AMD Vega 10 GPU Launching In 2022 - Features 4096 Cores & 16GB HBM2, Greenland Reincarnated

AMD's highly anticipated Vega 10 flagship 14nm FinFET GPU, previously known as Greenland, is launching in 2022 with up to 32 GB of HBM2. AMD unveiled its long-term GPU roadmap at Capsaicin yesterday for the first time, debuting its 2022 and 2022 architectures Vega & Navi. This is the very first time the visitor has ever discussed its graphics roadmap across Polaris, its extremely power efficient 14nm graphics compages launching this summer. Raja Koduri, AMD's Chief Architect of the Radeon Technolgies Group, revealed that the visitor is working on ii GPU architectures to succeed Polaris.

AMD Radeon Technologies Group Roadmap Vega Polaris Navi

The first is Vega. Which will be arriving next year and is the showtime from AMD to support second generation High bandwidth Memory. In 2022 AMD plans to introduce yet another graphics arachitecture, lawmaking named "Navi". With Navi Raja & the Radeon Technologies Group is planning to utilize vertically stacked HBM memory throughout its graphics product lineup. This is what's being referred to with "scalability".

By 2022 nosotros're potentially looking at third generation HBM, which hasn't been announced nor named yet. Whether it's HBM3 or something slightly or completely different, it's besides early to tell. else no one can really tell Even so, what we know for sure is that any memory standard comes after HBM2 will yet use vertical stacking in some form or fashion.

AMD Vega 10 / Greenland Flagship HBM2 GPU Specs

The specs for Vega 10, previously known as Greenland, have been spotted on the LinkedIn profile of an AMD engineer. The GPU volition feature 4096 GCN 4.0 cores, second generation HBM and a 4096bit memory interface.

AMD Vega 10 GPU Leak Yu Zheng R&D Managing director at AMD

Which sounds identical to the R9 Fury Ten but are far from it. The 4th generation GCN cores are ~l-55% faster than their predecessors and extraordinarily more than power efficient co-ordinate to several patents filed past the company that we'll share in an upcoming piece. Considering that the 14LPP process clocks considerably higher than 28nm nosotros're looking at upwardly of 60-seventy% operation improvement with Vega 10/ Greenland over the R9 Fury X.

Vega 10 Specs

WCCFTech	Polaris	Vega	Pascal	Pascal	Pascal
Year	2016	2017	2016	2016	2017
GPU	Polaris 10	Vega 10	GP102	GP104	GP102
Graphics Carte du jour	RX 480	TBA	GTX TItan X Pascal	GTX 1080	GTX 1080 Ti
Process	14nm	14nm	16nm	16nm	16nm
Transistors In Billions	v.7	TBA	12	7.2	12
Stream Processors	2304	4096	3584	2560	3328-3840
Functioning	five.8 TFLOPS	12 TFLOPS	11 TFLOPS	9 TFLOPS	x TFLOPS
Memory	8GB GDDR5	16GB HBM2	12GB GDDR5X	8GB GDDR5X	12GB GDDR5X
Memory Bus	256bit	4096bit	382bit	256bit	382bit
Bandwidth	256 GB/southward	512 GB/s	480 GB/s	320 GB/s	480 GB/s

AMD Vega x / Greenland Flagship HBM2 GPU Launching In 2022

Graphics generation lawmaking names aside, perhaps the nearly of import slice of information that nosotros got out of this is every bit follows : AMD's roadmap confirms that Vega will be the company's commencement architecture to feature second generation HBM. In other words, it confirms that Polaris will in all likelihood make use of GDDR5/10 retentiveness instead.

If you've been paying shut attention to the industry you will quickly betoken out that this makes sense in so many ways. Firstly at that place'due south the cost advantage of using GDDR5/X over HBM. GDDR5/Ten fries area cheaper to make because they don't accept to go through a meticulous procedure to stack them on superlative of each other. As a GPU maker you also don't need an interposer with GDDR5/10 like you lot do with HBM. Both of these factors give GDDR5/X a very existent cost advantage. Secondly there'south the question of HBM2's readiness. HBM is even so ramping and it would take sometime to for HBM2 to catch up.

