Horizon OC 2.0 - Most powerful overclock tool. CPU 2397MHz, GPU 1228MHz, RAM 3200+MHz

Flaya · Jun 1, 2026

verypedro said:
For some reason I just cannot get it to go over the stock/safe clock, no matter what I do. This was never a problem with sys-clk before

Same here. But then again, I couldn't with sys-clk either.

l7777 · Jun 2, 2026

verypedro said:
For some reason I just cannot get it to go over the stock/safe clock, no matter what I do. This was never a problem with sys-clk before

Just went through this myself on a new device.

Having CPU High UV set to something is a requirement before the higher clocks can be used. The guide recommends starting with 5 and adjusting from there with testing.

For GPU, you need to set the GPU Undervolt Table to "High UV Table" then set the voltage for 1267 and 1305 to Auto. Secondarily, if you have a lower SOC speedo rating you may have to increase the GPU Maximum Voltage.

ISHOWMANI · Jun 3, 2026

Hello everyone,
I spent most of the day trying to figure out why retro games downloaded through Tinfoil were no longer being installed. I use this feature frequently for older games, so it was quite frustrating.
At first, I suspected a Sys-patch issue because the games would download normally, but then get stuck on a black screen before the ROM installation process could begin.
One important detail is that all previously installed retro games continue to work correctly. The problem only affects newly downloaded games.
To troubleshoot, I performed a clean Atmosphère installation and started reinstalling my modifications one by one. After some testing, I found that the issue appears immediately after installing Horizon OC.
As a workaround, I created a separate boot entry in hekate_ipl.ini without the Horizon OC modifications. With that setup, retro game installations through Tinfoil work normally again.
At the moment, I switch between two boot entries: one with Horizon OC enabled for everyday use, and another without Horizon OC whenever I need to install new retro games.
I am posting this mainly as a bug report in case the developers would like to investigate it. Based on my testing, there seems to be a conflict between Horizon OC and Tinfoil's retro game installation process. Hopefully this information helps identify and resolve the issue in a future update.

Souldbminer · Jun 5, 2026

ISHOWMANI said:
Hello everyone,
I spent most of the day trying to figure out why retro games downloaded through Tinfoil were no longer being installed. I use this feature frequently for older games, so it was quite frustrating.
At first, I suspected a Sys-patch issue because the games would download normally, but then get stuck on a black screen before the ROM installation process could begin.
One important detail is that all previously installed retro games continue to work correctly. The problem only affects newly downloaded games.
To troubleshoot, I performed a clean Atmosphère installation and started reinstalling my modifications one by one. After some testing, I found that the issue appears immediately after installing Horizon OC.
As a workaround, I created a separate boot entry in hekate_ipl.ini without the Horizon OC modifications. With that setup, retro game installations through Tinfoil work normally again.
At the moment, I switch between two boot entries: one with Horizon OC enabled for everyday use, and another without Horizon OC whenever I need to install new retro games.
I am posting this mainly as a bug report in case the developers would like to investigate it. Based on my testing, there seems to be a conflict between Horizon OC and Tinfoil's retro game installation process. Hopefully this information helps identify and resolve the issue in a future update.

Piracy is not supported.

ChanseyIsTheBest · Jun 5, 2026

ISHOWMANI said:
Hello everyone,
I spent most of the day trying to figure out why retro games downloaded through Tinfoil were no longer being installed. I use this feature frequently for older games, so it was quite frustrating.
At first, I suspected a Sys-patch issue because the games would download normally, but then get stuck on a black screen before the ROM installation process could begin.
One important detail is that all previously installed retro games continue to work correctly. The problem only affects newly downloaded games.
To troubleshoot, I performed a clean Atmosphère installation and started reinstalling my modifications one by one. After some testing, I found that the issue appears immediately after installing Horizon OC.
As a workaround, I created a separate boot entry in hekate_ipl.ini without the Horizon OC modifications. With that setup, retro game installations through Tinfoil work normally again.
At the moment, I switch between two boot entries: one with Horizon OC enabled for everyday use, and another without Horizon OC whenever I need to install new retro games.
I am posting this mainly as a bug report in case the developers would like to investigate it. Based on my testing, there seems to be a conflict between Horizon OC and Tinfoil's retro game installation process. Hopefully this information helps identify and resolve the issue in a future update.

