After many fruitful years of partnerships with Apple, AMD is reportedly continuing the relationship with their latest Polaris based GPUs. Apple has alternated MacBook Pro suppliers between Nvidia and AMD in the past but tended towards AMD more with the Mac Pro. According to the source, the performance per watt of 14nm Polaris combined with the performance per dollar of the chips is what sold Apple.
AMD has long pursued a strategy os using smaller and more efficient chips to combat their biggest rival Nvidia. Prior to GCN, AMD tended to have smaller flagships that sipped less power and had lesser compute abilities. This all changed around with GCN where AMD focused on compute more while Nvidia did the opposite. This lead to Nvidia topping the efficiency charts and combined with their marketing soared in sales. If the rumours are true, Polaris 10 will be smaller than GP104, its main competitor.
With Polaris, AMD should be able to regain the efficiency advantage with both the move to 14nm and the new architecture. We may see Polaris based Macs as soon as WWDC in June, just after the cards launch at Computex. In addition to a ‘superior’ product, AMD is also willing to cut their margins a bit more in order to get a sale as we saw with the current-gen consoles. Perhaps, is AMD plays their cards well, we may see Zen Macs as well.
Battered on both the CPU and GPU fronts, consoles have been one of the few areas AMD has managed to outplay the competition. With competitive CPU and GPU architectures in one platform, AMD was able to secure Nintendo, Sony, and Microsoft’s current-gen consoles. Nintendo is also set to continue to use AMD chips for the Nintendo NX console and that device will reportedly use a 14nm Polaris like GPU.
From previous rumours, we’ve already learned that the Nintendo NX will use an x86 architecture chip paired with at least 6-8GB OF DDR4. What more, the new console will also feature 4K support via upscaling, streaming media and likely playback as well. To wrap it all up, AMD is reportedly supplying Nintendo with a 14nm Polaris-like GPU for their upcoming console. This is similar to how the PS4 and Xbox One used GPUs that were a merger of GCN 1.1 and 1.2. The Nintendo NX may use a something beyond the GCN 4 that is Polaris.The OS also will use Vulkan as it’s graphics API.
With a strong Polaris chip on 14nm, Nintendo will have a chance at seizing the performance crown for once. Nintendo consoles have proven weaker generally and have suffered from lesser third-party support as a result. With 4K support, the NX may well match the PS4K and the rumoured replacement for the Xbox One. Hopefully, we will finally get 1080p 60FPS with decent graphics on consoles soon enough.
Asynchronous Compute has been one of the headline features with DX12. Pioneered by AMD in their GCN architecture, Async Compute allows a GPU to handle both graphics and compute tasks at the same time, making the best use of resources. Some titles such as Ashes of the Singularity have posted massive gains from this and even titles that have a DX 11 heritage stands to have decent gains. In an update to Async Compute, AMD has added Quick Response Queue support to GCN 1.1 and after.
One of the problems with Async Compute is that it is relatively simple. It only allows graphics and compute tasks to be run at the same time on the shaders. Unfortunately, Async Compute as it stands right now will prioritize graphics tasks, meaning compute tasks only get the leftover resources. This means there is no guarantee when the compute task will finish as it depends on the graphics workload. Quick Response Queue solves this by merging preemption where you stop the graphics workload entirely, with Async Compute.
With Quick Response Queue, tasks can be given special priority to ensure they complete on time. Quick Response Queue will also allow the graphics task to run at the same time albeit with reduced resources. By providing more precise and dependable control, this allows developers to make better use of Async Compute, especially on latency/frame sensitive compute tasks. Moving on, we may see greater gains from Async in games as AMD allows more types of compute workloads to be optimized. Hopefully, this feature will reach GCN 1.0 cards but that depends on if the hardware is capable of it.
Right now, it has more or less been confirmed that Sony is working on an update to the PlayStation 4. Referred to as the PlayStation 4K by multiple sources, the console won’t be running any games at 4K anytime soon. Contrary to the expectations from its name, the new console probably won’t be able to game at 4K, with the update more of a performance bump rather than a giant leap.
From the various leaks, it looks like the update will be based on the new 14nm process AMD is using for Polaris. With the new process, Sony has a chance to use cheaper and more power efficient dies to either cut cost or grow performance for more or less free. Right now the PS4 has hardware similar to the 7870 based on GCN 1.1. With an update, Sony has a chance to jump to Polaris with GCN 1.3/4.0 along with a significant power efficiency increase, allowing the console to stuff a 7970 class GPU in the console. While not enough for 4K, this should allow many more titles to run at 1080p and at a higher level of quality. From what Sony has hinted at, the PS4K will simply be the same PS4 but with improved graphics and higher FPS. This should allow for a similar gameplay experience for users and ensuring games still work fine with the stock PS4. The update to a new architecture will also bring HDMI 2.0 support, which will allow for 4K at 60fps, a first for a console. This pairs up nicely with the Ultra HD Blu-Ray support we are expecting which will allow for 4K content at 60fps.
