3DMark API Overhead Benchmark Allows Direct Comparison of DX12, DX11 and Vulkan Performance

Written by Jon Sutton on Fri, Mar 24, 2017 11:16 AM

Up until now basically the only decent point of comparison for Vulkan was in id Software’s DOOM, and even then it was only plotted against OpenGL performance. Now though FutureMark has added Vulkan support to 3DMark, letting users directly compare performance between the three key graphics APIs - Vulkan, DirectX 11 and DirectX 12.

The chief benefits touted for both Vulkan and DX12 are that they provide high-efficiency low level access to the GPU and offload heavy demand from the CPU. They also claim to better use multi-core processors, which in turn allows for more draw calls. Draw calls are essentially what they sound like - commands send from the the CPU to the GPU to draw (or render) an object. The more draw calls can be sent, the greater the number of objects that can appear on screen at any one time.

The point of 3DMark’s API Overhead test then is to see how many draw calls the respective APIs can handle with any given CPU. The number of objects in the scene steadily increases until the benchmark can no longer be rendered above 30 frames per second. Once it drops below 30 FPS, the benchmark ends and you get presented with a score.

As far as graphics card performance goes, the API Overhead test isn’t going to tell us much. The benchmark makes the API overhead the performance bottleneck and minimises GPU load by not rendering textures and lighting effects.  “The point at which the test moves from being CPU-bound to GPUbound changes from system to system,” explains 3DMark. “It is not easy to tell from the test results whether the run was CPU or GPU limited. And what's more, it is difficult to isolate the relative impact of GPU performance and driver performance.”

This is all about comparing the respective APIs rather than for testing hardware, but it should give us a good idea of just where DX11, DX12 and Vulkan currently stand.

If you want to give the API Overhead benchmark test a go it’s now part of both 3DMark Advanced Edition and 3DMark Professional Edition.

Login or Register to join the debate

Rep
9
Offline
19:19 Mar-24-2017

Hi,
I have a weird GPU Load in the API Overhead test - 1080p.


My specs:
Win10 Pro
i5-4690k@4.5 Ghz
MSI GTX 1070 Gaming (stock)
16 Gb DDR3 1833 Mhz
(newest NVIDIA drivers)


This is the GPU usage from someone on game-debate:
https://postimg.org/image/mlea9hk3f/


And there is mine usage in DX11 and Vulkan test:

0
Rep
9
Offline
19:19 Mar-24-2017

What do you think about it? Is it normal? CPU bottlenecking the GPU in Vulkan test?

0
Rep
45
Offline
15:07 Mar-24-2017

This test has DX12 just beating Vulkan on my system but both are way ahead of DX11. This would only help in a cpu bound situation but it makes you wonder why DX12 isn't way out in front of DX11 on cpu intensive games or while using a weaker cpu.


http://www.3dmark.com/3dm/18818157?

1
Rep
179
Offline
admin approved badge
12:13 Mar-24-2017

Make it a part of the "free edition" and I'll give it a try ;)

3
Rep
67
Offline
admin approved badge
11:52 Mar-24-2017
1
Rep
49
Offline
admin approved badge
12:13 Mar-24-2017

Not bad, heres mine (going to try one with a different power profile) but my cpu is a bit weaker then yours and seems like for me dx12 has more draw calls haha.


http://www.3dmark.com/3dm/18815178?

1
Rep
179
Offline
admin approved badge
12:23 Mar-24-2017

"but my cpu is a bit weaker then yours"
I think you mean GPU?

1
Rep
49
Offline
admin approved badge
12:26 Mar-24-2017

Yea whoops haha. Running it in high performance power mode made a 10% difference in dx11 single threaded per min, you can see ryzen and w10 still need a bit of work.


http://www.3dmark.com/compare/aot/200519/aot/200511

1
Rep
179
Offline
admin approved badge
12:37 Mar-24-2017

I think the results you guys got speak for themselves.... DX11 needs to be phased out much faster and more games need to be written from the ground up for Vulkan and DX12..... DX11 is really holding everything back, and with devs just sloppily piling DX12 code ontop of DX11 code so that their games can work with either API... it's really not doing anyone any good.

4
Rep
179
Offline
admin approved badge
12:40 Mar-24-2017

And personally, I'd really love to see Vulkan just completely take over and dominate, since everyone will be able to benefit.... not just those running Windows 10.

1
Rep
49
Offline
admin approved badge
12:40 Mar-24-2017

Yea same thing i thought, its a bit weird though that in my case dx12 seems to be doing much better then vulkan. Not sure if its the ryzen cpu cause it or whats going on. I'm even oced and the power modes def made a difference in favor of going high performance, core parking must not be doing multi threading a favor in this case.

3
Rep
179
Offline
admin approved badge
12:43 Mar-24-2017

It's hard to say.... they're close enough though, and personally even if DX12 did perform a bit better I'd still rather see Vulkan take over simply because it doesn't rely on Windows 10.

3
Rep
319
Offline
admin badge
11:24 Mar-24-2017

so.. anyone gave this a run yet? sounds pretty interesting.

0
Rep
49
Offline
admin approved badge
12:32 Mar-24-2017

I did but wont let me reply haha.


http://www.3dmark.com/compare/aot/200519/aot/200511

1

Can They Run... |

| 60FPS, Medium, 1080p
Ryzen R5 1600 Radeon R9 280 Club3D royalKing 3GB Edition 8GB
| 60FPS, High, 1080p
Core i7-11800H 8-Core 1.90GHz GeForce RTX 3060 Mobile 16GB
0% No [1 votes]
| 60FPS, High, 1080p
Core i7-11800H 8-Core 1.90GHz GeForce RTX 3060 Mobile 16GB
| 60FPS, High, 1080p
Core i7-11800H 8-Core 1.90GHz GeForce RTX 3060 Mobile 16GB
100% Yes [1 votes]
| 60FPS, High, 4k
Core i5-11600 6-Core 2.8GHz GeForce RTX 3060 Ultra 16GB