I think it has to do with the loader.kip overwriting some stuff that old fowarders used as a workaround. You'ld have to update those old fowarders which would be a pain.

ISHOWMANI · Jun 5, 2026

ChanseyIsTheBest said:
I think it has to do with the loader.kip overwriting some stuff that old fowarders used as a workaround. You'ld have to update those old fowarders which would be a pain.

But it's not the forwarders of the games already installed that don't work.
It's the retro games just downloaded on tinfoil. But fortunately I found the solution to restart the console without the horizon.

Thank you for the answer, friend!

k0neer · Jun 9, 2026

[OLED Mariko] Ultimate Horizon-OC Profile: 90.8% Efficiency & Extreme Undervolt!
Custom Tuning for Samsung AB-MGCL Memory Chips

Hey everyone,

After many surgical stability tests, benchmark runs, and individual timing trial-and-errors using the Horizon-OC suite, I’ve managed to find the absolute "golden sweet spot" for my Nintendo Switch OLED.

I’m sharing my custom hybrid profile that successfully breaks the 90% efficiency barrier in Membench while operating at bone-stock factory memory voltages (1100mV/600mV). Thermals are incredible under continuous load (55 °C max during Furmark RAM stress test) and it is 100% stable after 30+ minutes running through Korok Forest in Zelda BotW.

1. Fixed Target Clocks & Efficiency Focus

CPU: 1963.5 MHz
GPU: 998.4 MHz / 921.6 MHz (Fully stable profiles)
RAM Clock: 2132 MHz (Actual configuration value: 2133000)
Handheld Mode Focus Trim: This specific profile is heavily optimized and targeted for Handheld/Portable mode, focusing on extreme battery longevity and ice-cold thermals.
The Sweet Spot: Pushing the Overclock any higher is completely unnecessary for the vast majority of Nintendo Switch games. This 2133 MHz setup delivers maximum real-world smoothness without melting your silicon or destroying your daily battery life.
Battery-Powered Testing (No Dock/AC): To fully confirm true portable stability, all synthetic stress tests and real-world gaming loops were performed strictly on battery power, without using the official AC adapter or dock. This proves the system maintains flawless electrical stability even under shifting battery voltage floors.

2. Custom Hybrid RAM Timings (The "Holy Grail")
Starting from the Common preset, I incrementally tightened each register one by one to find the absolute physical limit before encountering boot errors (finding the sweet spot between Common and Super Tight):

t1_trcd = 4 ➔ Hard physical limit (Setting to 5 causes boot error)
t2_trp = 4 ➔ Perfectly optimal and stable
t3_tras = 8 ➔ Super Tight (ST) value (Unlocks massive bandwidth)
t4_trrd = 5 ➔ The sweet spot (Setting to 6 introduces a minor MB/s penalty)
t5_trfc = 6 ➔ Stability ceiling (Setting to 7 triggers immediate boot error)
t6_trtw = 5 ➔ Crucial: Locked at 5 to completely bypass default ST black screens
t7_twtr = 7 ➔ Safe zone (Setting to 8 causes boot error)
t8_trefi = 6 ➔ Firmware/Suite hard-lock ceiling

Final Timing String:

Code:

4-4-8-5-6-5-7-6

3. CPU & GPU Overview (Aggressive Undervolt Analysis)

CPU Scaling Trim: My Mariko chip is running a very aggressive undervolt curve (
Code:
```
mariko_cpu_uv_high=10
```
). Thanks to excellent silicon lottery luck, my console sustains high boost frequencies like 1963 MHz at a mere 810 mV, dramatically dropping battery drain and thermal output compared to stock Nintendo behaviors. It also has a safety cap at 1.0V (
Code:
```
mariko_cpu_max_volt=1000
```
) to prevent random thermal spikes.
GPU Scaling Trim: The GPU curve is thoroughly optimized. The suite's dynamic voltage scaling accommodates the RAM clock perfectly to prevent bus signal collisions. Sustaining nearly 1 GHz (998 MHz) stable on just 635 mV puts this unit in the top tier of power efficiency.

