cancel
Showing results for 
Search instead for 
Did you mean: 

4.8GHZ (i7 3970X + RIVE) Overclock

Smiki007
Level 10
Here is my Rock solid 4.8GHz OC.


















Max Temps (LinX 0.64 Max Memory Stress 30min)


Trying to reach 5.0/5.2GHz, no success, need help. I've managed to get 5.0GHz, but max temps was 81-82C.


Always look beyond the limits...
20,793 Views
27 REPLIES 27

Myk_SilentShado
Level 15
ROG RealBench is all the stability testing you'll ever need again Chris 😉 it'll root out instability faster and lighter on resources than the rest

Gottex
Level 9
chrisnyc75, it's all about your preferences, e.g. I like snowboard and you like ski. 🙂
The same with stress testing, from my experience I can tell you that I've had errors in Prime after 2-3-4-5-6 hours ... I believe 1 hour is not enough, but again, its very individually
CPU: 3960X(C2)@4.8GHz (Offset +0.045, LLC-medium), HT-on, 24/7
MB: ASUS Rampage IV Extreme X79 (bios 4206)
GPU: GTX580 3-Way SLI@900/1800/2100 1.100v.
RAM: 16Gb G.SKILL RipjawsZ 2133Mhz 9-11-10-28(2T) 1.65v
SSD1: 128Gb SSD OCZ Vertex 4
SSD2: 512Gb SSD OCZ Vertex 4
HDD: 2Tb WD Caviar Black
PSU: Lepa G1600-MA 1600W
Case: Danger Dan Black Series - LDR Tower 29 - Double Wide
Cool: Water
Monitor: Dell 3007WFP-HC
OS: Windows 8.1 Pro x64 (UEFI)

Smiki007
Level 10
I'm also not a fan of large hours stressing, I've only use those stress programs for a max 30min - 1 hour just to see my max temps, and yes the LinX/IBT can detect an error in OC and launch BSOD or stop prematurely the LinX/IBT (showing the error), within the first 10-15min. Only once I've leave the stress program (Prime95), for more then 30 hours (what a waste of the time), and nothing happened, the max temps was the same like before 1, 2, 3... hours. Some OC gurus said that running LinX/IBT on 50% of your total RAM for 1 hour is more then enough to see if your OC is stable.


Always look beyond the limits...

Martigen
Level 7
Smiki007 wrote:
Here is my Rock solid 4.8GHz OC.

Hey, thanks for taking the time to post your setup. Mine is very similar, right down to using Medium LLC and keeping the VRM on Optimised 🙂 However, I'm currently at 4.6 with offset -0.005 (and VTT set to 1.1, VCCSA to 1.050).

Can I ask a few questions?:

* What's the Vcore like under load with +0.20?
* Whare are your VTT and VCCSA in Windows (as you've set them to Auto)?
* Why have you set the PCH to 1.3x?

Unless I'm mistaken, PCH is just SATA/USB/etc, and unless you're overclocking the BCLK, you don't need to up the switching frequency and increase VRM heat?

Here is a screen cap of AIDA64 with a 100 percent CPU load after 10 minutes:

28759
| Corsair 1000D | i9-10980XE @ 4.8 GHz 24/7 (All-Core) | Rampage VI EE |
| 32GB (4x8) Corsair Dominator RGB 3600 Kit (16-16-16-32 T1) |
| EVGA RTX 3090 FTW3 Ultra with EK block | Corsair AXi1600 PSU |
| 3x Samsung 970 PRO NVMe | Custom Liquid Cooling (Dedicated CPU/EK Monoblock)

Protocol
Level 7
Oh cool, nice!

4.8 GHz - Prime95 (AVX) 100% load @+0.055 offset = 1.392v - 1.40v (LLC=Regular)
4.5 GHz - Prime95 (AVX) 100% load @+0.020 offset = 1.342v - 1.35v - (LLC=Regular)

Interestingly enough if I try to run Smiki007's settings and a Medium LLC, I can't even fire up Prime95 without a hard system reset. Lowering the LLC to Regular and upping the offset got me 100 percent stable @ 4.8 GHz. Temps are ok with either LLC setting.

