High Bandwidth Memory
UPDATE: I have embedded an excerpt from our PC Perspective Podcast that discusses the HBM technology that you might want to check out in addition to the story below.
The chances are good that if you have been reading PC Perspective or almost any other website that focuses on GPU technologies for the past year, you have read the acronym HBM. You might have even seen its full name: high bandwidth memory. HBM is a new technology that aims to turn the ability for a processor (GPU, CPU, APU, etc.) to access memory upside down, almost literally. AMD has already publicly stated that its next generation flagship Radeon GPU will use HBM as part of its design, but it wasn’t until today that we could talk about what HBM actually offers to a high performance processor like Fiji. At its core HBM drastically changes how the memory interface works, how much power is required for it and what metrics we will use to compare competing memory architectures. AMD and its partners started working on HBM with the industry more than 7 years ago, and with the first retail product nearly ready to ship, it’s time to learn about HBM.
We got some time with AMD’s Joe Macri, Corporate Vice President and Product CTO, to talk about AMD’s move to HBM and how it will shift the direction of AMD products going forward.
The first step in understanding HBM is to understand why it’s needed in the first place. Current GPUs, including the AMD Radeon R9 290X and the NVIDIA GeForce GTX 980, utilize a memory technology known as GDDR5. This architecture has scaled well over the past several GPU generations but we are starting to enter the world of diminishing returns. Balancing memory performance and power consumption is always a tough battle; just ask ARM about it. On the desktop component side we have much larger power envelopes to work inside but the power curve that GDDR5 is on will soon hit a wall, if you plot it far enough into the future. The result will be either drastically higher power consuming graphics cards or stalling performance improvements of the graphics market – something we have not really seen in its history.
While it’s clearly possible that current and maybe even next generation GPU designs could still have depended on GDDR5 as the memory interface, the move to a different solution is needed for the future; AMD is just making the jump earlier than the rest of the industry.
Subject: Graphics Cards | May 17, 2015 - 12:04 PM | Ryan Shrout
Tagged: The Witcher 3, nvidia, hairworks, gameworks, amd
I feel like every few months I get to write more stories focusing on the exact same subject. It's almost as if nothing in the enthusiast market is happening and thus the cycle continues, taking all of us with it on a wild ride of arguments and valuable debates. Late last week I started hearing from some of my Twitter followers that there were concerns surrounding the upcoming release of The Witcher 3: Wild Hunt. Then I found a link to this news post over at Overclock3d.net that put some of the information in perspective.
Essentially, The Witcher 3 uses parts of NVIDIA's GameWorks development tools and APIs, software written by NVIDIA to help game developers take advantage of new technologies and to quickly and easily implement them into games. The problem of course is that GameWorks is written and developed by NVIDIA. That means that optimizations for AMD Radeon hardware are difficult or impossible, depending on who you want to believe. Clearly it doesn't benefit NVIDIA to optimize its software for AMD GPUs financially, though many in the community would like NVIDIA to give a better effort - for the good of said community.
Specifically in regards to The Witcher 3, the game implements NVIDIA HairWorks technology to add realism on many of the creatures of the game world. (Actually, the game includes HairWorks, HBAO+, PhysX, Destruction and Clothing but our current discussion focuses on HairWorks.) All of the marketing and video surrounding The Witcher 3 has been awesome and the realistic animal fur simulation has definitely been a part of it. However, it appears that AMD Radeon GPU users are concerned that performance with HairWorks enabled will suffer.
An example of The Witcher 3: Wild Hunt with HairWorks
One of the game's developers has been quoted as such:
Many of you have asked us if AMD Radeon GPUs would be able to run NVIDIA’s HairWorks technology – the answer is yes! However, unsatisfactory performance may be experienced as the code of this feature cannot be optimized for AMD products. Radeon users are encouraged to disable NVIDIA HairWorks if the performance is below expectations.
There are at least several interpretations of this statement floating around the web. First, and most enflaming, is that NVIDIA is not allowing CD Project Red to optimize it by not offering source code. Another is that CD Project is choosing to not optimize for AMD hardware due to time considerations. The last is that it simply isn't possible to optimize it because of hardware limitations of HairWorks.
I went to NVIDIA with these complaints about HairWorks and Brian Burke gave me this response:
We are not asking game developers do anything unethical.
GameWorks improves the visual quality of games running on GeForce for our customers. It does not impair performance on competing hardware.
