• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Stability testing with "speedstep" technologies

T

TechonNapkins

Weaksauce
Joined
Oct 20, 2010
Messages
94
[SCROLL DOWN TO THE BOTTOM OF THIS POST IF YOU WANT TO SKIP THE BACKSTORY AND JUST GET THE QUESTION OF THIS THREAD]

I put "speedstep" in quotations because it applies not only to Intel, but also AMD (both CPUs and GPUs) and Nvidia as well... I guess "clock profiles" would be more accurate, but I digress...

For the last few generations of CPUs and GPUs, I've noticed stability testing has become much more laborious and uncertain. It used to be you could run Prime95 or Intel Burn Test overnight and, if no errors, you could safely assume the system was stable. And if your GPU could withstand Furmark? Oh, done deal!

Well, the last time I remember than being the case was with my Core2Duo E8500. It all changed with my Thuban 1055T. Everyone was claiming 4GHz was a no brainer...

... a no brainer? Well, call me brainless, 'cause unless I pumped over 1.5V into that thing, I couldn't even boot into Windows at that speed. And running Burn Test for just 20 minutes at that speed I noticed the room temperature rise significantly. I stress tested it close to 3.8GHz at reasonable volts, but I would eventually run into odd crashes every so often. Backing it off to 3.7GHz eventually stabilized the system.

Early this year, I jumped on the 2500K bandwagon, and again, 5GHz was supposed to be "brain-dead-easy". Not for me it wasn't. I could stress test it at 4.6GHz, but I had to bump it down to 4.4GHz for day-to-day apps.

Then there's my recent GTX570 purchase... I can run Furmark for five minutes with no issue at 880 core, but can't play for 5 seconds without crashing in Just Cause 2.

What's going on here? Why would an app that stresses all cores at 100% show stability, yet a much more lightweight app eventually initiate a crash? I believe it's in the power/clock management. In a current CPU/GPU, the clocks and volts are bouncing all over the place to cope with demand. There's the "Load-Line Calibration" feature we can tweak to try to stabilize the voltage, but at the expense of exposing our CPUs to higher power spikes. I remember one guy with his Thuban made a custom profile for every clock stepping to ensure stability at any load.

I suppose an easy way out of this predicament is to go backwards - disable all the power saving features, set Load Line Calibration to high, and force a singular clock rate - and indeed, that is what many extreme overclockers have done... but what about us normal folk? I, for one, find a home PC idling over 200W to be unacceptable. I want to keep these power saving measures in place, and I'm sure many of you guys do too. So, how to we test overclocking within this new paradigm?

In short: what are the apps you've found that expose instability that others have failed to expose? CPUs or GPUs.
 
As hinted at in my story above, I found Just Cause 2 to be the quickest game to test an overclock on my GPU (I tested with max settings, except AA... left it at 4x). And that's actually playing the game. Just running the benchmarks don't reveal a bad overclock as fast. That being said, I have many more games to test.


Also, at least with the higher end Nvidia cards with 320-bit or 384-bit memory controllers that you can't run the RAM at it's rated speed (do to limitations both Nvidia and AMD have acknowledged about wide memory controllers), that the old way of determining max memory clock, that is, looking for weird sparkles/flickers/other graphical corruption, seems to no longer apply... now it just crashes after awhile just like a too-aggressive core/shader overclock.. probably due to the limitation being the memory controller now, and not the memory itself.
 
Last edited:
Call of Duty 4 is sensitive to instability of CPU.I got mine stable on P95 and a quick Linx test then getting BSOD in COD4.
 
Well, I found Crysis 2 with the new Ultra settings to really push things... it has just replaced Just Cause 2 as my quickest-way-to-test-an-overclocked-gpu test. 835 core passed the mustard in Just Cause 2, but had to back it down all the way to 823 on the overhauled Crysis 2
 
stop believing anyone who says XGHz is brain dead easy. Sounds like you know how to stability test so just do it like you have been and don't worry about others. I'm using Speedstep with no problems. I'm about 300MHz lower than the highest I got (stopped because I felt uncomfortable with voltages, didn't find the limit). Voltages are good LLC is on.
 
I've always used speedstep with my over clocks. My i7 930 runs at 4.32GHz with LLC off and speedstep and C1E on. I don't like the idea of my CPU running at full speed all the time and wasting all that power too.

My CPU drops down to 2.4GHz on 1.0v and ramps up the volts/MHz as needed.
 
SirMaster said:
I've always used speedstep with my over clocks. My i7 930 runs at 4.32GHz with LLC off and speedstep and C1E on. I don't like the idea of my CPU running at full speed all the time and wasting all that power too.

My CPU drops down to 2.4GHz on 1.0v and ramps up the volts/MHz as needed.
Click to expand...

What voltage are you running? That sounds like a pretty awesome overclock.
 
You must log in or register to reply here.
Back
Top