Question Raptor Lake - Official Thread

[Just Benching]

Blender?

Edit: And no pumping in winter air to get temps down

Tried it, doesnt draw more than 270 watts, peak temperature was 84c and average 81c. Ill try some overvolting tomorrow to make it break 300 watts and post the results if you are fine with that

 

[Hitman928]

Tried it, doesnt draw more than 270 watts, peak temperature was 84c and average 81c. Ill try some overvolting tomorrow to make it break 300 watts and post the results if you are fine with that

Is that at stock except unlimited power limit?

 

[DrMrLordX]

You realize the 13600 non-K is based on Alder Lake silicon and not Raptor Lake silicon, right?

1). Several of us tried to point that out earlier either in this thread or the Intel thread.
2). There is the suggestion that at least for the mobile parts, there may be silicon floating around that has 1.25MB L2/core but isn't Alder Lake, though it's hard to know the difference at that point.

 

[cortexa99]

Looks like the voltage setting is a mess. It's interesting to see what would happen next and how can users deal with it.

https://twitter.com/i/web/status/1619453625050492928

Can't undervolt i7 13700K

Hello, my i7 13700k runs pretty hot(under full load 100 degrees) so i want to undervolt it. The problem is, the voltage offset in XTU grayed out preventing me from making any changes. Any idea why and how can i bring the voltage of my processor down? i cant understand why i cant do this in an...

community.intel.com

 

[DrMrLordX]

Wait what? Oh wow, they're doing it to block Plundervolt on client CPUs? Wow that's terrible.

 

[coercitiv]

I'm surprised they're doing it to desktop units. It's still not clear to me whether this will result in undervolting option being removed from the desktop mobos BIOS though.

 

[Just Benching]

Is that at stock except unlimited power limit?

I changed lite load from 9 to 6, other than thah yeah, it's stock unlimited power

 

[Just Benching]

Is that at stock except unlimited power limit?

I overvolted it, managed to hit 310w with a temp limit of 95c. I don't think I can hit 350w, but 330w @ 100c might be possible. Want me to keep pushing?



At this point I don't think the cooler is the problem but more likely the ihs. I was thinking about getting a big AIO or something, but unless I tinker with the IHS (delid or w/e) ill hit the same limitations. I've tried the thermaltake mechanism but doesn't do much so I removed it.

 

[Just Benching]

Wait what? Oh wow, they're doing it to block Plundervolt on client CPUs? Wow that's terrible.

They can't. Even if they remove the voltage options, you still have LLC's to play with

 

[DrMrLordX]

They can't. Even if they remove the voltage options, you still have LLC's to play with

LLC doesn't do everything. You can introduce vdroop, that's it.

I'm surprised they're doing it to desktop units. It's still not clear to me whether this will result in undervolting option being removed from the desktop mobos BIOS though.

According to one of the linked posts, newer microcode updates are causing the CPU to ignore both XTU and UEFI voltage settings below stock.

 

[igor_kavinski]

According to one of the linked posts, newer microcode updates are causing the CPU to ignore both XTU and UEFI voltage settings below stock.

Wow, awesome! This HAD to happen since I bought a used 12700K CPU. Now I'm stuck with a hot CPU with no way ever to make it run cool

 

[DrMrLordX]

Wow, awesome! This HAD to happen since I bought a used 12700K CPU. Now I'm stuck with a hot CPU with no way ever to make it run cool

You can still lower power targets. That's about it.

 

[Just Benching]

Wow, awesome! This HAD to happen since I bought a used 12700K CPU. Now I'm stuck with a hot CPU with no way ever to make it run cool

Stop paying attention to nonsense. There was an article about it back in 2019, that intel was shippping motherboard makers with bioses that prevent UV. As of today, January of 2023, I can still undervolt just fine.

Unless you want to undervolt using XTU, but then the question is...WHY?

 

[Just Benching]

LLC doesn't do everything. You can introduce vdroop, that's it.

Which is basically undervolting. I'd argue - LLC tuning > undervolting in 13th gen, that's why Im just playing with LLC instead of voltage.

 

[DrMrLordX]

Which is basically undervolting.

Mmmhmm

 

[itsmydamnation]

how is LLC undervolting?

you can use LLC to run a more aggressive undervolted , by minimising Vdroop that would cause a crash with your lower voltage. but how the does LLC reduce overall voltage once your in a steady state of 100% utilisation.

funny thing is the best thing I have found to test if Vdroop is an issue is actually Path of Exile end game. induced vdroop related hard crashes that no other application or workload could replicate ( IBT , prime , etc )

 

[Just Benching]

how is LLC undervolting?

you can use LLC to run a more aggressive undervolted , by minimising Vdroop that would cause a crash with your lower voltage. but how the does LLC reduce overall voltage once your in a steady state of 100% utilisation.

funny thing is the best thing I have found to test if Vdroop is an issue is actually Path of Exile end game. induced vdroop related hard crashes that no other application or workload could replicate ( IBT , prime , etc )

Well, how do you think LLC stabilizes your undervolt? By providing more voltage. So by lowering LLC, you provide less voltage --> therefore undervolt.