4. My Custom Switch Overclock & Undervolt Tables

IMPORTANT NOTE ON VOLTAGES: These are extremely low, rare, and aggressive undervolt values for a Mariko chip. My silicon quality (and PMIC behavior) is so exceptionally pure that the hardware dynamically regulates power in ultra-fine steps of -10 mV / +5 mV depending on the real-time CPU load to ensure peak efficiency.

Real-world dynamic scaling example on my unit:

When running at 2397 MHz @ 955 mV, if the workload decreases, the system automatically drops the floor down by -10 mV to 945 mV to save power.
If a heavy computational load spike occurs, it instantly bumps the voltage up by +5 mV to 960 mV to preserve absolute stability.

CPU Frequencies & Voltages

Frequency (MHz)	Voltage (mV)
1021 MHz	590 mV
1122 MHz	595 mV
1224 MHz	620 mV
1326 MHz	645 mV
1428 MHz	675 mV
1581 MHz	690 mV
1683 MHz	720 mV
1785 MHz	750 mV
1887 MHz	780 mV
1963 MHz	810 mV
2091 MHz	855 mV
2193 MHz	890 mV
2295 MHz	940 mV
2397 MHz	955 mV

RAM Frequencies & Voltages by Profile

RAM Frequency	Profile 1600 MHz	Profile 1866 MHz	Profile 1996 MHz	Profile 2133/2166 MHz	Profile 2200 MHz	Profile 2233/2266 MHz	Profile 2333 MHz	Profile 2336/2400 MHz
460.8 MHz	500 mV	545 mV	560 mV	575 mV	580 mV	585 mV	600 mV	610 mV
537.6 MHz	500 mV	545 mV	560 mV	575 mV	580 mV	585 mV	600 mV	610 mV
614.4 MHz	520 mV	545 mV	560 mV	575 mV	580 mV	585 mV	600 mV	610 mV
691.2 MHz	540 mV	545 mV	560 mV	575 mV	580 mV	585 mV	600 mV	610 mV
768.0 MHz	565 mV	565 mV	565 mV	575 mV	580 mV	585 mV	600 mV	610 mV
844.8 MHz	585 mV	585 mV	585 mV	585 mV	585 mV	585 mV	600 mV	610 mV
921.6 MHz	615 mV	615 mV	615 mV	615 mV	615 mV	615 mV	615 mV	615 mV
998.0 MHz	635 mV	635 mV	635 mV	635 mV	635 mV	635 mV	635 mV	635 mV
1075.2 MHz	665 mV	665 mV	665 mV	665 mV	665 mV	665 mV	665 mV	665 mV
1152.0 MHz	690 mV	690 mV	690 mV	690 mV	690 mV	690 mV	690 mV	690 mV
1228.0 MHz	725 mV	725 mV	725 mV	725 mV	725 mV	725 mV	725 mV	725 mV

RAM VDD2 Voltage (

Code:

common_emc_mem_volt

): 1100 mV (Stock factory floor value—unbelievable efficiency for these timings!)
RAM VDDQ Voltage (

Code:

mariko_emc_vddq_volt

): 600 mV (Stock factory floor value)

5. Realized Benchmark Results & Power Draw

Membench GPU Write Silo Tyson: 30,892.3 MB/s [ 90.8% Efficiency ] with memory access latency sitting down at a swift 86.2 ns.
Furmark RAM Texture (Stress): Sustains a massive mixed real-world bandwidth output of 29,249 MB/s continuously for over 320 seconds (5+ mins).
Max Observed Thermals: CPU and GPU hold steady at 65.0 °C, while the RAM SoC reaches 55.0 °C under non-stop full synthetic stress.
Extreme Power Efficiency: Under full maximum synthetic load (1963 MHz CPU / 998 MHz GPU / 2133 MHz RAM), the total system power draw hovers strictly between -5.0W and -6.0W.
Horizon-OC Certified Safe Trim: Keeping the consumption around the -5W/-6W threshold confirms that this profile operates well within the official Horizon-OC safety guidelines. It delivers high-tier performance without forcing battery over-discharge or causing long-term component degradation.
Why 2133 MHz instead of pushing to 2700 MHz+? I intentionally chose 2133 MHz to achieve the absolute best performance-to-power ratio. While pushing the silicon to 2400 MHz or 2700 MHz is a fun benchmarking experiment, it is completely unnecessary for 99% of the Switch library and significantly degrades battery life while requiring dangerous voltages. By tightening the timings down to 4-4-8-5-6-5-7-6 at 2133 MHz, I achieved near-maximum real-world performance with balanced thermals and maximum power efficiency. Efficiency over raw speed is the real game-changer here!

