Discussion Raptor Lake Build Thread

Page 8 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

shorty1111

Junior Member
Jun 21, 2017
23
9
81
@shorty1111, you should run Aida64 memory and cache benchmark because if your cache is downclocked and undervolted like that it will affect the performance of the entire CPU.

It could also be why you're getting such low power draw despite setting PL values to 235w.
aha sure! ill try aida64 :D and ill try also to reset cache to default !!! ty !!!
 

shorty1111

Junior Member
Jun 21, 2017
23
9
81
I went ahead and ordered the Thermaltake contact plate. As advertised it did lower my temps about 6C. It allowed me to increase power from 175 to 200 watts while holding temps at 80C max.
niiice, im thinking of doing the same but im scared XD also im looking to get EK waterloop..
 

NGruia

Junior Member
Apr 29, 2019
12
11
51
Tried EK WB 240mm ARGB. Completely overwhelmed 101 Celsius Package, 100 core. Even with 4x120 Noctua iPPC 2000 rpm, push- pull.
360mm maybe...
 

NGruia

Junior Member
Apr 29, 2019
12
11
51
About 240W but with Enermax DF Storm 2x120mm. Beats Noctua 140mm iPPC 3000 rpm in sound levels.
220 W with Noctua 2x120 mm iPPC 2000 rpm, 240W with 4x Noctua.
Max temp 97 Celsius....
 

shorty1111

Junior Member
Jun 21, 2017
23
9
81
i was thinking of getitng 3000prm fans, instead of 1700 that i have atm for AIO. and my case fans are 1000rpm.. i think i can get lower temps if i change fans :D
 

Hulk

Diamond Member
Oct 9, 1999
3,695
1,299
136
All CPU cooling tests should be performed with a 13900K. Not that it's an honor by any means but it generally has the capability to push any non-custom loop cooler to the limit.
 

Kocicak

Senior member
Jan 17, 2019
763
796
136
I wouldn't underclock the cache, but undervolting it should be fine as long as its stable. It's amazing that the 13900K can drop the voltage so much and still be capable of hitting those clock speeds.
I have been examining how 13900K can perform at lower clocks and undervolted, at this moment I have -100 mV offset, all P cores at 5400 MHz, E cores at 4000 MHz, I can reach 37000 CNB R23 score at 210W power limit, that is the absolute maximum my air cooler can handle short term. You can see resulting voltage under load (8 and all threads) in the screenshots.

underclock volt 54 40 -100mV 8T xT 210W.png
 
  • Like
Reactions: Carfax83

Hulk

Diamond Member
Oct 9, 1999
3,695
1,299
136
I have been examining how 13900K can perform at lower clocks and undervolted, at this moment I have -100 mV offset, all P cores at 5400 MHz, E cores at 4000 MHz, I can reach 37000 CNB R23 score at 210W power limit, that is the absolute maximum my air cooler can handle short term. You can see resulting voltage under load (8 and all threads) in the screenshots.

View attachment 71556
5GHz P, 4GHz E will get you over 37,000 so I don't think your P's are actually holding 5.4GHz.
 

Kocicak

Senior member
Jan 17, 2019
763
796
136
You just described the screenshot! :) You can see that most P cores were at 5 GHz and E cores at 3.9. The highest score I could get was 37 342.

I will try to set the frequency limits to 5000 and 3900 and set a higher voltage drop to see what happens.

EDIT: I tried that frequencies with 150 mV drop, but the score was a just below 37K, it seems that the voltage drop it too high and the CPU cannot function 100% at such low voltage.
 
Last edited:

Hulk

Diamond Member
Oct 9, 1999
3,695
1,299
136
You just described the screenshot! :) You can see that most P cores were at 5 GHz and E cores at 3.9. The highest score I could get was 37 342.

I will try to set the frequency limits to 5000 and 3900 and set a higher voltage drop to see what happens.

EDIT: I tried that frequencies with 150 mV drop, but the score was a just below 37K, it seems that the voltage drop it too high and the CPU cannot function 100% at such low voltage.
Ah! Got it!
That's about where I'm at as well. 180W PL1 with a 25 second 225W PL2 "burst."
Allows my 13900K to "go nuts" for a bit with bursty workloads.
 

DrMrLordX

Lifer
Apr 27, 2000
20,518
9,605
136
All CPU cooling tests should be performed with a 13900K. Not that it's an honor by any means but it generally has the capability to push any non-custom loop cooler to the limit.
Do 13900ks have hotspots? If not, then it wouldn't be any better than generic hotplate tests.
 

Carfax83

Diamond Member
Nov 1, 2010
6,805
1,494
136
You just described the screenshot! :) You can see that most P cores were at 5 GHz and E cores at 3.9. The highest score I could get was 37 342.

I will try to set the frequency limits to 5000 and 3900 and set a higher voltage drop to see what happens.

EDIT: I tried that frequencies with 150 mV drop, but the score was a just below 37K, it seems that the voltage drop it too high and the CPU cannot function 100% at such low voltage.
That doesn't sound right at all. I recently lowered my offset voltage even more. Used to be --0.100mv, now it's --0.165mv with the same clock speeds, 5.2ghz for the P cores and 4.3ghz for the E cores at PL1 and PL2 set to 215w. According to CPU-Z, my voltage at max load is 1.066v. And when I run Cinebench R23, my score is 39,480.

So clearly my CPU is still operating normally despite this ultra low voltage and I haven't crashed not once. I think Raptor Lake is current limited and not voltage limited.

