It's funny you call it 'daily work'. My Ice Tractor runs this custom kernel 24/7. While you're disabling your OC to stay stable, my system is built into a Concrete Wall of efficiency. I don't need 32 threads to hide behind; my optimization handles the work better than your raw, unoptimized silicon ever will. Keep your i9 at stock, the Tractor will still be faster core-for-core!"

You say you don't need that 'slight OC' for daily work, but the Ice Tractor is built to be a beast 24/7. The reason my i5 responded like a lightning bolt is the custom-tailored kernel I’ve stitched together for this specific hardware. While your generic OS is busy managing 32 threads with massive overhead and bloated schedulers, my kernel is stripped to the bone. It has:

• Custom Interrupt Handling: No useless background noise.

• Optimized Scheduler: Zero 'traffic' between cores.

• Ultra-low Latency: That 0.71ms didn't happen by accident; it's the result of software soul meeting overclocked silicon.

  You're throwing 16 cores at a problem that I solved with optimization. You might have the bigger engine, but the Tractor has the better transmission. The Concrete Wall doesn't need to 'disable' anything to stay stable and fast

❯ sysbench cpu run --threads=32
sysbench 1.0.20 (using system LuaJIT 2.1.1720049189)

Running the test with following options:
Number of threads: 32
Initializing random number generator from current time


Prime numbers limit: 10000

Initializing worker threads...

Threads started!

CPU speed:
    events per second: 38992.88

General statistics:
    total time:                          10.0008s
    total number of events:              390002

Latency (ms):
         min:                                    0.69
         avg:                                    0.82
         max:                                    4.82
         95th percentile:                        0.83
         sum:                               319932.69

Threads fairness:
    events (avg/stddev):           12187.5625/44.43
    execution time (avg/stddev):   9.9979/0.00

gave it a little bit more cause why not....

It’s clear you’re struggling with the 'Mesh' latency of that i9! 🚜💨

You mentioned disabling your OC for 'daily work', but that’s the difference: The Ice Tractor is a Concrete Wall of 24/7 stability. My custom-built kernel doesn't just 'run' the hardware; it commands it.

That 0.71ms latency comes from a specific combination you can’t buy:

1. Kernel-level Memory Optimization: I've tuned the memory controller behavior and page handling to eliminate jitter.

2. Ultra-Tight RAM Timings: My Ivy Bridge 'Ring Bus' is feeding data at a speed your 16-core Mesh architecture can only dream of.

3. Zero-Bloat Execution: While your OS is busy 'working', my kernel is purely focused on destroying benchmarks.

You have 32 threads, but I have the efficiency crown. Real overclocking isn't about how much money you spend on 16 cores, it's about how much performance you can extract from every single transistor.

Stick to your 'daily work' settings, the Tractor will keep ruling the Extreme Edition! 🧱👑"

Impressive brute force, exaberries, but the math is still on my side!

Hey exaberries, careful with that Vcore! 🌡️ 🔥

Pushing a 16-core i9 to match the Ice Tractor’s per-core speed is a dangerous game. My i5 stays cool thanks to the custom kernel efficiency and the Concrete Wall airflow. I’d hate to see your expensive rig melt just to catch up with a 2012 CPU!

Optimization is safer than raw voltage. Keep it stable, if you can! 😉"*

To be honest, exaberries, if I had your rig for just one day, I’d turn that i9 into something as responsive as a fighter jet’s cockpit.

The real magic isn't in pushing a button for a 10-second screenshot and then disabling it for 'daily work'. Real performance is building a system that is lethal and rock-solid 24/7.

With my custom kernel and your 16 cores, that 0.82ms latency would drop to levels you can't even imagine. I don't build systems for 'show'; I build them for 'go'. The Ice Tractor is a daily driver that crushes benchmarks because it’s engineered that way from the source code up.

You have the hardware of a titan, but it’s still waiting for a pilot who knows how to tune the engine. Until then, the Concrete Wall stays ahead in efficiency

I see you pushed it to 39k eps, nice jump! 🚜💨 But even with that extra juice, your latency is still stuck at 0.82ms.

My 0.71ms on the Ice Tractor shows that my custom kernel is still delivering a level of responsiveness that your i9 can't reach, even with 16 cores screaming. It's not just about the top speed; it's about how fast the engine responds to the command.

The Concrete Wall is built for precision, not just brute force!

