6851 MHz AMD Ryzen 9000 with Liquid Nitrogen
We beat a 7.1 GHZ Core i9 with an AMD Ryzen 9000 to take the record in the OCCT AVX Single Thread benchmark.
Table of Contents
Introduction
Today, we’re having a shot at some overclocking records with the Ryzen 7 9700X “Granite Ridge” processor and liquid nitrogen cooling. I will walk you through my process and show some of the benchmark results I was able to achieve.
Before making this video, I made a SkatterBencher guide for the Ryzen 7 9700X in which I laid out how to overclock the processor in five unique overclocking strategies. I followed that up with a special episode in which I took that configuration and eliminated all thermal constraints to maximize the system’s performance.
Today’s video is a little different because we’re not aiming for stable overclocks. Instead, we’re aiming to achieve the #1 spot in benchmark leaderboards.
Ryzen 9000 LN2 Hardware Setup
After wrapping up the SkatterBencher guide, I pinged Jon from ElmorLabs and asked if he had space in his office to try liquid nitrogen. Not only did he have a space, but he already had a system prepared for Ryzen 9000 liquid nitrogen overclocking. In fact, he used that very system to break the 3DMark CPU Profile 16 Thread record with the Ryzen 9 9950X.
The system we used is nearly identical to the one from SkatterBencher #78 and #78X with here and there a little change.
Item | SKU | Price (USD) |
CPU | AMD Ryzen 7 9700X | 359 |
Motherboard | ASUS ROG Crosshair X670E Gene | 600 |
CPU Cooling | ElmorLabs Volcano LN2 Container ElmorLabs HOT300 Heater Controller ElmorLabs HOT300 Heater CPU Backplate | 260 20 30 |
Thermal Paste | Thermalright TFX Extreme | 25 |
Memory | G.SKILL Trident Z5 DDR5-6400 32GB | 110 |
Power Supply | Enermax PlatiGemini 1200W | 250 |
Graphics Card | ASUS ROG Strix RTX 2080 TI | 490 |
Storage | AGI 512GB NVMe M.2 Gen3 | 75 |
For example, I’m using the Thermalright TFX Extreme thermal paste which is suitable for extreme cooling conditions.
I also use the ElmorLabs Volcano LN2 container. That’s the same pot used for the 9 GHz CPU Frequency world record, so I figured if it could do that, it’ll suit the 9700X too.
Lastly, I use an ElmorLabs HOT300 Heater Controller and Heater Backplate. This is a nifty device that heats up the back of the motherboard. This helps prevent the cold of the nitrogen from spreading across the PCB, which could create patches of condensation.
Objectives
The main objective of any “World’s Fastest” video is to have a legitimate claim to the title. So, wanted to achieve at least one top spot in a global benchmark leaderboard. Apart from that, I also wanted to see the maximum CPU frequency I could achieve with Precision Boost.
Ryzen 9000 LN2 BIOS Configuration
I used the BIOS configuration from SkatterBencher #78X as my baseline settings. Let me quickly go through the settings and highlight the key elements.
First, we enable EXPO to improve the memory performance
Then we use asynchronous ECLK to increase the Precision Boost frequency. By setting it to 110 MHz, it means that we increase the frequency for every VFT point by 10%. So, 4 GHz becomes 4.4 GHz, 5 GHz becomes 5.5 GHz, and so on.
Next, based on the performance data we gathered in SkatterBencher #78, I rely on an ASUS Memory Preset to finetune the memory sub-timing. From previous testing, sub-timing tweaking does almost as much as increasing the memory frequency.
Now we switch to the AMD Overclocking submenu and manually set the FCLK and UCLK frequency to 2200 and 3200 MHz respectively.
Since we’re relying on the Precision Boost technology to determine the CPU operating frequency, most of the tuning will be done in the Precision Boost Overdrive submenu.
First, we unleash the power limits by setting the PBO Limits to Motherboard. That effectively removes any motherboard power or current limit.
Then, we set Scalar to 10X which increases the voltage limit in all-core workloads from 1.35V to 1.375V.
Next, we set the Fmax Boost Override to +200. On the 9700X, that means the new Fmax is 5750 MHz. But with our ECLK of 110 MHz, the actual new Fmax is 6325 MHz (5750 x 1.1).
Now, the most important part of extreme overclocking with Precision Boost is voltage tuning.
To increase the voltage, we use a positive Curve Optimizer across all cores. That shifts the entire V/F curve upwards and forces the CPU to run about 100 mV higher for every VFT point. In daily overclocks, that would significantly reduce the operating frequency. But with ECLK and much lower temperatures, it actually gives more overclocking headroom.
Then, I also used Curve Shaper to address minor stability issues. I found that my overclock was stable at sub-zero temperatures but not at ambient. By setting the medium and high temperature to a positive Curve Shaper, we increase the voltage at ambient temperatures and thus make it more stable.
Lastly, we made some additional small changes to the SoC Uncore configuration, including enabling OC Mode and setting the voltage to 1.3V.
