Haha totally. I have an M1 Air with 16GB RAM and I love it, but it’s far from the 100% replacement for my 16-inch Pro that many M1 fans would have others believe.
@rondocap your tests have encouraged me to give a dual W5700X setup a try.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
MP 7,1 W6800x Duo benchmarks vs Vega ii duo, 2x w5700x, 2x 6900xt
- Thread starter rondocap
- Start date
- Sort by reaction score
@rondocap your tests have encouraged me to give a dual W5700X setup a try.
Agreed. I gave m1 air base. 16 fully maxd. And 7.1 base with 5700x. I'm going to put in 2nd 5700x and upgrade cpu to 24 core myself.
How would fcpx contend with a 7,1 having a w6800x Duo and Vega II Duo in one machine? I wonder if there would be significant bottleneck. Somehow I think that one of them wouldn't even be accessed for the job....?
I just swapped one of my Vega II duos with a W6800X duo; only the following improvements in fcpx can be expected:
-some decrease in system power usage;
-improvement in 8k Redcode processing.
However:
No improvements in 10 bit HEVC processing are available. To make matters worse, 4k ProRes exporting appears to have slowed down. So the extra horsepower is of no use at all, at least to me (since I do not work with 8k Redcode material).
Thus, I am returing the W6800X duo.🥲
-some decrease in system power usage;
-improvement in 8k Redcode processing.
However:
No improvements in 10 bit HEVC processing are available. To make matters worse, 4k ProRes exporting appears to have slowed down. So the extra horsepower is of no use at all, at least to me (since I do not work with 8k Redcode material).
Thus, I am returing the W6800X duo.🥲
Based on my experience with mismatched cards, you can select one to be used in the FCP Preferences, and the other sits idle.
Yeah in fcp I wouldn’t recommend using two different gpus , I’m using two of the w6800x duos and fcp uses them, but doesn’t scale perfectly so one is the best option
Wonder if the W6900x would be a better option for your workflow? Of course you need to stomach the $$$.
I saw an improvement with 4K pro res raw with the W6800x Duo vs 2 Vega ii. Of course with afterburner it's the fastest
You really need two or more of the same for FCP, if you have different GPUs in there it can even give you slower results
I do not have bootcamp; but I may download "Tomb Raider" (2014 reboot) and Deus Ex Human Revolution, just to see how those perform. Big maybe though; hard to justify all that download time when I have a PC gaming rig.
What about speed of the background render in fcpx? Fcpx asked for a choice of which gpu to use for background rendering of the timeline. Would a duo be significantly faster than two 6800xt’s for this task? I’m upgrading from two Radeon VII’s and trying to decide if it is worth going for the mpx module.
I’m keen to know about background rendering performance too. Any comparisons between w6800x / w6800 duo and w5700x / w5700x x2 will be very welcome.
Yeah that is a good question; I am not a FCP X user; but that would be cool to have one 6800 rendering while the other one is handling whatever you are doing in the timeline live (having a duo card that is).
It might be a situation where having a W6900X and a W800X Duo might be a good combo.
It might be a situation where having a W6900X and a W800X Duo might be a good combo.
I wish Apple had programmed FCPX to allow background rendering while playing the timeline or editing, but the background rendering stops while the timeline is playing. The timeline background render is the key to FCPX playback speed, though.
Oh that sucks, in that case then if FCPX can utilize all 4 GPUs if you have the W6800x DUO * 2; then that would probably be the best config. Hard to say.
Thank you so much for doing these tests I know it takes a ton of time and all of the mac users on here greatly appreciate you and your time.
yes in Resolve you can choose any GPU set you want, so in case of dual MPX card you can choose to use one for display only and second half of the gpu just for the compute. But in general I recommend to use separate GPU for displays and separate GPU/s for compute. It's not much obvious with smaller datasets, but if you will have full GPU memory thanks to driving your displays and even start let's say 3D rendering on same GPU performance would go significantly down.
Last edited:
Hi - we chatted on your video too, classical technology here.
What tests did you run the W6800x? Maybe I can run the same tests too, I have two W6800x Duos so I can compare those results vs yours to see if there is any difference.