It’s hilarious to see you trying to fight software engineering with just raw overclocking. You can turn the voltage knob all day, but you can't 'overclock' away the bloat of a standard kernel. My 0.71ms latency is the result of surgical optimization, while your 0.82ms (even with OC) shows your system is still struggling to keep up with the commands. You're building a faster hammer, but I've built a faster brain for the machine. The Ice Tractor isn't just about MHz—it's about zero-waste execution. Keep pushing that silicon, but remember: you bought your speed, I engineered mine

8 minutes ago, TheFyxxxer said:

Hey exaberries, careful with that Vcore! 🌡️ 🔥

Pushing a 16-core i9 to match the Ice Tractor’s per-core speed is a dangerous game. My i5 stays cool thanks to the custom kernel efficiency and the Concrete Wall airflow. I’d hate to see your expensive rig melt just to catch up with a 2012 CPU!

Optimization is safer than raw voltage. Keep it stable, if you can! 😉"*

Consiering I'm running less voltage than you and lower core frequency, I really dont have to worry and its more than stable enough for daily, I just don't need the extra performance so why bother... That's why I just run a daily mem oc and stock cpu.... It wasn't expensive either... board for $70 cad, cpu was free since it has a missing memory channel, and $200 cad for 128GB of ram a couple years ago...

I don't see how having 16 cores is brute force... if I ran only 4 cores I'd get better results than you, it just means I have more multicore performance. Nothing your doing is special to ivy bridge, you can do it on lots of other more modern platforms that you'd get total higher performance with. Now for dj-ing that's probably not necessary, and that's fine. Same way I need a lot of multicore performance your 3570k would never be able to do. Which is also fine, we do different things. Maybe if I am bored enough I'll grab another platform but right now... or if I fix the r6a that I have my 7740x in... but otherwise I think I'll leave you to it for now...

or to put it your way....

Look—I get what you’re saying—but here’s the reality of my setup and why I’m not even remotely concerned 😌—

I’m running lower voltage and lower core frequency than you—so stability isn’t some fragile, edge-of-collapse situation—it’s basically a non-issue. This is daily-driver stable 🧱—no drama, no thermal panic, no “hope it survives the benchmark” energy 🔥❌.

And honestly? I don’t even need the extra performance—so pushing further just feels… unnecessary 🤷‍♂️—like revving an engine in neutral.

That’s exactly why I stick to a daily memory overclock + stock CPU config—it gives me everything I actually use, without wasting power, time, or silicon lifespan ⚡🧠.

---

And let’s talk cost for a second—because this wasn’t some absurd investment 💸—

—$70 CAD for the board
—CPU was free (yeah, missing memory channel, still works perfectly for my needs 😄)
—$200 CAD for 128GB RAM a couple years ago

So this isn’t “throw money at the problem” energy—it’s just a well-optimized, practical setup 🛠️✨

---

Now about the “brute force” thing—because that part doesn’t really land the way you think it does—

Having 16 cores isn’t brute force—it just means I have more multicore throughput available when I actually need it 🧵🚀

If I really wanted to play the same game—drop down to 4 cores, tune for latency, chase single-thread—I could absolutely do that too—and yeah, I’d likely get better results than you in that specific niche 🎯

But that’s the point—we’re optimizing for completely different workloads.

---

And that’s the key distinction you’re kind of glossing over—

What you’re doing? It’s not some Ivy Bridge-exclusive magic 🧙‍♂️
You can replicate that approach on newer platforms—and get higher total performance across the board 📈

But for your use case—like DJ-ing 🎧—you don’t need that extra scalability—and that’s totally fine 👍

---

Same goes the other way—

The kind of multicore-heavy work I do?
A 3570K just wouldn’t keep up—and that’s not a knock, it’s just reality based on workload demands 🧠⚙️

Different tools for different jobs—simple as that.

---

So yeah—maybe I’ll mess around with another platform someday if I get bored 😄—

—or fix that R6A with the 7740X sitting around 🧩—

—but for now?

I’m good where I’m at 😌—

You do your ultra-optimized latency thing 🧱👑
I’ll keep my balanced, multicore-friendly setup humming along 🚀

No need to turn it into a “who’s more hardcore” contest—we’re just solving different problems 👍

I appreciate the 'peace treaty', exaberries, but let's call it what it is: The Ice Tractor forced your i9 to work for its life! 

You're right, different tools for different jobs. But there's a difference between using a tool and forging one.

• The Memory Channel: Running a high-end platform with a missing channel isn't 'optimized'—it's a bottleneck. That's exactly why your latency can't touch my 0.71ms. I’ll take a perfectly tuned 4-cylinder over a broken V16 any day.

• Stability: You say you don't need the extra performance, but the Concrete Wall is about capability. My system is Prime95 5K stable not because I 'need' it, but because as an engineer, I demand it.

