The AMD Radeon RX Vega 64 & RX Vega 56 Review: Vega Burning Bright
by Ryan Smith & Nate Oh on August 14, 2017 9:00 AM ESTThe Vega Architecture: AMD’s Brightest Day
From an architectural standpoint, AMD’s engineers consider the Vega architecture to be their most sweeping architectural change in five years. And looking over everything that has been added to the architecture, it’s easy to see why. In terms of core graphics/compute features, Vega introduces more than any other iteration of GCN before it.
Speaking of GCN, before getting too deep here, it’s interesting to note that at least publicly, AMD is shying away from the Graphics Core Next name. GCN doesn’t appear anywhere in AMD’s whitepaper, while in programmers’ documents such as the shader ISA, the name is still present. But at least for the purposes of public discussion, rather than using the term GCN 5, AMD is consistently calling it the Vega architecture. Though make no mistake, this is still very much GCN, so AMD’s basic GPU execution model remains.
So what does Vega bring to the table? Back in January we got what has turned out to be a fairly extensive high-level overview of Vega’s main architectural improvements. In a nutshell, Vega is:
- Higher clocks
- Double rate FP16 math (Rapid Packed Math)
- HBM2
- New memory page management for the high-bandwidth cache controller
- Tiled rasterization (Draw Stream Binning Rasterizer)
- Increased ROP efficiency via L2 cache
- Improved geometry engine
- Primitive shading for even faster triangle culling
- Direct3D feature level 12_1 graphics features
- Improved display controllers
The interesting thing is that even with this significant number of changes, the Vega ISA is not a complete departure from the GCN4 ISA. AMD has added a number of new instructions – mostly for FP16 operations – along with some additional instructions that they expect to improve performance for video processing and some 8-bit integer operations, but nothing that radically upends Vega from earlier ISAs. So in terms of compute, Vega is still very comparable to Polaris and Fiji in terms of how data moves through the GPU.
Consequently, the burning question I think many will ask is if the effective compute IPC is significantly higher than Fiji, and the answer is no. AMD has actually taken significant pains to keep the throughput latency of a CU at 4 cycles (4 stages deep), however strictly speaking, existing code isn’t going to run any faster on Vega than earlier architectures. In order to wring the most out of Vega’s new CUs, you need to take advantage of the new compute features. Note that this doesn’t mean that compilers can’t take advantage of them on their own, but especially with the datatype matters, it’s important that code be designed for lower precision datatypes to begin with.
Facebook
Twitter
RSS
213 Comments
View All Comments
tipoo - Monday, August 14, 2017 - link
For a while those APUs were floating them while their standalone CPUs and GPUs struggled. Maybe they've gotten too slim for three strong tentpoles, alas, and one will always suffer.Da W - Tuesday, August 15, 2017 - link
Buldozer core sucked next to ''ok'' GCN igpu. It was very bandwith dependant and the igpu was most of the time starving for data. There was no point of pushing another dozer apu. They were waiting for Ryzen core, and infinity fabric to feed the gpu. Vega is just launching now and if you noticed AMD is only making one 8 core monolitic die sold in multiple package (1 for ryzen-2 fro treadripper-4 for epic). They have yet to cut that 8 core die in half and integrate their new Vega core in there, which is, i believe, what R&D is doing as of this morning....Yaldabaoth - Monday, August 14, 2017 - link
So, the TL;DR is that the Vega 64 competes on (relatively) cheap computing power and perhaps 4K gaming, and the Vega 56 competes on (relatively) very cheap computing power and being a value for 1440p gaming? Neither seem to compete on efficiency.tipoo - Monday, August 14, 2017 - link
Vega 56 seems well positioned for now. 1070 performance at a decently lower price. Question is if Nvidia can/will drop that price on a whim with enough margin (with a smaller die in theory they could, but AMD is probably getting low margins on these). Vega 64 is a far less clear value prospect, in one way it's similar to the 1070 vs 1080, but with Nvidia you're actually getting the best, which 64 can't claim.Jumangi - Monday, August 14, 2017 - link
Thats the big unknown. I suspect Nvidia is playing with much better margins than AMD is when looking at the chips to compete with them here. If Nvidia can lower prices on the 1070 to squeeze AMD if they want and still make a good profit.webdoctors - Monday, August 14, 2017 - link
The Vega56 is so cheap for the hardware you get I wonder if its being sold for a loss. I commented earlier that I thought these chips would be selling for double what they released at, and if they're profitable at this price point AMD might have some secret low cost manufacturing technology that is worth more than their entire company right now.As a consumer I'm practically getting paid to take it LOL.
tipoo - Monday, August 14, 2017 - link
I doubt it's at a loss, but it's probably at a very slim margin. Nvidia could potentially split the difference with a 50 dollar drop and still have the smaller cheaper die (presumably, if TSMC/Glofo cost similar).Drumsticks - Monday, August 14, 2017 - link
Great review Ryan and Nate. I totally agree with your comment at the end about where Vega was designed. Relative to Nvidia, it's a further step back in almost every metric you can measure - perf/w, perf/mm^2, absolue perf of high end flagship...You really have to hope AMD can find one more rabbit in their hat a year or two from now. Nevertheless, the Vega 56 looks like an impressive product, but you can't be happy about getting 8% more performance out of something >50% larger in silicon.
Morawka - Monday, August 14, 2017 - link
yup and next generation memory to boot.. AMD need better gpu designers. If not for Crypto, AMD would be in serious trouble.Threska - Thursday, April 4, 2019 - link
Hello. I'm writing from the future and I bring important news about Google Stadia." To make it possible on its servers, Google has combined an x86 processor (likely an Intel one) with hyperthreading that runs at 2.7GHz, with 16GB of RAM, and a custom AMD graphics chip. It’s said to uses HBM 2 and has 56 compute units, delivering enough raw horsepower for 10.7 TFlops.
That sounds like a modified Vega 56, although it’s equally possible that it’s one of AMD’s upcoming Navi line of graphics cards."
https://www.digitaltrends.com/gaming/google-stadia...