Assessing an X5670 overclock

[HirschbergT]

Hi, all. I'm new here on the forums, but not to overclocking in general.

I thought I'd post my OC results for my X5670 (Westmere LGA 1366) to get some feedback from the community. I'd like to know if there is anything any of the veterans out there would try to push the chip even higher or optimize the current settings.

Here's what I'm running:

Mobo - Asus P6T Deluxe V2 crossflashed to P6WS Pro to overcome TDP throttling
RAM - Ripjaws DDR3 1600 (3x4GB)
CPU: Xeon X5670

And here are my current settings:

CPU Multiplier: 24x
Bclk: 175
DRAM Frequency: 1403 MHz (underclocked, I know)
Uncore: 2807 MHz
QPI Link: 6316 MT/s
Vcore: 1.352 volts according to CPUID and HWinfo
CPU PLL: 1.80 volts
QPI/DRAM core voltage: 1.25 volts
IOH voltage: Auto
ICH voltage: Auto
DRAM voltage: 1.5 volts

I was able to successfully run Intel BurnTest for 10 consecutive runs at max memory consumption, and my peak core temperature didn't go above 56 °C. So my temperatures are well under control, and so far the system appears to be stable. I'm running OCCT for 3 hours to double check stability right now.

I'd like to hear from the community if there are any suggested optimizations or further tweaks I could make. I'd like to either drop my Vcore a little bit and maintain the 4.2 GHz cpu speed or further push the chip up to 4.3 GHz and still maintain stability.

Thanks!

 

[agent_x007]

Too slow UnCore. Aim for 3,2-3,5GHz clock on UnCore.

Use lower multi (21) and higher BCLK (200MHz), to get memory up to 1600MHz.
You may need to slow QPI speed at that point.

 

[HirschbergT]

Great! Thank you for the suggestion. I'm running IBT with that configuration to see if it's stable. I noticed with that my Vcore occasionally peaks at 1.36V. Do you foresee this being an issue long term? I've heard that I should be staying at or below 1.35V...

 

[agent_x007]

It depends more on temps than Voltage.
But first try to get to UnCore higher and check temps.

 

[HirschbergT]

Well it looks like the system crashed at some point during IBT with those settings. Any suggestions? Last I saw, my peak core temperature was up to 58 °C. Should I try increasing the Vcore one increment? Interesting to see that a higher bclk with a lower multiplier might require a higher voltage...

 

[agent_x007]

Try 1,3-1,35V on QPI/DRAM.
What Timings are you using (CPU-z "Memory" tab) ?

 

[HirschbergT]

I assume that the added QPI/DRAM voltage is needed since the RAM is now running at a higher speed? I am running IBT right now to assess. My RAM is running at its factory timings: 9-9-9-24.

 

[agent_x007]

UnCore = L3 + IMC.
You ran higher UnCore, more QPI/DRAM is needed (since it's ASUS version of "UnCore Voltage").

 

[Regeneration]

Underclock everything and then find the maximum frequency of each component.

Some of G.Skill Ripjaws modules can reach 2400 MHz with timings 10-12-12-31 and maximum voltage of 1.65v.

If your memory is capable of running at 2000 MHz and above, keep uncore ratio of 1.5. If it can't, push uncore to ratio of 2.x.

MemTest64 is the preferred software to test DDR stability.

Use Prime95 v26.6 (custom 512 to 1024 with 80% of RAM) with LuxMark (GPU stress test) to test QPI/BCLK/IMC stability.

Its better to disable Turbo boost and stick to high clock on all cores, unless you're limited by BCLK, and forced to use turbo multiplier.

And finally, use Linpack Xtreme (or IBT) to test CPU stability, but stick to a normal problem size (4GB or 8GB), it doesn't work well with 'maximum memory'.

 

[steen]

The only thing I'll add is to try odd multipliers. I've only played with W3670/80s, but was able to fairly easily get 4.4GHz with 1.375V (IIRC) on a Sabertooth. I've been running the W3680s @ 4Ghz all core in HP Z400s for 4+yrs.

 

[HirschbergT]

Thank you all for your suggestions. I was able to successfully pass IBT for 10 consecutive runs along with a 5 hour run of OCCT with the 21x200 settings. Granted, I was using the "maximum memory" setting in IBT. So I might yet want to re-run IBT with the normal problem size as Regeneration has suggested. Temperatures were below 60 °C in all tests so far. So thermal concerns are a non-issue at this point.