Best CPU Stressor Out There

[mcrumiller]

In case anybody's wondering, I've since backed down to a 1700 FSB, which I can run at 1.44375v--I feel safe at this level. Dropped from 28 GFlops to 25, but I guess I'll save myself some hardware problems down the road.

 

[Gillbot]

I still think 1.44v is excessive.

 

[BonzaiDuck]

Originally posted by: Gillbot
I still think 1.44v is excessive.

Ditto. Then, the question is "What is the reported voltage from the sensor at idle and load, respectively?"

 

[mcrumiller]

Idle - 1.392v.

Load - 1.344v.

 

[Gillbot]

Originally posted by: BonzaiDuck

Originally posted by: Gillbot
I still think 1.44v is excessive.

Ditto. Then, the question is "What is the reported voltage from the sensor at idle and load, respectively?"

IMHO, irrelevant. If 1.4v is max for 45nm, you should not exceed 1.4v from BIOS. If you use 1.4 after load, the chip can still see in excess of 1.4v when not loaded.

 

[BonzaiDuck]

Originally posted by: Gillbot

Originally posted by: BonzaiDuck

Originally posted by: Gillbot
I still think 1.44v is excessive.

Ditto. Then, the question is "What is the reported voltage from the sensor at idle and load, respectively?"

IMHO, irrelevant. If 1.4v is max for 45nm, you should not exceed 1.4v from BIOS. If you use 1.4 after load, the chip can still see in excess of 1.4v when not loaded.

Some motherboard BIOS' that I've seen firsthand show a greater and greater disjunction between the amount of voltage specified in the "setting" and the amount reported by BIOS monitor -- greater as you increase the voltage. This would be corrected to some extent with BIOS revision. I haven't seen this discussed with great clarity, although I -- myself -- should probably go back to Graysky and see how much it's mentioned there.

But -- at least in my humble view of things -- the ladder of settings for VCORE are at minimum just a set of voltage increases which may or may not correspond with any precision to the true voltage. And, again at minimum, they probably just represent incremental voltage increases.

That leaves the sensors. I once possessed a book on control systems -- absorbing only small parts of it. But even the sensors would be subject to some inaccuracy, and we know the only way to get accurate measurements is to apply a multi-tester or some similar device to the right pin-outs.

I'm also just a bit confused as to whether you mean "1.4V is max for 45nm" as applied to the VTT voltage, or the VCORE voltage. The Intel documentation is very specific: the "safe" range is 0.85 to 1.3625V. There is a wider range of VCORE, for which Intel says the processor will still "operate," having an upper bound of 1.45V. Intel states that operation above the 1.3625V boundary and below the 1.45V boundary for any length of time will degrade the CPU -- which would continue to operate at lower voltages if the VCORE is lowered.

Anandtech proved in a short article last year that increasing the VTT voltage for 45nm CPUs above 1.4 will simply burn out the CPU -- or that the risk of a quick burn-out increases dramatically above 1.4.

Anyway, Gilbot, it is my personal practice -- a matter of choice and caution -- to assure that my idle voltage doesn't exceed 1.3625V. Intuitively, I'd think that the greatest stress would result under load, though. Then, there's the matter of the "vOffset" to account for the instantaneous spike in voltage when the CPU returns from load to idle. Supposing that the spike rises above the intel spec (1.3625), and that the spike is nominally 0.02V above the equilibrium idle value, I wouldn't know how much damage might result there. But for 0.02V -- why bother with the risk? With nothing more to go on than the VCORE sensor reading, I'm content to let my E8600 idle VCORE max out at 1.34V.

 

[Gillbot]

Most people accept that 1.4v is max for vcore on the 45nm chips. There is misconception that it should be 1.4v via Windows vs. BIOS. I always say via bios because you are basically overvolting to achieve 1.4v via windows.

 

[Idontcare]

Originally posted by: Gillbot
Most people accept that 1.4v is max for vcore on the 45nm chips. There is misconception that it should be 1.4v via Windows vs. BIOS. I always say via bios because you are basically overvolting to achieve 1.4v via windows.

Yep, the gap you specifically refer to is called Voffset.

Among other things (like eliminating Vdroop), LLC (loadline calibration) is tuned to make Voffset = 0. You can do this manually by upping the Vcc in the BIOS until the Vcore shown at idle in Windows is equal to your target Vmax (1.4V in this example) but then you have expose your CPU to the peak voltage transient during overshoot when the cpu goes from a loaded state to an unloaded state.

