Overclocking Raptor Lake - 13900k,13700k, 13600k etc results, bins and discussion

affxct said:

13700K 79-89-59 💀☠

Yes, I attempted the Falkentyne method to no avail. It’s a legit SP 79 on two different boards. 🥹🔫

Click to expand...

acoustic said:

Yep, that's actually what I'm trying right now. Stock P+E-Cores, with Ring @ 49x. I pulled a CLOCK_WATCHDOG_TIMEOUT BSOD 13minutes into Realbench.

I don't know how good this chip is. I know it can do 57/43/45 @ 1.250v VR VOUT in CB23, so I'd imagine the P-Cores don't need 1.215v VR VOUT for 5.5Ghz. I could have weak E-Cores or weak Ring; just a matter of narrowing it down

EDIT: 16minutes in now, and still going. I think it's safe to pin the blame on the Ring. Going to let this go to 25min before I put the E-Cores back to 45x.

Just for others reference, CB23 30min passed @ 50x Ring and 45xE-Cores with lower VR VOUT than this. I don't think it's a good way to test any form of stability. Realbench seems much more functional, and it's also slightly lower power usage than CB23. I'm using 240watts right now with 1.215v VR VOUT. Seems to test the entire package better than CB23.

I don't remember what caused me to download and start using it again, but I'm glad I did. I used to use Realbench as my main stability test with prior generations.

Click to expand...

tps3443 said:

I’ve been testing a new 13900K today. This one is Batch X241M888 (4th chip) Force rating is like 143. So definitely nothing spectacular.

But hey! It still runs 5.7-6.2Ghz on the P-cores. Average ring, average E-cores

This is a Slightly above average chip. I could definitely rock this daily. All chips perform within like 1-3% right whether it be SP98 or SP115 lol

(This thing is nothing like my other CPU, but it’s amazing for just an average chip, regardless it’s still a 13900K at the end of the day lol)

View attachment 2585776

Click to expand...

Falkentyne said:

The problem is, those VF points that you see are based on 2 core load, Those 1.408/1.418v VIDs that you see are based on two cores turbo boosting to 5.8 ghz on a light workload running on no more than 4 threads. And also based on Intel stock TDP and electrical limits being used. Your fallacy here is assuming that just because those cores say 1.408/1.418v VID, it means that all 8 of the P-cores are designed to run at 1.418v load. Not only are they NOT designed to, if you tried to do it, not only would you seriously degrade your CPU, but you would be unable to cool it!

Click to expand...

Ichirou said:

The 13700 (and 13600 and so on) have different SP tiers; they can't really be compared. They'll always be somewhat worse overall compared to the 13900.

CLOCK_WATCHDOG = too low Vcore at that moment in time.

Great, now you can rock this chip and give me your CPU Force 117 chip
Just let me know your price (dead serious)

Click to expand...

affxct said:

13700K 79-89-59 💀☠

Yes, I attempted the Falkentyne method to no avail. It’s a legit SP 79 on two different boards. 🥹🔫

Click to expand...

Ichirou said:

The 13700 (and 13600 and so on) have different SP tiers; they can't really be compared. They'll always be somewhat worse overall compared to the 13900.

CLOCK_WATCHDOG = too low Vcore at that moment in time.

Great, now you can rock this chip and give me your CPU Force 117 chip
Just let me know your price (dead serious)

Click to expand...

HemuV2 said:

Try getting a replacement? Or is tht not possible in za?

Click to expand...

Krautmaster said:

i've heard the V/F table is bugged below BLK of 100.0, I can also set like more than -200mV on many points with any issue and no change when i check the voltages in HWinfo.
If I limit to like 120W, and i set the ring to eg 51 or 49 whatever, the performance is lower than with ring on auto. Give it a try. CB23 / CB21 for example.

Example:
View attachment 2585707

I can run my FFMPEG benchmark with 8x58 + 16x45. Load Line Calibration on Mode 1, Lite Load Control on Mode 5.
Adaptive 1.45V
HWInfo shows me 1.35V on load, its even temperature throtteling time to time with slightly over 300W.

I can reboot, set 56x8 instead and set Load Line Calibration to Mode 3, Lite Load Control to Mode 1.
Adaptive 1.45V
That gives me >1,45V, struggeling to get 5600Mhz stable with that then and it tells me 80°C at 350W which seems to be less voltage showing 1.45V as it might have been in the first setting which showed 1.35V.

See here. Load Line 3 and Lite Load Auto
View attachment 2585708

Adaptive 1.55V, no throtteling, <300W? Other than the above screen its running but not 100% Stable, temps are lower. The real voltage might dip a lot dies to LLC Mode 3 ? So even while showing 1.51V it might be lower than above where HW Info tells me 1.35V? Temps seem to be lower too so ... 0.2V reading difference?

Click to expand...

HemuV2 said:

so i dont think i have done this yet, but here is my cpu running cinebench at locked 55/43/45x stock configuration, as you can see the package power and vrm power are different. i have set it to 1.36V llc5 in bios and this is around 1.206V die sense for this config to run without complaints. would like @Falkentyne @RobertoSampaio @Ichirou input on this as they seem to have seen tons of samples at this point, when i lowered voltage to like 1.34 it ran only 2-3 loops and errored(not WHEA but CB crashed). this is 109/73 sample btw

View attachment 2585265

Click to expand...

Krautmaster said:

Yeah if u push the cooling I'm round 90-95 too and 300W and no throttling. Did not try to set more negative offset. It's adaptive on Auto -0.01V
Digi LLC mode 2, Lite load 20/89

But as said. Depending on the LLC I can get same temperature and consumption with 1.4V+
Edit: Like urs
View attachment 2585728

Edit. That way I started but if I lower the MSI Digi LLC to any below 2 it gets rapidly instable. You may try to tweak it by going to multi 60, use TVB to set eg 80 °C and -3x offset. That way I hold 8x6Ghz in Games easily.
With LLC Mode 3 stable but i had to do like +0.01V on the highest two A/V Points
View attachment 2585733

Click to expand...

Falkentyne said:

1.34v Bios set + LLC5 is about 1.190v die sense load.
I think that you're crashing because of the poor E-cores.
if you set them to x42, I bet you would pass this.
And since you don't BSOD and R23 just completely crashes, if it's never a BSOD that's a pretty good reason

Click to expand...

7empe said:

Does SP79 keeps advertised speed of 55x all core at all? If so, what's the die sense vcore needed?

Click to expand...

Falkentyne said:

1.34v Bios set + LLC5 is about 1.190v die sense load.
I think that you're crashing because of the poor E-cores.
if you set them to x42, I bet you would pass this.
And since you don't BSOD and R23 just completely crashes, if it's never a BSOD that's a pretty good reason

Click to expand...

HemuV2 said:

Regardless of SP any cpu will do the stock configuration at reasonable voltage, beyond that even 200mhz might be an issue if your SP is super low, but for i7 i dunno how the numbers work as we haven't seen many i7 sp numbers on here

Click to expand...

7empe said:

Well, my colleague's 13900k had SP71 (can't remember p-cores score) and it could not maintain 55x all-core at reasonable cooling. It required something like 1.30V die-sense, which without water cooling is impossible to cool down to avoid throttling. He just gave it back claiming that product does not meet advertised criteria and he got a new one, much better chip. Every chip has to run at advertised speed with stock (air?) cooling.

Click to expand...