What's likely holding you back is your PL1 and PL2 values are too low. Try 215w to 220w instead. I may reduce my voltage even more in the future, but for now, I'm good :D
 
  • Wow
Reactions: igor_kavinski

shorty1111

Junior Member
Jun 21, 2017
23
9
81
That doesn't sound right at all. I recently lowered my offset voltage even more. Used to be --0.100mv, now it's --0.165mv with the same clock speeds, 5.2ghz for the P cores and 4.3ghz for the E cores at PL1 and PL2 set to 215w. According to CPU-Z, my voltage at max load is 1.066v. And when I run Cinebench R23, my score is 39,480.

So clearly my CPU is still operating normally despite this ultra low voltage and I haven't crashed not once. I think Raptor Lake is current limited and not voltage limited.

What's likely holding you back is your PL1 and PL2 values are too low. Try 215w to 220w instead. I may reduce my voltage even more in the future, but for now, I'm good :D
niiiiiiiiiiiiiiiiiiiice, temps? and cooler
 

Tech Junky

Platinum Member
Jan 27, 2022
2,177
752
96
It's difficult to cool
You can make anything difficult if you try hard enough. I did a stress test of my 12700k the other day pushing to 100% with a cheap 6 pipe air cooler and it didn't hardly break a sweat. The fans on the cooler ramped up to 1800 and it stayed under 50C. The other fans stayed at their normal speeds and everything was stable.
 

Hulk

Diamond Member
Oct 9, 1999
3,695
1,299
136
I had a 12700K and now have a 13900K. As far as cooling they are two completely different animals. The 12700K will crap out due to frequency at MUCH lower frequency than the 13900K. The 13900K will just keep going and going if you feed it enough voltage and can keep it cool. That's why you see them pulling well over 300 Watts with good cooling solutions.

Now this is admittedly an academic exercise as you aren't really getting much performance in return for that extra 100-150W. But if you do want maximum performance for bursty workloads and want all of those P's at 5.5 or 5.6GHz and the E's at 4.3 or 4.4 GHz then you are going to have to work at it.
 

Hulk

Diamond Member
Oct 9, 1999
3,695
1,299
136
BTW, does anyone know at what temp HWinfo will report "thermal throttling" in red font.

Seem like it's something over 80C but I'm not sure exactly?
 

DrMrLordX

Lifer
Apr 27, 2000
20,518
9,605
136
Immaterial. It's difficult to cool making it a great test subject.
No, really. Anyone using TSMC nodes seems to get hotspots with their end products. My entire current machine is a case study in that phenomenon. Also die size and positioning matters. Some manufacturers (*cough*AMD*cough*) make cooling more "interesting" by positioning heat sources in very specific parts of the package. It can, at the very least, change where you want to position your TIM during the mount. It can make HDT coolers worse since some of the pipes might not line up well with generally-hot areas of the IHS. Etc.

Just because something is good at cooling a 13900k doesn't mean it's good for cooling everything else.
 

Hulk

Diamond Member
Oct 9, 1999
3,695
1,299
136
No, really. Anyone using TSMC nodes seems to get hotspots with their end products. My entire current machine is a case study in that phenomenon. Also die size and positioning matters. Some manufacturers (*cough*AMD*cough*) make cooling more "interesting" by positioning heat sources in very specific parts of the package. It can, at the very least, change where you want to position your TIM during the mount. It can make HDT coolers worse since some of the pipes might not line up well with generally-hot areas of the IHS. Etc.

Just because something is good at cooling a 13900k doesn't mean it's good for cooling everything else.
It's been a few years since Heat Transfer in college... but here goes, let's discuss.

There is of course validity in what you are saying. As you know there are 3 types of heat transfer, conduction, convection, and radiation. Heat is transferred from the CPU to the cooler contact plate via conduction. Assuming a high conduction material such as copper and good contact between the surfaces there will be limit to the heat transfer coefficient between these two surfaces.

Next the heat must be removed from the system. With both air and water eventually air is used to disperse the thermal energy from cooler.

The greater the temperature difference between the cooler contact plate to the CPU the greater the heat transfer. And a cooler that can dissipate more thermal energy will maximize this difference thus maximizing heat transfer. Hot spots or no hot spots thermally a more efficient cooler will work better on both a homogeneous and heterogeneous hot silicon.

Assuming the contact between the cooler and CPU is optimal I assert that a cooler that can move more joules will be better on both an evenly heated CPU and one that has hot spots as the heat moving in the cooler will immediately "flow" to the cooler spots, thus the ability of the cooler to dissipate heat will ultimately be the limiting factor on it's effectiveness. Again this assumes optimum heat transfer from the CPU to the cooler in both cases.

Now I will admit there are many other factors to be involved here. If you take a cooler that can move a lot of heat on a CPU that is evenly heated but has shitty thermal contact to the CPU that's going to be a bigger problem than a CPU that can't move a lot of heat but has great thermal contact. This condition would be better for a CPU like Zen 4, which has hot spots yet doesn't need to remove as much heat as the 13900K.

But in reality most high end coolers have very good thermal contact and are made of materials with high thermal conductivity so as I wrote above when dealing with real world examples where thermal contact to the CPU is optimal the 13900K based on it's thermal demands is a great test unit for CPU cooling.

For this not to be the case we'd have to be talking about a theoretical niche case on the far end of the spectrum of what actually happens. ie find a cooler that can handle the 13900K but can't handle the 7950X both operating at max manufacturer specs? Won't happen. But I bet you can find a cooler that will handle the 7950X but not the 13900K. Hot spots are much less important than overall cooling capacity. If you have pick a CPU to test on I stand by my assertion that the 13900K is a great one.
 

ASK THE COMMUNITY