Blackmagic Forum

I just recently picked up a 2013 8 core Mac Pro. D700 graphics cards and 32gb ram.

I started running F65 footage through it (no nodes) and noticed that after about 3TB of footage to prores 422 HQ I started seeing an artifacting across the viewer in the delivery pane of resolve 10.

I let the render finished and noticed the same artifacting (across the center of the frame) was also on the transcodes.

I rebooted and all was well for about another 10 minutes. I then noticed that if i went back into the COLOR window of resolve I would see it there as well.

RAW footage was fine when viewed in sony raw viewer.

I changed my render destination (originally rendering FROM and TO my pegasus2 R6 RAID) from my Pegasus RAID to my 1TB internal flash.. the problem then didn't present itself again and I was able to get through another 1TB of footage.

Yesterday I was rendering some 4k HD 5:1 RED EPIC footage and the problem again started.. Sometimes I'll see it in the color window before I even start rendering any footage. Very similar to when your VRAM is full but not as bad.

I rendered out a 30 minute render and the problem didn't happen but then started again later in the day.

I can grab a screen shot i'm sure when I get back to my system.

system has been tried with AUTO GPU and OPEN CL as well as all the past 3 version of resolve 10.

I'm a bit stumped here. could this be a faulty GPU?

Can you post a frame of the problem image?

What does the artifact look like? This sounds very similar to an issue we saw on a new Mac Pro, that would manifest as one or two pixel tall flickering horizontal lines, various colors, across some or all of the width of the frame. It would show up intermittently in the viewer, on external monitoring, and in renders.

Probably was a faulty GPU. It went away after Apple exchanged the unit for us. Bad news is, these machines are in such short supply the exchange took about two and a half weeks. (This was about a month ago; maybe it's not as bad now.)

I'm also having similar problems with resolve since the last system software update.
In 10.9.1 using resolve 10.1.2 with Desktop Video 9.9.3 the problem does not exist.
Updating to 10.9.2 the problem manifests itself. Reverting to 10.9.1 the problem goes away.

The error is present in the exported files, in the output from the decklink while working and in the viewer on the resolve interface.
The problem does not exist in any other software that I have tested, and that uses the decklink output like Blackmagic MediaExpress or Final Cut Pro.


Also with the update to 10.9.2 the GPU that was identified as a D300 inside resolve is now identified as a AMD Radeon Pitcairn PRO Prototype Compute Engine...

System Specs
Mac Pro 3,1
OSX 10.9.2
8GB RAM
Radeon R9 270X 4GB
Resolve 10.1.2
Desktop Video 9.9.3
Decklink HD Extreme 3D+

Interesting.

Is this a hardware issue ? Software issue ?

What Chris and americano are explaining is what I'm seeing as well.

Here's what I'm seeing.

it jumps around as the clip plays. This is in the Render and also apparent in resolve and SDI out.

Just did a complete clean install and will report back tomorrow.

It could be caused by hardware, but only manifest itself with software that accesses that hardware in certain ways. For instance, if it's caused by some bad VRAM somewhere, it would only show up when software tried to use that specific region of VRAM, and those sort of access patterns could hypothetically change between app and OS versions, and could be different enough between different apps that you'd see a problem regularly in one app and seldom or never in another.

There logically does seem to be some hardware component to this, as our previous Mac Pro showed the issue quite regularly under OS X 10.9.1 and with Resolve 10.1.1, even after I totally wiped and re-installed the system and Resolve, while the replacement unit with the same software versions does not show the issue. On the other hand, if it is hardware it's odd there are so many people seeing the issue within the small population of this forum. How much bad VRAM can there be out there?

Yup.

Did a clean install of mavericks last night. Got through about 1 hour of Alexa footage before the problem started again.

Reset the PRAM today. Transcodes went from 115fps to 85fps. Interesting right ?

Could this be bad ram or is it VRAM for sure ?

Do you have AppNap disabled for Resolve? I'm still waiting for my nMP to arrive (next week!), but I and many others suffered problems on other Mavericks platforms that showed up after 30-60 minutes of rendering and App Nap was to blame.

Hey Michael,

I did disable it..

It wasn't something that I would see after a period of rendering. Sometimes it happens immediately when I start the render sometimes not until 5-10 minutes in.

Matthew,

Are you on 10.9.2 now? Did this problem exist with 10.9.1?

Don't know what Matthew is seeing, but when I had the unit that exhibited this problem it showed it under 10.9.1. The replacement unit didn't show it under 10.9.1 and still doesn't show it under 10.9.2.

