AMD Radeon RX 5700 / RX 5700XT Linux Gaming Benchmarks

[josh_walrath]

Confirmed here (albeit a little late); my 5700 XT also gets way less frames than the Phoronix benchmarks, and I also ran the Heaven benchmark manually.

Unigine Heaven Benchmark 4.0

FPS:	60.9
Score:	1534
Min FPS:	13.4
Max FPS:	116.4

System

Platform:	Linux 5.10.2-2-MANJARO x86_64
CPU model:	AMD Ryzen 9 3900X 12-Core Processor (3792MHz) x24
GPU model:	Unknown GPU (256MB) x1

Settings

Render:	OpenGL
Mode:	2560x1440 8xAA fullscreen
Preset	Custom
Quality	Ultra
Tessellation:	Extreme

HOWEVER, when running with the Phoronix Test Suite (phoronix-test-suite benchmark unigine-heaven), my FPS goes from 60.9 to 101.10, so a significant difference definitely seems to exist from the manual benchmarks:

heaven_bench

AMD Ryzen 9 3900X 12-Core testing with a ASUS ROG CROSSHAIR VIII HERO (WI-FI) (1001 BIOS) and XFX AMD Radeon RX 5600 OEM/5600 XT / 5700/5700 8GB on ManjaroLinux 20.2.1 via the Phoronix Test Suite.
Unigine Heaven 4.0
Resolution: 2560 x 1440 - Mode: Fullscreen - Renderer: OpenGL
Frames Per Second > Higher Is Better
Unigine Heaven benchmark . 101.10 |============================================

[josh_walrath]

For some reason I'm not getting the same results as the AMD RX5700 XT benchmarks are showing. I'm assuming I did something wrong?

Copy/Paste from the Generated HTML file:

Unigine Heaven Benchmark 4.0

FPS:	59.5
Score:	1498
Min FPS:	18.9
Max FPS:	120.4

System

Platform:	Linux 4.18.0-25-generic x86_64
CPU model:	Intel(R) Core(TM) i7-4770K CPU @ 3.50GHz (3499MHz) x8
GPU model:	Unknown GPU (256MB) x1

Settings

Render:	OpenGL
Mode:	2560x1440 8xAA fullscreen
Preset	Custom
Quality	Ultra
Tessellation:	Extreme

Confirmed here (albeit a little late); my 5700 XT also gets way less frames than the Phoronix benchmarks, and I also ran the Heaven benchmark manually.

Unigine Heaven Benchmark 4.0

FPS:	60.9
Score:	1534
Min FPS:	13.4
Max FPS:	116.4

System

Platform:	Linux 5.10.2-2-MANJARO x86_64
CPU model:	AMD Ryzen 9 3900X 12-Core Processor (3792MHz) x24
GPU model:	Unknown GPU (256MB) x1

Settings

Render:	OpenGL
Mode:	2560x1440 8xAA fullscreen
Preset	Custom
Quality	Ultra
Tessellation:	Extreme

[ic3man5]

its not broken. its a feature and i'm being serious about that. amd purposefully made it that the fan will never go above ~2000-2300rpm as that is the max rpm. out of 5000. they designed the fan profile to keep the core around 85c and 115c hotspot. they chose to kill performance rather than increase the fan because they still have ptsd vietnam flashbacks over the 290 series with their loud fans running at 10,000rpm and how they got ridiculed heavily for it.

I picked up a 5700xt myself and put a morpheus on it. keeps the core around 55-57c and hotspot around 79-81c. had to use 0.3mm metal washers from home depot on the mounting nuts connecting to the cold plate. along with using the amd retention bracket after messing around 7 times with different mounting techniques. kinda feel like i should have just picked up a 2070 super fe instead but i wanted open source linux drivers that just werked. though the card is pretty beastly at 1900mhz range its running at on its own since the thermals are under control. really keeps up with a 2070 super and in some games with a 2080. like bf:v. wish it didn't cost the same after the morpheus mod. well i wish amd had AIB's release at launch instead of waiting one to two months out. ridiculous.

on a side note the windows drivers have been ridiculously awful. so damn buggy. from green screens during driver installation, to random TDR's causing the screen turning black at anytime and directx errors of cannot find graphics device. so many complaints. amd seriously needs to get their act together and learn to code. they make great hardware but seriously shoot themselves in the foot with the software experience.

I agree with everything you've said here, except the first sentence. AMD might not consider it broken but it is.

I can't justify the Morpheus II because that that point I should just by a 2070 Super. I did manage to get it working pretty decent in windows right now:

-Disable freesync
-Disable HDCP
-Don't touch the Virtual Super Resolution (leave it disabled)
-Undervolt / Underclock the GPU (I'm actually putting it at the game clock AMD says it is suppose to run at).
-Force close all applications that poll sensors (anything related to amd and radeon in task manager).

I'm tempted to return this and wait for the aftermarket coolers, Mine isn't stable without the undervolt / underclock to keep temperatures in check.

[coder]

Because I am not "sub-$200 GPU range" and I game at 1080p!

Looking at the benchmarks he did run, it seems like even the base RX 5700 should pretty much max your refresh rate at 1080p, except for Deus Ex: Mankind Divided.

I'll grant you that it would've been helpful for him to run that one at 1080p.

[digitalsin]

So, why did you complain about no 1080p benchmarks

Because I am not "sub-$200 GPU range" and I game at 1080p!

[coder]

Actually I game at 144hz, and for that you are pushing over twice as many pixels as 60hz. So 2560x1440 @ 144hz would even be more than 4k @ 60hz.

So, why did you complain about no 1080p benchmarks and seemingly criticize him for too many @ 4k (which there weren't)?

[digitalsin]

Most of his benchmarks aren't 4k! 2560x1440 is about half as many pixels, and about twice that of 1080p.

Most people using 1080p are probably down in the sub-$200 GPU range. These are nearly twice as expensive. You don't need such fast GPUs for 1080p.

Actually I game at 144hz, and for that you are pushing over twice as many pixels as 60hz. So 2560x1440 @ 144hz would even be more than 4k @ 60hz.
I prefer faster refresh rates over increase in resolution...

[ic3man5]

That just shows that reference cooling is underperforming (at given dBA, to boot), not that the chip itself is hot. Non-ref can easily solve this.

I can tell you the fan profile is broke under windows, I was hitting 90C because the fan wasn't spooling up.

[ic3man5]

What res is your monitor?

I've noticed a bug with the unigine benchmarks on linux, at least on my setup if you run them fullscreen it completely ignores the resolution you ask for and just uses the current desktop resolution. When trying to run 1080p fullscreen I get the same results as 1440p fullscreen, but if I run in 1080p windowed mode all of a sudden the results are way faster than 1440p, as expected.

1440, I was running the PTS towards the end and not just straight from the app. I'll keep the fullscreen bug in mind if I run more tests with it.

[unit327]

I just got done installing windows (sadly) and running the suite again with the same settings as I ran in Linux:

FPS:	105.3
Score:	2652
Min FPS:	8.0
Max FPS:	218.0

Linux was way smoother (as indicated by my min FPS here) but the average FPS is way lower.

What res is your monitor?

I've noticed a bug with the unigine benchmarks on linux, at least on my setup if you run them fullscreen it completely ignores the resolution you ask for and just uses the current desktop resolution. When trying to run 1080p fullscreen I get the same results as 1440p fullscreen, but if I run in 1080p windowed mode all of a sudden the results are way faster than 1440p, as expected.