Contribute
Register

[SUCCESS] Gigabyte Designare Z390 (Thunderbolt 3) + i7-9700K + AMD RX 580

Joined
Apr 19, 2018
Messages
54
Motherboard
Giga Z390 Designare
CPU
i7-9700k
Graphics
RX 580
Mac
  1. MacBook Pro
Update on HEVC encoding with iGPU QuickSync
1min 10 seconds in Resolve 16
38 seconds in FCPX
I see, thanks for the test

Can you do a comparison between iMac 19,1 and iMacPro 1,1 with the same software since we are trying to figure out which SMBIOS configuration yields the fastest encoding result?

As for my result above, according to the monitor report, AME is fulling utilizing my RX 580 on both 19,1 and 1,1. This was the setting i had used.
Screen Shot 2020-06-22 at 4.06.52 PM.png


HEVC encoding in FCPX (Again, note that I am using RX 580)
iMac 19,1 - 46 sec
iMacPro 1,1 - 49 sec

Screen Shot 2020-06-22 at 5.49.57 PM.png
 
Last edited:
Joined
Aug 11, 2011
Messages
92
Motherboard
Gigabyte Z490 Vision D
CPU
i9-10900K
Graphics
RX 6800 XT
Mac
  1. iMac
  2. MacBook Air
  3. MacBook Pro
I see, thanks for the test

Can you do a comparison between iMac 19,1 and iMacPro 1,1 with the same software since we are trying to figure out which SMBIOS configuration yields the fastest encoding result?

This has been a very interesting experiment. I changed my SMBIOS to iMacPro1,1 and here is what I am getting.

iMacPro1,1
H.264 - 32s in FCPX
HEVC - 50s in FCPX
H.264 - 35s in Davinci Resolve
HEVC - 48s in Davinci Resolve

Now what is baffling me is the result I got when I went back to iMac19,1 and emptied the render cache, reran the test.
iMac19,1
H.264 - 46 in FCPX
HEVC - 50s in FCPX
H.264 - 45s in Davinci Resolve
HEVC - 50s in Davinci Resolve

Somehow I was not able to reproduce the time I was getting this morning. My only guess is that somehow the morning run had background render on, or I must have set the frame rate wrong. Ran several times and this is pretty much the result I got. What is also puzzling is my H.264 took longer while HEVC shorter time to export than the last test I ran earlier this after. I have to see what kind of result hylkepylke will get with HEVC decoding, but seems like there is not a huge difference in HEVC, but in H.264, the result is not flopped.
 

CaseySJ

Moderator
Joined
Nov 11, 2018
Messages
19,511
Motherboard
Asus ProArt Z690-Creator
CPU
i7-12700K
Graphics
RX 6800 XT
Mac
  1. MacBook Air
  2. MacBook Pro
  3. Mac Pro
Classic Mac
  1. Quadra
Mobile Phone
  1. iOS
** The end of Hackintosh as we know it **

Today's momentous announcement of the transition to Apple Silicon means that Macintosh will eventually become a completely closed and proprietary system.

But this won't happen right away. There is plenty of time. The chart below is one way of forecasting the future. It is not an official chart (obviously), but it does show three inevitable phases for Hackintosh, in green, orange, and red.


End of Hackintosh v2.png
 
Last edited:
Joined
Oct 20, 2014
Messages
31
Motherboard
HP 8470P
CPU
I5-3380M
Graphics
HD 4000
Mac
Classic Mac
Mobile Phone
M2 drives, PCI lanes and SATA ports.

Hi there. So, I am planning a dual boot Hack along the lines of this build and had a question about M2 drives. I've been delving through the gigabyte forums but everyone seems to have a different opinion. If I run Win in the top M2 and OSX in the bottom (both PCI SSD's) I lose SATA ports 0, 4 and 5. This I can handle. With PCIx4 set to PCH the GPU still has the full 16 lanes but I'm trying to figure out what I have left for PCI cards. Is it just 4 lanes shared across the x8, x4, two x1's and the M2's. As the M2's are each different OS they'll only be running one at a time so not so much of a bandwidth issue but I wonder if there are still knock ons to do with addressing.