Demanding source code access to all our cool technology is an attempt to deflect their performance issues. Giving away your IP, your source code, is uncommon for anyone in the industry, including middleware providers and game developers. Most of the time we optimize games based on binary builds, not source code.
GameWorks licenses follow standard industry practice. GameWorks source code is provided to developers that request it under license, but they can’t redistribute our source code to anyone who does not have a license.
The bottom line is AMD’s tessellation performance is not very good and there is not a lot NVIDIA can/should do about it. Using DX11 tessellation has sound technical reasoning behind it, it helps to keep the GPU memory footprint small so multiple characters can use hair and fur at the same time.
I believe it is a resource issue. NVIDIA spent a lot of artist and engineering resources to help make Witcher 3 better. I would assume that AMD could have done the same thing because our agreements with developers don’t prevent them from working with other IHVs. (See also, Project Cars)
I think gamers want better hair, better fur, better lighting, better shadows and better effects in their games. GameWorks gives them that.
Interesting comments for sure. The essential take away from this is that HairWorks depends heavily on tessellation performance and we have known since the GTX 680 was released that NVIDIA's architecture performs better than AMD's GCN for tessellation - often by a significant amount. NVIDIA developed its middleware to utilize the strength of its own GPU technology and while it's clear that some disagree, not to negatively impact AMD. Did NVIDIA know that would be the case when it was developing the software? Of course it did. Should it have done something to help AMD GPUs more gracefully fall back? Maybe.
Next, I asked Burke directly if claims that NVIDIA was preventing AMD or the game developer from optimizing HairWorks for other GPUs and platforms were true? I was told that both AMD and CD Project had the ability to tune the game, but in different ways. The developer could change the tessellation density based on the specific GPU detected (lower for a Radeon GPU with less tessellation capability, for example) but that would require dedicated engineering from either CD Project or AMD to do. AMD, without access to the source code, should be able to make changes in the driver at the binary level, similar to how most other driver optimizations are built. Burke states that in these instances NVIDIA often sends engineers to work with game developers and that AMD "could have done the same had it chosen to." And again, NVIDIA reiterated that in no way do its agreements with game developers prohibit optimization for AMD GPUs.
It would also be possible for AMD to have pushed for the implementation of TressFX in addition to HairWorks; a similar scenario played out in Grand Theft Auto V where several vendor-specific technologies were included from both NVIDIA and AMD, customized through in-game settings.
NVIDIA has never been accused of being altruistic; it doesn't often create things and then share it with open arms to the rest of the hardware community. But it has to be understood that game developers know this as well - they are not oblivious. CD Project knew that HairWorks performance on AMD would be poor but decided to implement the technology into The Witcher 3 anyway. They were willing to sacrifice performance penalties for some users to improve the experience of others. You can argue that is not the best choice, but at the very least The Witcher 3 will let you disable the HairWorks feature completely, removing it from the performance debate all together.
In a perfect world for consumers, NVIDIA and AMD would walk hand-in-hand through the fields and develop hardware and software in tandem, making sure all users get the best possible experience with all games. But that style of work is only helpful (from a business perspective) for the organization attempting to gain market share, not the one with the lead. NVIDIA doesn't have to do it and chooses to not. If you don't want to support that style, vote with your wallet.
Another similar controversy surrounded the recent release of Project Cars. AMD GPU performance was significantly lower than comparable NVIDIA GPUs, even though this game does not implement any GameWorks technologies. In that case, the game's developer directly blamed AMD's drivers, saying that it was a lack of reaching out from AMD that caused the issues. AMD has since recanted its stance that the performance delta was "deliberate" and says a pending driver update will address gamers performance issues.
All arguing aside, this game looks amazing. Can we all agree on that?
The only conclusion I can come to from all of this is that if you don't like what NVIDIA is doing, that's your right - and you aren't necessarily wrong. There will be plenty of readers that see the comments made by NVIDIA above and continue to believe that they are being at best disingenuous and at worst, are straight up lying. As I mentioned above in my own comments NVIDIA is still a for-profit company that is responsible to shareholders for profit and growth. And in today's world that sometimes means working against other companies than with them, resulting in impressive new technologies for its customers and push back from competitor's customers. It's not fun, but that's how it works today.
