Vivante Mobile GPU Architecture Gains Traction 29
MojoKid writes "Over the past few years, a handful of mobile graphics companies have emerged but the top dog, by far, has been Imagination Technologies, with Qualcomm, Nvidia and ARM all picking up significant businesses of their own as well. But now, there's a new kid on the block — a company with a tiny, highly customized GPU, a number of recent design wins, and a strong product portfolio. Vivante got started in 2004 and started licensing its GPU designs in 2007. The company's early wins have been in Eastern markets, but this past year, it's begun to show up in devices intended for the West, including the Samsung Galaxy Tab 3 and Google's Chromecast. Vivante has taken a different approach to core design from most of the other companies that play in this space. All modern GPUs are explicitly designed to be modular and scalable. Typically what that means is that a company like Nvidia or AMD defines a single compute unit that can be duplicated throughout the GPU design. Vivante's GPUs are modular as well, but with a much finer level of granularity. Each of the three shaded blocks (3-D Pipeline, Vector Graphics Pipeline, 2-D Pipeline) can be segmented or stacked into various configurations. A GPU core, in other words, could contain more ultra-threaded shaders, or additional vector graphics engines, up to 32 cores in total. One of the advantages of this tiny, modular architecture is that you can clock the cores like gangbusters. According to Vivante, the 28nm high performance silicon variant of the Vivante architecture can clock up to 1GHz at full speed, but fall back to 1/64th of this in power saving mode, or roughly 16MHz."
It's an article (Score:3, Informative)
It fairly benchmarks the current Vivante device against an iPhone4s, against which it loses, but the article points out the design is modular and has plenty of scope for performance improvements and scaling.
It does appear as though the rest of the journalism is a little on the lazy side though
Vivante only looks good on paper (Score:5, Informative)
From one of the etna_viv open source Vivanete drivers:
https://blog.visucore.com/2013/3/12/opencl-on-i-mx6 [visucore.com]
afair its something like: 512 instructions per kernel, no atomics, 1K memory (!!!), 64B cache(!!!!!!!), and shitty narrow databus :)
Theoretical peak 16Gflops, real world results 600Mflops
Vivante is a JOKE when it comes to GPGPU. They try to paint themselves as pioneers and leaders (of of the first mobile gpus with working opencl driver), but its only a token gesture. Same goes for the graphics part of the GPU. They are pretty notorious about advertising features that DO NOT WORK or are UNIMPLEMENTED.
Of course they arent the only ones, whole mobile GPU market is littered with CRAP
https://dolphin-emu.org/blog/2013/09/26/dolphin-emulator-and-opengl-drivers-hall-fameshame/ [dolphin-emu.org]
Re:It's an article (Score:4, Informative)
As the author: No, they weren't. I don't own them.
But furthermore, the iPhone 4S is the best comparison against the Galaxy Tab 3 7.0. Both are Cortex-A9. Both are dual-core. The Tab 3 7.0 has a faster chip. The two displays have *exactly* the same number of pixels. And the iPhone 4S generally wins.
While I'd have included the additional devices if I *owned* those devices, the point was not to simply dogpile a new chip by dumping far more powerful chips at it, particularly when the focus was on the GPU architecture itself and not the Tab 3 7.0 as a device.
Re:Pay-for-Play PR on Slashdot (Score:5, Informative)
Alright, so, I don't normally comment on this sort of thing, but I'm going to make an exception. I'm the author of the article. I've written for hardware IT sites for 12 years. I've written for Hot Hardware since 2009.
Vivante contacted us in August and asked if we'd be interested in doing an article on the architecture. I looked over the slides and data, and said I'd like to see an example of the chip in action. They arranged for us to test a Samsung Galaxy Tab 3 7.0. This was done in exactly the same fashion as any other product review.
The Galaxy Tab 3 7.0 is objectively a pretty lousy tablet relative to other products in the market. It's an 8GB device with a dual-core Cortex-A9 at $199, when you can buy a 16GB Nexus 7 for $229. But remember -- the point of the article was to talk about Vivante, not the mediocre device the GPU is currently paired with. To be clear, the major problem here is *price.* If the Tab 3 7.0 was selling at $129, I'd call it a much stronger product and it would compare passing well against the non-advertising supported version of the Kindle Fire.
If I'd had more devices to compare against, I would have compared against them, but the only iOS product I own is an iPhone 4S that I bought out of my own pocket. I wrote the article to discuss an interesting, more modular approach to the GPU market, backed it up with standard benchmarks, and noted that faster, larger versions of the core were available. If I had access to those versions of the core, I would have benchmarked them.
I will be paid at a standard rate for this work, with no additional kicker from Vivante or Hot Hardware. I was not asked to take a particular stance on the product or its performance, and I chose my own benchmarks.
I maintain that it's interesting to see a new GPU vendor come on the scene, that building midrange parts that are designed for modest products is a viable strategy, and that a company doesn't need to deliver the fastest performance to build an acceptable product for a given price point. Vivante's work is still relatively new and so I am not surprised that its driver stack or OpenCL support is still in early stages.
Anyone claiming that this chip is automatically a Kepler, or PowerVR, or Adreno killer is mistaken. I did not evaluate the chip under such criteria and feel no obligation to rate it against chips it was not designed to face.