Basically I need to be confident that I can get away with 2x PCI M2's, Fenvi wifi, MOTU PCIe424 (connected to 2 x 24IO boards), Genenric Firewire 400/800 card connected to Fireface 400. They're all x1 cards and the MOTU won't be running at the same time as the fireface. If the TB3 > 2 > 1 conversion works I can scrap the generic firewire card. Don't want to bank on that though. In terms of what will be operational at any given time I think 3 lanes to the active M2 would allow full speed operation but that leaves 1 lane shared across audio, FW and wifi/bluetooth. Having written all this down it doesn't seem like a lot, I've crammed a lot more into a board before but NVME is new to me and I don't want to shoot myself in the foot before I begin.

Many thanks,

Jim
 
Last edited:
Joined
Sep 10, 2012
Messages
113
Motherboard
Gigabyte Z390 Designare
CPU
i9-9900KS
Graphics
RX Vega 64
Ok, let's try and make this more readable.

SETUP
This is test is to see how different SMBIOS with iGPU either enabled, disabled or in headless mode have an effect on render speeds.
iMacPro1,1 = iGPU disabled, only dGPU (Vega 64)
iMac19,1 = iGPU enabled, headless (using OC 0.5.9, see code below), in combination with dGPU (Vega 64)
Code:
<key>DeviceProperties</key>
    <dict>
        <key>Add</key>
        <dict>
            <key>PciRoot(0x0)/Pci(0x2,0x0)</key>
            <dict>
                <key>AAPL,ig-platform-id</key>
                <data>AwCYPg==</data>
                <key>AAPL,slot-name</key>
                <string>Internal@0,2,0</string>
                <key>device_type</key>
                <string>Display controller</string>
                <key>framebuffer-patch-enable</key>
                <data>AQAAAA==</data>
                <key>model</key>
                <string>Intel UHD Graphics 630</string>
            </dict>

METHOD
Download the footage as described in this post by @verendus. Make a new project in DR 16.2.3 (unfortunately I don't have FCPX), create a new timeline (3840x2160, 60 fps), drop the clip in there, and export out, either to QuickTime (MOV) or MP4, and time the export (DR shows how long the export took after it's done).

My machine: i9-9900KS, 64GB RAM, Vega 64 8GB

RESULTS
H.264 H.265 (HEVC)
iMacPro1,1 0:34 0:57
iMac19,1 0:38 1:12


DISCUSSION
I noticed that GPU utilization (as measured by looking at Activity Monitor) is near 100% on iMacPro1,1; but only ~50% on iMac19,1 during rendering, which makes sense I guess? I'm not sure how I can see QSV utilization other than looking at CPU utilization, which was ~200% for both iMacPro1,1 and iMac19,1.
As this is why I built this setup, I'd love to dig deeper, and I'm willing to do a lot more testing if needed. Hope this contributes in some way!
 
Last edited:
Joined
Jun 13, 2012
Messages
117
Motherboard
Gigabyte Z390 Designare
CPU
i9-9900K
Graphics
Vega 64 | R9 270X
Mac
  1. MacBook Air
  2. MacBook Pro
  3. Mac Pro
Mobile Phone
  1. Other
Thanks for taking the time to run the test. Just ran the same test with the same setting. I am on 32GB, iMac19,1, iGPU headless (0x3E980003), DR16.2.3.

21 seconds. That is roughly 60% (or 40% increase in speed) of using dGPU. The only difference would be my Radeon VII vs your Vega 64, but as I said, the majority of encoding work in our case is carried out by the CPU (QuickSync vs intel cores). I'd be interested in seeing how T2 chip in new MBP or iMac might fare in this test.

For those who use FCPX, the export was 22 seconds with the same setting.

So the moral of the story: If you can, enable iGPU headless mode and use iMac19,1 to maximize your export. QuickSync DOES make a big difference in encoding performance in DR and FCPX.

I also wanted to give your test a shot:
I'm running 19,1 in headless mode
vega64
and an r9 270x
64mb ram
i9 9900k

compressor 4k H264 = 48 secs
DR 4k H264 = 25 secs
FCPx 4k H264 = 27 secs

so not bad!
 
Joined
Sep 10, 2012
Messages
113
Motherboard
Gigabyte Z390 Designare
CPU
i9-9900KS
Graphics
RX Vega 64
...
DR 4k H264 = 25 secs
...
so not bad!
Wait, that's a great score! Are you running DR Studio by chance, so both GPU's get used? Can you try clearing DR cache, and double checking the frame rate? @verendus got 22 seconds on the first run but got 45 after double checking. Otherwise, we have to dig deeper to figure out how you got such a great score!
 
Top