Normal Voltage for E6750

[profJim]

Re: Normal Voltage for E6750

I've found MaxxMem2 to have greater variation with its memory benchmark scores than Everest or other programs, especially when I repeat a test several seconds after the previous test.
When I run benchmarks, I always boot and then wait 10 - 15 minutes for all the start up programs to complete.

Your "MushkinSean" timings are reasonable, but probably not at the optimal settings.
There won't be a big difference in memory performance with "reasonable" versus "optimal" settings.
You will have a bigger performance improvement if you can run your system with somewhat higher cpu/memory speeds, even if it's only +5MHz FSB.

When I compared my 1-3-3-50-1 results with my 3-3-5-50-3 settings, the improvement was less than 0.05%, that is less than 1/20 of 1%.

[Ken429]

Re: Normal Voltage for E6750

Oops #3. I got caught up in Microsoft W7 induced niddle quiver! ProfJim accept my apology for questioning you tight memory timings. The attached spreadsheet is not real official but it was enough to convience me I was incorrect. Looks like I'll have to break down and buy the Everest program if I want to run with the Big Dogs.

ProfJim, I don't know if you remember all the problems we had with the Q9550 overclock but I suspect updating that system BIOS to FE will solve a lot of those "slow learner" problems too?! It looks like you may be weakening a bit on the LLC Enabled setting. Should I start another thread on that can of differing opinions?

I'm going to let this thing rest awhile so thanks to all you TweakTown guys for the help and comments from the contrary E6750 system.

[profJim]

Re: Normal Voltage for E6750

No need to apologize.
Your spreadsheet shows that tightening some advanced timing controls doesn't always have a significant effect on memory performance. MushkinSean's suggested settings are a bit looser than "normal" and this is done to help ensure that the timings will be stable. Settings will vary somewhat from system to system, even when two systems have identical hardware.

If you update your Q9550 system to the FE bios, you will need to re-enter your saved bios profiles and repeat your stability tests. You will probably need to perform additional tweaks with the new bios, as there's no guarantee that your current stable settings will be completely stable with an older or newer bios.

I've seen many threads where the OP overclocked their systems to 4.20GHz stable or higher, only to settle on a final o/c in the upper 3GHz range for everyday use. Some of these systems higher vcore voltages were not stable unless they enabled load line calibration. Intel intentionally designed vdroop to help keep cpu over- and under-shoot voltages within a specific voltage range. Some systems with LLC enabled passed all extended stability tests, but would freeze or BSOD sometimes at idle. There are some excellent articles on the web that explain the advantages and disadvantages when LLC is enabled. Some brands/models of motherboards have significantly lower vdroop values than others.

Check out both of Conner's Q9550 threads for a wealth of information in his quest for speed.

Before you flash to the new bios, use Q-Flash to save a copy of your current FD bios profiles to USB and/or hard drives. Even with saved bios profiles, it's best to have a hard copy for each of your current saved bios profiles. We can add to your former Q9550 thread or you can create a new thread.

I'll re-visit your Q9550 thread and PM you if I have any suggestions for a higher overclock. If you find that the newer FE bios has any problems, you can re-flash to your current FD bios where your current stable settings should still be okie-dokie.

With my C2D systems, I've found that an overclock of xx% will generally yield a performance increase that will be about half of the xx% o/c increase.

Glad to have helped out with both of your systems.

[Ken429]

Re: Normal Voltage for E6750

Just for information - using the settings in post #113 and ProfJim's tighter memory timings in post #114 the system ran IBT for 10 passes successfully @ 1.4250V, 1.43125V and 1.43750V. Prime95 ran unsuccessfully for 3+ hours @ 1.4250V, 8+ hours @ 1.43125V and finally successfully(?) for 10+ hours @ 1.43750V. Is the 1.43750V too high for 24-7 usage? The maximum temperature does not seem to be a problem since it reached 45C during Prime95 but that is with the case cover off and an ambient temperature of 18C. Going from FSB 440 to FSB 450 required a pretty big jump in voltage and I'm not sure if the relatively small increase in speed is worth the voltage risk.

