Post Your BaseMark GPU Scores!

[xkm1948]

Wait a minute, GPU-Z does not show Vulkan support level?

 

[biffzinker]

Huawei Mate SE
SoC Kirin 659
CPU - 4xA53 2,360 MHz + 4xA53 1,700 MHz
GPU - Mali-T830 MP2 900 MHz (40.8GFlops)
Mem - LPDDR3 64-bit Dual Channel (933MHz)

 

[MrGenius]

Does the benchmark need to connect to the internet to function?

With the free version, yes.
https://www.basemark.com/products/basemark-gpu/

Wait a minute, GPU-Z does not show Vulkan support level?

It should. It does for me. Under the Advanced tab.

 

[Nuckles56]

A slightly higher score with my phone,after I put it in the fridge for 10 minutes to try and keep the clocks up higher for longer

I'm pushing the absolute limit on the card under air here, I'm going to give it another run tomorrow with my rig outside as we're expecting -4 degrees tomorrow and see if I can squeeze a tiny bit more out of the card then (and crack the 4k mark)

 

[T4C Fantasy]

for clarification, for phones, CPU i will only list the SoC name, for gpu ill list whats inside the SoC. it gets too complicated with the dual x4 cores A53 57 crap mainly its the space needed.

 

[looniam]

will this work for a sub?

btw, i got 5030 running the card (980ti) at ~1318Mhz that score of 5430 is @1448. i'm sitting on a H61 mobo so no OCing the CPU ATM.

 

[IceShroom]

One thing though, why does smartphone gpus score more than desktop gpus???

 

[T4C Fantasy]

One thing though, why does smartphone gpus score more than desktop gpus???

1080p which is why it has its own table xD

desktop is 4k tests

 

[defiiancecp]

After the fun of DDU'ing and reverting back to 18.5.1 drivers, finally got this running in vulkan on my Vega 64. Posting results from OpenGL and from Vulkan, but it's pretty clear something's very wrong with the OpenGL results. In Vulkan, my GPU utilization stays near-continuously pegged at 99%, and the clocks are pegged in the 1600s (see hwinfo screen snip from that session- Avg GPU Utz is 91% in spite of the dead period as the benchmark is loading, average clock is 1543, etc). In the OpenGL session, however, the utilization is terrible - average under 70%, and because it's not pushing the GPU the clocks stay low as well (avg just under 1200, and never even bumps into the 1500 range).

 

[xkm1948]

After the fun of DDU'ing and reverting back to 18.5.1 drivers, finally got this running in vulkan on my Vega 64. Posting results from OpenGL and from Vulkan, but it's pretty clear something's very wrong with the OpenGL results. In Vulkan, my GPU utilization stays near-continuously pegged at 99%, and the clocks are pegged in the 1600s (see hwinfo screen snip from that session- Avg GPU Utz is 91% in spite of the dead period as the benchmark is loading, average clock is 1543, etc). In the OpenGL session, however, the utilization is terrible - average under 70%, and because it's not pushing the GPU the clocks stay low as well (avg just under 1200, and never even bumps into the 1500 range).

Nothing wrong. GCN based GPU was never good at OpenGL. Vulkan was MADE to address this. All GCN GPU are pretty bad at keeping the compute engine fully utilized due to bad scheduler efficiency. Vulkan and DX12 removes the burden and allows close to max stream processor utilization, hence why a lot higher scores.

 

[defiiancecp]

Nothing wrong. GCN based GPU was never good at OpenGL. Vulkan was MADE to address this. All GCN GPU are pretty bad at keeping the compute engine fully utilized due to bad scheduler efficiency. Vulkan and DX12 removes the burden and allows close to max stream processor utilization, hence why a lot higher scores.

I see very different behavior in OpenGL games. Vulkan is much faster, of course, but it's not because the GPU isn't being utilized in OpenGL, like this benchmark shows. I just tried it in Doom, which supports both OpenGL and Vulkan. Set to 4k, ultra, load Kadengir Sanctum, observe framerate, load, and GPU clocks for 1:30->

Vulkan: 99-100 FPS, 1627 avg GPU clock, 99.0% avg GPU Utilization

OPENGL: 65-66 FPS, 1629 avg GPU clock, 97.3% avg GPU Utilization



To add, looks like there's something funky with AMD's OpenGL implementation specifically - see: https://community.amd.com/thread/206176 -- mentions the relatively very poor performance, as well as talking about the low GPU clocks I'm seeing. In that thread there's some theory that maybe it's being locked to a CPU-bound thread, but that doesn't seem to be the case here (I'm not showing any CPU threads maxed out), and it looks like they came to the same conclusion..

But I guess id found a way to work around it? Doom OpenGL definitely isn't suffering from the same problem.

I guess the fact that I couldn't find any other games running openGL in my library means I shouldn't be too worried about it, though.

