My 13700k went 9C over normal 100C tjmax without throttling.

[chrcoluk]

Attaching the csv hwinfo log, curious for thoughts, is this expected with temp throttling feature?

Board is Asrock Z690 Steel Legend
Bios is mostly defaults, exceptions below.
Cooler set to air cooler.
PL1 set to 125w (auto configured by above)
Disabled Driver auto update
Advanced settings default bios screen.
Post delay boosted to 3 secs.

Vcore peaked at 1.400v during multi core.
Clocks seemed to stay at max, whilst over 100C (109C). 5278.1mhz

Rows 5 to 8 multi core load. and around 8/9-13 for single core load.

Going to set PL2 to 125w as well as dont trust the thermal throttle.

--

Looking at the tjmax delta for the first run it looks like tjmax is set to 115C instead of 100C, in the bios the setting for it is set to 'Auto' The core that hit 109C, at the same point was 6C to tjmax. I will change that value to 100 manually, pl1/2 both to 125w before I move on.

Attaching data from fix vcore of 1.3v also. Temp peaked at 97C (18c to tjmax so again tjmax looks to be set to 115c). Peak power consumption down 30w which seems a little underwhelming.

 

[OolongGoat]

Need your full system specs. It would appear that your current system specs are outdated.

 

[chrcoluk]

Yeah sorry, they not outdated in terms of my PC, I cam currently testing this board/cpu outside of a case with spare components. Here is the connected hardware list.

4x 8 gigs team force ddr4 3000CL16 sticks, currently running at jedec, XMP is off.
samsung 830 128gig ssd so can use windows
evga 500w power supply, dont know exact model off hand (no discrete gpu)
cheap arctic alpine 17 co
using igpu
m.2 wifi intel ax210

Obviously high temps are expected with this cooler, thats not why i made the post, but rather that the board on its default auto setting allowed the cpu to above 100C without throttling, I have now set the tjmax to 100c manually, and it appears to function correctly. I think my thread title is bit misleading on this I will edit.

pl1 and pl2 are now both set to 125w, the undervolt on second run as I was just curious if it would boot and get through cpuz on it, I will look at undervolting again when its under the actual cooler I will be using. But until then voltage is back on default.

The first run which hit 109C without throttling was on stock voltages.

 

[wolar]

From what i understand the limitations are set on the motherboard not on the CPU, you can try to update bios maybe ?

 

[phanbuey]

1.4V at 5300 mhz is ridiculously high -- the motherboard is both overvolting that chip and set the 115C tjmax -- pretty agressive.

There's not a lot of thermal paste that survives that kind of abuse for more than 6 months.

 

[chrcoluk]

Yep, I posted it here as was lots of interest in exploding AMD cpus, I felt this was similar issue as 115C seems dangerous.

115 is also the max allowed in the bios, its set to 100 now and I might lower it to 95.

I will update my generic testing in the 13700k review thread, I have ended up doing more undervolting testing. (or am I just pushing it to more normal normal voltages for a 13700k).

 

[phanbuey]

Yep, I posted it here as was lots of interest in exploding AMD cpus, I felt this was similar issue as 115C is outside of spec right?

115 is also the max allowed in the bios, its set to 100 now and I might lower it to 95.

I will update my generic testing in the 13700k review thread, I have ended up doing more undervolting testing. (or am I just pushing it to more normal normal voltages for a 13700k).

115C is the upper limit of the silicon, but the chip won't explode, and you shouldn't lower it to 95 as 100C limit is perfectly fine. Might want to use tfx/PTM 7950 or a kryosheet if you plan on regularly hitting 100C with it (or just set TDP limits to intel standard).

The AMD chips are exploding due to the failsafes on the motherboards failing entirely.

 

[chrcoluk]

115C is the upper limit of the silicon, but the chip won't explode, and you shouldn't lower it to 95 as 100C limit is perfectly fine. Might want to use tfx/PTM 7950 or a kryosheet if you plan on regularly hitting 100C with it (or just set TDP limits to intel standard).

The AMD chips are exploding due to the failsafes on the motherboards failing entirely.

Intel® Core™ i7-13700K Processor (30M Cache, up to 5.40 GHz) - Product Specifications | Intel

Intel® Core™ i7-13700K Processor (30M Cache, up to 5.40 GHz) quick reference with specifications, features, and technologies.

www.intel.co.uk

This says 100C

Even that intel engineer with debaur mentioned 100C as the safety limit.

I think this should be locked to max in microcode if bios tries to set higher.

 

[phanbuey]

Intel® Core™ i7-13700K Processor (30M Cache, up to 5.40 GHz) - Product Specifications | Intel

Intel® Core™ i7-13700K Processor (30M Cache, up to 5.40 GHz) quick reference with specifications, features, and technologies.

www.intel.co.uk

This says 100C

Even that intel engineer with debaur mentioned 100C as the safety limit.