Fans of AMD will point to G-Sync, GameWorks, CUDA, PhysX, FCAT and even SLI as indications of NVIDIA's negative impact on open PC gaming. I would argue that more users would look at that list and see improvements to PC gaming, progress that helps make gaming on a computer so much better than gaming on a console. The truth likely rests somewhere in the middle; there will always be those individuals that immediately side with one company or the other. But it's the much larger group in the middle, that shows no corporate allegiance and instead just wants to have as much fun as possible with gaming, that will impact NVIDIA and AMD the most.
So, since I know it will happen anyway, use the comments page below to vent your opinion. But, for the benefit of us all, try to keep it civil!
Subject: Graphics Cards | May 15, 2015 - 11:36 PM | Scott Michaud
Tagged: windows 10, geforce, graphics drivers, nvidia, whql
The last time that NVIDIA has released a graphics driver for Windows 10, they added a download category to their website for the pre-release operating system. Since about January, graphics driver updates were pushed by Windows Update and, before that, you would need to use Windows 8.1 drivers. Receiving drivers from Windows Update also meant that add-ons, such as PhysX runtimes and the GeForce Experience, would not be bundled with it. I know that some have installed them separately, but I didn't.
The 352.84 release, which is their second Windows 10 driver to be released outside of Windows Update, is also certified by WHQL. NVIDIA has recently been touting Microsoft certification for many of their drivers. Historically, they released a large number of Beta drivers that were stable, but did not wait for Microsoft to vouch for them. For one reason or another, they have put a higher priority on that label, even for “Game Ready” drivers that launch alongside a popular title.
For some reason, the driver is only available via GeForce Experience and NVIDIA.com, but not GeForce.com. I assume NVIDIA will publish it there soon, too.
Subject: Graphics Cards, Processors, Displays, Systems | May 15, 2015 - 03:02 PM | Scott Michaud
Tagged: Oculus, oculus vr, nvidia, amd, geforce, radeon, Intel, core i5
Today, Oculus has published a list of what they believe should drive their VR headset. The Oculus Rift will obviously run on lower hardware. Their minimum specifications, published last month and focused on the Development Kit 2, did not even list a specific CPU or GPU -- just a DVI-D or HDMI output. They then went on to say that you really should use a graphics card that can handle your game at 1080p with at least 75 fps.
The current list is a little different:
- NVIDIA GeForce GTX 970 / AMD Radeon R9 290 (or higher)
- Intel Core i5-4590 (or higher)
- 8GB RAM (or higher)
- A compatible HDMI 1.3 output
- 2x USB 3.0 ports
- Windows 7 SP1 (or newer).
I am guessing that, unlike the previous list, Oculus has a more clear vision for a development target. They were a little unclear about whether this refers to the consumer version or the current needs of developers. In either case, it would likely serve as a guide for what they believe developers should target when the consumer version launches.
This post also coincides with the release of the Oculus PC SDK 0.6.0. This version pushes distortion rendering to the Oculus Server process, rather than the application. It also allows multiple canvases to be sent to the SDK, which means developers can render text and other noticeable content at full resolution, but scale back in places that the user is less likely to notice. They can also be updated at different frequencies, such as sleeping the HUD redraw unless a value changes.
The Oculus PC SDK (0.6.0) is now available at the Oculus Developer Center.
Subject: Graphics Cards | May 14, 2015 - 07:00 AM | Scott Michaud
Tagged: tonga, radeon, R9, pitcairn, Fiji, bonaire, amd
Benchlife.info, via WCCFTech, believes that AMD's Radeon R9 300-series GPUs will launch in late June. Specifically, the R9 380, the R7 370, and the R7 360 will arrive on the 18th of June. These are listed as OEM parts, as we have mentioned on the podcast, which Ryan speculates could mean that the flagship Fiji XT might go by a different name. Benchlife.info seems to think that it will be called by the R9 390(X) though, and that it will be released on the 24th of June.
WCCFTech is a bit more timid, calling it simply “Fiji XT”.
In relation to industry events, this has the OEM lineup launching on the last day of E3 and Fiji XT launching in the middle of the following week. This feels a little weird, especially because AMD's E3 event with PC Gamer is on the 16th. While it makes sense for AMD to announce the launch a few days before it happens, that doesn't make sense for OEM parts unless they were going to announce a line of pre-built PCs. The most likely candidate to launch gaming PCs is Valve, and they're one of the few companies that are absent from AMD's event.
