Graphics Test: Bruce X

[MoisesConstantino]

My Hackintosh with Catalina 10.15.3

ASUS ROG Maximus XI Hero (Wi-Fi) Z390
Intel Core i9-9900K
Corsair CMW32GX4M2C3200C16 Vengeance RGB PRO 32GB 64 total
2x AMD Vega Frontier 32gb Gddr5 (16gb each)
Sabrent 1TB Rocket Nvme PCIe 4.0 M.2 2280 Internal SSD Maximum Performance
Power Supply ASUS Rog Thor 1200
Water Cooler, the closest one that I have is this one Thermaltake Liquid Cooling Kit

My takes 6.03 seconds

 

[MattRobertson]

My Hackintosh with Catalina 10.15.3

ASUS ROG Maximus XI Hero (Wi-Fi) Z390
Intel Core i9-9900K
Corsair CMW32GX4M2C3200C16 Vengeance RGB PRO 32GB 64 total
2x AMD Vega Frontier 32gb Gddr5 (16gb each)
Sabrent 1TB Rocket Nvme PCIe 4.0 M.2 2280 Internal SSD Maximum Performance
Power Supply ASUS Rog Thor 1200
Water Cooler, the closest one that I have is this one Thermaltake Liquid Cooling Kit

My takes 6.03 seconds

That's extremely impressive! I feel like we're nearing the limits of how quickly this test can be completed haha. I'm struggling to imagine it can get much faster than this.

 

[oreoapple]

Sapphire Rx 580 8GB
i5-6400 16G DDR3
Catalina 10.15.3 Final Cut Pro X 10.4.8
ProRes 422
13s

 

[Jevens]

Same. Well, sort of, right now I am getting around 8-9secs until I repeat the test the 4th or 5th time, by then it's around 22secs. Why my initial results are faster than yours when I have a single Vega 64 and you have dual 5700 XT's I have no ideia why, shouldn't it be much faster for you?

The Activity Monitor GPU history window never shows more than 20-45% usage tho, the usage report there starts getting lower and lower with each test which explains the longer render times.. iStats, on the other hand, always reports 99% GPU usage, temps are fine (50C)... The usage is so low that the GPU fans dont even ramp up. All this is baffling to me. I wonder if this is a Hackintosh issue and doesn't happen on real Macs or if its the same.

It looks like macOS is using the GPU conservatively to keep noise/fans down, to an absurd degree.
View attachment 440514

Results change in LOT depending on FCP and macOS versions too, on some versions it was 22secs, now it's been 8 secs, the previous version was 12sec, over the last few years I've seen it go up and down...

DaVinci Resolve, on the other hand, shows quite a bit more GPU usage under Activity Monitor but still doesn't go above 65% give or take, with occasional higher spikes.

The only thing I can think of is trying the new Mac Pro SMBIOS, maybe since it has superior cooling macOS uses the GPU's more aggressively.

Unfortunately, I can also confirm this happens on my Hackintosh with Mac Pro 7,1 SMBIOS.

 

[D-One]

Unfortunately, I can also confirm this happens on my Hackintosh with Mac Pro 7,1 SMBIOS.

Thanks very much, this saves me the trouble of trying it. A 580 and Vega64 give pretty much the same result, baffling.

Yeah... I'm kind of giving up now while I dont have a bunch of work in FCP that will trap me in it, moving on to Da Vinci Resolve which seems much more consistent.

I also believe Bruce X as a benchmark is extremely flawed since in modern hardware it finishes too fast for the computer to even warm up, so it doesn't really mean much.

 

[jaymonkey]

I also believe Bruce X as a benchmark is extremely flawed since in modern hardware it finishes too fast for the computer to even warm up, so it doesn't really mean much.

@D-One,

If the single 5K Bruce X test is too short for you you can either use the 5K x 5 test or the 8K version of the test.
All versions of the test are in the attached FCPX bundle library.