Johnny S,

I was seeing it in both 10.9.1 and 10.9.2.

Hello,

100% same problem. With MacPro 12core, 64ram, D700.
Did you find a solution?

Robin

First thing that occurs to me with this kind of intermittency is thermal.

One would think that cold-solder assembly issues were a thing of the past, but there is an electronic engineering axiom that 90% of circuit failures are power-supply related, and the other 90% are bad connectors. The rest are operator errors.

Oh and the other observation is that we used to get this kind of failure a lot with Final Touch/COLOR, but mostly with SAN-type media distribution. It was a momentary disruption that caused the renderer to just fill in the blanks with gack.

jPo

Hello,

It could be thermal. But in my case, I managed to reproduce the problem just 5 minutes after the computer starts. And I tested this with render on internal disk and also external.

Best
Robin

Hello,

I managed to create 3 videos capture screen to show you the problem.
Tested on R3D files 4K from Red One. Resolve 10.1.3.
Everything is downloadable with this wetransfer link :

http://we.tl/1Vlqixbq1P

You will see in Demo1 and 2 that the problem appears in Color page, Edit
Page and Render page and Render Result. In Demo 3, you will see the problem
in Color page but not in the render. I didn't change anything between these Demo
1,2,3. The problem is that's not constant. Sometimes it appends sometimes
not.

Here is an other example of the problem that could show a GPU problem : http://we.tl/feP7gEEHG8

JPOwens wrote:Oh and the other observation is that we used to get this kind of failure a lot with Final Touch/COLOR, but mostly with SAN-type media distribution. It was a momentary disruption that caused the renderer to just fill in the blanks with gack.

Pretty sure this one isn't I/O related, as we were able to reproduce it with 1080p ProRes 4444 clips on the internal SSD, which provides over an order of magnitude more bandwidth than is required for that format.

I recommend that anyone seeing this problem who hasn't already done so immediately begin the process of getting Apple to replace the unit. That solved the problem for us after nothing else (full system re-install, upgrading to 10.9.2, trying lots of different settings in Resolve, testing with lots of different source/render formats, testing with lots of different drives, etc.) did.

Hey guys.

Got my 8 core switched out to a 12 core with D700s.. haven't seen the problem since. bad hardware all the way.

We are seeing this too, with 4k red (scarlet) files on latest resolve on mac pros with d700 cards. And to make matters worse, project export fails too so it was a pita to move it to an old mac. Red-cine was able to export the same file without these render issues. SDK issue with new mac-pros and resolve/red combination? Image is a crop of a 2k prores render (even tried at lower de-bayer quality)

Marcus

Follow-up. Moved project to an older Mac Pro also running 10.1.4, renders without issue. Footage coming from same source (SAN)

Also, on a second new Mac Pro running 10.1.3 also seems to render fine. Will upgrade this other mac to 10.1.4 and test to see if it's software issue or this specific mac that is the culprit

Hi everyone I am also having the same issue. Dont bother talking to Apple Support they will just direct you to the SDK developers site. Here is what I posted to apple. Will see what they say.
Summary:
DaVinci resolve has a issue when playing or rendering files while using OpenCL. Similar issues also occur in Adobe Primer.
COmputer Specs are as follows:
Model Identifier: MacPro6,1
Processor Name: 8-Core Intel Xeon E5
Processor Speed: 3 GHz
Graphics: AMD FirePro D700
OS: 10.9.2 (13C64)
DaVinci Resolve 10.1.4.009
Image playback does not seem to be codec dependant however it does seem to happen more using Red .r3d and prores.mov files than say a Phantom .cine file.
Fault appears as a Row of horizontal coloured pixels showing up for only a single frame and is only a pixel tall and takes up only a portion of the frame horizontaly. This can occur many times and in different places while playing the source video. The event is some what random and does not seem to be heat dependant.
Image corruption is seen in GUI of software while playing video and also on external DSDI connected monitor. After render of the video file the fault is backed into video.

Steps to Reproduce:
1. Boot up system.
2. OpenDavinci Resolve
3.Load some video files.
4. Play or scrub threw video glitch will be seen.
5. Rendering of video files also shows the same issue.

Expected Results:
Glitch tends to come and go often and in some cases will only go away after reinstall. Expected result should be no video corruption.

Actual Results:
Same as Expected results.