I think this should be locked to max in microcode if bios tries to set higher.

100C is a very safe limit to run at regularly -- that's why setting it at 95C doesn't really do much. Motherboards CAN set it to 115C but it's pretty aggressive and not necessary at 5.3ghz.

8.2 GHz and Beyond: Core i9-13900K is An Extreme Overlocker’s Dream | Tom's Hardware (tomshardware.com)

 

[chrcoluk]

Thats 100C, I was talking about 115C. But if you think 115 is fine go ahead and set your tjmax to that and use a fan that hits those temps.

The 95C consideration is my personal preference I just dont like hot running parts. I may or may not do it, not decided yet. Probably wont make that decision until I am actually running the chip in the case and the proper cooler.

 

[ir_cow]

Straight from the Intel engineer. 100C is fine. Even 115C is too. Which is officially supported by the KS CPU. Same Silicon, so that would mean 115c for Alder Lake and Raptor Lake CPUs.

 

[freeagent]

My third gen used to reboot at about 110c.. she's still good

 

[Calenhad]

The 95C consideration is my personal preference I just dont like hot running parts.

Soo, 100C is hot and 95C is not? I would leave this to the experts. See video two posts above mine.

 

[nguyen]

Wow my 13700K has never touched 80C at all (5.4 all core 1.26Vcore), I might be over-protective of my PC

 

[chrcoluk]

I am about to post some testing results in the review thread. (chip cant stretch its legs at high tdp at the moment, and tests were done at jedec ram speed)

Bear in mind though in regards to thermals I am not running a proper cooler at the moment.

 

[P4-630]

Max temp I see with my i7 12700K all P-Cores @ 5Ghz is just over 70C..
I'm good.

I am not running a proper cooler at the moment.

Noctua NH-D15S ? (should be good enough)

 

[chrcoluk]

Max temp I see with my i7 12700K all P-Cores @ 5Ghz is just over 70C..
I'm good.


Noctua NH-D15S ? (should be good enough)

Thats the cooler I will use when I swap the hardware in to my case. This is been tested externally at the moment.

I am going away tomorrow, so PC wont be changed over until the earliest late next week.

--

Ok a nice update.

Someone was helping me on reddit and he diagnosed the ac/dc loadlines were whacked. Set very aggressively by asrock.

I have now updated to the latest bios and there is more sane defaults, the stock vcore is way lower than before, in addition the stock tjmax is now 100. IA AC/DC is now also configurable in bios like other vendors.

Getting low on time, so probably it for today. Thanks to those who replied.

 

[phanbuey]

Wow my 13700K has never touched 80C at all (5.4 all core 1.26Vcore), I might be over-protective of my PC

According to intel you're leaving 20C of performance on the table.

 

[nguyen]

According to intel you're leaving 20C of performance on the table.

That would be like another 200mhz OC

 

[chrcoluk]

I came back to it, posted new results in 13700k review thread, much better on new bios.

Stock vcore only hit 1.280v now under load. tjmax also 100c default.

Also I found reducing cache offset is more effective than core offset.

 

[phanbuey]

I came back to it, posted new results in 13700k review thread, much better on new bios.

Stock vcore only hit 1.280v now under load. tjmax also 100c default.

Also I found reducing cache offset is more effective than core offset.

Do you mean in terms of power consumption or temperature (or both?)?

It actually sounds like you have an OK chip and asrock just went nuts with the bios defaults on loadline and tjmax. Now that you have the voltages dialed in there, is your chip stable with a +1 turbo ratio offset?

 

[chrcoluk]

I havent done what would be my normal run it through the mill testing, but its done dozens of cinebench runs now, 3 hours of memory testing, no whea errors, no signs of instability (but bear in mind because of cooler I am not fully stressing for long), so I think I should wait until it is connected to the kit it will be used for daily before I do proper testing, I do have a starting point now though and hopefully where it is now bears fruition.

Yeah at first I thought I had a disaster of a chip, but as you said it seems at least ok now.

Both power and heat, I consider the two usually link up.

 

[chrcoluk]

For anyone who might be curious I have finally got windows up to a more modern version, but had to fight some quite nasty performance regressions.

As an example on 21H2, my AMD hosted VM can run a specific task at 0.7 seconds, 13700k at 0.2 seconds, 9900k was 0.35 seconds on 1809, but regressed unexpectedly massively to a whopping 5-7 seconds, since i plan to move the install over to the 13700k, I had to diagnose this issue, it was even slower than my broadwell laptop which has been doing the task at about 4 seconds. There was other slowness as well, like slow UAC prompts, random delays for menus to show etc.

When I switched the 9900k to its igpu for the desktop all of the performance loss went in a flash, so seems to be WDDM/Nvidia related, I bumped the drivers, installed clean, stripped down with nvcleanstall and its now much better but still not where it should be, however now fast enough I feel I can move on and see if the cpu change fixes it, if not its only slightly slower at this point.