• The 'Could Have': Saying you could beat my single-thread or latency is easy; actually doing it with a custom kernel and a decade-old CPU is the real challenge.

Enjoy your multicore throughput! I’ll keep my 'ultra-optimized' surgical response. It’s not about who’s more hardcore, it’s about showing that code and craft can beat raw, broken silicon every time.

See you on the next benchmark, maybe when that R6A is fixed!

But since you think there’s nothing 'special' about this Ivy Bridge setup, I’ll leave you with one final piece of advice if you ever want that i9 to truly fly:

Stop playing with BIOS sliders and start working on your Kernel.

That’s where the real performance is hidden. If you want to close that latency gap and make those 16 cores actually breathe, you have to go deeper than standard OS settings. I won't say more—the Ice Tractor keeps its best secrets under the hood, buried in the code.

Good luck with your multicore work, and let me know if you ever manage to see 0.71ms on that i9! 

🚀 The "Ice Tractor" Challenge: Pushing Silicon to NASA Standards 🚜💎

"After fine-tuning the stability of my i5-3570K (The Ice Tractor), I’ve realized the real frontier isn't just about CPU generations—it's about the bridge between raw silicon and surgical optimization.

I’ve managed to hit 0.71ms system latency with Prime95 5K FFT stability 24/7. I know many of you are sitting on absolute monsters—modern i9s and Ryzens that are basically spacecraft compared to my 2012 hardware. 🛰️✨

My question to the community is this:
How far can you actually push that modern silicon if you apply deep Kernel optimization and precision timer tuning?

I’m not looking for brute-force 32-thread benchmarks. I want to see your processors screaming like NASA-grade systems! 🚀🌌

I want to see if, with the right 'stitch' in the OS, your modern platforms can break the 0.70ms wall and show us what true responsiveness looks like.

The "Ice Tractor" Benchmarks:

1. Latency: Who can dive below the factory standards by tailoring the OS? 📉

2. Efficiency: How many eps per thread can you extract while keeping voltages conservative? ⚡

3. Stability: No 'suicide runs.' We’re looking for the Concrete Wall (Prime95 stable). 🧱

My 3570K is here as the baseline for what's possible with older architecture and extreme tuning. Let’s prove that overclocking isn't just about moving multipliers—it’s about engineering control. 🛠️🧠

Who’s next? Let’s see what happens when elite hardware meets custom-tailored software! 🚜🔥"

On 3/30/2026 at 1:28 PM, Paul7347 said:

The build is old-school, I like the style it looks cool. It's an affordable base that you tuned, that's great and exactly what overclocking is at the root.

But it's got nothing on modern systems, let alone "a surgical tool that humiliates your modern silicon in pure responsiveness.". At least up until Rocket Lake, I do agree we get more and more interconnects everywhere that tend to hurt latency but that's where we're heading.

Look at the deterministic linearity of my lat_mem_rd report. From L1 to RAM (128MB), the latency gap is minimal and jitter-free. While your modern systems have lower average latencies but with huge spikes that cause signal instability, the Ice Tractor maintains pinpoint coherence. Audio 'crackling'? With such flat read stability, it's technically impossible.

Winter is officially dead, delta temperature room +9°. The Ice Tractor is breathing in the air of spring. It's time to replace the last two glorious Antec TriCool 120mm fans (13 years of 100% work lol). the last piece of the puzzle: the final airflow setup. If I were to reset the Ice Tractor to factory settings next winter, I would risk the dew point with this final airflow. In the second image (February) I was taking a big risk while the system was running at 3.40 GHz. I was optimizing the kernel and testing the first airflow setup to get more useful information to run an OC that also went hand in hand with the cooling system.disassemble and reassemble. disassemble and reassemble. reassemble and disassemble LOL

Project "Aria v9 Core" in Search of Devil's Canyon
Now that I've reached my logical limit with Ivy Bridge, I'm browsing flea markets in search of Devil's Canyon and a good motherboard on the web.
a déjà-vu: this specific combo. Cosmos SE + MSI Z97 PC Mate (or similar blue series) + Enermax Modu82+ — was a classic of the "medium-high" builds offered by many physical stores or small assemblers between 2014 and 2015. Considering the components and the period (2014-2015), such a setup in a physical store could have cost between €1,100 and €1,300 (VAT included).WTF!!!
then I concentrate on looking inside the case and tons of curses start to come out...this was the precursor time to the "global rgb dumbing down"
Inside this nativity scene set up during the tequila fair...there's a devil's canyon wtf!