But I wasn't talking about loadline calibration, which is LLC, but about lite load calibration, which is different. One handles heavy load vdroop, the other one lighter load vdroop. It's better to tinker with these two instead of the traditional undervolting, but that's just me

 

[igor_kavinski]

But I wasn't talking about loadline calibration, which is LLC, but about lite load calibration, which is different. One handles heavy load vdroop, the other one lighter load vdroop. It's better to tinker with these two instead of the traditional undervolting, but that's just me

I don't see any option for lite load calibration in my ASROCK Z790 BIOS. Can you describe your process of UV using LLC?

 

[mikk]

I couldn't undervolt my i5-12500 without performance loss and found two settings on my Asus Z690 which gives me the desired effect without performance loss.

SVID Behavior -->Best-Case Scenario
CPU Load-Line Calibration -->Level 1

It reduced the power consumption in CB R23 from 100W to 73W. Because the boxed cooler struggles with 100W it is very important to lower the default consumption. Of course I also could lower the power limit but this will lower the performance which I don't want.

 

[Hitman928]

I overvolted it, managed to hit 310w with a temp limit of 95c. I don't think I can hit 350w, but 330w @ 100c might be possible. Want me to keep pushing?
View attachment 75526


At this point I don't think the cooler is the problem but more likely the ihs. I was thinking about getting a big AIO or something, but unless I tinker with the IHS (delid or w/e) ill hit the same limitations. I've tried the thermaltake mechanism but doesn't do much so I removed it.

So you are hitting thermal throttle as you push into 300W territory. And yes, of course the IHS causes additional thermal resistance which makes the chips harder to cool compared to bare die, this has been known since they were first introduced. Adding a beefier cooler will help overcome this by a bit but this is a big reason why I said even 360mm AIOs have real trouble cooling a CPU above 300C, the largest thermal resistance in the chain remains the thermal resistance between the bare die to the IHS. Even when the thermal interface inside the IHS is solder, instead of paste, you still have almost an order of magnitude less thermal conductivity than copper and the thickness of the solder is relatively large to allow for manufacturing tolerances and to protect the die. Add to that another (but relatively thin) layer of thermal paste (even worse thermal conductivity than solder) and contact issues due to manufacturing variations, and it becomes nearly impossible to cool a CPU pushing over 300W with anything that's not sub ambient.

If you risk taking the IHS off and going bare die with something like liquid metal, things can change drastically. Even then, I think your cooler would probably have a hard time with a 350W CPU just because of thermal hotspots in modern CPUs, but it would have a much better shot than with the IHS on. A tower cooler with heatpipes and large fans could theoretically dissipate a ton of power, probably 350W+ under ideal conditions (i.e. the heat is widely spread across the base of the heatsink with even contact and an excellent thermal interface) but trying to cooler modern CPUs is very far from ideal.

 

[Just Benching]

So you are hitting thermal throttle as you push into 300W territory.

I thermal throttle cause i set the temperature limit at 95c instead of 100c. But I mean all this is pointless, the CPU doesn't scale that high at all in performance, even with 24cores - 300+ watts are overkill.

 

[Hitman928]

I thermal throttle cause i set the temperature limit at 95c instead of 100c. But I mean all this is pointless, the CPU doesn't scale that high at all in performance, even with 24cores - 300+ watts are overkill.

A better cooler would allow you to sustain higher frequencies for much longer, even if both were pushing 300W and experiencing thermal throttling. I agree, the difference wouldn't be significant, maybe like 5% best case, but that's not really what we were discussing.

 

[Just Benching]

A better cooler would allow you to sustain higher frequencies for much longer, even if both were pushing 300W and experiencing thermal throttling. I agree, the difference wouldn't be significant, maybe like 5% best case, but that's not really what we were discussing.

I might test it cause im curious myself, but I doubt that even the best AIO would give me more than 10-20w with a 95c limit. I have a couple of spare AIOs here, when I feel like it I might give it a go. Tom's hardware tested the LT720, which is the best AIO they ever tried, he hit 100c at 315 watts, which is basically the same ballpark / worse than the results I get with the u12

 

[Hitman928]

I might test it cause im curious myself, but I doubt that even the best AIO would give me more than 10-20w with a 95c limit. I have a couple of spare AIOs here, when I feel like it I might give it a go. Tom's hardware tested the LT720, which is the best AIO they ever tried, he hit 100c at 315 watts, which is basically the same ballpark / worse than the results I get with the u12

Due to how modern CPUs handle boosting, it's power and frequency.

 

[Just Benching]

There is a thread over at ocuk were we are testing rpl with zen 3 and zen 4 at those heavy cyberpunk areas. Results are.. eyeopening. The 7600x and the 3d struggle to hit 40 - 50 fps, while a 13600k casually hits 120+.