My Simple Guide to Overclocking and Fine-Tuning an Intel Core Ultra 7 270K Plus

I got a good deal on a not-so-good 270K Plus (SP 84), but it still kicks ass. It also took the greater part of the day of undervolting, overclocking, and fine-tuning. Motherboard used was an ASUS TUF Gaming Z890-Pro WiFi updated to BIOS version 3012. Memory used was a Corsair Vengeance RGB 2x16GB 6000MHz CL36 DDR5 kit. CPU Cooler used was a darkFlash DV360S MAX AIO (White).

There's 2 things to think about with the 270K Plus.
1. The typical P-Core, E-Core, and Memory.
2. The Ring, NGU, and D2D.

If you want speed, you'll want #1. Cinebench will be your stability tester. It will take 10+ minutes.
If you want faster latency and snappiness, you'll want #2. Y-Cruncher Benchmark Pi, Multi-Threaded (all cores) will be your stability tester. If it errors or your score is worse by a good bit, dial it back down.

Ordered steps:

1. Set memory overclock
2. Set P-Core voltage and overclock
3. Set E-Core voltage and overclock
4. Set CPU D2D voltage and overclock
5. Set NGU voltage and overclock
6. Set Ring Ratio voltage and overclock
7. Re-adjust NGU overclock
8. Re-adjust CPU D2D overclock

Benchmarks:

• Cinebench 2024 (best to test stability)
• Geekbench 6 (best to see scores getting better or worse)
• CPU-Z (least effective)
• Y-Cruncher (best to test memory stability)

CPU Temp software used was Core Temp.

1. Set memory overclock
Performance Preferences to "ASUS Advanced OC Profile"
Ai Overclock Tuner to "XMP I" or "XMP II" or "Manual"
I prefer manual, for example, setting to 6400MHz at CL36 timings, Gear 1:2 (my options are only 1:2 or 1:4), and leave the secondary timings to "Auto"
NPU Boost left on "Auto"
BCLK Mode to "Auto"
CPU BCLK Frequency left at "100.00"

Tested with Y-Cruncher, Benchmark Pi, Multi-Threaded (all cores), 2,500,000,000 Decimal Digits, 23.1GB.

2. Set P-Core voltage and overclock
ASUS MultiCore Enhancement to "Enabled - Remove All Limits"
SVID Behavior left on "Auto"
Performance Core Ratio to "By Core Usage", but can also be set to "Sync All Cores"
Set all cores to "57"
Specific Performance Core -> Set all Ratio Limits to "57"
DIGI + VRM -> CPU Load-Line Calibration to "Level 5"
Long Duration Package Power Limit to "250"
Package Power Time Window to "128"
Short Duration Package Power Limit to "250"
Power Limit 4 Switch to "Enabled" and "425"
Thermal Velocity Boost -> TVB Voltage Optimizations to "Enabled"
Enhanced TVB to "Disabled"
Overclocking TVB to "Disabled"
CPU P-Core Voltage to "Manual Mode"
CPU P-Core Voltage Override to "1.28000"

Tested with Cinebench 2024 Multi-core. Highest core temp was 88C. If unstable, increase CPU P-Core Voltage Override by 0.01V and test again. If temps get too high, set all cores to "56" and adjust voltage by -0.01V, instead. Keep testing until stable, and lower voltage until it results in instability. Highest VID observe in Core Temp was 1.2612V, with Power as high as 199.2W.

3. Set E-Core voltage and overclock
Efficient Core Ratio to "Sync All Cores"
All-Core Ratio Limit to "48"
Specific Efficient Core -> Set all Ratio Limits to "48"
CPU E-Core Voltage Override to "Manual Mode"
CPU E-Core Voltage Override to "1.17000"

Tested with Cinebench 2024 Multi-core. Highest core temp was 88C. If unstable, increase CPU E-Core Voltage Override by 0.01V and test again. If temps get too high, set all cores to "48" and adjust voltage by -0.01V, instead. Keep testing until stable, and lower voltage until it results in instability.

4. Set CPU D2D voltage and overclock
CPU D2D Ratio to "30"
VNNAON Voltage to "1.00000"

Tested with Y-Cruncher, Benchmark Pi, Multi-Threaded (all cores), 2,500,000,000 Decimal Digits, 23.1GB. If stable, increase CPU D2D Ratio by "1". If unstable, decrease CPU D2D Ratio by "1" until stable.

5. Set NGU voltage and overclock
NGU Ratio to "32"
CPU System agent Voltage to "Manual Mode"
CPU System agent Voltage Override to "1.31000"

Tested with Y-Cruncher, Benchmark Pi, Multi-Threaded (all cores), 2,500,000,000 Decimal Digits, 23.1GB. If stable, increase NGU Ratio by "1". If unstable, decrease NGU Ratio by "1" until stable.

6. Set Ring Ratio voltage and overclock
Ring Down Bin to "Disabled" (if you have it)
Min. CPU Cache Ratio to "41"
Max. CPU Cache Ratio to "41"
Ring DLVR Voltage to "Manual Mode"
CPU Ring Voltage Override to "1.20000"

Tested with Y-Cruncher, Benchmark Pi, Multi-Threaded (all cores), 2,500,000,000 Decimal Digits, 23.1GB. If stable, increase Min and Max CPU Cache Ratio by "1". If unstable, decrease Min and Max Cache Ratio by "1" until stable.