One of the headline features for AMD’s GCN architecture is their superb support for Asynchronous Compute. One of the showcases games for Asynchronous Compute and DX12, Ashes of the Singularity has used the feature to provide AMD with a significant boost in that title. However, it looks like Asynchronous Compute will be harder to manage if Hitman is anything to go by.
According to IO Interactive, Hitman only gained about 5-10% performance boost on AMD cards after implementing Asynchronous Compute. This is a bit lower than expected given all the hype around the feature. Furthermore, IO Interactive had to do a lot of tuning and it was ‘super-hard’ to implement properly in order to no suffer any performance loss. This is still much better than the issue faced by Nvidia cards which virtually gained nothing from Asynchronous Compute.
For an early implementation of Asynchronous Compute in what is still mainly a DX11 title, 5-10% isn’t that bad. If AMD is able to get Asynchronous Compute into game engines and integrate it into game development on the ground up, easier and more plentiful gains will likely follow. For now, Asynchronous Compute is still enough to put AMD cards ahead of their various competitors and 10% isn’t something to scoff at, especially combined with other DX12 gains.
One of the biggest changes DX12 brings to the table is the increased reliance on developers to properly optimize their code for GPUs. Unlike DX11, there will have fewer levers to tweak in the GPU driver, with more work being needed in the game engine and the game itself. To address this, AMD has announced a partnership with multiple game engine and game developers to implement DX12.
To kick start the effort, AMD is headlining 5 games and engines they are partnering with to ensure DX12 works smoothly with Radeon GPUs with the software. These are Ashes of the Singularity by Stardock and Oxide Games, Total War: WARHAMMER by Creative Assembly, Battlezone VR by Rebellion, Deus Ex: Mankind Divided by Eidos-Montréal and the Nitrous Engine by Oxide Games. These titles span a wide range from RTS to RPG and FPS which gives a sense that AMD is trying to cast as wide a net as possible.
In addition to this, AMD will also be working with EA and Dice to get the Frostbite 3 engine to enable DX12. This engine is of particular importance due to the many AAA EA and other titles using it. AMD is also hoping to push Asynchronous Compute and to make sure games will squeeze the most out of GCN using DX12.
This year, both AMD and Nvidia will be launching their new Polaris and Pascal based GPUs. Unfortunately, it looks like the flagship chips won’t be arriving till next year. Set to arrive in early 2017, Vega, also known as Greenland, is to be the flagship replacement for Fiji. According to information 3DCenter dug up, Vega will feature 4096 GCN shaders, the same amount as Fiji currently has.
With Polaris and Vega, there are suggestions that AMD has managed to improve GCN 4.0’s performance by 30% compared to current GCN offerings. This alone should allow a significant performance increase over the Fury X. Fiji was also limited due to the design of GCN being unoptimized for massive chips with too many shaders and if AMD has managed to fix this, Vega will perform better than expected.
Furthermore, Vega will utilize HBM2 which will finally remove the 4GB cap faced by HBM GPUs as well as reduce latency. The use of 14nm as well and other Polaris improvements will also allow for a cooler and less power hungry die. We can also expect Vega to come in at a die size similar to Hawaii rather than Fiji, with a true Fiji size successor to come later on in the process cycle.
With just about 2 months to launch, it is inevitable that more details about Polaris are going to leak out. According to a patch submitted by AMD, it appears that Polaris 10 and 11 are the same chips previously leaked as Ellesmere and Baffin respectively. Furthermore, it appears that Ellesmere will be an R9 390/390X configuration replacement but offer Fury X like performance.
According to the patch, Polaris 10, the top end Polaris chip is codenamed Ellesmere while Baffin is Polaris 11. This collaborates previously released information that Ellesmere, Baffin and Greenland would be launching soon. While Baffin has been spotted shipping, Ellesmere has not, suggesting it is either a bigger more complex die or the higher end product AMD wants to keep under wraps.
