Boost your gaming experience without any cost.
Boost your gaming experience without any cost.
The setup consists of an i9-7920X with 12 cores, paired with a Corsair Vengeance LPX RGB 3000, 8x8GB RAM, and a MSI 3070 Gaming monitor at 1440p/144Hz with G-sync. It's primarily intended for gaming (over 90%), streaming, and AVX-512 compute tasks. I'm evaluating each CPU generation to estimate gaming performance. The Skylake-X architecture struggles with single-core speed due to low clock speeds (under 4 GHz) and limited IPC. Even newer chips might gain significant speed on clocks alone, not accounting for IPC improvements. While I don't require peak FPS, certain titles feel sluggish compared to older high-end CPUs like the 11700K. I'm considering tweaking performance options but am cautious about stability. Disabling Hyper-Threading could help if it improves core utilization, though I've used a similar workaround before. It won't boost raw speed but might ease thread contention. AVX optimization seems unlikely to help since this board doesn't support AVX2 and lacks AVX-512 features. Raising the CPU power limit isn't expected to yield noticeable gains, and testing the default 140W TDP setting is advisable. Software tools for monitoring CPU power are unreliable here. Other adjustments remain uncertain.
I've been using daily overclocked CPUs since I built my first PC. It's easier than it looks to get a stable overclock for continuous use. If your setup can handle an hour of Prime95 Small FFTs, you're in good shape. You might extend testing to 3 or 6 hours for longer stability under full load. From what I've experienced, it's usually enough to test once daily without needing more. I haven't faced any problems with my overclocked CPUs because I always check their stability before finalizing the change.
My processing demands match Prime95's performance. On other machines I rarely notice system issues—sometimes just once a month—which I find too low. My usage here is less frequent these days. Because Prime95 (and my specific workload) supports AVX-512, it operates at a completely different benchmark than AVX2 or regular loads. Confirming Prime95 stability doesn't guarantee stability in other forms. Unfortunately, the CPU power monitoring hasn't functioned well on this machine. The data from software gives confusing 1.x range numbers. When I run Prime95 with AVX-512 it does drop to base frequency, hinting at TDP constraints, which is the default BIOS setting. I suspect the maximum allowed for all cores might be around 3.8, though I'm not certain. I might consider raising the power limit just to test. This seems like a reasonable adjustment. Before committing, I checked past hwbot entries with this CPU before choosing it as my primary machine. It appeared stable for x265 at about 4.5 GHz. Perhaps I can try a setup with 12 cores disabled (to maximize per-thread performance), AVX-512 off, and see if I can reach around 4.4 GHz—still better than the current 4.5 GHz, possibly with extra gains by skipping HT. Skylake wasn't particularly fast compared to newer chips. The consumer version likely wouldn't exceed 4.7 or 4.8 GHz on non-AVX tasks without heavy cooling, given outdated AVX knowledge and the necessity of AVX now.
Did you try them using Prime95 Small FFTs? The stability of an overclock matches its testing reliability. I’ve never faced problems with this approach—it’s highly dependable. Prime95 excels at revealing instabilities and reporting them, making it by far the most trustworthy method I know of, and I’ve had no issues with it.
The tools I use share the same math library as Prime95. It's comparable to a basic FFT stress test, but applied to real tasks instead of just benchmarks. Finding errors after weeks of operation is crucial. Proving system stability isn't straightforward—it depends on how often issues arise. My acceptable limit might be stricter than typical standards. Prime95 demands heavy FP64 calculations and can be very power-hungry at a fixed clock speed. In the past, I've seen systems that worked well with Prime95 but failed in other programs. Relying on one workload isn't enough; I've handled extreme overclocking before, so I understand the trade-offs. Whether 10% effort is justified depends on whether it's worth the risk, especially since I haven't done this in a while. Distinguishing between gaming and critical stability testing is helping me decide if the challenge is worth taking.
Similar to production specifications, certain components have tighter limits while others allow more flexibility. This applies to how boost and clock rates respond. I haven’t faced this issue since I always keep the clock speed fixed, ensuring my static all-core overclocks consistently exceed the single-core maximum boost clock (and turning off boost). I understand the desire—you want to push performance further.
The optimal choice seems to be a deli item paired with some liquid metal to manage the chip’s heat. Keeping the temperature under control makes it easier to boost voltage without sacrificing performance. A 1.25v, 4.5ghz setup could offer consistent stability around the clock, delivering a clear performance lift. Small overclocks haven’t affected reliability in my experience and usually only raise power use slightly. Modern chips are closer to stable operation than older models, especially with adequate cooling. For maximum assurance, aim for sustained P95, IBT, Linpack, and R23 scores of at least two hours. You might still face rare cosmic interference events, but you’ll need to define what stability means for your needs. Running the system on a solid UPS will help it handle the load.