And this is where I run out of ideas. Launching a line of OEM parts at E3 is weird unless it was to open the flood gates for OEMs to make their own announcements. Unless Valve is scheduled to make an announcement earlier in the day, or a surprise appearance at the event, that seems unlikely. Something seems up, though.
Subject: Graphics Cards | May 12, 2015 - 12:11 PM | Ryan Shrout
Tagged: GTX 980, giveaway, contest, asus, anniversary
I bet many of you are going to feel a little old as you read this, but ASUS is celebrating its 20th anniversary of being in the graphics card market! Starting in 1996, when ASUS launched its first discrete graphics product based on the S3 Virge/DX GPU and running through 2015 with the release of updated GTX 900-series and AMD Radeon 200-series cards, ASUS has been a pivotal part of the GPU landscape.
To celebrate this achievement, ASUS built a new, gold, 20th anniversary edition of its GeForce GTX 980 product, limited to only 200 units in all of North America! And the best part? You can win one right here on PC Perspective for FREE!
This amazing graphics card has some killer features:
- GTX 980 GPU
- 4GB GDDR5 memory at 7.0 GHz
- 1317 MHz Base / 1431 MHz Boost clocks
- 8+8-pin power connections
- DirectCU II 0db Fan Technology
- Memory Defroster
- "Safe Mode" VBIOS reload
- 14-phase Super Alloy power
How do you enter for this prize? Pay attention, there are some specifics!
- First and foremost, you must be subscribed to the ASUS PCDIY email list, which you can sign up for at http://pcdiy.asus.com/. Repeat: this is a requirement!
- You must leave a comment on the news story here and tell us what you love about ASUS graphics cards or the new 20th anniversary GTX 980 Special Edition!
- This is a global contest - so feel free to enter from anywhere in the world!
- You can gain other entries by utilizing the Gleam.io form below!
The contest will end on May 15th at 9pm ET so be sure to get your entries in early!
Subject: Graphics Cards | May 6, 2015 - 02:21 PM | Ryan Shrout
Tagged: amd, hbm, radeon, gpu
During today's 2015 AMD Financial Analyst Day, CEO Dr. Lisa Su discussed some of the details of the upcoming enthusiast Radeon graphics product. Though it wasn't given a name, she repeatedly said that the product would be announced "in the coming weeks...at upcoming industry events."
You won't find specifications here but understanding the goals and targets that AMD has for this new flagship product will help tell the story of this new Radeon product. Dr. Su sees AMD investing at very specific inflection points, the most recent of which are DirectX 12, 4K displays and VR technology. With adoption of HBM (high bandwidth memory) that sits on-die with the GPU, rather than across a physical PCB, we will see both a reduction in power consumption as well as a significant increase in GPU memory bandwidth.
HBM will accelerate the performance improvements at those key inflection points Dr. Su mentioned. Additional memory bandwidth will aid the ability for discrete GPUs to push out 4K resolutions and beyond, no longer limited by texture sizes. AMD's LiquidVR software, in conjunction with HBM, will be able to improve latency and reduce performance concerns on current and future generations of virtual reality hardware.
One interesting comment made during the conference was that HBM would enable new form factors for the GPUs now that you now longer need to have memory spread out on a PCB. While there isn't much room in the add-in card market for differentiation, in the mobile space that could mean some very interesting things for higher performance gaming notebooks.
Mark Papermaster, AMD CTO, said earlier in the conference call that HBM would aid in performance but maybe more importantly will lower power and improve total GPU efficiency. HBM will offer more than 3x improved performance/watt compared to GDDR5 while also running more than 50% lower power than GDDR5. Lower power and higher performance upgrades don't happen often so I am really excited to see what AMD does with it.
There weren't any more details on the next flagship Radeon GPU but it doesn't look like we'll have to wait much longer.
Subject: Graphics Cards | May 5, 2015 - 12:46 PM | Sebastian Peak
Tagged: The Witcher 3, nvidia, GTX 980, GTX 970, gtx, geforce, batman arkham knight
NVIDIA has announced a new game bundle for GeForce graphics cards starting now, and it’s a doozy: Purchase a qualifying GTX card and receive download codes for both Witcher 3: Wild Hunt and Batman: Arkham Knight!