Souldbminer · Jun 9, 2026

k0neer said:
[OLED Mariko] Ultimate Horizon-OC Profile: 90.8% Efficiency & Extreme Undervolt!
Custom Tuning for Samsung AB-MGCL Memory Chips

Hey everyone,

After many surgical stability tests, benchmark runs, and individual timing trial-and-errors using the Horizon-OC suite, I’ve managed to find the absolute "golden sweet spot" for my Nintendo Switch OLED.

I’m sharing my custom hybrid profile that successfully breaks the 90% efficiency barrier in Membench while operating at bone-stock factory memory voltages (1100mV/600mV). Thermals are incredible under continuous load (55 °C max during Furmark RAM stress test) and it is 100% stable after 30+ minutes running through Korok Forest in Zelda BotW.

1. Fixed Target Clocks & Efficiency Focus

CPU: 1963.5 MHz

GPU: 998.4 MHz / 921.6 MHz (Fully stable profiles)

RAM Clock: 2132 MHz (Actual configuration value: 2133000)

Handheld Mode Focus: This specific profile is heavily optimized and targeted for Handheld/Portable mode, focusing on extreme battery longevity and ice-cold thermals.

The Sweet Spot: Pushing the Overclock any higher is completely unnecessary for the vast majority of Nintendo Switch games. This 2133 MHz setup delivers maximum real-world smoothness without melting your silicon or destroying your daily battery life.

2. Custom Hybrid RAM Timings (The "Holy Grail")
Starting from the Common preset, I incrementally tightened each register one by one to find the absolute physical limit before encountering boot errors (finding the sweet spot between Common and Super Tight):

t1_trcd = 4 ➔ Hard physical limit (Setting to 5 causes boot error)

t2_trp = 4 ➔ Perfectly optimal and stable

t3_tras = 8 ➔ Super Tight (ST) value (Unlocks massive bandwidth)

t4_trrd = 5 ➔ The sweet spot (Setting to 6 introduces a minor MB/s penalty)

t5_trfc = 6 ➔ Stability ceiling (Setting to 7 triggers immediate boot error)

t6_trtw = 5 ➔ Crucial: Locked at 5 to completely bypass default ST black screens

t7_twtr = 7 ➔ Safe zone (Setting to 8 causes boot error)

t8_trefi = 6 ➔ Firmware/Suite hard-lock ceiling

Final Timing String:

Code:

4-4-8-5-6-5-7-6

3. CPU & GPU Overview (Aggressive Undervolt Analysis)

CPU Scaling Trim: My Mariko chip is running a very aggressive undervolt curve (

Code:

mariko_cpu_uv_high=10

). Thanks to excellent silicon lottery luck, my console sustains high boost frequencies like 1963 MHz at a mere 810 mV, dramatically dropping battery drain and thermal output compared to stock Nintendo behaviors. It also has a safety cap at 1.0V (

Code:

mariko_cpu_max_volt=1000

) to prevent random thermal spikes.

GPU Scaling Trim: The GPU curve is thoroughly optimized. The suite's dynamic voltage scaling accommodates the RAM clock perfectly to prevent bus signal collisions. Sustaining nearly 1 GHz (998 MHz) stable on just 635 mV puts this unit in the top tier of power efficiency.

4. My Custom Switch Overclock & Undervolt Tables

IMPORTANT NOTE ON VOLTAGES: These are extremely low, rare, and aggressive undervolt values for a Mariko chip. My silicon quality (and PMIC behavior) is so exceptionally pure that the hardware dynamically regulates power in ultra-fine steps of -10 mV / +5 mV depending on the real-time CPU load to ensure peak efficiency.

