The Polaris Architecture
Since 2011, AMD has been using the 28nm die on their graphics card, and after 5 long years, they are using the new 14nm FinFET die from Samsung and Global Foundries. Even the Fury series were using 28nm fab process, one thing new with the Fury is the incorporation of High Bandwidth Memory. But expect we will see more graphics cards from AMD with HBM in the near future.
AMD’s 4th generation GCN (Graphics Core Next) brings a number of new features and these features can be enjoyed with the new Radeon RX 480.
AMD has also improved geometric processing with the Primitive Discard Accelerator which cuts triangles early in the pipeline with zero area or no inclusive sample points.
Shader efficiency is also another huge improvement in the Polaris architecture compared to the previous architectures. The details on the improved shader efficiency are already highlighted on the above and below slides.
With the Polaris architecture, memory controller has also been improved to support GDDR5 memory and speeds up to 8Gbps and up to 256GB/s memory bandwidth. We won’t be seeing any GDDR5X with the Polaris architecture though, but we may be seeing GDDR5X or HBM with the Vega architecture.
The GPU’s L2 cache is doubled delivering up to 40% power savings on memory transactions and improves power efficiency.
AMD has been pushing asynchronous compute and this result in a significant improvement in gaming experience.
With the new Polaris architecture, there are two Hardware Schedule enabled. HWS is present on the Fury cards, but the RX 480 (which costs significantly less) now also features this technology thanks to the new Polaris architecture.
The new AMD Polaris also supports DisplayPort 1.3 HBR3 and DisplayPort 1.4 HDR, enabling the RX 480 to be compatible with more output monitors at higher resolutions and faster refresh rates.
The next two slides below pretty much sums up what AMD’s Polaris architecture has to offer. And all of these improvements and features are present in the new Radeon RX 480 graphics cards. On the next page, let’s take a closer look at the RX 480 itself.
Why didn’t you OC the memory when you OC’d the Card?
Memory Oc is probably even more important that Core OC on this card
I did try to OC the memory. Unfortunately, I can’t make the card stable enough to finish the entire benchmark suite with its memory OCed. Some reviewers where able to OC the memory of their RX 480, probably mine was an isolated case. I am hoping to get a non-reference RX 480. Then we can compare them side by side.
Did you raise the power limit and Have the VRAM voltage at 1V?
Also, you could undervolt the Core a bit to acheive better clocks and thermals
I’ve got higher scores within Heaven at 1080p Ultra https://uploads.disquscdn.com/images/60d821e79e9b2cd767ae2b2e8240687d8a8d10d7d0998a941e9270b985ea775f.png with my Red Devil and A10-7870K APU (OC to 4.2GHz) Memory OC to 2400. I know it’s not fair to compare the reference card with an Aftermarket solution, but I have an APU for CPU, which it’s supposed to be a bottleneck right?