Needless to say, both of these titles have been highly anticipated, and neither have been released just yet with Witcher 3: Wild Hunt due to be released on May 19, and Batman: Arkham Knight arriving on June 23. So which cards qualify? Amazon has a page specifically created for this new offer here, and depending on which card you select you’ll be eligible for either both upcoming games, or just Witcher 3. NVIDIA has this chart on their promotion page for reference:
So GeForce GTX 980 and GTX 970 cards will qualify for both games, and GTX 960 cards and the mobile GPUs qualify for Witcher 3 alone. Regardless of which game or card you may choose these PC versions of the new games will feature graphics that can’t be matched by consoles, and NVIDIA points out the advantages in their post:
"Both Batman and The Witcher raise the bar for graphical fidelity, rendering two very different open worlds with a level of detail we could only dream of last decade. And on PC, each is bolstered by NVIDIA GameWorks effects that increase fidelity, realism, and immersion. But to use those effects in conjunction with the many other available options, whilst retaining a high frame rate, it’s entirely possible that you’ll need an upgrade."
NVIDIA also posted this Witcher 3: Wild Hunt behind-the-scenes video:
In addition to the GameWorks effects present in Witcher 3, NVIDIA worked with developer Rocksteady on Batman: Arkham Knight “to create new visual effects exclusively for the game’s PC release.” The post elaborates:
"This is in addition to integrating the latest and greatest versions of our GPU-accelerated PhysX Destruction, PhysX Clothing and PhysX Turbulence technologies, which work in tandem to create immersive effects that react realistically to external forces. In previous Batman: Arkham games, these technologies created immersive fog and ice effects, realistic destructible scenery, and game-enhancing cloth effects that added to the atmosphere and spectacle. Expect to see similar effects in Arkham Knight, along with new features and effects that we'll be talking about in depth as we approach Arkham Knight’s June 23rd release."
Of note, mobile GPUs included in the promotion will only receive download codes if the seller of the notebook is participating in the "Two Times The Adventure" offer, so be sure to check that out when looking for a gaming laptop that qualifies!
The promotion is going on now and is available “for a limited time or while supplies last”.
DirectX 12 Has No More Secrets
The DirectX 12 API is finalized and the last of its features are known. Before the BUILD conference, the list consisted of Conservative Rasterization, Rasterizer Ordered Viewed, Typed UAV Load, Volume Tiled Resources, and a new Tiled Resources revision for non-volumetric content. When the GeForce GTX 980 launched, NVIDIA claimed it would be compatible with DirectX 12 features. Enthusiasts were skeptical, because Microsoft did not officially finalize the spec at the time.
Last week, Microsoft announced the last feature of the graphics API: Multiadapter.
We already knew that Multiadapter existed, at least to some extent. It is the part of the specification that allows developers to address multiple graphics adapters to split tasks between them. In DirectX 11 and earlier, secondary GPUs would remain idle unless the graphics driver sprinkled some magic fair dust on it with SLI, CrossFire, or Hybrid CrossFire. The only other way to access this dormant hardware was by spinning up an OpenCL (or similar compute API) context on the side.
Subject: Graphics Cards | April 30, 2015 - 08:28 PM | Sebastian Peak
Tagged: PC Gamer, gpu, Fiji, E3 2015, amd
We haven’t had much more than rumor and speculation about upcoming AMD graphics for a while now, but there is more than enough fresh fuel for the GPU fire today to ignore completely. It seems that AMD and PC Gamer magazine have teamed up to announce a special (what else) PC gaming event at this year’s E3 show on June 16, and this would be the perfect place for some new hardware announcements.
Not enough for you? Well, while the AMD Fiji GPU rumors are nothing new to followers of industry news, it has now been indirectly announced that the upcoming Fiji GPU from AMD will in fact feature 2.5D high-bandwidth memory (HBM). As reported by tech news/rumor site wccftech the announcement came via the official schedule for the upcoming Hot Chips symposium, which is slated for August 23-25 in Cupertino, California.
This screenshot was taken this morning from the official online event schedule
(Note: This part of the day 2 schedule has now been changed to read “AMD’s Next Generation GPU and Memory Architecture”, with all mention of Fiji and HBM removed.)
Whether this gives us insight into the actual release date of the long-awaited Fiji GPU from AMD is unclear, but new AMD GPU products certainly seem to be imminent as we move into the summer months. Speculation is fun (for a while), but hopefully the PC gaming event at E3 in June will provide at least some official news from AMD on the new GPU products we've been waiting for.