Real-world dynamic scaling example on my unit:

When running at 2397 MHz @ 955 mV, if the workload decreases, the system automatically drops the floor down by -10 mV to 945 mV to save power.

If a heavy computational load spike occurs, it instantly bumps the voltage up by +5 mV to 960 mV to preserve absolute stability.

CPU Frequencies & Voltages
[TABLE=full]
[TR]
[th]Frequency (MHz)[/th][th]Voltage (mV)[/th]
[/TR]
[TR]
[td]1021 MHz[/td][td]590 mV[/td]
[/TR]
[TR]
[td]1122 MHz[/td][td]595 mV[/td]
[/TR]
[TR]
[td]1224 MHz[/td][td]620 mV[/td]
[/TR]
[TR]
[td]1326 MHz[/td][td]645 mV[/td]
[/TR]
[TR]
[td]1428 MHz[/td][td]675 mV[/td]
[/TR]
[TR]
[td]1581 MHz[/td][td]690 mV[/td]
[/TR]
[TR]
[td]1683 MHz[/td][td]720 mV[/td]
[/TR]
[TR]
[td]1785 MHz[/td][td]750 mV[/td]
[/TR]
[TR]
[td]1887 MHz[/td][td]780 mV[/td]
[/TR]
[TR]
[td]1963 MHz[/td][td]810 mV[/td]
[/TR]
[TR]
[td]2091 MHz[/td][td]855 mV[/td]
[/TR]
[TR]
[td]2193 MHz[/td][td]890 mV[/td]
[/TR]
[TR]
[td]2295 MHz[/td][td]940 mV[/td]
[/TR]
[TR]
[td]2397 MHz[/td][td]955 mV[/td]
[/TR]
[/TABLE]

RAM Frequencies & Voltages by Profile
[TABLE=full]
[TR]
[th]RAM Frequency[/th][th]Profile 1600 MHz[/th][th]Profile 1866 MHz[/th][th]Profile 1996 MHz[/th][th]Profile 2133/2166 MHz[/th][th]Profile 2200 MHz[/th][th]Profile 2233/2266 MHz[/th][th]Profile 2333 MHz[/th][th]Profile 2336/2400 MHz[/th]
[/TR]
[TR]
[td]460.8 MHz[/td][td]500 mV[/td][td]545 mV[/td][td]560 mV[/td][td]575 mV[/td][td]580 mV[/td][td]585 mV[/td][td]600 mV[/td][td]610 mV[/td]
[/TR]
[TR]
[td]537.6 MHz[/td][td]500 mV[/td][td]545 mV[/td][td]560 mV[/td][td]575 mV[/td][td]580 mV[/td][td]585 mV[/td][td]600 mV[/td][td]610 mV[/td]
[/TR]
[TR]
[td]614.4 MHz[/td][td]520 mV[/td][td]545 mV[/td][td]560 mV[/td][td]575 mV[/td][td]580 mV[/td][td]585 mV[/td][td]600 mV[/td][td]610 mV[/td]
[/TR]
[TR]
[td]691.2 MHz[/td][td]540 mV[/td][td]545 mV[/td][td]560 mV[/td][td]575 mV[/td][td]580 mV[/td][td]585 mV[/td][td]600 mV[/td][td]610 mV[/td]
[/TR]
[TR]
[td]768.0 MHz[/td][td]565 mV[/td][td]565 mV[/td][td]565 mV[/td][td]575 mV[/td][td]580 mV[/td][td]585 mV[/td][td]600 mV[/td][td]610 mV[/td]
[/TR]
[TR]
[td]844.8 MHz[/td][td]585 mV[/td][td]585 mV[/td][td]585 mV[/td][td]585 mV[/td][td]585 mV[/td][td]585 mV[/td][td]600 mV[/td][td]610 mV[/td]
[/TR]
[TR]
[td]921.6 MHz[/td][td]615 mV[/td][td]615 mV[/td][td]615 mV[/td][td]615 mV[/td][td]615 mV[/td][td]615 mV[/td][td]615 mV[/td][td]615 mV[/td]
[/TR]
[TR]
[td]998.0 MHz[/td][td]635 mV[/td][td]635 mV[/td][td]635 mV[/td][td]635 mV[/td][td]635 mV[/td][td]635 mV[/td][td]635 mV[/td][td]635 mV[/td]
[/TR]
[TR]
[td]1075.2 MHz[/td][td]665 mV[/td][td]665 mV[/td][td]665 mV[/td][td]665 mV[/td][td]665 mV[/td][td]665 mV[/td][td]665 mV[/td][td]665 mV[/td]
[/TR]
[TR]
[td]1152.0 MHz[/td][td]690 mV[/td][td]690 mV[/td][td]690 mV[/td][td]690 mV[/td][td]690 mV[/td][td]690 mV[/td][td]690 mV[/td][td]690 mV[/td]
[/TR]
[TR]
[td]1228.0 MHz[/td][td]725 mV[/td][td]725 mV[/td][td]725 mV[/td][td]725 mV[/td][td]725 mV[/td][td]725 mV[/td][td]725 mV[/td][td]725 mV[/td]
[/TR]
[/TABLE]