Finally, we are getting some more specifications for Polaris 10. According to benchmarks spotted online, one Ellesmere configuration will be 2304 GCN cores in 32CUs running at 800-1000mhz. The chip will also have a 256bit memory bus and 8GB of VRAM running at 1500Mhz (7500Mhz effective). This places the card as a solid replacement for the 390/390X segment with a 2560 core unit being the top model. The reduced memory bus width is offset by AMD’s lossless texture compression and increased GDDR5 clocks.
Given the gains from 14nm and the new Polaris GCN architecture, it wouldn’t be unexpected for a 2560 core unit to beat Fury X as shown in the Hitman demo. This means AMD has managed to bump GCN performance by about 30%, the biggest from them since GCN was originally launched. With Pascal not expected to give much of a boost, AMD may finally seize the performance/efficiency crown from Nvidia.
Looking back, AMD missed a big opportunity to get into the mobile phone and tablet market. According to Raja Koduri, SVP for RTG, AMD may be contemplating getting back into the mobile graphics market, provided the circumstances are right.
Originally a part of ATI, the mobile graphics division, Imageon was acquired by AMD along with the parent company. After running into severe financial hardship, AMD decided to sell the mobile division off to Qualcomm which renamed it Adreno, an anagram of Radeon. Combined with their custom ARM CPUs, Qualcomm has managed to become the largest mobile SoC vendor, putting Adreno into millions of devices. The only other major competitors are Imagination and Mali from ARM itself.
By considering the mobile GPU market if the right customer comes by, AMD is opening yet another market for them to enter. Right now, Adreno is still largely based on the VLIW architecture that ATI and AMD left in 2011. GCN, on the other hand, is a more complex and advanced architecture with arguably better performance per watt. With the rise of GPU based compute being used in gaming, GCN may be a potent force in tablets.
Seeking more custom chip customers makes sense of AMD given that their consoles deals are helping keep the firm afloat as other sources of revenue are dropping. There is a large measure of risk however as Nvidia has demonstrated with their flagging Tegra lineup. By securing a customer first, AMD can pass on the risk and run a much safer course. Perhaps, the next PSP or DS will be running GCN.
In the few days after AMD first demoed Polaris 10 to us at Capsaicin, more details about the upcoming graphics cards have been revealed. Set to be the big brother to the smaller Polaris 11, the better performing chip will drop sometime after June this year.
First off, we’re now able to bring you more information about the settings Hitman was running at during the demo. At Ultra Settings and 1440p, Polaris 10 was able to keep to a constant 60FPS, with VSync being possible. This means the minimum FPS did not drop below 60 at any point. This puts the card at least above the R9 390X and on par if no better than the Fury and Fury X. Of course, the demo was done with DX12 but the boost is only about 10% in Hitman.
Another detail we have uncovered is the maximum length of the engineering sample. Based on the Cooler Master Elite 110 case used, the maximum card length is 210mm or 8.3 inches. In comparison, the Nano is 6 inches and the Fury X 7.64 inches. Given the small size, one can expect Polaris 10 to be as power efficient as Polaris 11 and potentially be using HBM. Given that Vega will be the cards to debut HBM2, Polaris 10 may be limited to 4GB of VRAM. Finally, display connectivity is provided by 3x DP 1.3, 1x HDMI 2.0 and 1 DVI-D Dual Link though OEMs may change this come launch unless AMD locks it down.
After Samsung and Nvidia had their recent legal spat, more light has been shed on the world of GPU patents and licensing. While Intel holds their own wealth of patents, no doubt some concerning GPUs, Nvidia and AMD, being GPU firms, also hold more important patents as well. With Intel’s cross-licensing deal with Nvidia set to expire in Q1 2017, the chip giant is reportedly in negotiations with AMD to strike up a patent deal.
Being one of the big two GPU designers, AMD probably has many important and critical GPU patents. Add in their experience with APUs and iGPUs, there is probably quite a lot there that Intel needs. With the Nvidia deal expiring, Intel probably sees a chance to get a better deal while getting some new patents as well. Approaching AMD also makes sense as being the smaller of the two GPU makers, AMD may be willing to share their patents for less. It’s also a way to inject some cash into AMD and keep it afloat to stave off anti-trust lawsuits.
AMD also has a lot to offer with the upcoming generation. The GPU designer’s GCN architecture is ahead of Nvidia’s when it comes to DX12 and Asynchronous Compute and that could be one area Intel is looking towards. Intel may also be forced into cross-licencing due to the fact with some many patents out there, there have to be some they are violating. The biggest question will be if AMD will consider allowing their more important and revolutionary patents to be licensed.
With the Nvidia deal being worth $66 million a quarter or $264 million a year, AMD has the chance to squeeze out a good amount of cash from Intel. Even though $264 million wouldn’t have been enough to put AMD in the black for 2015, it wouldn’t have hurt to have the extra cash.