Overall, at -50 degrees Celsius, this configuration yielded a geomean benchmark performance improvement of +16.24% over stock.
Ryzen 9000 LN2 Challenges
Apart from the BIOS configuration, there were two additional challenges to using extreme cooling with the Ryzen 9000.
First of all, we had to boot the system at positive temperatures. Otherwise, the temperature sensor would give trouble, and the maximum frequency was locked to less than 5 GHz. However, lowering the operating temperature in the operating system was no problem.
Second, the temperature delta between the CPU core and the bottom of the LN2 container was surprisingly large. When running the OCCT SSE Stress Test at 6 GHz, the CPU core temperature was -25 degrees Celsius but the LN2 container temperature was -80 degrees Celsius.
Ryzen 9000 LN2 Overclocking Results
OCCT Benchmark
The headline result is of course the OCCT AVX Single Thread record. The AVX performance of the Ryzen 9000 CPUs is something that really stood out in the hardware reviews. I highly recommend Alex’s article at Numberworld on the topic. I found the Ryzen 9000 CPUs perform exceptionally well in the OCCT AVX benchmark and I almost took the number one spot with just AIO cooling.
The number one spot is also actually mine as I achieved it during the Computex 2024 activities with ASUS. You can check out my 7.1 GHz OCCT stability certificate content for more information.
With liquid nitrogen, it was relatively easy to break the record set by the Core i9-14900KF. My final score was 269.35 points. That same system also got close to beating the SSE Single Thread record, however, we’re just missing a little more frequency to make that happen.
- OCCT Benchmark Leaderboard: https://www.ocbase.com/benchmark
- 127.79 SSE 1T: https://www.ocbase.com/benchmark/cpu/66c42b30bc593f4530427d4c
- 251.93 AVX 1T (AIO): https://www.ocbase.com/benchmark/cpu/66b4db14d97be318684310d4
- 269.35 AVX 1T (LN2): https://www.ocbase.com/benchmark/cpu/66c42b30bc593f4530427d4e
CPU-Z Benchmark & Validation
Next up is CPU-Z, where I tried both the regular version and the XOC version.
The main difference between the two versions is that the regular version runs a heavy workload while validating the CPU frequency, whereas the XOC version just captures the frequency. I did this with the same configuration from the SkatterBencher guide at 6325 MHz. As I demonstrated in SkatterBencher #78X, I could reach 6325 MHz for all cores at -50 degrees Celsius.
That also gave a nice score of 1003 1T and 10805 nT in the CPU-Z benchmark. It’s cool to see that puts it past a Core i9-14900KS in a 1T workload and past the 16-core Ryzen 9 3950X in a 32-thread workload.
Switching to the XOC version, I could also validate a frequency of 6.8 GHz by disabling SMT and pushing up the ECLK frequency.
- Ryzen 7 9700X CPU-Z Leaderboard: http://valid.x86.fr/top-cpu/414d442052797a656e203720393730305820382d436f72652050726f636573736f72
- Ryzen 7 9700X 6325 MHz + benchmark: https://valid.x86.fr/nxn7lq
- Ryzen 7 9700X 6851 MHz: http://valid.x86.fr/a4y0c2
- Ryzen 7 9700X 1T compare http://valid.x86.fr/bench/nxn7lq/1
- Ryzen 7 9700X 32T compare: http://valid.x86.fr/bench/nxn7lq/32
Geekbench 6 Benchmark
Then, I also ran Geekbench 6, which some Twitter users discovered before I published this guide.
This result also put me at the top of the leaderboard for Single-Core and Multi-Core.
- Geekbench 6 score: https://browser.geekbench.com/v6/cpu/7410926
- Single-Core Leaderboard: https://browser.geekbench.com/v6/cpu/search?dir=desc&q=Ryzen+7+9700X&sort=score
- Multi-Core Leaderboard: https://browser.geekbench.com/v6/cpu/search?dir=desc&q=Ryzen+7+9700X&sort=multicore_score
Of course, I also ran a couple of games with an LN2-cooled system for the SkatterBencher 78X guide. No records here, but I thought it was a cool sight.
Ryzen 9000 LN2 Conclusive Thoughts
Overall, this was a pretty cool project to undertake. It’s unusual to stick with Precision Boost for extreme overclocking because it’s a very dynamic clocking technology. Extreme overclockers prefer predictability when finetuning for maximum performance. But after a couple of days of experimenting, I now feel more confident to use Precision Boost even with extreme cooling.
The OCCT 1T AVX performance is of course a stand-out result for the Ryzen 9000 as it highlights its strength in AVX workloads. The other results are pretty sweet too, though I wouldn’t say they’re remarkable. Furthermore, there are much more talented overclockers at HWBOT pushing the Ryzen 9000 to even higher frequencies.
Anyway, that’s all for today. I want to thank ElmorLabs for the hospitality, the Patreon supporters for supporting my work, and you for watching the video. If you have any questions or comments, please drop them in the comment section below.
See you next time!