RAM VDD2 Voltage (

Code:

common_emc_mem_volt

): 1100 mV (Stock factory floor value—unbelievable efficiency for these timings!)

RAM VDDQ Voltage (

Code:

mariko_emc_vddq_volt

): 600 mV (Stock factory floor value)

5. Realized Benchmark Results

Membench GPU Write Silo Tyson: 30,892.3 MB/s [ 90.8% Efficiency ] with memory access latency sitting down at a swift 86.2 ns.

Furmark RAM Texture (Stress): Sustains a massive mixed real-world bandwidth output of 29,249 MB/s continuously for over 320 seconds (5+ mins).

Max Observed Thermals: GPU 55.5 °C / CPU 54.0 °C / RAM 46.7 °C under non-stop full synthetic stress.

[*]Why 2133 MHz instead of pushing to 2700 MHz+? I intentionally chose 2133 MHz to achieve the absolute best performance-to-power ratio. While pushing the silicon to 2400 MHz or 2700 MHz is a fun benchmarking experiment, it is completely unnecessary for 99% of the Switch library and significantly degrades battery life while requiring dangerous voltages. By tightening the timings down to 4-4-8-5-6-5-7-6 at 2133 MHz, I achieved near-maximum real-world performance with freezing cold thermals (46 °C RAM) and maximum power efficiency. Efficiency over raw speed is the real game-changer here!

AI. Use as high ram as you can, mem bottleneck persists. 1175mV voltage is optimal. Timing Harmony may not be optimal with this setup. Also, TOTK is generally better for stability test

k0neer · Jun 9, 2026

Souldbminer said:
AI. Use as high ram as you can, mem bottleneck persists. 1175mV voltage is optimal. Timing Harmony may not be optimal with this setup. Also, TOTK is generally better for stability test

Thanks for the feedback! I totally agree that the Switch is fundamentally memory-bottlenecked and TotK is the ultimate stability test (I'm currently benchmarking it).

However, the main goal of this specific profile is extreme power efficiency for pure Handheld play. While 1175mV and 2400MHz+ yield higher raw numbers, they draw significantly more wattage and increase thermals. On my specific 'Golden' silicon, 1100mV/600mV has proven to be perfectly stable so far. In fact, I've already stress-tested this setup in Zelda BotW running at a locked 720p / 60 FPS in Korok Forest with heavy combat explosions in the most demanding areas, resulting in zero stutters, artifacts, or corrupted saves. This allows me to achieve the absolute best balance between battery life and performance.

Post automatically merged: Jun 9, 2026

Update on Samsung AB-MGCL (OLED Mariko) - 2132 MHz @ 1100mV VDD2 Rock Solid Stability

Hey @Souldbminer, first of all, thanks for the amazing insight regarding the memory bottleneck on Mariko and your work on Horizon OC. You are 100% right about TOTK being the ultimate stability test. I took your advice and put this profile through the ultimate stress test.

While I acknowledge that 1175mV is the optimal sweet spot for absolute stability and higher frequencies on most chips, I wanted to push the absolute floor of my specific Silicon Lottery winner unit.