(source)

 

[Gillbot]

Originally posted by: Idontcare

Originally posted by: Gillbot
Most people accept that 1.4v is max for vcore on the 45nm chips. There is misconception that it should be 1.4v via Windows vs. BIOS. I always say via bios because you are basically overvolting to achieve 1.4v via windows.

Yep, the gap you specifically refer to is called Voffset.

Among other things (like eliminating Vdroop), LLC (loadline calibration) is tuned to make Voffset = 0. You can do this manually by upping the Vcc in the BIOS until the Vcore shown at idle in Windows is equal to your target Vmax (1.4V in this example) but then you have expose your CPU to the peak voltage transient during overshoot when the cpu goes from a loaded state to an unloaded state.

(source)

I'm going to copy this and make it a sticky. I'm tired of repeating this over and over.

 

[Pelu]

Originally posted by: n7
Welcome to the forums.

I'm calling BS though, sorry.

Your E8500 will shut off due to thermal protection well before it hits 120C, unless your sensors are wrecked.

I hate to be the hater, but until i see screenies showing over 100C with CoreTemp or RealTemp, i don't believe.

Run LinX or IBT set to max RAM to see high temps.
http://www.xtremesystems.org/f...howthread.php?t=201670
http://www.xtremesystems.org/f...howthread.php?t=197835

I was thiking the same.. lol.!!! maybe is F temperature and not C

 

[mcrumiller]

Well, I did have CPU thermal control turned off in the BIOS, so it is entirely possible that I was hitting temps that high momentarily. Surely if I had kept it up my chip would have been fried. Here's the reasons I think I might have actually hit those temps:

-I can hit 85C when running at 1.45v on max load.
-I can hit 97C hwen running at 1.5128v on max load.
-I was running on 1.56v on max load with a >1800 FSB and higher clock speed during the temp spike. Extrapolating from the first two points, temps > 100C are pretty much inevitable. I'm just glad I didn't burn it out. Thanks again for the help guys.

 

[Idontcare]

Originally posted by: mcrumiller
Well, I did have CPU thermal control turned off in the BIOS

I'm not judging you, pm me the answer if you prefer to avoid public scrutiny on your response, but I'm really curious why did you turn off the thermal control?

I've no doubt you had your reasons, but for the life of me I can't fathom what they would be, so help satiate my curiosity and fill me in on why you elected to do this

 

[mcrumiller]

Well when I initially started OCing I just disabled pretty much everything...XD, C1E, virtualization, etc. Thermal Temp control went with it.

 

[Idontcare]

Yeah I could see doing that, makes sense.

So it wasn't so much a "I went out of my way to go into the bios and selectively disable thermal protection because of XYZ" but more of a "I disabled anything that looked power-saving related while I was tuning up my rig's OC".

I do the same thing, sans shutting off the thermal protection...at least I have always assumed I left that feature active, I could have deactivated without realizing it I suppose.

 

[daw123]

I would restore the CPU to stock settings until your liquid cooling set up is sorted out and your temps are reasonable.

 

[Idontcare]

Originally posted by: daw123
I would restore the CPU to stock settings until your liquid cooling set up is sorted out and your temps are reasonable.

+1

Sound advice.

 

[i7guy1]

Without thermal protection, you could melt your cpu, motherboard. It could get ugly. :-(

 

[Avalon]

Originally posted by: ShawnD1

Originally posted by: Cheex
OCCT 3 (built-in Linpack test).

/thread

I checked this out after reading this thread, and I'm very impressed. It can somehow make my computer draw 480W of power. The most I have ever seen, while doing GPU and CPU folding, was only 350W. That's one hell of a stress test.

No kidding. I just tried it out myself and it had my E8400 peaking at 10C higher than Orthos. Coretemp was telling me I was at 73C full load, OCCT saying 67C. Using Orthos, Coretemp was only reporting 63C max.

 

[Odysseus145]

I don't have enough memory for 10000x10000 but 5000x5000 only brings me from 38c to 45c on an air cooled 2.7ghz x2 3800, so something's definitely wrong with your setup.

 

[mcrumiller]

Oddysseus, make sure it's a (5000x5000 matrix) * (5000x5000 matrix). How much mem do you have?

 

[mcrumiller]

Also, I max out at 75C for Linx now.