[profJim]

Re: Normal Voltage for E6750

Nice work!!
The maximum VID for your cpu is 1.50 volts.
Enabling LLC will let you lower your voltage a bit, it's your call.
Your Prime95 temps at 18 (C) are very good. Usually these temperatures are in the lower 50 (C) range where the ambient temperature is probably about 22 - 23 (C).

My personal preference is to overclock with "reasonable" temperatures up to the point just before I need a big jump in cpu voltage.
I've seen several posts where the OP has summer and winter bios profiles for their extreme setups to help keep temperatures at a safe level.
You might want to have two stable bios profiles, one at 450MHz and one a bit lower at 440 or 445MHz.
I run each of the three Prime95 tests separately (blend, small FFT, and large FFT) for 24 hours, followed by the OCCT PSU test for 1 - 2 hours.
Call it overkill, but I like my system to be more stable than I am.

[Ken429]

Re: Normal Voltage for E6750

Well I thought I was home free with the Vcore set at 1.43750V LLC Disabled. IBT was able to run 10 passes successfully from 1.42500V to 1.43750V. OCCT too was happy all along the way. The crazy part was Prime95 would run anywhere from a few minutes before posting a rounding error to 10+ hours with no errors. I guess that is what is called UNSTABLE? I gave up on the LLC Disabled setting for now and am running a Prime95 at 1.42500V LLC Enabled. My question(s) are: with LLC Enabled OCCT reports 1.36V+/-almost nothing when Vcore is set to 1.42500V. OCCT reports 1.36-8V (mostly 1.36V) with Vcore set to 1.43750V. There in lies my problem, I think. Why is the voltage drop reported by OCCT slightly more at Vcore 1.43750V than Vcore 1.4250V? It looks like this thing wants the OCCT reported 1.39V with LLC Enabled and the Vcore set at 1.42500V to run at FSB 450. Next question, how do you measure the VTT setting of 0.05-0.10V lower than Vcore? Is it the Vcore set in the BIOS or what the Vcore is under load? Last question, is there a "how to" to test the max the memory is capable of without getting the CPU FSB involved? I got desperate and ran MemTest86+ V4.0 for four hours thinking I might have had a memory issue. I know MemTest86+ does not push the CPU like Prime95 but is it possible to get the OC CPU out of the equation?

[Lsdmeasap]

Re: Normal Voltage for E6750

Some voltages vary in the amount of vdrop and vdroop that gets applied, LLC enabled or not. So you may need to try the next few up or down and see if they work better for you, these also vary from BIOS to BIOS.

Vtt measurement in regards to your actual question should really be done with a DMM, since most users cannot do this it is best to keep it under actual idle voltages (CPU-z) because that is where the CPU runs most often.

Yes you can lower the CPU multi to the minimal but keep strong voltage to it, then raise your FSB to as high as you need with whatever memory multi you are trying to use. Then tinker with your memory overclocks, keep in mind Vtt, NB, and tRD come into play here.

[Ken429]

Re: Normal Voltage for E6750

Thanks. Lowering the multiplier on the CPU was the part I was missing.

I ran a 14 hour Prime95 session (no errors) with LLC Enabled and the Vcore set at 1.425V and VTT set at 1.34V. CPUz says the voltage is 1.392V at idle so based on ProfJim's rule I'm ok. Probably too much Vcore. Now I have lowered the Vcore to 1.4125V and lowered the VTT to 1.32V with LLC Enabled. CPUz says the voltage is 1.376V at idle so I'm Ok with the lower VTT? I'm trying to find the lowest stable voltage with LLC Enabled. Then I'll try to find what Vcore setting is needed with LLC Disabled to match the LLC Enabled stable voltage reported by CPUz. Sound like a plan?

[Lsdmeasap]

Re: Normal Voltage for E6750

You may or may not be ok with Vtt, only testing will tell you if it is enough or not. 65nm CPU's can handle higher Vtt (1.4-1.5) anyway, so yes that is fine in itself.

Might as well just leave LLC enabled, no need to test with it disabled.