinebench R24. Since that goes for 10 min, you can see where the temp issue is. If all the cores are equally getting to 100c at default power limit, you have mounting problem or cooling.

Lastly you can change the max temp to 115c. The KS CPUs have this by default. Same silicon. Of your hitting 115c, than I would say your cooling's broken.

 

[ShrimpBrime]

Why not just leave the Intel Defaults in place?

Also XTU is bad news. It holds your PC even when it's "off" ive found. Only solution I had was to load default and apply. If you uninstall, it should remove whatever is applied. I suggest not using any software to OC. They all suck and can easily kill the CPU. Not the first time this has happened to someone.

The other option is to put a contact frame on. It will drop your temps 5-10,c. But no matter what you do, the CPU will try to go to 100c or hit the wattage limit (PL1/PL2). Whatever comes first.

If you want to get extra mad, set it all to defaults and run prime95. Watch it go to 100c.

For troubleshooting, I would first see if one core is pretty far off while running Cinebench R24. Since that goes for 10 min, you can see where the temp issue is. If all the cores are equally getting to 100c at default power limit, you have mounting problem or cooling.

Lastly you can change the max temp to 115c. The KS CPUs have this by default. Same silicon. Of your hitting 115c, than I would say your cooling's broken.

Must learn XTU before calling it bad. It's the only software that actually works correctly. Yes, must be used at defaults. It holds because have checked to start with windows. Unchecked that, should solve this minor issue.