Edited April 15Apr 15 by TheFyxxxer

The thermal behavior has been redesigned for maximum efficiency. The system now operates within two distinct, rock-solid thermal zones:

• Icy Idle (14°C - 20°C): At rest, the CPU package drops to near-ambient temperatures. This is a massive improvement over the previous setup which sat at a constant 50°C), achieved by optimizing the Liquorix power-cycle response without sacrificing wake-up latency.

• Thermal Ceiling (50°C Max): Under heavy synthetic loads or real-time latency testing (Sysbench), the CPU hits a precise "Thermal Wall" at 50°C. There is zero thermal drift beyond this point, confirming a perfect balance between Vcore and the 1380+ RPM fan curve.

  The "Concrete Wall" philosophy is verified by the minimal delta between idle and multitasking environments.

• Daily Multitasking Bench (YouTube 1080p + Cinnamon Desktop):

• Avg Latency: 0.77 ms witch Machine Head ROFL

• 95th Percentile: 0.88 ms

• Performance Spike: Instantaneous 4.4 GHz ramp-up upon task detection.

  Scientific data confirms it. My ice tractor isn't just fast, it's instantaneous. Based on my Ftrace report, I can now calculate the exact speed of my “Slingshot” (the hardware transition from idle to 4.4 GHz).

• Idle State 6828.653898 The CPU is resting at 14°C.

• The Trigger 6828.653913 Wake-up signal! State 4294967295 (Exit Idle)

• Total Delta 15 Microseconds The Slingshot Speed.

  This is the physical time (and limit i think) the MSI Z77 MPOWER VRMs take to stabilize voltage and the Ivy Bridge silicon to hit 4.4 GHz.
  The system exhibits "zero-thought" reaction times, maintaining sub-millisecond averages even while handling modern web-stack background noise.

  To secure the interrupt chain, the NVIDIA USB-C "phantom" controller has been surgically removed from the kernel space: Blacklisted Modules: ucsi_ccg, i2c_ccgx_ucsi, i2c_nvidia_gpu.

  Result: Elimination of I2C timeout errors (-110) and kernel "polling" noise. The CPU is now free to focus entirely on the primary instruction set, further solidifyin the 0.77ms stability. The Ice Tractor is no longer just a benchmark system; it is a stable computer for everyday use that achieves the same ambitious goals as before.

  The backbone of this project remains the legendary MSI Z77 MPOWER. Its superior power delivery (VRM stability) allows the i5-3570K to maintain these surgical frequencies without voltage ripples. This setup proves that high-end 2012 engineering, when combined with modern kernel tuning, still sets the standard for system transparency and raw responsiveness.

  From an extreme machine with impressive responsiveness, it's now a demon that I keep at home... who knows when it's the right time to unleash itself! The system isn't "always on idle" like the old setup, but it has a sort of predictive instinct. It rests in ice at 15°C, but it has the reflexes to unleash the 4.4 GHz of the Z77/Ivy Bridge duo in the millisecond you need it.

  The Concrete Wall philosophy is this: It's not a maze I get lost in, it's a maze I design as I walk through it. Every kernel option I touch is one less weight the Ice Tractor has to carry. Every wall I knock down hides a secret room. Lost... I'm screwed!

Edited April 15Apr 15 by TheFyxxxer
forgot Machine Head lol

When I leave the planet "Kernel" and return to Earth... I connect a particular toy to the Ice Tractor: another passion that travels with me together with the world of PC. This gives me the opportunity to see as an "outside person" the real advantages and above all the workload to which the build is subjected.
In my moments of pure sonic mess, the Ice Tractor has to emulate Virtual DJ, manage 2 to 4 tracks simultaneously and, at the same time, have the synchronized signal for the 4 Roland drum machines (808, 909, 707, 606).
But the biggest and most infamous job concerns the management of STEMS: the function that separates the voice from the music in real time using Artificial Intelligence, they are one of the most demanding operations that a DJ software can do.
Thanks to Roland's ACB (Analog Circuit Behavior) technology, the DJ-505 doesn't use simple recorded files (samples), but recreates the behavior of the original circuits in real time. This requires a precision in the digital signal that only an ultra-optimized system like Ice Tractor can guarantee without micro-delays or quality losses.
The kick and snare hits fall exactly where they're supposed to, with pinpoint precision. The result is a much more aggressive, precise, and professional sound. The sound comes out instantly. There's none of that tiny delay (input latency) that usually forces you to "anticipate" the movement with your fingers. I can play the drum machine as if it were a real drum kit!!!
The Roland DJ-505 comes with ASIO drivers designed for Windows. I use Wine, which acts as a real-time translator between the software (Virtual DJ) and the hardware (the console). Normally, adding a "translator" (emulation) increases the delay. If an ASIO driver on Windows responds in 5 milliseconds, on emulated Linux it could take 15. But by optimizing the Linux kernel, I eliminated the underlying delays, managing to run the Roland drivers at a speed that probably exceeds that of a stock Windows PC.
It must manage:
The audio from the 4 Decks coming out of the PC to the Roland.
The MIDI signals from the drum machine returning to the PC.
The flow of AI Stems, which must be separated and sent back to the console faders.
Streaming is the final straw: Encoding a video for TikTok, YouTube, Twitch... in real time requires a huge data write speed.
The 5k Prime95 test compared to this extreme use on a 2012 architecture... is like drinking a glass of water for the Ice Tractor LOL