Thirdly, there's Raja'southward revelations terminal year that AMD has ii 14nm GPUs coming out this year, non iii, two.

Patrick Moorhead, Semiconductor Financial Analyst - Forbes
Raja also talked nigh how Avant-garde Micro Devices' RTG will demand to execute on their architectural designs and create brand new GPUs, something that Advanced Micro Devices has struggled with lately. He promised two brand new GPUs in 2022

That is Polaris 10 and Polaris 11. Both of which AMD has actually shown to journalists, we're talking about the actual physical dies. Those who have seen them - we've but seen Polaris 11, but AMD has shown a Polaris x dice to visitors of its suite at CES - reported that neither of the dies sported an interposer or HBM like Fiji.

Nosotros after confirmed with AMD that indeed the Polaris ten and Polaris 11 GPUs showcased were configured for GDDR5.

AMD helped atomic number 82 the development of HBM, was the beginning to bring HBM to market place 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 price-effective, efficient and feasible memory technology.

This doesn't necessarily exclude that either of these fries can end up with HBM. Later-all AMD might have been trying to be really coy and just showed off GDDR5 configured die. Simply our money is that they weren't trying to exist coy and that both of these fries will end up GDDR5/X. For the elementary fact that any GPU would accept to be designed from the very beginning with one memory standard in mind. As a designer you either cull to equip your GPU with an HBM memory controller or a GDDR5/X memory controller. Non both, that would exist utterly wasteful to the extremely limited flake area you have.

Designing a GPU with both types of memory technologies is theoretically possible. Only as the GPU volition just end upwards with one retentiveness standard or the other, either HBM or GDDR5/X, not both at the same fourth dimension. Ane of the two memory controllers volition repetitively end up every bit useless as white space on a die. So it just doesn't brand any sense to blueprint whatsoever such production.

With all of that said, this wasn't actually unexpected. AMD has really hinted to usa back in December at Sonoma, when we were get-go introduced to Polaris, that at that place are indeed plans for HBM2 in the future. Yesterday those plans accept been made public. HBM2 once once again ,just as start generation HBM did, will debut at the high-stop. Nvidia likely has very similar plans to AMD for 2022 and 2022. And that is to launch two Pascal GDDR5/X GPUs and introduce GP100 with HBM2 next year.

AMD Vega 10, Polaris 10, Polaris 11 Feature

AMD's 2.5X Polaris Power Efficiency Jump Is More Impressive Than You Think

This is perhaps one of the less expected realizations to come up to later yesterday'south Capsaicin event. A 2.5X - 250% - power efficiency jump is impressive in its own correct. Simply actually, it'due south even more than so when you lot consider the fact that Polaris achieves this without HBM.

The power savings from HBM were the primary driver backside Republic of the fiji islands's considerable ability efficiency gains over Hawaii and the R9 290X. In fact HBM played a pivotal role in allowing AMD to design a GPU that not just had 45% more GCN cores than the R9 290X - 4096 vs 2816 - but, also burned less ability on boilerplate. Co-ordinate to AMD, Polaris GPUs will deliver ii.5x the perf/watt of AMD's 2022 28nm GPUs. Those include the Radeon 300 series and the R9 Fury series.

AMD Radeon Technologies Group Roadmap Vega Polaris Navi
It's quite hard to pin-bespeak exactly how far ahead each architecture is of the i earlier it on AMD'due south roadmap. But Navi looks to be sitting at double Polaris's ability efficiency, making it 5x as ability efficient equally the current lineup of 28nm GPUs. Vega sits roughly in the middle between Polaris and Navi, slightly closer to Polaris. So nosotros might be looking at a 3X improvement over 28nm GPUs, which puts it 20% ahead of Polaris in perf/watt. That's about what you lot'd wait from HBM2.

