13th Gen Intel stuff

[Trig]

What is it that's killing them?

I've just picked up a i5-13500 which, when running BIOS defaults will happily sit at 65w according to HWMonitor, with a brief stint at 135w when you initially kick Cinebench in.
I've got a beta BIOS on there with the new 0x129 microcode, which apparently fixes whatever issues there are, with that installed and the BIOS set to the 'Intel Performance' setting, rather than the defaults, it seems to run at 135w happily for the entire 10 minute Cinebench run and doesn't go above 71oC...

I'm guessing that sticking over 50% of its rated 65w wont be great for it, so will most likely knock off the overboost or whatever it is doing it, but it got me thinking what the actual issue is with these things...

 

[jigger]

Intels design.
Power level.
Materials.
Possibly node/production issues.

 

[Deleted User 456458]

What is it that's killing them?

I've just picked up a i5-13500 which, when running BIOS defaults will happily sit at 65w according to HWMonitor, with a brief stint at 135w when you initially kick Cinebench in.
I've got a beta BIOS on there with the new 0x129 microcode, which apparently fixes whatever issues there are, with that installed and the BIOS set to the 'Intel Performance' setting, rather than the defaults, it seems to run at 135w happily for the entire 10 minute Cinebench run and doesn't go above 71oC...

I'm guessing that sticking over 50% of its rated 65w wont be great for it, so will most likely knock off the overboost or whatever it is doing it, but it got me thinking what the actual issue is with these things...

With eTVB the Voltage requested for boost was too high and depending on the voltage tolerance of the chip, could lead to degradation.

 

[Rroff]

These are covered in the 14th gen thread and the "Intel has a Pretty Big Problem.." thread - but short version is there still isn't definite information.

There appears to be at least 3 different issues with one of those probably having 2 different incarnations - some supposedly small number of 13th gen have a via oxidation issue in production, a small number of 13th gen have an issue with stability out the box separate to the bigger issue which has gained prominence with the 14900s and then this big issue which is kicking off which at least partly seems to be related to incorrect voltage levels for certain conditions.

But so far nothing has been confirmed as to the root cause(s) and scale of the problem.

 

[Tetras]

I've just picked up a i5-13500

I'm pretty sure those CPUs are always C0 dies, so I think you're fine whatever.

I was talking to someone with a C0 die earlier and they confirmed that their CPU's microcode wasn't even updated with the fixed BIOS (C0 and B0 don't share microcode), though would like to get more confirmation on that.

I'm guessing that sticking over 50% of its rated 65w wont be great for it, so will most likely knock off the overboost or whatever it is doing it, but it got me thinking what the actual issue is with these things...

Those numbers are fine, using the CPU's normal boost within PL1/PL2 is unlikely to cause degradation. If it did, we'd have a lot of dead CPUs from previous gens.

In terms of broken 13th-14th gen CPUs that have the overvolting microcode or are from the bad batch, that's a different question.

 

[Trig]

Ok, so we've not had enough info from Intel to narrow down a particular set of 'things' that will cause a certain batch or chip from a certain fab to die then, typical..
I've just come from a 5800x to a 13500, just built a machine for a mates kid with a 13600KF in it and another machine that seemed to be throwing random BSOD's out running a 14900K that keeps coming back to me from another mates kid...

 

[Blackbadger]

I understand some asus boards provide a SP score,in the bios which is effectively is a silicon quality calculation made by reading the vid tables for that chip. Which effectively is the voltages required for stable operation the VID settings are unique to the CPU (programmed by intel at factory).
The higher the sp score the better, they give readings for the e and p cores, ranging from 70 for bad, 90 moderate, 100 good, 110+ being golden samples. Although I have seen people getting scores as low as 30, which would mean by default the chip requires more voltage.


I wonder if their is a correlation between the lower sp scores and failures as they would be requesting a higher voltage.


Does anyone know if there is a simple tool for users that can access the chips vid tables if they don't have a special motherboard that has such calls built in ? As I believe a lot of tools only provide the data after the motherboard etc has messed with it.

 

[jigger]

I understand some asus boards provide a SP score,in the bios which is effectively is a silicon quality calculation made by reading the vid tables for that chip. Which effectively is the voltages required for stable operation the VID settings are unique to the CPU (programmed by intel at factory).
The higher the sp score the better, they give readings for the e and p cores, ranging from 70 for bad, 90 moderate, 100 good, 110+ being golden samples. Although I have seen people getting scores as low as 30, which would mean by default the chip requires more voltage.


I wonder if their is a correlation between the lower sp scores and failures as they would be requesting a higher voltage.


Does anyone know if there is a simple tool for users that can access the chips vid tables if they don't have a special motherboard that has such calls built in ? As I believe a lot of tools only provide the data after the motherboard etc has messed with it.

Intel might have something like that, but I can’t see how reading that data could be useful even if the reported score was accurate which it almost certainly isn’t.

 

[teejay73]

Slightly off topic, but I've noticed my 13700k when doing a cpu stress test in occt maxes out at 3.4ghz on the p cores & hits the tdp of 253 watts, whereas when trying an undervolt of 0.100 I get 4.4, though again maxing out at 253w. I've noticed this since I've updated the microcode today. Does this happen to anyone else, or is this normal?
In Cinebench 24 I hit 5.3ghz and the tdp is lower with the undervolt, which makes sense of course, though with no undervolt it fluctuates between 4.8 & 5.3ghz. I guess what I'm seeing is a tdp throttling, but I would have thought that 3.4ghz is too low or am i reading this wrong?

Edit - I should also add that the temps are fine. With no undervoltage applied, I'm hitting 92c, with a water temp of 36-37c. With the undervolt applied at 0.100 temps are lower by about -5c.

 

[spooks]

Slightly off topic, but I've noticed my 13700k when doing a cpu stress test in occt maxes out at 3.4ghz on the p cores & hits the tdp of 253 watts, whereas when trying an undervolt of 0.100 I get 4.4, though again maxing out at 253w. I've noticed this since I've updated the microcode today. Does this happen to anyone else, or is this normal?
In Cinebench 24 I hit 5.3ghz and the tdp is lower with the undervolt, which makes sense of course, though with no undervolt it fluctuates between 4.8 & 5.3ghz. I guess what I'm seeing is a tdp throttling, but I would have thought that 3.4ghz is too low or am i reading this wrong?

Edit - I should also add that the temps are fine. With no undervoltage applied, I'm hitting 92c, with a water temp of 36-37c. With the undervolt applied at 0.100 temps are lower by about -5c.

I'm pretty sure my 13700k & 14900k boost to lower frequencies when in OCCT vs Cinebench. In fact, I am positive. I wondered the same but didn't really worry about it as its just a synthetic tool/benchmark.

 

[teejay73]

I'm pretty sure my 13700k & 14900k boost to lower frequencies when in OCCT vs Cinebench. In fact, I am positive. I wondered the same but didn't really worry about it as its just a synthetic tool/benchmark.

Thanks for letting me know. Tbh, it seems weird that a progarm you'd test for stability, isn't able to maximise the full power of the cpu because of reching the tdp of 253w!? It's making me second guess if my undervolt is stble or not.

 

[Trig]

My 13500 runs pretty cool tbh, in Cinebench at least, I'm about to do a reinstall of Windows 11 and then I'll see whats what on the CPU front.
At the mo PL1 is 252w, PL2 300w, temps are low 70's, so would presume it has more headroom, but I've spent the last good while on AMD so are more used to PBO stuff than whatever Intels auto boosty stuff is or does...