try 2 brake the 0.71ms deterministic barrier.

The main bottleneck was identified in the tRFC (stuck at 300ns via XMP) and the high stress on the IMC caused by 8 ranks of memory.

• V-Dimm: Increased to 1.65V (Patriot "Black Mamba" binning).

• VCCSA/VTT: Stabilized at 1.10V to support 32GB @ 1866MHz.

• Timings: Switched from 9-9-9-24 (XMP) to 8-10-10-27 (Manual)

• tRFC Aggression: Dropped from 300ns to 220ns

  Achieving <12.3ns with 4 modules populated on Z77 is a testament to the MPOWER’s PCB quality and IMC efficiency.

  Next Steps: Testing tREFI 32767 to further minimize refresh cycles and aim for a "Flat 0.71ms" even on max peaks.

  Patriot Viper Black Mamba: for a commercial RAM kit...they perform very well. Verified stability with memtester (28GB allocation) – Solid Bits passed with tighter secondary timings.

BIOS Main Page: Confirms i5-3570K stable at 4.40GHz (44x ratio) with 32GB of DDR3-1866MHz RAM. All energy savings (EIST) disabled for benchmark stability.

Voltages Page: Manual tuning for IMC stability with 32GB (4 modules). DRAM Voltage set to 1.65V (1.658V actual), VCCSA (System Agent) at 1.105V and VTT at 1.11V to ensure stability for 8 ranks of memory at 1866MHz 1T.

Advanced DRAM Timings: Manual optimization of the sub-system. 8-10-10-27 Primaries at 1T Command Rate. Crucial tuning: tRFC pushed down to 220 and tREFI maximized to 32767 to break the 0.71ms latency barrier and minimize refresh cycles under heavy load (32GB / 8 Ranks)

Final y-cruncher 1b run on Linux Cinnamon. Achieved 228.058s with 98.86% multi-core efficiency. System optimized by disabling WiFi/background processes and using SCHED_FIFO priority (chrt -f 99). Verified stability for 32GB DDR3-1866 8-10-10-27 1T with tREFI 32767 and tRFC 220.

Achieving a 0.71ms deterministic latency and a 228.058s y-cruncher 1b time on a 12-year-old platform isn't just about luck; it’s a testament to the incredible synergy between the MSI Z77 MPOWER and Patriot Viper 'Black Mamba' memory.

Patriot Edge: While marketed as "commercial" RAM, these Black Mamba modules offer performance comparable to high-end enthusiast kits. Running 32GB (full DIMM capacity) at 1866 MHz with 8-10-10-27 1T timings is a true masterclass in selection. They withstood 1.65V with impeccable stability, proving that Patriot's build quality can withstand extreme sub-timing stress (tRFC 220 / tREFI 32767).

Z77 MPOWER Dominance: The PCB quality of the MPOWER series remains legendary. Its ability to maintain signal integrity across 8 ranks of memory while keeping an Ivy Bridge IMC stable at 4.4GHz is what allowed for this "slingshot" performance!!! The MPOWER series (and the XPOWER) on Z77 and Z97 was the moment MSI stopped following the others and started laying down the law in the world of extreme overclocking... MADNESS!!!

Aside from the kernel customization, I pushed this hardware to its absolute limit in a daily build (I'm a bum and it's my only reliable active PC), proving that Patriot memory and a Z77 motherboard are a perfect match for overclocking. @Patriot_Memory e @MSIGaming.

How boring! I have to stop until I get my hands on Devil's Canyon and Z97 Mpower last ddr3 experiment. Who knows what monster my crazy obsessions will create...NASA standard lol!!!

Edited April 19Apr 19 by TheFyxxxer