But does this hateful that AMD's fabled Vega 10 / Greenland flagship GPU will be three times more power efficient than the Fury X. I'd say possible, but more likely nosotros're looking at a 3X efficiency jump over the R9 390X. In case you missed our exclusive report a couple of months back , Vega ten is Greenland. The ~18 billion transistor 32GB HBM2 behemoth that's been the subject field of many leaks over the by year.

AMD Radeon Graphics Architectures

WCCFTech	Year	Consumer Production	GPU	Procedure	Transistors In Billions	Operation	Memory	Bandwidth
Southern Islands	2012	HD 7970 Ghz	Tahiti	28nm	four.3	four.ane TFLOPS	3GB GDDR5	264GB/s
Volcanic Islands	2013	R9 290X	Hawaii	28nm	half-dozen.2	5.6 TFLOPS	4GB GDDR5	320GB/s
Caribbean Islands	2015	R9 Fury Ten	Fiji	28nm	8.9	viii.6 TFLOPS	4GB HBM1	512GB/due south
POLARIS	2016	RX 480	Polaris 10	14nm	five.7	5.ii TFLOPS	8GB GDDR5	256 GB/southward
VEGA	2017	RX Vega 64	Vega 10	14nm	12.5	xiii TFLOPS	8GB HBM2	480GB/s
VEGA Refresh VEGA ii	2018 2022	TBA TBA	TBA Vega twenty	12nm 7nm	12.5 TBA	TBA TBA	TBA 32GB HBM2	TBA 1 TB/south
NAVI	2019	TBA	Navi 10	7nm	TBA	TBA	Nextgen Retention	TBA

No HBM2 This Year?

Information technology's truthful that we will have to wait until next year for HBM2 GPUs from either Nvidia or AMD. Nvidia'southward GP104 powered GTX 1080 and GTX 1070 cards - not official names - are reportedly configured for GDDR5X. And at that place's most no take a chance of neither Nvidia or AMD risking product of large GPUs similar GP100 or Vega 10 this early on in the nodes' lifecycles. Neither 16nm or 14nm volition be mature plenty for large GPU launches this year. This is all assuming HBM2 production has already sufficiently ramped up, which hasn't all the same.

Despite that, what we're getting this yr is still going to deliver a very significant jump in performance, performance per watt and operation per dollar. This is thanks to the much needed inflow of FinFET manufacturing engineering besides equally GDDR5X.

This update to the GDDR5 retention spec is simple plenty. Double the capacity and double the information rate of GDDR5 with minimal alterations to the protocol. That in turn means existing GDDR5 memory controllers volition only need pocket-sized blueprint updates to exist compatible with GDDR5X. Which saves Nvidia and AMD a lot of engineering science endeavour and cost. Double the bandwidth also ways GDDR5X will do more than an adequate chore of keeping faster, bigger, next generation FinFET GPUs happy and well fed until HBM2 arrives in 2022.

At double the speed of GDDR5, a GPU configured with a 256bit memory interface can have access to up to 448GB/s of bandwidth and 8GB of retentiveness. That's double the bandwidth available to the GTX 980 and twice the capacity. It'southward also 33% more bandwidth than what'south available to the GTX Titan Ten. With that in perspective GDDR5X is clearly going to be more than enough for Nvidia'due south GP104 GPU and AMD's Polaris 10 GPU.

It'southward non every twenty-four hours that we get GPUs with two.5x better perf/watt, in fact not ever. This is the biggest jump from one generation to the side by side that we have ever seen. Just information technology's non all about efficiency, Raja is also promising the "nearly revolutionary jump in functioning then far".

AMD appear that Polaris graphics cards are going to be released mid 2022, so this summer. AMD's CEO Lisa Su and so elaborated further, saying that Polaris graphics cards are to be expected on shelves and in notebooks earlier the back to schoolhouse flavour this yr. At that place'southward also been a lot of buzz effectually Nvidia potentially demoing Pascal at GTC in April, with GTX 1080 and GTX 1070 - not official names - graphics cards launching in the summer. AMD has already demoed Polaris eleven at CES and Polaris 10 yesterday at GDC, and then nosotros tin can't expect to see Pascal in activity. Hopefully adjacent month.