Dynamic RAM preview max other than 0 for rendering still haunts me

bitman wrote on 9/20/2017, 2:13 PM

I am still plagued by random rendering issues like frame hiccup, frame flash back and forth or other anomalies or even complete Vegas crash with rendering in either Vegas 14 or 15 on my system if I dare keep Dynamic RAM preview max at it's default 200 or higher. All this issues are solved when keeping it at zero. The rendering issues or crashing seem more dominant on my more complex videos (say with time stretching speeding up, and Newblue plugins added).

The rendering format chosen does not seem to influence it.
Putting GPU on/off also not.
It is just that darn Dynamic RAM preview max ... why, why, why???? Is it 12GB of video memory of Titan-x that collides with Vegas or my 16 GB of system memory? Newblue Plugins that use the same memory area as reserved for Vegas?

The thing is I am not stuck or blocked here, I can just leave the dynamic mem to 0, but I would like to assign some dynamic mem as it renders a bit faster (every thing I can squeeze out extra rendering speed is a bonus).

I would appreciate the Vegas team would look into this, I am sure I am not the only one (still) suffering from this. Or am I?

Comments

Kinvermark wrote on 9/20/2017, 3:20 PM

I am sure you are NOT alone, but it seems like a very difficult problem to diagnose - so many permutations.

What I see, is that a dynamic ram preview is not always accurate. Sometimes it is obvious (eg not even close to full frame rate playback), but more dangerous is when it looks like it is accurate but the render then does not match.

OldSmoke wrote on 9/20/2017, 3:43 PM

Mine is and has always been since SVP11 on 200 and I never had an issue with it. I rather suspect that some user have RAM timing issues in their system and have to reduce it to zero.

Kinvermark wrote on 9/20/2017, 3:47 PM

I recently read a post by astar over on the COW forum that basically suggests the same thing. He wasn't specifically referring to this problem, but was explaining how frames were passed back and forth in CPU/GPU memory and just how fast that has to happen before frames start dropping.

set wrote on 9/20/2017, 4:48 PM

When I first tried VP15, I keep DRAM default 200..., but on long rendering, I still have 'rendering stuck' issue, and finally drop it to 0 and problem went away.

bitman wrote on 9/22/2017, 4:48 AM

I am not convinced it is a timing issue with my memory, first I have a rock stable system I build myself with the finest components running for more than 3 years. No overclock, memory is Corsair DDR3 16GB (1600-999) Vengeance Pro, power supply is Corsair 1200W AX1200i Professional series and the mobo is Gigabyte G1.Sniper 5 (rev. 1.0). I was not even running the memory at it's rated XMP-1600 until recently, but to the lower, standard default speed. And I have the same issues with default, non xmp memory speed. I did not try to under clock my memory, I may try this some time.

Kinvermark wrote on 9/22/2017, 7:39 AM

Just curious, but do Intel HEVC renders give you "random rendering issues like frame hiccup, frame flash back and forth" ?

OldSmoke wrote on 9/22/2017, 10:20 AM

I am not convinced it is a timing issue with my memory, first I have a rock stable system I build myself with the finest components running for more than 3 years. No overclock, memory is Corsair DDR3 16GB (1600-999) Vengeance Pro, power supply is Corsair 1200W AX1200i Professional series and the mobo is Gigabyte G1.Sniper 5 (rev. 1.0). I was not even running the memory at it's rated XMP-1600 until recently, but to the lower, standard default speed. And I have the same issues with default, non xmp memory speed. I did not try to under clock my memory, I may try this some time.

Maybe this needs a more methodical approach. The 4790k has 16PCIe lanes and all are used by the graphic card. Depending on the motherboard, there might be some resource sharing going on with USB ports or SATA controllers. Even my 3930k with 40PCIe lanes would share one of the PCIe x4 sockets with some of the USB3 ports if I fully populate all PCIe sockets, which I don't.

Maybe a spreadsheet that users can share and input their system data may reveal which systems have issues with the default 200MB setting but it would require system data that are usually only found in the manual or BIOS settings.

bitman wrote on 9/22/2017, 12:17 PM

Just curious, but do Intel HEVC renders give you "random rendering issues like frame hiccup, frame flash back and forth" ?

@Kinvermark

I just did an intel HVEC render with default 200MB setting (with nvidea in preference), but still horrible renders. According GPU-Z Nvidea was barely used (I suppose just for the screen updates), Intel GPU was also not used, so it seems a pure sw render which is not abnormal as i7-4790K Haswell generation does not support HW acceleration of HVEC yet. Render times are 3x longer than magix nvec.

Note that Intel HVEC render with dyn mem=0 was OK