Here are some 6k Red footage results with up to 4 GPUs:
Hi "classical technology"
I prepared for people who still asks for that Geekbench, Luxmark, Basemark, GXFBench, Resolve Candle, My Resolve real past project test, Resolve 8K RED, Resolve R5 8K footage, some gaming test, Blender 3D rendering and my beloved Houdini SideFX 3D viewport performance and rendering in Maxon's Redshift renderer. I have some shocking results. Tomorrow I will shoot some addons and I want to release it as soon as possible - hopefully Monday or Tuesday
Last edited:
That's great!Hi "classical technology"
For using the Redshift benchmark with 4x W6800x, I was able to get 1 min 46 sec, which is the fastest Mac score ever, and compares nicely to some pretty high end Nvidia GPUs too. I also many of the same tests as you too, Red 6k, 8k, candle, etc.
After testing both W6800x Duo and two of them for a total of 4 GPUs, I have some interesting findings, some that also compare to the Vega ii Duos. I find double GPUs more interesting than single, since you generally get nice gains with 2.
First, let's get a W6800x Duo vs two W6800x Duo out of the way below.
Export Red 8k to 4444 XQ - surprisingly the two single Vega ii got a time of 4 min 18 sec, which beat 4x W6800x Duo that had 4 min 47, just 5 seconds faster than a single W6800x Duo at 4.52.
Testing on a 28 core Mac Pro, Red set to both Debayer and Decompress in settings for best results
Remember, the W6000x GPUs, while newer, do have one difference vs the older Vega ii - slower memory speeds, even if they have the same amount of memory.
The (932mb) in parenthesis is just the write speed on the NVME drive, wanted to keep track of it to make sure it wasn't a bottleneck.
The two Duos are faster than 1 Duo, but often by only seconds for most exports. What's going on here? Possibly driver issues, because 4 GPUs are not scaling like they should.
Resolve nor Final Cut have had an update yet since these GPUs are so new, so we may need more drivers and software optimizations.
Oddly, the two solo Vega ii are winning in this test. They do have more PCIE bandwidth, at 16x each vs 16x for a single Duo, plus they have faster memory of 1TB vs 512 of the GDDR6. Could that be a factor? Possibly
I mean, the W6800x Duos are cheaper than the Vega ii Duo was by $800 or so, so I guess it isn't terrible, but it's very close in this test.
How about in red 6k instead? Here the results are more what we would expect, 4 W6800x do win at 7 min 51 seconds, but only slightly ahead of 2 W6800x. It Is a nice improvement over the 9 min 29 sec of the Vega ii duo, though - so perhaps 6k allows the GPUs to work a bit harder. I did notice the 8k had higher CPU utilization, reducing the impact of the GPU potentially.
So a nice gain, but the single Duo seems to be the sweet spot, not much gained with adding another.
How about in Final Cut Pro?
Here once again it seems like 6k does well, 8k was not a good result for these, losing to the Vega ii duo once again.
Notice how the W5700x Dual was basically the same as the W6800x Duo - while the 2x 6900xt and 4 W6800x were farther out ahead. The Vega ii seem to do better in 8k vs 6k.
The W6800x Duo, especially two of them for 4 GPUs, are really meant for 3D work and not video editing, like the below shows them scaling really well with 3D the more GPUs that are added.
So in conclusion, maybe future software updates may improve performance, but for now, it really isn't a huge difference vs the last gen GPUs for video work, but for 3D it certainly is.
I would say the best configurations for video editing in terms of pure performance:
1. Two W6900x for FCP and Resolve
2. two W6800x Duo for 4 GPU
3. two w6800x solo or a Vega ii duo, sort of close right now
4. two W5700x
In terms of price to performance:
1. two W5700x
2. W6800x Duo
3. Two vega ii or a vega ii duo if you get a cheap used price - this may even be close to #1 if prices fall a lot
The W6900x is a terrible value, even if it has great performance for a single GPU. As you can see, real world results don't scale as you would think due to various limitations with the software and other bottlenecks in the system, so while these are the fastest technically, I'd rather have 4 W6800x for cheaper as they can be more flexible than 2 W6900x.