In the end, I got 57P/48E/41R/32N/30D2D. However, just because they pass benchmarks doesn't mean they're entirely stable in real-world use. Make sure you got your E-core overclocks stable before you proceed with overclocking the Ring, NGU, and D2D.

Cinebench 2024 Single Core = 148
Cinebench 2024 Multi Core = 2468


Geekbench 6 Single Core = 3404
Geekbench 6 Multi Core = 23538


CPU-Z Single Core = 954.6
CPU-Z Multi Core = 19785.1


Y-Cruncher = 175.755 seconds


Photo of the build:


Final thoughts:
Initial login to Windows 10 Pro is slow. The startup apps take a while, and I'm not able to fully use my system until after a full minute and 15 seconds, which is very annoying. Also, while my cores are set to 57P/48E, during full loads, it sometimes downclocks to 5.3 or 5.4GHz. I don't know why, yet. I was hoping it would stay at 5.7GHz for all cores. However, after the startup apps are done getting settled, it's a very snappy, quick, and fast system. Input latency is low, and gaming on it has been okay to great, depending on the tuning.

It does not game as well as a 14900K, 14700T, 13600K, 13600T, or 9600X at stock, but when fine-tuned and stable, I think it's better. The problem is, if you don't fine-tune it to where it's stable, you definitely notice how bad it is. But, when you do get it fine-tuned, you definitely notice how good it is. However, in some games, the frametime just isn't great, no matter how well oiled you get it. This is a veteran builder's CPU geared more towards production than games.

It's definitely faster in production tasks when it's fine-tuned and stable as well. It can also be pretty efficient. I haven't seen it get over 199.2W's, with voltages not getting higher than 1.2612V. Max core temps is typically 88C when benchmarking, but in normal tasks, it doesn't get nearly that hot. I've seen it at 92C at times when I'm constantly benchmarking and the heat starts to build up under my desk. TJ Max is 105C, so there's still overclocking headroom.

If you have an i9-14900K with a DDR5 motherboard and memory, I wouldn't feel the need to move to a 270K Plus. The 14900K is a great all-around CPU, and at stock vs stock, I think it's better than the Core Ultra 7 270K Plus. But, when you get the 270K Plus undervolted, overclocked, and fine-tuned, it's an incredible CPU.

I will take a look at mine later. I have done zero tweaking to the 270k. It'll outdo my 14700k and 13600k pretty easy as it sits.

This is an exceptional guide with proven results. A bit more efficient that a 14900K, that's for sure. Mine has some really weak e-cores lol.

shrimpbrime said:

This is an exceptional guide with proven results. A bit more efficient that a 14900K, that's for sure. Mine has some really weak e-cores lol.

Click to expand...

Lol, at least it won't be lower than stock. The P-Cores can almost always get to 5.8GHz, but with increased voltage. Even the best ones typically stay at 5.7GHz, while E-Cores can still get up to 5.2GHz. I've got 2 more to eventually clear, but my worst one (this one so far), only gets up to 4.7GHz (EDITED) 4.8GHz.

El_Capitan said:

Lol, at least it won't be lower than stock. The P-Cores can almost always get to 5.8GHz, but with increased voltage. Even the best ones typically stay at 5.7GHz, while E-Cores can still get up to 5.2GHz. I've got 2 more to eventually clear, but my worst one (this one so far), only gets up to 4.7GHz.

Click to expand...

That's not bad. My 14700K will do 4.7ghz E-cores, but they are less efficient then that 270, so it doesn't mean much. The P-cores are good with HT disabled. So I can see why Intel abandoned the technology for Core Ultra. On that note, if I'm gaming or benchmarking, I will disable HT and run 8P and 8E instead. Seems to have worked out to a decently watted and powerful combination, replacing the HT with some E-cores. I just wish they added 2 P-cores for Core 200 though. I think 10 P-cores would have made for an interesting niche product.

shrimpbrime said:

That's not bad. My 14700K will do 4.7ghz E-cores, but they are less efficient then that 270, so it doesn't mean much. The P-cores are good with HT disabled. So I can see why Intel abandoned the technology for Core Ultra. On that note, if I'm gaming or benchmarking, I will disable HT and run 8P and 8E instead. Seems to have worked out to a decently watted and powerful combination, replacing the HT with some E-cores. I just wish they added 2 P-cores for Core 200 though. I think 10 P-cores would have made for an interesting niche product.

Click to expand...

Interesting. I think I'll try disabling HT on my next reboot. That's something I never really played around with, much.

El_Capitan said:

Interesting. I think I'll try disabling HT on my next reboot. That's something I never really played around with, much.

Click to expand...

For sure. Should see a bit lower temps and better high frequency stability.

El_Capitan said:

Interesting. I think I'll try disabling HT on my next reboot. That's something I never really played around with, much.

Click to expand...

on skylake disabling HT could gain you 1-200mhz at same voltage, haven't tried it on the newer CPUs though