Being the fastest single-card graphics card to date, we all know that AMD’s new Radeon Pro Duo is fast. Just how fast though is the dual-Fiji giant we don’t yet know though the 16TFOPs number and similar performance to 2 FuryX’s do give a rough estimate. To shed some light on the card, we do have some internal benchmarks of 3DMark AMD has run with their latest and great graphics card.
Testing conducted by AMD Performance Labs as of March 7, 2016 on the AMD Radeon Pro Duo, AMD Radeon R9 295X2 and Nvidia’s Titan Z, all dual GPU cards, on a test system comprising Intel i7 5960X CPU, 16GB memory, Nvidia driver 361.91, AMD driver 15.301 and Windows 10 using 3DMark Fire Strike benchmark test to simulate GPU performance. PC Manufacturers may vary configurations, yielding different results. At 1080p, 1440p, and 2160P, AMD Radeon R9 295X2 scored 16717, 9250, and 5121, respectively; Titan Z scored 14945, 7740, and 4099, respectively; and AMD Radeon Pro Duo scored 20150, 11466, and 6211, respectively, outperforming both AMD Radeon R9 295X2 and Titan Z.
According to AMD, the Radeon Pro Duo is undoubtedly the fastest card, at least according to 3dMark Firestrike. At Standard (1080p), the Pro Duo manages to have 134% of the Titan Z’s performance, a card that Nvidia priced at $2999 at launch. The lead only grows at Extreme and Ultra with 148% and 152% respectively.
Against the R9 295X2, the Pro Duo still manages a decent lead, with about a decent 120% lead across all settings. While lower than the 140% you might expect from a pure hardware standpoint, the 4GB of HBM1 and the limits of GCN do play a role. It does mean there won’t be any surprises fo users running 2 Fury or FuryX’s in CFX as they won’t have anything to worry about. The biggest question is if the card is worth the premium over running your own CFX solution, a question many dual-GPUs cards have faced.
Right before the Capsaicin event at GDC was about to begin, AMD teased everyone that they will reveal Polaris 10 running a demo for the Valve SteamVR benchmark. Unfortunately, that did not come to pass on the live stream, those of us at home still got a demo of Polaris 10 gameplay in the end.
“Showcasing next-generation VR-optimized GPU hardware – AMD today demonstrated for the first time ever the company’s forthcoming Polaris 10 GPU running Valve’s Aperture Science Robot Repair demo powered by the HTC Vive Pre. The sample GPU features the recently announced Polaris GPU architecture designed for 14nm FinFET, optimized for DirectX® 12 and VR, and boasts significant architectural improvements over previous AMD architectures including HDR monitor support, industry-leading performance-per-watt2, and AMD’s 4th generation Graphics Core Next (GCN) architecture.”
Running the latest Hitman title, Polaris 10 seemed to handle itself well enough. Performance, however, is hard to ascertain given the poor quality of the stream, unknown FPS count and unknown settings. For now, we can only speculate whether or not Polaris 10 is big Polaris or not and how it will perform in the end. Luckily, we only have to wait till June before the first Polaris chips arrive in our waiting hands.
With the reddit AMA now behind us, we can share with you some of the answers that we felt were the highlight of the question period. As expected, AMD was a bit light on details and specifics about Polaris but there were a number of important and new pieces of information.
The biggest news is that Polaris will be using the 14nm LPP FinFET process from GlobalFoundries, not a mix of 14nm LPP and 16nm FinFET+ from TSMC as previous leaked. This means Nvidia and AMD will no longer be sharing a process node for their CPUs. It also means that AMD’s GPU and CPU lineups will now be using the same process, simplifying things for APUs. Furthermore, the Taiwan earthquake that hit TSMC won’t impact Polaris yields and timeline as well. Polaris is also confirmed once again for a mid-2016 launch. Polaris will also bring Display Port 1.3 support as well.
Another confirmation is the move that AMD started in 2014 with a big annual driver release with major feature additions spaced out with point releases for specific fixes and optimizations. Those hoping that RTG would speed up driver updates to implement more features faster will be disappointed. For hardware, Fiji Gemini has already debuted for B2B customers and shipped to them but consumer launch is still waiting for HMD VR, a mistake in my mind.
Other tidbits include the fact that the LiquidVR SDK has support Affinity Multi-GPU which will allow a dedicated GPU for each eye in VR. VR is also expected to make use of TrueAudio, something that PC has shunned but consoles have picked up a bit on. There are also 25 million active daily users for AMD’s Gaming Evolved application.
