Precision 7530 & Precision 7730 owner's thread

@kittenlips, as @Aaron44126 mentioned, it'll be very clear where your limit for undervolting is—if you exceed it, there will be a noticeable point where you get multiple BSODs daily. For example, my Xeon at -140 mV is unstable. I've currently kept it at a similar undervolt to yours, -125 mV, which was a very lazy one-hour job, and I haven't thoroughly benchmarked it. It works stably—no BSODs with my present use, which is sufficient for me.

The best way to find out, is to run a suite of computationally-intense benchmarks with error-checking, that check calculated values against known-good results. Also, get ECC memory if your CPU supports it.

As in the long debate, there is an inherent. but slight, and I daresay negligible, increase in risk of an unstable system with an undervolt, but the benefits greatly outweigh the risk.

And @Michiko, the debate ended nearly a day ago...

 

I would recommend running programs such as Prime95 and AIDA64 system stability test to check for undervolt stability. Like @Ionising_Radiation mentioned, they perform calculations to check to see if your CPU is making errors.

 

Would be apt to note that Prime95 is best performed in a cool environment such as an air-conditioned room (or better, Antarctica) given how much it pushes the vector extensions of Intel and AMD CPUs.

I am thoroughly envious of forum members like @Papusan who live in Scandinavia, and can boast extremely cool ambient temperatures, while I'm stuck bang smack on the Equator...

 

Aggressive undervolts are just as likely to fail during idle or low load when the voltage drops as during stress tests. And don't forget to keep an eye out for WHEA correctable errors. HWiNFO sensors has a handy section for that. The clock speed and voltage I thought was stable for 4 months threw a WHEA correctable error at me recently so I had to up the voltage.

 

whenever he feels like benching it's just a 1 minute trip outside to get negative temperatures

 

True. At -140 mV my notebook randomly BSODed while scrolling through Chrome. 120 and 125 mV (on AC and battery) respectively have given no WHEA errors (correctable or otherwise) for the past week; let's see how it goes over time.

 

Keep in mind that pushing the system at 100% for an extended period of time isn't the only way you should test for an unstable undervolt. Running at lower power states could also present issues on a seemingly stable one when pushed. For my system, I had issues at -100mV when the laptop is on battery, in and out of sleep mode.

On the topic of undervolting, what's the best practice to set an undervolt profile in XTU upon boot?

 

Consider using ThrottleStop instead; it's also easier to implement a startup profile there by manually creating a log-on task in Task Scheduler.

 

This is false.

You can have calculation errors, cache problems, and other weird phenomena. A BSOD is a “major” (in relative CPU terms) mistake that crashes the whole system.

EDIT: As mentioned, Prime95 is a good tool to test your CPU for errors. It also will cause it to work very hard, so watch the cooling.

ECC memory has nothing to do with CPU misses due to abnormal voltages.

ECC memory corrects bit flips in large memory caches due to radioactive and cosmic occurrences.

 

For all its negatives, there's one really bright spot with the Quadro P3200: its drivers, though equivalent to the GeForce ones, don't include GeForce Experience. Even if the GeForce driver is installed in place of the ODE driver, by the very virtue of the fact that the GPU itself isn't a GeForce. Nice; we could all do without yet another bit of software clogging up our background tasks.