Before committing, I actually did a side-by-side comparison between your recommended auto-configurations and my tight manual timings. Interestingly, when testing with both the Common and ST (Standard) profiles at 1100mV, the performance was noticeably more unstable—the FPS bar was fluctuating constantly and introduced heavy micro-stuttering. It seems that under the absolute stock voltage floor, the relaxed auto timings loosen the memory controller sync too much for TOTK's aggressive asset streaming.

To fully validate your advice, I also tested your recommended 1175mV** ceiling combined with the Common/ST presets. Surprisingly, even with the extra voltage headroom, the auto timings still introduced the same frame-pacing fluctuations during the dive. My custom tight hybrid suite out-performed the standard presets at both 1100mV and 1175mV, flattening out the FPS bar completely.

However, reverting back to my custom tight hybrid suite completely stabilized the frame pacing. Here is the final telemetry:

Component Target Frequency Voltage / Config
CPU 1963.5 MHz 810 mV fixed
GPU 998.4 MHz / 921.6 MHz 635 mV / 615 mV
RAM 2132 MHz VDD2: 1100 mV | VDDQ: 600 mV
Timings Suite 4-4-8-5-6-5-7-6 t6=5 (Bypasses boot loops/black screens)

TOTK Stress Test Benchmarks (Handheld Mode / Official Charger Connected):

Target Profile: 60 FPS Fixed (FPSLocker) + Dynamic Resolution maxed out at 896p.

Prologue (Caves & Miasma): Locked 60 FPS at native 896p. Heavy camera panning showed zero artifacting, zero micro-stuttering, and zero graphical glitches.

The Great Sky Island Dive (Asset Streaming Test): Drops to 50-55 FPS during free fall due to CPU bottlenecking. The engine dynamically scaled resolution down to 720p as expected to protect frame pacing, but the memory controller handled the aggressive texture streaming flawlessly at 1100mV.

Thermals under max stress: CPU/GPU peaked at 64°C | RAM SoC stayed cool at 55°C.

Stability: 0 crashes, 0 system freezes, no error codes (2162-0002).

It seems that keeping the RAM around 55°C preserves enough electrical efficiency to prevent data corruption even at the stock 1100mV floor. The custom t6=5 timing bypass is definitely doing some heavy lifting keeping the memory controller happy and steady compared to the Common/ST presets.

Let me know what you think! I might step up to 2400 MHz or 2766 MHz later using your recommended 1175mV to break that 50 FPS dive bottleneck, but for pure power-efficiency at 2132 MHz, this golden sample is holding up like a champ.

RumbelBoss · 2026-06-18T04:02:52+0100

KIP patches not working on Atmosphere 1.11.2 sadly, I got a black screen everytime I try to boot onto emuMMC. Safe mode boot works just fine.
Had to go back to 1.11.1 in order to boot correctly with my undervolt profile.

aidia_dayo · 2026-06-18T05:09:40+0100

RumbelBoss said:
KIP patches not working on Atmosphere 1.11.2 sadly, I got a black screen everytime I try to boot onto emuMMC. Safe mode boot works just fine.
Had to go back to 1.11.1 in order to boot correctly with my undervolt profile.

Did you update to Horizon OC 2.4.2 yet? I'm on fw22.5.0 with Atmosphère 1.11.2 and it's been working fine for me on both sysMMC and emuMMC.

IFunkymonkey · 2026-06-18T09:24:31+0100

Quick question: Is all that even worth the hustle for an v1 NS and BotW with the goal to run it at 60fps?

Component	Target Frequency	Voltage / Config
CPU	1963.5 MHz	810 mV fixed
GPU	998.4 MHz / 921.6 MHz	635 mV / 615 mV
RAM	2132 MHz	VDD2: 1100 mV \| VDDQ: 600 mV
Timings Suite	4-4-8-5-6-5-7-6	t6=5 (Bypasses boot loops/black screens)

Horizon OC 2.0 - Most powerful overclock tool. CPU 2397MHz, GPU 1228MHz, RAM 3200+MHz

Well-Known Member

Well-Known Member

Member

Active Member

Well-Known Member

Member

Active Member

Active Member

Active Member

Member

New Member

Active Member

Similar threads

Popular threads in this forum