Finally, AMD revealed that the optimal tessellation amount for GCN is 8-16x. Beyond that, there will be a heavy performance hit for no real gain in visual fidelity *cough* HairWorks/GameWorks. VSR or virtual super resolution is also performance free with GCN. There will be more Polaris details as we get closer to launch so stay tuned.
Stanard practice in the technology world is for established firms to keep everything super secret till the day they launch. AMD however, is bucking the trend and will be hosting an AMA (Ask Me Anything) on reddit in just a day, on Thursday Mach 3rd. The AMA will happen between 10 AM and 5 PM US Central time and will be focused on the GPU side of things or Radeon Technologies Group (RTG).
AMD Robert from RTG will be the main host and there will likely be some big secrets to be unveiled that day. Given that AMD is confident enough to start talking more about their upcoming Polaris GPUs hopefully means the launch will be imminent as some leaks as suggested. Major expected topics include Vulkan, FreeSync, GPUOpen, Polaris, Fury X2, VR and DirectX 12 as expected. The conversation will likely coalesce around Polaris though as that is the upcoming GCN architecture.
Unfortunately for CPU lovers, questions about Zen are banned and will not be answered as it is still a super secret. We will be following the AMA closely on Thursday and will bring any new juicy details AMD reveals as they happen. Here’s hoping FuryX2 finally gets off the ground.
AMD looks to be preparing things behind the scenes for their Radeon 400 series of GPUs. After a number of AMD GPUs have been spotted shipping out of Hong Kong and India, a number of new GPUs have shipped out of Canada where AMD has their Markham office with the telltale AMD labelling. Of course there is the flagship Polaris part, the Baffin XT but also the yet to be revealed Banks Pro and Weston[XT] and Weston Pro. Both Weston and Banks are northern islands in the Artic, right in line with the Northern Islands lineup for the 400 series.
Carrying the C981 label, Baffin XT is exactly what I suspected it to be, an R9 390X replacement part. This is because it carries a G5 moniker with 4GB of memory, pointing to either GDDR5X or GDDR5. The most likely configuration in my mind is GDDR5X with 256bit bus which should be enough considering AMD’s new delta colour compression techniques. The 4GB VRAM buffer does cast doubt though if it will be a Hawaii class chip, rather it may be a Tonga replacement.
Moving on we have two Weston hips, both based on the C729 die, with one being the XT and the other the Pro variant. The Banks is based off the C728 die instead but all three chips utilize GDDR5(X) and have 2GB’s worth. These 3 actually first appeared back in September 2015 and the C7xx moniker gives it away as either 28nm parts or pre-GCN 4. The 4500Mhz GDDR5 clock gives these cards away as either Oland or Cape Verde, probably making them rebrands for the budget segment. Either way, it looks like AMD will be making some major announcements at GDC next month.
With Polaris only months away, more details about what Polaris 10 and 11 will look like are coming out. According to an ex-AMD employee profile on LinkedIn, a dGPU codenamed Project F is set to be built on GlobalFounrdies and Samsung 14nm LPP process. What’s more, the Project F GPU will be 232mm2, making the chips similar in size to the 232mm2 of Pitcairn and Curacao from the 28nm generation of GCN.
At 232mm2, Project F is a relatively small chip for a new process. This should give AMD a low-risk option to evaluate the 2nd generation 14nm process from GlobalFounrdies and Samsung focused on performance. With the use of the new process, the number of transistors will grow massively due to the increased transistor density over 28nm. This should let Project F to pack in at least 4-5 million transistors, putting it at least into the R9 380 and 380X (359mm2) performance range before considering any architectural improvements.
Finally, by using 14nm, Project F is most likely the GPU that AMD showed off back at CES that severely trounced the GTX 950 in performance per watt. With the larger die size, it stands to reason that the demo unit was likely underclocked a bit or not allowed to turbo, letting it be much more efficient. Even at full power, AMD likely has a card that will be much more efficient than Maxwell while offering significantly improved performance over the last generation at a similar cost.
Ever since HBM1 was revealed and launched with Fury X, many have been looking forwards to what HBM2 would bring along in 2016. While HBM1 brought large power savings and a major boost in memory bandwidth, it was largely limited to a relatively low 4GB capacity. HBM2, however, is set to provide a boost in capacity and bandwidth by increasing the number of stackable dies. We’re now getting reports that AMD’s upcoming Polaris chips will utilize HBM2.