My plan is something like this. Partially due to my health, not as fit as I used to be.

OS backup.
Keep PC on its current table, but just lay it down so can work on it.
Remove cooler, and move it to the current test setup.
Do more testing, to see if actually stable under sustained load and transitions, even if ok I will likely back out a bit on the voltages for a day to day config.
After I am satisfied with the new tests, I will then proceed with the swap over, so removing everything off current board, removing board from case, then installing new board (installing my NVME before its in case) and then finally reattach everything to finish it off. Boot it up, remove ghosted drivers from old board, make sure drivers are good for new one.
On the new OS, I did integrate the drivers, so the chipset/ime/network stuff wont need any downloading. iGPU drivers install themselves from windows update.
Do some final testing including testing the ram, as I wont be using the ram I have been tested with, will use the ram from my existing build.

I have concerns with how hard the cpu power slots are to access, with the cooler pre installed before going in case, it might make it unworkable, so if I feel the cooler will make it too hard, the new testing wont happen until its in the case, as I am not installing the cooler twice.

After that I will update the community with how things are going.

 

[chrcoluk]

So plan changed, as I observed when I held the cooler over the socket, the cpu power connectors would be impossible without a severe fight.

So board is now in case (luckily my main psu has 2 8 pin cpu cables so even spread on VRM), currently all drives not connected, and using the test OS I setup to do more tests. Ended up replacing back exhaust fan, washed the noctua heatsink which ended up annoying me as was a fight to dry it so could use it. But it was gross with the amount of dust on it and clogged in fins.

So regarding stability and benches, the temps are lower which isnt a surprise, but not sure if I think they seem too high still, on normal multi core cinebench r23 runs it tops out around 78C with pl2 high enough to have no power throttling. If I let it run for a sustained period, I did 10 minutes, to see if where the temperature starts flatlining it eventually settles at around 80-81C. For this I had to raise pl1 also.

No whea errors or app crashes or bsod's as of yet.

Now something interesting on the cache, I applied a undervolt to the cache, which is still configured in the bios, was checking the throttling part of hwinfo data, and it is reporting cache clock is throttled due to a too low SVID, so I assume thats down to undervolt, looks like some kind of undervolt protection, nothing like that for the core clock, this is just for the cache clock. hwinfo reports the highest peak cache clock at 4.8ghz, it sustains 4.5ghz during load. The cache clock is throttled nearly all the time according to hwinfo. Its mostly on a yes state. Not sure what the actual load cache clock is supposed to be, still looking. Not on this page.

Intel Core i7-13700K Specs

Raptor Lake-S, 16 Cores, 24 Threads, 3.4 GHz, 125 W

www.techpowerup.com

In addition the bios with the samsung B die in, didnt auto configure 3200mhz, had to manually set it, it just set the timings only. Target clock speed said 3200mhz in bios, but the live speed in bios and OS was 2133mhz.

The score I recorded from first run of r23 multi core is 30743, dont have my older notes accessible (although did post it on here in other thread) but I think thats about 1k higher.

Will remove the cache undervolt in XTU and observe cache clock behaviour and hwinfo throttle state.

Also design flaw (in my opinion) on the board for first m.2 socket the heatsink cover, they moved one of screws upwards of centre due to x1 pcie slot in way, and that slight shift means with the noctua cooler in (this is the offset model as well) means I can only do it with one of those tiny screwdrivers at a slight angle as well.

--

Had a power cycle (first time ever had a PC do this) during long period of load (cinebench r23), with cpu at stock settings it did it as well but quicker, unstable power delivery?

PSU Antec HCG (gold) 750, supposed to be rebadged seasonic, still no GPU in the system, load was about 205-210w. Its handled loads over 300w routinely on old spec so could a incapable VRM also be cause of this?

--

Is ram, when I did my first tests I kept at jedec just to prevent ram instability affecting things, seems IMC (or board) cant handle XMP on this B die. Found a few posts on reddit indicating IMC quality seems highly variable, but the SA voltage seems really low as well, so could just be bad asrock bios defaults. It is now fine at 2133mhz on stock and undervolts.

Will come back to this tomorrow, for reference the SA voltage is 0.784v. (auto default). Ram is 4x 8 gigs 3200CL14 Samsung B die. @ir_cow curious of your experience of 4 dimm B die on raptor lake and voltages?

 

[chrcoluk]

--

Tried a few things including raising SA voltage, but after some hours I noticed what appears to be dram voltage droop, 1.35v set in bios, 1.31v in hwinfo, boosted it and I think might be the issue.

It was failing within 20 seconds on actual memory tests, extremely quick, working at 3000, but now past that immediate at 3200 failure after fixing the voltage.