AMD’s Flagship Vega 10 GPU Secret Development & Radeon’s New Architectural Naming Scheme
AMD’s Raja Koduri, head of Radeon Technologies Group, revealed that the company will be naming its future GPU architectures after stars, star systems and galaxies. The company’s head of the newly founded Radeon Technologies Group explained that the team is heading towards a new direction with a fresh look on all things graphics and there was a need to reflect this change of winds.
AMD Radeon Technologies Group Bids A Somber Farewell To Island Code Names And Begins Its “Journey Into Space”
The hallmark of AMD’s next generation Polaris architecture is the highly anticipated move to the revolutionary FinFET process technology and considerable engineering focus on innovations to push architectural efficiency.
AMD describes Polaris as an architecture that will deliver a “historic” leap in performance per watt, and that’s where the naming convention stems from. As stars are the most efficient photon generators in our universe, thus AMD found it only fitting to call its most ambitious graphics architecture to date “Polaris” after the brightest star seen from earth.
Excerpt from AMD’s Official Press Release :
AMD’s Polaris architecture-based 14nm FinFET GPUs deliver a remarkable generational jump in power efficiency. Polaris-based GPUs are designed for fluid frame rates in graphics, gaming, VR and multimedia applications running on compelling small form-factor thin and light computer designs.
“Our new Polaris architecture showcases significant advances in performance, power efficiency and features,” said Lisa Su, president and CEO, AMD. “2016 will be a very exciting year for Radeon fans driven by our Polaris architecture, Radeon Software Crimson Edition and a host of other innovations in the pipeline from our Radeon Technologies Group.”
Raja Koduri, vice president at AMD and chief architect of the Radeon Technologies Group holds a very ambitious goal of powering 90% of the world’s pixels. To achieve that goal efficiency is key and just as stars are the world’s most efficient photon generators, Koduri wants AMD graphics technology to be the world’s most efficient pixel generators.
One of the more notable changes that Raja talked about was giving each new graphics architecture a unifying code name, something which the company hasn’t been doing for the past several years, not in any official capacity at least. Raja explained that this change will make things easier for everyone internally and externally and that future GPU architectures will be given astronomical code names, underlining the beginning of Radeon Technologies Group’s new “journey into space”.
We have some exciting hardware announcements as well. This is designed for FinFET. Our guiding principle for the Polaris architecture was power efficiency. We have the new naming scheme for our architectures. It’ll be based on galaxies, star systems, and stars. You’ll see more of this coming in the future. Polaris is the beginning of our journey through space.
VB: Does the Polaris brand supplant the Radeon brand?
Koduri: It’s an architecture codename. It’ll still be Radeon something something on the box. But we didn’t have a consistent architecture name like our competitors do. It was hard, because for people, including yourselves and some of the press and enthusiasts—This family of chips has this architecture and a similar class of features. You can group them easily together.
[Exclusive] AMD’s Greenland Has Been Known As “Vega10” Internally For Some Time
Vega is the brightest star in the Lyra constellation and the second brightest star in the northern hemisphere. Vega along with Arcturus and Sirius, is one of the most luminous stars close to our earth’s own sun. Vega has been extensively studied by astronomers and as such was described as “arguably the next most important star in the sky after the Sun”. Furthermore, it was the northern pole star around 12,000 BCE.
Vega was the first star other than the Sun to be photographed and the first to have its spectrum recorded. But perhaps the most significant fact of all about Vega is that it served as the baseline for calibrating the photometric brightness scale, and was one of the stars used to define the mean values for the UBV photometric system.
This short backstory explains why Vega is such an important codename for AMD and why the company has decided to give it to its highly anticipated next generation GPU, formerly known as Greenland of the Arctic Islands family. All of this plays into AMD’s brand new strategy of powering better “brighter pixels”.
It’s important to note that the “Polaris” code name is a relatively new one and it was given to a GPU architecture that had already been code named Arctic Islands according to our sources. The decision to rename it was one made by the Radeon Technologies Group to underline just how important and different this new 14nm FinFET based “Polaris” architecture is to AMD.
So far AMD has unveiled two 14nm FinFET Polaris GPUs, given the names Polaris 11 and Polaris 10. We’re not yet sure if any of these two is “Greenland” or “Vega 10”. However we know that one is very small, estimated to be around the same size as AMD’s Cape Verde which is 123mm² large. The goal for this chip was to deliver console-class gaming performance in thin and light notebooks. While the other is a large GPU, described as a successor to the Radeon R9 Fury X.
With two identical systems running Star Wars Battlefront, a reading of the active power usage of each system was recorded via a power meter. The Polaris – small Polaris i.e. Polaris 10 – equipped system recorded an average of 86W, while the Maxwell based GTX 950 machine recorded an average of 150W in the actual demonstration that saw, and 140W in AMD’s labs. This amounts to a delta of 53W if we go by the lower recorded average power of 140W and a 63W delta if we go by the reading the we saw during the demo.
Intel Core i7 4790K with 4x4GB DDR4-2600 MHz memory, Windows 10 64bit.
Star Wars Battlefront, X-wing Training
Med Preset @1080p. AMD Polaris GPU tested with 16.10 beta drivers . The GTX 950 tested with 359.06 drivers.
In other words the Nvidia GTX 950 system used 63-74% more power to deliver the same performance. Additionally, if we take into account that a GTX 950 would normally average less than 100W on its own excluding other system components. This indicates that the Polaris video card is only using roughly 50W, making it twice as power efficient as the Nvidia Maxwell based GTX 950 graphics card. Finally, the Polaris graphics chip tested was an early engineering sample and both hardware and driver optimizations are still on-going so additional power efficiency improvements can be expected.
Interestingly, AMD confirmed to us that while it remains committed to the High Bandwidth Memory technology it co-invented with SK Hynix and brought to the market last year, Polaris remains compatible with both HBM/HBM2 and GDDR5 memory standards. Technical marketing lead at AMD Robert Hallock explained that they have the flexibility to use either technology as the need arises. As such each memory technology will be employed in the market segments where it makes sense.
AMD helped lead the development of HBM, was the first to bring HBM to market in GPUs, and plans to implement HBM/HBM2 in future graphics solutions.
At this time we have only publicly demonstrated a GDDR5 configuration of the Polaris architecture.It’s important to understand that HBM isn’t (currently) suitable for all GPU segments due to the current HBM cost structure. In the mainstream GPU segment, GDDR5 remains an extremely cost-effective, efficient and viable memory technology.
Polaris graphics cards are set to launch next summer, before the back to school season, on both desktops and gaming notebooks.
AMD Radeon Graphics Architectures
|GPU||Process||Transistors In Billions||Performance||Memory||Bandwidth|
|Southern Islands||2012||HD 7970 Ghz||Tahiti||28nm||4.3||4.1 TFLOPS||3GB GDDR5||264GB/s|
|Volcanic Islands||2013||R9 290X||Hawaii||28nm||6.2||5.6 TFLOPS||4GB GDDR5||320GB/s|
|Caribbean Islands||2015||R9 Fury X||Fiji||28nm||8.9||8.6 TFLOPS||4GB HBM1||512GB/s|
|POLARIS||2016||RX 480||Polaris 10||14nm||5.7||5.2 TFLOPS||8GB GDDR5||256 GB/s|
|VEGA||2017||RX Vega 64||Vega 10||14nm||12.5||13 TFLOPS||8GB HBM2||480GB/s|
|NAVI||2019||TBA||Navi 10||7nm||TBA||TBA||Nextgen Memory||TBA|