As a major revamp of the GCN architecture, a Polaris flagship GPU would be the natural product to debut HBM2. A flagship GPU much more powerful than current generation chips due to the new architecture and process node would likely require more memory bandwidth to feed it and a high memory capacity as it would be meant for VR and 4K gaming. Being the largest chip in the lineup, the flagship would also benefit from the major power savings, helping offset its core power consumption. The confirmation of HBM2 also suggests that we will be getting high-end Polaris chips this year.
At the same time, AMD is also confirming that they will continue to use GDDR5 and likely GDDR5X as well. At CES, AMD showed off a low powered Polaris chip using GDDR5 that was able to provide the same performance as Nvidia’s GTX 950 but with a significantly lower power consumption. With such a leap in efficiency, the HBM2 chips will likely be light years ahead of current cards in terms of efficiency if GDDR5 already shows such massive gains.
When AMD unveiled their new Polaris architecture last week, details about the timeline were a bit scarce. Other than previous confirms about a mid -2016 launch for one of the Polaris cards, AMD had been mum about when Polaris would actually arrive. Now information has surfaced about Polaris shipments has been spotted. According to a shipment tracking site, 4 Polaris cards have already been shipped around, with the earliest from Jan 2015.
The more recent models were the C981 and C980 which shipped in late December and early January. Valued at 47,578 and 40,790 INR respectively, these should fall into the product lineup around the R9 390/390X. Surprisingly, the earliest Polaris shipment, the 913, was in January 2015, nearly a year ago. this means AMD has had working Polaris silicon for nearly a year. A 924 was also shipped in August. Given their reported value, these 2 chips will likely be mid-tier about where the R7 370 and R9 380 fall.
If AMD had Polaris working for them for nearly a year, it suggests that a launch may come sooner than expected. This bolstered by the fact that AMD already has development advanced enough to show off a working demo at CES. Whatever the case, Polaris is fast approaching and will likely launch within 6 months.
First leaked yesterday, we’re now able to bring you the full AMD presentation on their upcoming Polaris GPU architecture. Set to ship mid-2016, Polaris will be using a 14/16nm FinFET process and bring massive power consumption and efficiency improvements. According to RTG SVP Raja Koduri, AMD says the massive potential that would come out of moving to FinFETs and a lower process node at the same time and decided to design a new architecture just for that.
Not surprisingly, the biggest focus is on the efficiency side of things. AMD is claiming a historic leap in performance per watt that any Radeon GPU has ever seen. The key to this is the new compute units for Polaris, or GCN 4.0. While we can expect much remains similar to past GCN revisions, the new release will be adding more parts to make the chip more efficient. A more efficient hardware scheduler, primitive discard accelerator and improved memory compression are expected to help as well. While HBM(2) will help reduce power consumption, GDDR5X models will be launched first.
To show off their improved efficiency, AMD showed off working Polaris silicon with presumably GDDR5X. Facing off against a GTX 950 in an identical system, the Polaris part managed to pull 1.65x less power from the wall as a whole. Using only 86W total compared to the 140W on the Nvidia card. This was while both cards were running Star Wars Battlefront at 1080p 60fps. Even if some of the gains are coming from the FinFETs and die shrink, that is still pretty impressive as AMD has been lagging behind on this front. The demo GPU was made using Globalfoundries 14nm FinFET process fo what it’s worth but expect some 16nm parts from TSMC as well.
Just as AMD has done with previous GCN releases, Polaris will also see other parts of the modular system upgraded. These include the Command Processor, Geometry Processor, Multimedia Cores, Display Engine, L2 Cache and Memory Controller. For the Multimedia Cores, the biggest additions are support for 4K HEVC (h.265) encoding and decoding at 60 FPS which should be welcome as well as AMD continues to push HDR. On the connectivity side, DisplayPort 1.3, and, at long last, HDMI 2.0a are supported.
For AMD, 2016 will be a critical year as both their GPU and CPU get a major architectural overhaul at the same time they get a major process upgrade. If executed well, this may finally pull AMD out of it’s slow decline and bring the firmly back into the black. Whether that happens remains to be seen.
AMD is on the roll with their latest GPU announcements this half of the year. After spinning off their graphics department into the Radeon Technologies Group, AMD announced their new GPUOpen initiative and a new Crimson Edition for their Catalyst drivers. On the hardware side of things, we’re now getting more information on their upcoming Greenland GPUs. Set to debut with their first post-GCN design; the new architecture is codenamed Polaris.
Named after the North Star, Polaris will be the “guiding lights [that] power every pixel on every device efficiently. Stars are the most efficient photon generators of our universe. Their efficiency is the inspiration for every pixel we generate.” Notably, the statement hits home on two key targets AMD is aiming for, namely better pixels in the form of HDR and improved efficiency in order to reduce power consumption.