Version:
Model Identifier: MacPro6,1
Processor Name: 8-Core Intel Xeon E5
Processor Speed: 3 GHz
Graphics: AMD FirePro D700
OS: 10.9.2 (13C64)
DaVinci Resolve 10.1.4.009


Notes:
Originally this was reported to Apple Support and they said they could not help with me this issue.
Case number for references is 609#######
This bug is occurring to other users of this software and other software as well. I am trying to make sure this is not a Mac Pro hardware issue as some people are reporting it as such.
"This sounds very similar to an issue we saw on a new Mac Pro, that would manifest as one or two pixel tall flickering horizontal lines, various colors, across some or all of the width of the frame. It would show up intermittently in the viewer, on external monitoring, and in renders.

Probably was a faulty GPU. It went away after Apple exchanged the unit for us. Bad news is, these machines are in such short supply the exchange took about two and a half weeks. (This was about a month ago; maybe it's not as bad now.)"




Configuration:
Model Identifier: MacPro6,1

OS: 10.9.2 (13C64)
DaVinci Resolve 10.1.4.009
Name: DeckLink HD Extreme 3D+
Type: Video
Slot: Thunderbolt@159,0,0



Hopefully Apple will look into this more as it's been awhile with this problem.

I've spoken to Apple and since I have 1/2 of my 1TB drive free they are telling me to partition my hard drive, do a clean install on the free portion, then install the software and try to replicate the problem. If the problem persists then bring the computer in for diagnostics. Do you guys think a tech will be able to diagnose a D700 etc. hardware issue?

Hello,

I hope the problem has been solved. Apple changed my GPU D700. But I need more tests to be sure.

Robin

Hello,

Maybe I spoke to fast... the problem isn't solved even with new graphic cards.

Robin

cinerard wrote:Hello,

Maybe I spoke to fast... the problem isn't solved even with new graphic cards.

Robin

Are you running 10.9.3?

Hello,

No the problem exists also on 10.9.2. I'm still on 10.9.2.

Robin

Same render bugs here with a new Mac Pro on dual D700. Little green or red lines in rendered footage. Can anybody from blackmagic tell if it is a apple problem oder an resolve bug? For now the mac pro is nothing more than a machine for conforming and a very expensive paper weight.

BlackMagic said this isn't a bug of Resolve. Adobe Premiere has the same bug... it looks more like a openCL bug.

Have you tried rendering using the same footage with After Effects, Premiere or FCP?

If you did the same kinds of transforms and render to the same kind of file as Resolve, that could tell you whether the glitch is from Resolve or the GPU.

Hello,

I didn't try that. But I know the problem exists also on Adobe Premiere.
I think it's not a BM issue only, but their job is also to find a solution with Apple.

Robin

I can confirm that latest 10.9.4 Beta 3 on MacPro 6.1 D700 did not help to solve the problem. Again green and black lines in rendered footage. Especially in footage which is bigger than HD.

This user reported a specific beta that works for them.
viewtopic.php?f=11&t=23490

Using the same Beta (13E19) fixes the crashes. That is right! But not the Render Bug!!

Any resolution to this issue? Has a hardware swap been a fix for some? I'm having the same problem and not sure where to go from here..

pixelworker wrote:Any resolution to this issue? Has a hardware swap been a fix for some? I'm having the same problem and not sure where to go from here..

Yes, hardware swap immediately fixed this problem for me. No recurrence in the ~6 months since. This makes me suspect that the people for whom a hardware swap didn't fix this just happened to get unlucky twice.

Chris Kenny wrote:

pixelworker wrote:Any resolution to this issue? Has a hardware swap been a fix for some? I'm having the same problem and not sure where to go from here..

Yes, hardware swap immediately fixed this problem for me. No recurrence in the ~6 months since. This makes me suspect that the people for whom a hardware swap didn't fix this just happened to get unlucky twice.

Well, I made a test, copied a video to my internal SSD (full now) and rendered a long movie.. No artifacts, I moved the source to an USB3 disk, and tried again, random artifacts. Unfortunately that proves nothing, as sometimes I can render 10x a file with no artifacts.. but I would have seen something on the SSD..

so my question is : people who are seing artifacts, are they using RAID ? USB3 ? TB ? TB2 ?

Thx

Hello Pierre-Emannuel,

On my side (since march), the bug exists even if I render on my internal SSD. I tested everything.

Best
Robin

I also have the issue with Neat Video : the curious thing is that it happens EVEN if I set to CPU only
I don't get it ..

