ARM was showing off two graphical demos featuring Mali at GDC, one fairly tame the other technically interesting. Of the two, the depth culling implementations are an unsettled question, 4K gaming is a foregone conclusion.
Lets start with the easy one, ARM was showing off their driving game demo on a 4K screen. 4K gaming from a mobile device is not a big trick, it is something that will happen sooner or later as even low-end screen resolutions grow to iPad/Retina levels. Expect this to trickle down from the current high-end all the way to sub-$100 devices in a few years. For the moment the demos are running at 25/30FPS because the HDMI outputs are incapable of pushing more even if the GPUs can. When DPnext and HDMInext come out, look for full 4K/60 from mobile devices shortly thereafter.
More impressive is the depth culling demo, shown below on the tablet connected to the ARM graphical debugger tools on the laptop. On the tablet itself is a rainbow-colored cube made up of tiny polygons. The colors represent the depth of the polygons and although it spins, the depths are fixed so it takes a lot of sorting and culling to make it render correctly. In short you can’t do this with a physical object but 3D renderings aren’t constrained by such things.
Update March 26 @ 11:30am: The colors aren’t depth exactly, they represent the position of the triangles in the vertecies buffer. You can see how this would affect sort load as it spins.
Rainbow cubes and debugging consoles make for a technical demo
At the angles where there is a lot of sorting and culling to do, remember that Mali doesn’t have depth sorting/culling/Z in hardware like Imagination, the CPU and GPU have more work to do. You can see this in the load graphs on the right, as the cube rotates the compute load goes up and down cyclically.
Update March 24, 2014 @ 8pm: ARM has pointed out that Mali does indeed have depth/Z hardware, just a different approach than Imagination. These differences more than anything are the relevant bits, but the details are out of the scope of this article. Basically Mali takes a very different approach but does not lack hardware Z.
Update March 26 @ 11:30am: The graphs show the load on the hardware used for depth calculation, so it is one step removed from power but a good proxy for it.
What is more interesting is that CPU and GPU load in mobile is pretty directly related to energy used, and therefore to battery life. Stepping a bit further out into big picture land, Imagination went with depth hardware which costs area but returns performance and battery life in 3D graphical loads. ARM does depth in software and ends up with a smaller GPU with lower performance and higher energy use under 3D loads.
Note: ARM disagrees with this point.
Which one is better? Depends on what you are trying to accomplish and what your target market is. For mobile gamers and premium devices where die costs come out in the wash, depth hardware may be a good choice. For the high volume, low-cost devices that make up the majority of sales, software is likely the better choice.
Things get interesting in the middle, where the crossover lies is the multi-billion dollar question and no right answers, only more questions. The demo shows that both approaches result in a correct rendering so which one you choose is more a question of economics than tech, approaches instead of correctness. 4K was the polar opposite of this, it is coming from top to bottom and there is no doubt there.S|A
Have you signed up for our newsletter yet?
Did you know that you can access all our past subscription-only articles with a simple Student Membership for 100 USD per year? If you want in-depth analysis and exclusive exclusives, we don’t make the news, we just report it so there is no guarantee when exclusives are added to the Professional level but that’s where you’ll find the deep dive analysis.
Latest posts by Charlie Demerjian (see all)
- Broadcom fills out StrataDNX with three new chips - Nov 30, 2016
- Qualcomm’s Quick Charge 4.0 us based on USB-PD - Nov 28, 2016
- Intel outs lots of AI focused bits at AI day - Nov 17, 2016
- Qualcomm unveils the first 10nm phone SoC, Snapdragon 835 - Nov 17, 2016
- Intel preferentially offers two customers Skylake Xeon CPUs - Nov 17, 2016