While the Pixel initiative will undoubtedly be interesting, the biggest part will likely be the efficiency. AMD has stuck with GCN quite a long time and right now, Nvidia’s Maxwell trumps GCN 1.2 in the efficiency category. With the whole new Polaris architecture to build upon, it will be interesting to see what approach AMD has taken to gain better efficiency. Combined with the new 14/16nm process nodes and HBM2, 2016 should be a pretty good year for AMD.
AMD has largely been building off of GCN since its initial debut in 2011, with incremental improvements made via GCn 1.1 and 1.2. Combined with the limitations of having to use the 28nm process, the GPU space hasn’t seen any great gains of late. However, that is set to change soon with the arrival of AMD’s new Greenland and Nvidia’s Pascal architectures. We’re now getting some information that Greenland will launch in Summer 2016 and what’s more, jump straight to 14nm.
Greenland was rumoured to use the 14nm process from Samsung/Globalfoundries, eschewing the 16nm from TSMC, their usual supplier. If this is the case, AMD and Nvidia’s cards will not only have different architectures but finally different process nodes. Greenland manufacturing is to begin in June with the launch to be held sometime in late summer, in time for the back to school and holiday sales.
AMD had previously confirmed that the 14nm process would be used for GPUs but this is a major change in strategy. Globalfoundries and Samsung are also expected to use the 14nm process to make AMD’s next-gen Zen CPU as well. From an integration standpoint, it does make it easier for AMD to produce APUs since both the CPU and GPU will target the same node. Combined with HBM2, next years GPU and CPU launches should be pretty eventful.
AMD recently overhauled Catalyst Control Center for a more feature-rich and streamlined experience via the Radeon Crimson software suite. If you’re interested to see the new functionality or performance improvements, please check out our full review here. Unfortunately, AMD have decided to cease official support for non-GCN graphics cards and focus on making the best possible gains on the GCN architecture. As a result, various cards like the 5870 and more will not receive any future driver updates. According to AMD, this is because the GPU optimizations are maximized and clearly, they want to focus on their modernized driver user-interface and GCN enhancements. Here is the official statement from AMD in full:
“AMD Radeon HD 8000 Series (HD 8400 and below), Radeon HD 7000 Series (HD 7600 and below), Radeon HD 6000 Series, and Radeon HD 5000 Series Graphics products have reached peak performance optimization as of November 24th, 2015.
These products have been moved to a legacy support model and no additional driver releases are planned. This change enables us to dedicate valuable engineering resources to developing new features and enhancements for graphics products based on the GCN Architecture.
As a courtesy to our valued supporters we are providing a final “As Is” Beta driver together with the release of the AMD Radeon Software Crimson Edition. The AMD Radeon Software Crimson Edition Beta driver includes the redesigned user interface as well as some enhanced features. More information can be found in the AMD Radeon Software Crimson Edition Beta Driver Release Notes.
For users looking for a WHQL certified driver, the AMD Catalyst 15.7.1 driver will continue to be available for users running Windows® 10, Windows® 8.1, and Windows® 7. Users running Windows® Vista and older Microsoft® operating systems can use the AMD Catalyst 13.12 driver. These drivers are available for download on the Drivers + Download Center page.
For users passionate about gaming, you may wish to consider upgrading to a modern GPU and benefit from exciting new features, improved performance, lower power consumption and broader compatibility with the latest PC games. For information about resellers and product availability, visit: http://shop.amd.com.”
With more leaks than the Fury launch received, we’re finally getting word on the launch for AMD’s R9 380X. According to varioussources, the R9 380X will launch tomorrow on November 19th. This jives with information obtained earlier which pointed to a November 20th launch in Japan. Given the time zone differences, this all fits in for an early launch in North America for the 19th.
As expected, the 380X features a full Tonga/Antigua die with 2048 Shader Cores, 128 TMUs and 32 ROPs. Surprisingly, the core clocks are pretty low given how mature 28nm is, at only 970 base, with boost determined by the AIB partner. Memory bandwidth comes in as expected with a slight bump to 5700Mhz for 182GB/s for 4GB of GDDR5. Compared to the 280X it is replacing, the newer card manages to shave off 60W for 190W TDP.
