MP 7,1 What GPU tests would you guys like to see - Mac Pro 6800XT, w5700x

[rondocap]

Wanted some input on consistent test methodologies to gauge the performance of these gpus, to compare the 6800xt and if multiple gpu setups are worth it.
1. My primary use is fcpx and Resolve - Red Raw, Canon R5 raw/4k60, S1H prores raw, a7siii 10 bit

2. Mac Pro: 24 core, 96gb ram - possible gpu setups:

Two W5700x mpx modules
reference 6800 xt with 11.4 beta
vega 64 reference

at the same time I can fit the dual w5700x modules and 1 pcie gpu like the 6800xt, for a total of 3. Fcpx seems to use all 3 and spread workload amongst them in many cases.

3. iMac Pro base vega 56; but with egpu setup (already tested 6800xt in this and it’s stable and very fast, but I’d like test suggestions to quantify that)


also have a Threadripper 3960x PC with dual 3090s for resolve, but likely no need to test this as the performance is on another level entirely not approchable by Mac Pro hardware currently, and it’s not mac os.

 

[elfamosisimoJON]

Bruce X benchmark FCPX please, also export 10bit h265 in FCPX

also if possible, playback of 10 bit 4.2.2 HEVC files like que ones in the canon r5 or a7s iii

 

[rondocap]

'.

Bruce X benchmark FCPX please, also export 10bit h265 in FCPX

also if possible, playback of 10 bit 4.2.2 HEVC files like que ones in the canon r5 or a7s iii

Sony plays back fine, at highest bitrates - but canon r5 still choppy. You need an m1 Mac to decode that

 

[elfamosisimoJON]

'.

Sony plays back fine, at highest bitrates - but canon r5 still choppy. You need an m1 Mac to decode that

damn, i tough RDNA 2 would have hardware decoding for this.

 

[OkiRun]

It's understanding that FCPX takes load off CPU mainly.
Also, FCPX includes streamlined and GPU optimized demosaicing algorithm. So optimized for all Mac Pro 7,1 GPU offered through order.
So am wondering, if will be any important difference between GPU performance and tests you run with PRO RES and FCPX.
Are you expecting with pre-theory?
And Afterburner Card helps with CPU so have it or not - not much difference in GPU numbers?

 

[rondocap]

Bruce X, 5k Pro res 4444 HQ seems to be about 5.59 seconds for the best time I got it with the 6800 xt

 

[rondocap]

It's understanding that FCPX takes load off CPU mainly.
Also, FCPX includes streamlined and GPU optimized demosaicing algorithm. So optimized for all Mac Pro 7,1 GPU offered through order.
So am wondering, if will be any important difference between GPU performance and tests you run with PRO RES and FCPX.
Are you expecting with pre-theory?
And Afterburner Card helps with CPU so have it or not - not much difference in GPU numbers?

I think afterburner will only really help with doing a few streams of Pro Res Raw, - then it takes the load off the CPU. Most video codecs in FCPX are very GPU heavy, like red raw, using metal, etc - but pro res raw still uses a lot of CPU power too, which can be reduced by the afterburner card, basically its main use.

But I have a 24 core, and doing just one stream of pro res raw seems to be handled fine, only if doing more I think the afterburner is worth it.

 

[elfamosisimoJON]

Bruce X, 5k Pro res 4444 HQ seems to be about 5.59 seconds for the best time I got it with the 6800 xt

i think is a new record, my W5700X does it in 9 sec, great!

 

[rondocap]

i think is a new record, my W5700X does it in 9 sec, great!

By any chance do you know the Bruce X score of the Pro Vega II?

The 6800xt seems pretty fast - for some reason when it is in an eGPU on an iMac Pro, it only gets 8.25 seconds or so in Bruce X, but much faster when directly inside the Mac Pro.

 

[elfamosisimoJON]

By any chance do you know the Bruce X score of the Pro Vega II?

The 6800xt seems pretty fast - for some reason when it is in an eGPU on an iMac Pro, it only gets 8.25 seconds or so in Bruce X, but much faster when directly inside the Mac Pro.

I think Vega II does it in 8 secs, but I’m not sure. That’s the problem with eGPU, it doesn’t take full advantage of the card, It was a good experiment for some time, but i will never replace internal and now with apple silicon, it’s hard to beat the latency of the new systems.

 

[rondocap]

I think Vega II does it in 8 secs, but I’m not sure. That’s the problem with eGPU, it doesn’t take full advantage of the card, It was a good experiment for some time, but i will never replace internal and now with apple silicon, it’s hard to beat the latency of the new systems.

So I tested out my dual w5700x in Bruce x, seems like it gets 5.92 so very fast too when both work together

 

[flowrider]

I would just love to buy an RX6800XT for a reasonable (list) Price

Lou

 

[DFP1989]

Bruce X, 5k Pro res 4444 HQ seems to be about 5.59 seconds for the best time I got it with the 6800 xt

I get 6.75 secs on my 16-core with the 6900 XT. The CPU is maxed out for the test, GPU only about 30%.

Displays connected to both my W5700X and 6900 XT (main 4K on 6900 XT DisplayPort, 1080p display on W5700X HDMI) makes no difference, zero usage of the W5700X.

 

[rondocap]

I get 6.75 secs on my 16-core with the 6900 XT. The CPU is maxed out for the test, GPU only about 30%.

Displays connected to both my W5700X and 6900 XT (main 4K on 6900 XT DisplayPort, 1080p display on W5700X HDMI) makes no difference, zero usage of the W5700X.

Mine is the 24 core, maybe that accounts for my 6800xt time being a little faster.

surprisingly dual w5700x is 5.92, not bad - one w5700x usually around 9 seconds

 

[rondocap]

Davinci Resolve does not seem to be playing nicely yet with the 6800xt, but that's to be expected since it is beta. Nit shows up fine, but performance is really slow vs using a slower older GPU. FCPX does seem at least more responsive to it - but I think we need more optimizations before we can judge the performance for video editing on these new gpus.

Stuff like Geekbench 5 metal numbers are impressive though, 145,000+ so that is promising.