Microstutters, odd ground popups and WHEA-uncorrectable _error

[SledDriver 39]

This video demonstrates the first two issues listed in the title.

https://www.youtube.com/watch?v=tenimhbP_JI

I'm running P3D V4.4 +ASP4 +ASCA +PTA on powerful hardware with decent GFX. Having just moved from FSX to P3D, I was rather hoping the days of microstutters were behind us. Admittedly I am pushing the settings quite hard, but does anyone have advice from experience of what are the most likely things to cause microstutters. All drives are SSD. CPU is i7-8700K overclocked and delidded at 5GHz. Watching TaskMan meters, system is not overloaded it would seem.

Secondly, in the video, there are a number of horrid vertical 'popups' happening at various times. Any ideas what is causing these and if there is a fix?

Lastly, about once or twice a session the machine is BSODing while running P3D with a WHEA-uncorrectable _error code. This only happens in P3D. The machine has been rock solid for 15 months with FSX with all 12 cores loaded heavily, and I have previously soaktested the CPUs and GPU at max load to ensure my 5GHz clock is stable. There have been NO BSOD's in 15 months since new until P3D.

The only difference I can see now is that the CPU's are hit less hard by P3D than in FSX, and the GFX card is hit harder, but I have chosen settings such that the GFX card is not fully loaded, not is it running at too high a temperature.

So any advice on these three issues will be gratefully received.

Cheers.

[WillsB3 1]

Step 1 - Temporarily remove your overclock and see if the BSODs go away. If they do then your machine isn't stable for the type of load P3D is imparting on your hardware.

[Bob Scott 7,714]

I've largely solved my stuttering by using VSync with a 30Hz monitor.  G-sync could also be a solution.

WHEA errors, especially if accompanied by stopcodes 101 or 124, are hardware faults, most often caused by an unstable overclock.  As they say in the stock market, "past performance is no guarantee of future results"

Sometimes hardware slowly degrades from its initial condition, and voltages that may have worked at first (but were probably on the ragged edge of stability) no longer do, or perhaps temps are now running just enough higher to induce errors.

Time to re-run your stability testing as if you were doing a new overclock.  If there's room, bumping up voltages may help, or possibly pulling the overclock back a notch may bring it back into stability.  And run a memory test such as MemTest 86+ to make sure you don't have a DIMM going bad.

Regards

[djbully 210]

With regard to OC testing, you should really do a solid test, 20-30 runs of intel burn test or 24hrs of prime95 and cpu should stable if there are no errors. As for memory 6-8 runs of memtest86 or 3000% hci memtest will confirm memory stability. There are no shortcuts! Just an hour or 2 does not cut it, in my experience. Some people will say otherwise. 

[djbully 210]

Further to the above, once you rule out cpu and ram, then graphics card is next. Often they can be unstable at stock clocks, and may need a nudge down, particularly if you have “factory overclock” version. If that is the case you should probably rma it. Or just nudge it down and settle for that. 

[SledDriver 39]

I seem to have got the system stable again now. The trick was to increase the AVX offset by 200MHz. Was set at 2, now at 4 and stable.

[newtie 533]

WHEA error for me turned out to be a bad (old) video card.

Upgraded to a 1080i and all is well.

[SledDriver 39]

Yeah yeah! 🙂

Those hardware faults, forcing you to buy shiny new toys! 😉

Enjoy! 

[AnkH 2,233]

7 hours ago, SledDriver said:

I seem to have got the system stable again now. The trick was to increase the AVX offset by 200MHz. Was set at 2, now at 4 and stable.

You are aware what you now did? You basically told your CPU to run AVX based operations with 4.6GHz instead of 4.8GHz. Means: for whatever reason, you did not have any issue with AVX offset 2, now suddenly you can maintain only AVX offset 4. This clearly indicates that something is wrong with your OC or, I bet on it, with your cooling. Did you check the temperatures after 20min of P3D? Did you check if maybe your cooler is obstructed by dust or something? Did you think about renewing the thermal compound?

[SledDriver 39]

11 hours ago, AnkH said:

You are aware what you now did? You basically told your CPU to run AVX based operations with 4.6GHz instead of 4.8GHz. Means: for whatever reason, you did not have any issue with AVX offset 2, now suddenly you can maintain only AVX offset 4. This clearly indicates that something is wrong with your OC or, I bet on it, with your cooling. Did you check the temperatures after 20min of P3D? Did you check if maybe your cooler is obstructed by dust or something? Did you think about renewing the thermal compound?

Geez, you gotta lotta opinions with knowing few of the facts.

Are you an expert overclocker, able to offer specific tuning advice? If so, happy to discuss this.

I know exactly what I did. I have a machine which was previously not running any AVX hungry apps, but was running non AVX stuff at 5GHz reliably - for 15 months, since new.

I introduced P3D, and the machine started BSODing. Not immediately. Sometimes it would run for an hour on P3D with all the sliders turned up, before it would BSOD.

So I thought, the only obvious difference is the AVX stuff is now getting hammered more than before, so I turned the AVX clock down a bit and haven't had a BSOD since in many hours of continuous P3D, never mind your suggested 20 mins test.

I constantly checked the temps throughout this process and they were all well within spec 33degC idle on the CPU cores, 55-60degC on full load - delidded CPU. The fans (7x140mm case, 2x140 CPU, 1x140mm PSU + the GPU fans) are all hardware monitored, are all spinning fine, and the fan speed is adjusting appropriately with the temps, each fan being individually controlled and monitored. The CPU lid and heatsink are mounted using Thermal Grizzly Kyronaut heat paste. The GPU is not exceeding 77degC under full load. All temps return to idle values within 3-4 seconds of the load being removed - to me that indicates very good cooling.

The case is clean, the filters are not blocked, the fans all draw air in the same direction through the case. And the computer is not in a hot room.

I don't claim to be an overclocking expert, but the guy who did all the settings for this machine definitely is - like literally the world champion over overclocking multiple times, no other than 8pack of Overclockers.co.uk. https://www.overclockers.co.uk/8pack The only deviation from his settings I have now made is to slow the AVX down. I'd be very interested in your knowledge of how I might be able to reliably take that back up to it's original 8pack setting.

So, all that said, what do you think is wrong with my overclock? I'm really all ears?

Edited April 1, 2019 by SledDriver

[kaha 205]

What CPU is it? 

[SledDriver 39]

i7-8700K professionally delidded.

[kaha 205]

HT on or off?

[SledDriver 39]

On.

[kaha 205]

Then 5.0 with AVX offset 0 would be quite good! I assume your CPU is on water and not on air.

 

Icannot achieve this with my aircooled 9900K. I have HT switched off, this allows for much cooler temps.

 

You never know what happened. Do a prime95 run (at least V29.4 build8 for small FFTs, max heat). If this is ok for an hour you can pretty sure that it works for P3D.

 

My CPU (stock, not delidded) is running at 5.0 with AVX offset 0 and HT off. It's aircooled. Max CPU temp after 5 hours flying is below 70°. In prime95 CPU temp goes as high as 82. In addition there are 2 overclocked 1080TI GPUs in the rig also producing heat.

 

Some voltages are of course important. I use:

- vcore 1.35V

- vccio 1.13125

- system agent 1.13125

 

But I do not know if Voltages are similar to the 8th generation CPUs.

 

Karl