OC 3570k cause Quick Sync instability

cjny71

Distinguished
Jan 26, 2010
7
0
18,520
Summary: Stable moderate overclock settings (4.5 and 4.6 tested) with Prime95, good temps, no in game crashes, but MediaEspresso fails using Quick Sync if OC is greater than 4.2.

Setup: I have an ASRock Z77 Extreme 4 MB, 8 GB 1600 RAM (CL9), Hyper 212EVO cooler, and EVGA 560Ti Superclocked edition. I installed the latest BIOS and drivers from the web site. I have Virtu MVP virtualization on using D-mode (LCD connected to 560Ti), I ran this at stock settings with an overnight run of MemTest 86+ v4.2 and another overnight run of Prime95 blend test, and a 10 minute test with Intel Burn Test. I then used the computer for a week and the Platform was very stable.

Overclocking: I followed the "Definitive Ivy Bridge OC Guide" and the one at FlyingSuicide. I tried a fixed 1.25v OC to see what to set the LLC on the ASRock to (L2/L3 worked about the same), and worked my way up to 4.5 GHz. I paid attention and made sure to set the C-states for fixed/offset as per the guides. I then tried offset OC using a very safe +5mv offset/+4mv turbo voltage, and worked my way up to 4.6 GHz. I could not go higher than 4.6 without Prime95 crashes even using up to 4 voltage bin settings higher on offset/turbo, but that is ok as I was not really looking for more than 4.5 anyway, and the voltages were getting up to 1.3 according to CPU-Z. At 4.6 Prime temps were averaging 75-80C, with 90C peak on the 8k tests about 50 minutes in. I let Prime run for another hour, and then I played about 3 hours of Batman: Arkham city without a crash.

Problem: I wanted to see just how much faster/better Quick Sync encoding would be. I ran MediaEspresso (ME), which detected QuickSync as long as Virtu MVP was turned on. ME was able to convert a video file with HW acceleration turned off (CPU), and using CUDA (MVP off). However switching to Quick Sync caused either: a HW Error message from ME, ME to crash, the display driver for Intel IGP to crash, and one BSOD 0x1A (memory).

Note that I did NOT change the default BIOS settings for the Intel iGPU (HD4000), and I did not OC the RAM (voltage and timings match spec). After some more testing I settle at a fixed 1.25v OC at 4.2 and ME appears stable (converted 1 entire DVD, about 2:45 of content). 4.3 allows some short clips to convert but there are crashes on longer clips, and above that ME crashes about 15 seconds in. I did not notice any additional vDroop when ME was running, and CPU load was 25-55%. The Intel Extreme Tuning utility showed the iGPU at load at 1150 MHz, and TDP Watts under the 77 ceiling during ME runs.

Question: How is a stock clocked iGPU/Quick Sync/HD4000 becoming unstable when the CPU is OC? Is there a dependency on RAM speed or CPU-RAM ratio than might be causing this? Does the CPU require more voltage when the iGPU is loaded?

Any help is appreciated. Please do not reply regarding opinions on Virtu, Ivy vs Sandy, my HW selection vs yours, or achieving crazy high OC results on the same platform. I am just looking for a stable OC at 4.5 with ability to use Quick Sync/ME.
 

cjny71

Distinguished
Jan 26, 2010
7
0
18,520
Update: I read that to OC the iGPU (which I am not attempting) required raising vCore on the CPU, so I thought maybe the reverse was true. I set the iGPU voltage to manual, which defaulted to 1.10v. I had to raise this to 1.25 for Quick Sync to be stable when the CPU was at 4.3 GHz (43x100), 1.325 at 4.4, and 1.350 at 4.5. I was surprised at the increase required since the iGPU is still at stock clock 1150. I don't think I have to worry about the voltage as the Intel specs allow for up to 1.52. Thermals and TDP seemed to increase an inconsequential amount while transcoding. Should I be concerned about adding +.25v to the iGPU? Has anyone here run similar voltages on HD4000 for any length of time?