PierreEmmanuel wrote:Well, I made a test, copied a video to my internal SSD (full now) and rendered a long movie.. No artifacts, I moved the source to an USB3 disk, and tried again, random artifacts. Unfortunately that proves nothing, as sometimes I can render 10x a file with no artifacts.. but I would have seen something on the SSD..

so my question is : people who are seing artifacts, are they using RAID ? USB3 ? TB ? TB2 ?

On the unit that had this problem I saw it with footage both from drives in a USB 3 drive dock and from an SAS RAID attached via an SAS controller in a Thunderbolt PCIe enclosure. So, I don't think it's related to the storage interface.

PierreEmmanuel wrote:I also have the issue with Neat Video : the curious thing is that it happens EVEN if I set to CPU only
I don't get it ..

If you're talking about the Neat Video Resolve plugin, then even if Neat Video is processing on the CPU, everything still ends up flowing through Resolve's GPU pipeline at some point.

That said, I'm not sure we've conclusively demonstrated this issue is GPU related. That's certainly the way I'd bet, but in my case Apple swapped out the entire computer, so conceivably the issue could have been caused by some other hardware.

Chris Kenny wrote:

PierreEmmanuel wrote:I also have the issue with Neat Video : the curious thing is that it happens EVEN if I set to CPU only
I don't get it ..

If you're talking about the Neat Video Resolve plugin, then even if Neat Video is processing on the CPU, everything still ends up flowing through Resolve's GPU pipeline at some point.

That said, I'm not sure we've conclusively demonstrated this issue is GPU related. That's certainly the way I'd bet, but in my case Apple swapped out the entire computer, so conceivably the issue could have been caused by some other hardware.

I'm talking about the Preview window in Neat : I don't see the artifact in Resolve, I see it in Neat, even if disabling GPU.. well, ok I agree, at one point the image is displayed, and so the GPU is used.

If I click on the image to show the unfiltered image and the the filtered one, I see the artifacts on the filtered one and not the other (Resolve's output), with GPU disabled.

I realized that, after updating to 10.9.4 I did not update Cuda.. there is a 6.0.51, I updated this night, and re-rendered my projects : for the moment, no artifact. Let`s cross fingers.

Anyone who has solved the issue of the RGB-artifacts (lines) without handing in the Mac Pro for a hardware swap? It shows up at least every time i turn on Optical Flow in Davinci Resolve.




Specs: Mac Pro late 2013 6-Core D700

Rendered a >3h movie, with :
- 30fps->24
- Optical Flow
- best motion estimation
- Neat Video
- 4k->HD

between 3 and 5 fps

Checked this morning : not a single artifacts. It does not prove anything, as the artifact is random, but it loos like updating to Cuda 6,0,51 made the thing.

Again, I could be wrong, as I already got the impression that my problem was solved, but never for a so long movie.

danielsenphoto wrote:Anyone who has solved the issue of the RGB-artifacts (lines) without handing in the Mac Pro for a hardware swap? It shows up at least every time i turn on Optical Flow in Davinci Resolve.




Specs: Mac Pro late 2013 6-Core D700

Are you on 10.9.4 OS?

Peter Chamberlain wrote:Are you on 10.9.4 OS?

Yes. Also just upgraded to Resolve 11 Beta (licensed). Seems like people with v10 have experienced the same problem reading this thread.

I confirm that the problem seems solved : 4h12 rendered with HEAVY processing and no artifact at all.
What I did : 10.9.4 AND latest cuda.

PierreEmmanuel wrote:I confirm that the problem seems solved : 4h12 rendered with HEAVY processing and no artifact at all.
What I did : 10.9.4 AND latest cuda.

I'm running the latest OS 10.9.4. Isn't Cuda irrelevant when running the dual AMD D700 that are in the nMP?

danielsenphoto wrote:

PierreEmmanuel wrote:I confirm that the problem seems solved : 4h12 rendered with HEAVY processing and no artifact at all.
What I did : 10.9.4 AND latest cuda.

I'm running the latest OS 10.9.4. Isn't Cuda irrelevant when running the dual AMD D700 that are in the nMP?

yes, you are right, cuda is for nvidia, but this is the only changes I made .. mmm
Oh, I read on a premier forum that installing Cuda was a bad idea on a OpenCL machine, it created some issues with some Premiere users, that was solved by removing Cuda, so.. even if not used.. does it updates other libraries also used by Open CL ? no idea, but it looks like its not as innocent as it may look.

Can't say more than what happened (until I see artifacts again )

Please see : viewtopic.php?f=3&t=20134#p125128