Overall, this should make the 380X a pretty card for AMD. With pricing at $229-249 USD, the card slots right into the gap between the 960 and the 970. Efficiency should be improved enough to bring it inline with at least Kepler levels which is good for those concerned with that. Performance should be decent as well though that depends on the final clock speeds. With a 10% architectural improvement, the 380X might not be a worthy successor to the 280X just yet, but it will be enough to get those on the 660/760/960 and 265/270X to upgrade.
While much of the talk around DX12 recently has been around the reduced CPU/driver overhead and AsyncCompute, another feature is getting its first real world test. Dubbed Explicit Multi-Adapter in Microsoft’s material, the feature allows multiple GPUs that support DX12, irrespective of vendor, to work together on the same task. Developer Oxide has created a Multi-Adapter demo from their now famous Ashes of the Singularity title, using the in-house Nitrous engine.
While DX12 continues to allow the GPU drivers to allow multi-GPU setups like SLI and Crossfire, Microsoft has decided to build in a more powerful feature right into DX12. This means if the developer takes the time and effort to implement it, any DX12 title will allow any 2 DX12 card work together and boost performance. This is exactly what Anandtech tested when Oxide provided a custom build of Ashes of the Singularity with early support.
Using the built-in DX12 Multi-Adapter, top end cards like the Fury X, 980Ti and Titan X were able to show gains of between 46 to 72%. While lower than what Crossfire can offer at about 80% gains, this is pretty crazy considering the fact that it is using two cards with vastly different architectures at times from 2 different vendors. Interestingly enough, combinations with the Fury X as primary card out did those with the Nvidia card as the main one, even when the Titan X was used. This held true of older cards like the 7970 vs the 680, with the 680+7970 doing worse than just the 680 or 7970 alone. This may be due to the inherent nature of AMD’s GCN architecture being better suited to the task, but it’s still early in the game.
If developers choose to make use of this feature later on, it could make big performance boosts in teh future. Instead of having to buy two of a card, gamers can just use 1 higher performance card with a lower end one. When it comes time to upgrade, the weaker card can be tossed out and a new top-tier card takes control of the old master card. This extends to pairing up mostly unused iGPUs to get that extra bit of eye candy and fps. With control in the hands of developers and not hardware vendors, it will be interesting to see if this feature takes off.
When we first found out about AMD’s limitations for the R9 Nano, one of the biggest questions was how were the various AIB partners going to differentiate their cards. While restrictions were nothing new, AMD has traditionally been more lax. Today, we getting our first glimpse on the ASUS “custom” R9 Nano White Edition.
With major PCB and heatsink changes barred, ASUS has contented themselves with what appears to be a mere color scheme change. With a white colored stock heatsink shroud, we are given the “White Edition” of the R9 Nano. Unfortunately, the PCB is not white though that may change in future models. There are some hints that the PCB is custom but we have no confirmation of that yet. Any changes likely will be targeted towards the coil whine faced by the Nano but the layout of components should still be reference.
Even with the restrictions, the stock R9 Nano PCB and cooler are pretty good. With the addition of the White Edition, it looks like ASUS is trying to appeal for those looking for a white and black theme without having to void their warranties by modding. It will be interesting to see if there is a market for these kinds of models with minor aesthetics changes. In the end, we may be better off served by true custom Nano and Fury X cards.
From the GPU-Z screenshot, we pretty much get a good idea of the card’s performance. The 2048 shader cores, 128 TMUs and 32 ROPs all clock in at a good 1070 Mhz. Pixel fill rate comes in at 34.2 GPixel/s which is pretty much expected given the Tonga configuration. Texture fillrate is 137 GTexel/s which is much better than what the R9 380 and R9 280X boasted. 4GB of 6125Mhz GDDR5 VRAM wrap it up by giving 172GB/s via the 256bit bus. Overall, these specs place the card solidly between the R9 380 and R9 290/390.
In 3DMark 11 Extreme, the card managed to score 4024 overall with a relatively weak Intel Core i5 and 3768 in graphics. The R9 290 scores around the 4200 mark and the R9 280X at about 3300. Based off our estimates from extrapolating Tonga/GCN1.2 improvements over the R9 290X/GCN1.0, we would expect the 380X fall a bit short of the R9 290 but still surpass the GTX 780 in most cases. This is despite the 780 scoring about 3600 in 3DMark 11 since that test tends to favour Nvidia cards more.
Overall, AMD looks to have winner int he midrange with this card. Depending on the price, the 380X can steal some marketshare back from Nvidia which has a sizable gap between the 970 and 960 in terms of performance. Given some of the limitations of the 960, Nvidia may want to consider a cut-down 970 that is not memory bottlenecked in order to do battle. As one of the last 28nm and GCN cards, AMD is making sure to go out with a bang.