Cheers
Jay

 

[Jevens]

@D-One,

If the single 5K Bruce X test is too short for you you can either user the 5K x 5 test or the 8K version of the test.
All versions of the test are in the attached FCPX bundle library.

Cheers
Jay

Me again....
I have tried this enhanced BruceX benchmark.
Regardless MacPro 7,1 or iMacPro 1,1 smbios, fist run result of 8k or 5k x3 is normal.
But as soon as I try the second or 3rd runs, the result will deteriorate quickly.
To get the reasonable result back, I had to quit FCPX and open it again.
Sth must be wrong.
Maybe someone with genuine apple computer should verify this, whether it is a hackintosh problem or universal for apple device.

 

[D-One]

@D-One,

If the single 5K Bruce X test is too short for you you can either use the 5K x 5 test or the 8K version of the test.
All versions of the test are in the attached FCPX bundle library.

Cheers
Jay

Hi Jay. It's not really about being too short for a specific person/system, it's more about it being inaccurate due to the way FCP behaves, the fact that this changes every other update either from macOS or from FCP doesn't help with the confusion either. I am not sure if it's the way the test is designed that is outdated or if FCP is just a weirdo.

Where dd the 5x version come from? Is it official?
Is the point of 5x timeline to divide the result later by 5? Those results can't be compared with the normal 1x version of the benchmark.

1. If I run normal BruceX I get 10 seconds.
2. If I multiply the timeline 10x I get 420 seconds, 420 divided by 10 is 42....

There is a gigantic gap between 10s and 42s, I can't just say my benchmark result is 42 because I chose to do the test differently, that will induce others in error.
The methodology is important when comparing results with others, if everyone does it differently then there's no real point.

Even with the 10x timeline, my GPU doesn't even turn the fans on / warm-up significantly during the test, that's an odd thing to happen in a benchmark, all this matches up to the fact that people with a much lower GPU than mine have the same results thus making it pointless to have a high-end GPU for FCP. I want to invest in a Radeon VII but nothing indicates that I will get better results than someone with a $200 580... that's kind of sad.

But as soon as I try the second or 3rd runs, the result will deteriorate quickly.
To get the reasonable result back, I had to quit FCPX and open it again.
Sth must be wrong.

Something must be very wrong indeed. In order to work with FCP every time we export a project, we would have to relaunch the app to ensure the fastest export.
For science (lol) I did the normal 1x 4k test around 9 times in a row without quitting the app, I started at 10 seconds and ended at 38.

Maybe someone with genuine apple computer should verify this, whether it is a hackintosh problem or universal for apple device.

That's a great question, unfortunately only someone with an aftermarket GPU and a MacPro can answer that.
My MBP has consistent results with multiple runs but it doesn't have a dedicated GPU so it doesn't count as confirmation.

 

[mlross1]

Osx: 10.13.6
Mobo: ga-z77-ds3h
CPU: i5 3570k 3.4ghz
8GB RAM
120GB Sandisk extreme
2x1TB WD green drives (owned well before the build)
Not overclocked
Stock intel heatsink
Gpu Nvidia gt 640 metal enabled

5k Brucex: 1:49s

I just updated to high Sierra, and fcpx works much better. Timeline isn’t choppy with 4K anymore and render times increased. Davinci resolve works like a dream with cDNG files (again surprised lol) Pretty slow but not bad given it’s age and price. I was expecting like 5 minutes+ haha. Probably time to upgrade lol. Just posting in case someone has this card or setup and wondering if it will handle some editing. It’s good for timeline navigation and simple renders of effects/luts just slow export conversion times. Especially for a 2 second clip. I edit with pro res anyways so the BruceX it’s pointless for my needs. Good to have everyone on the same page for a bench though.

 

[mlross1]

New Update:

Went to Mojave and from fcpx 10.4 to 10.4.7 and my bruceX time went from 1:49s to 1:25s. Still slow but 10.4.7 is actually faster. Damn.