VTT = 1.084v - 1.106v in Windows (AIDA64)
VCCSA = 1.093v - 1.112v in Windows (AIDA64)

I should post a correction to the VCCSA also... "AUTO" forced an "Extreme" VCCSA LLC, which is way too high for DDR1866, so I bumped it down to regular as well : )

VTT, VCCSA and PLL are all hard-set in the BIOS now.

I do run 1.3x for everything. It appears to generate a tad more heat, but much more stable and predictable fluctuations. My loop handles all of this extra heat extremely well though, so I personally try to focus more on the voltages than temps since I will only run 24/7 safe settings. What is amazing is the heat deltas between standard and extreme power regulations.

I did have a weird problem with the PLL set at 1.8 (default). It's strange, but my USB ports would stop responding all together. A disconnect and reconnect did nothing. The system was still responsive as processing continued... even the reset button wouldn't work!

I also force memory voltage at 1.6v (1.615v in BIOS) down from what XMP wants (1.65v @ 2400 MHz). I do not use the XMP settings since 2400 MHz calls for a CR of 2T and 1866 MHz can do 1T. CL also gets bumped down to 9 instead of 11.
| Corsair 1000D | i9-10980XE @ 4.8 GHz 24/7 (All-Core) | Rampage VI EE |
| 32GB (4x8) Corsair Dominator RGB 3600 Kit (16-16-16-32 T1) |
| EVGA RTX 3090 FTW3 Ultra with EK block | Corsair AXi1600 PSU |
| 3x Samsung 970 PRO NVMe | Custom Liquid Cooling (Dedicated CPU/EK Monoblock)

Protocol
Level 7
UPDATE:

Hitting 4.8 GHz stable @ 1.376v - awesome 😄

This was after my Corsair Vengeance memory kit died, which forced me to purchase the G.Skill kit as a replacement. I believe memory was giving me false positives before, making determining CPU versus memory errors difficult in turn making adjusting for the errors troublesome.

VTT/VCCSA/PLL are very low for this frequency, but it works!

AIDA64 SensorPanel screenshot:

29882
| Corsair 1000D | i9-10980XE @ 4.8 GHz 24/7 (All-Core) | Rampage VI EE |
| 32GB (4x8) Corsair Dominator RGB 3600 Kit (16-16-16-32 T1) |
| EVGA RTX 3090 FTW3 Ultra with EK block | Corsair AXi1600 PSU |
| 3x Samsung 970 PRO NVMe | Custom Liquid Cooling (Dedicated CPU/EK Monoblock)

Protocol
Level 7
UPDATE:

Hitting 4.8 GHz stable @ 1.376v - awesome 😄

This re-clocking was prompted after my Corsair Vengeance memory kit died, which forced me to purchase a G.Skill kit as a replacement. I believe memory was giving me false positives before, making determining CPU versus memory errors difficult in turn making adjusting for the errors troublesome.

VTT/VCCSA/PLL are very low for this frequency, but it works!

I'd be inclined to just to 4.9 or 5.0 GHz, but given my temps sitting here with P95, I think 4.8 is going to be the limit. If I was running a 480mm (or perhaps two in-line) then I might be able to hit the 5.0 mark safely.

AIDA64 SensorPanel screenshot: If you are wondering why the VGA and CPU pumps are running at the same RPM and only the CPU is under load... that's RIVE PWM for you, well the 2 true PWM headers that exist anyhow.
| Corsair 1000D | i9-10980XE @ 4.8 GHz 24/7 (All-Core) | Rampage VI EE |
| 32GB (4x8) Corsair Dominator RGB 3600 Kit (16-16-16-32 T1) |
| EVGA RTX 3090 FTW3 Ultra with EK block | Corsair AXi1600 PSU |
| 3x Samsung 970 PRO NVMe | Custom Liquid Cooling (Dedicated CPU/EK Monoblock)