I plan on doing more complex color grading tests to try to find areas where the new GPUs may be superior, like noise reduction, many nodes, etc. Right now my first trials of these tests have revealed results pretty close to the last Gen Vega ii duo, so I need to test more before coming to a conclusion on that.
What are your thoughts?
First, let's get a W6800x Duo vs two W6800x Duo out of the way below.
Export Red 8k to 4444 XQ - surprisingly the two single Vega ii got a time of 4 min 18 sec, which beat 4x W6800x Duo that had 4 min 47, just 5 seconds faster than a single W6800x Duo at 4.52.
Testing on a 28 core Mac Pro, Red set to both Debayer and Decompress in settings for best results
Remember, the W6000x GPUs, while newer, do have one difference vs the older Vega ii - slower memory speeds, even if they have the same amount of memory.
The (932mb) in parenthesis is just the write speed on the NVME drive, wanted to keep track of it to make sure it wasn't a bottleneck.
The two Duos are faster than 1 Duo, but often by only seconds for most exports. What's going on here? Possibly driver issues, because 4 GPUs are not scaling like they should.
Resolve nor Final Cut have had an update yet since these GPUs are so new, so we may need more drivers and software optimizations.
Oddly, the two solo Vega ii are winning in this test. They do have more PCIE bandwidth, at 16x each vs 16x for a single Duo, plus they have faster memory of 1TB vs 512 of the GDDR6. Could that be a factor? Possibly
I mean, the W6800x Duos are cheaper than the Vega ii Duo was by $800 or so, so I guess it isn't terrible, but it's very close in this test.
How about in red 6k instead? Here the results are more what we would expect, 4 W6800x do win at 7 min 51 seconds, but only slightly ahead of 2 W6800x. It Is a nice improvement over the 9 min 29 sec of the Vega ii duo, though - so perhaps 6k allows the GPUs to work a bit harder. I did notice the 8k had higher CPU utilization, reducing the impact of the GPU potentially.
So a nice gain, but the single Duo seems to be the sweet spot, not much gained with adding another.
How about in Final Cut Pro?
Here once again it seems like 6k does well, 8k was not a good result for these, losing to the Vega ii duo once again.
Notice how the W5700x Dual was basically the same as the W6800x Duo - while the 2x 6900xt and 4 W6800x were farther out ahead. The Vega ii seem to do better in 8k vs 6k.
The W6800x Duo, especially two of them for 4 GPUs, are really meant for 3D work and not video editing, like the below shows them scaling really well with 3D the more GPUs that are added.
So in conclusion, maybe future software updates may improve performance, but for now, it really isn't a huge difference vs the last gen GPUs for video work, but for 3D it certainly is.
I would say the best configurations for video editing in terms of pure performance:
1. Two W6900x for FCP and Resolve
2. two W6800x Duo for 4 GPU
3. two w6800x solo or a Vega ii duo, sort of close right now
4. two W5700x
In terms of price to performance:
1. two W5700x
2. W6800x Duo
3. Two vega ii or a vega ii duo if you get a cheap used price - this may even be close to #1 if prices fall a lot
The W6900x is a terrible value, even if it has great performance for a single GPU. As you can see, real world results don't scale as you would think due to various limitations with the software and other bottlenecks in the system, so while these are the fastest technically, I'd rather have 4 W6800x for cheaper as they can be more flexible than 2 W6900x.
I plan on doing more complex color grading tests to try to find areas where the new GPUs may be superior, like noise reduction, many nodes, etc. Right now my first trials of these tests have revealed results pretty close to the last Gen Vega ii duo, so I need to test more before coming to a conclusion on that.
What are your thoughts?
CommandPost let’s you get around several limitations like this.
Wonderful work thank you. I think these two statements sum up the uncertainty of where we are right now. Things may change in time with software/driver updates. Thanks again.
This is what I’m most keen to see. Background rendering of colour grading and other effects in Final Cut is where I’m hoping to see significant improvements.
I don't think Resolve or FCPX updates would change anything on performance of the 6xxx GPUs at all.