 

[Psychoholic]

Quick run with everyday settings..
EDIT: CPU is at 4.9.. although you wouldnt know it looking at cpuz.

 

[Zyll Goliat]

After the fun of DDU'ing and reverting back to 18.5.1 drivers, finally got this running in vulkan on my Vega 64. Posting results from OpenGL and from Vulkan, but it's pretty clear something's very wrong with the OpenGL results. In Vulkan, my GPU utilization stays near-continuously pegged at 99%, and the clocks are pegged in the 1600s (see hwinfo screen snip from that session- Avg GPU Utz is 91% in spite of the dead period as the benchmark is loading, average clock is 1543, etc). In the OpenGL session, however, the utilization is terrible - average under 70%, and because it's not pushing the GPU the clocks stay low as well (avg just under 1200, and never even bumps into the 1500 range).
View attachment 102896View attachment 102895View attachment 102897View attachment 102894

Indeed...very strange and a BIG gap between Open GL and Vulkan results with AMD cards as I already commented few pages before now you just confirmed.....this should be investigated furthermore....

 

[T4C Fantasy]

Quick run with everyday settings..
EDIT: CPU is at 4.9.. although you wouldnt know it looking at cpuz.
View attachment 102904

I will add in morning

 

[MrGenius]

I will add in morning

My Vulkan score too please(plus the link to my OpenGL needs changed, since I combined both in the same post). Thanks!

 

[Nuckles56]

OpenGL scores are slightly higher for a phone than the Vulkan ones by the looks of things

I also ran a 4k test just for fun with my phone and it outperforms my gt 745m in my laptop by almost 2 to 1 (though it uses a different texture compression method) and is half as fast as my RX 480 when using openGL and the etc2 texture compression


I did encounter a strange issue with my laptop when I was doing runs with that, I could only run the vulkan tests on the nvidia GPU, as soon as I tried the OpenGL test it would revert to the intel iGPU for some reason.

And also a funny fail run which says my gt 745m is equal to 2 GTX 1080ti's assuming perfect SLI scaling

 

[Sirillya]

8700@5.2 and GPU 1080TI ASUS STRIX OC 2113 Under Air

 

[Arctucas]

With the free version, yes.
<SNIP>

I will have to pass, then.

 

[fray_bentos]

1. Samsung Galaxy S8 (EU) SM-G950F
CPU: Exynos 8895 Octa-core (4x2.3 GHz & 4x1.7 GHz)
GPU: Mali-G71 MP20



2.

I did encounter a strange issue with my laptop when I was doing runs with that, I could only run the vulkan tests on the nvidia GPU, as soon as I tried the OpenGL test it would revert to the intel iGPU for some reason.

I solved this problem on my 970M laptop by going to the Nvidia control panel\manage 3D settings\program settings tab, and adding C:\\Program Files\Basemark GPU\binaries\BasemarkGPU_gl.exe and then setting this to use "High-performance NVIDIA processor". Similarly, you can force the Integrated graphics to be used instead, as I did under point 4. below.


3. Gigabyte Aero 14 laptop with i7-6700HQ using 970M @1172 Mhz clock / 1303 MHz memory.



4. Gigabyte Aero 14 laptop with i7-6700HQ using iGPU Intel HD Graphics 530 (undervolted by -130mV)

 

[T4C Fantasy]

how cheap is that the average FPS is the score for phones.

well and for desktop aswel

 

[fray_bentos]

how cheap is that the average FPS is the score for phones.

well and for desktop aswel

No it's not, it's the average FPS x100! They would have wiser just to report the average FPS as the score without the x100 factor. Scores are arbitrary, but an FPS is something that people can relate to and understand.

 

[T4C Fantasy]

No it's not, it's the average FPS x100! They would have wiser just to report the average FPS as the score without the x100 factor. Scores are arbitrary, but an FPS is something that people can relate to and understand.

they removed decimal for desktop tests but i bet they would be 27.27 score 2727, but for phone its exactly that

 

[fray_bentos]

they removed decimal for desktop tests but i bet they would be 27.27 score 2727, but for phone its exactly that

Yep, dropped, and not even rounded the decimals on desktop. Also, using that x100 relationship, I spotted some typos in the table:

Buznoob 24 ave. fps vs 3871 score
defiiancecp 63 ave. fps vs. 3254 score

 

[T4C Fantasy]

Yep, dropped, and not even rounded the decimals on desktop. Also, using that x100 relationship, I spotted some typos in the table:

Buznoob 24 ave. fps vs 3871 score
defiiancecp 63 ave. fps vs. 3254 score

Buzznoob is coorect with a 2470 score which would be 24.70 avg they didnt round xD

 

[fray_bentos]

Buzznoob is coorect with a 2470 score which would be 24.70 avg they didnt round xD

The bottom entry for Buzznoob is correct, but there is another one higher up the table that is wrong.