NVIDIA Announces Hybrid SLI Technology
You've certainly heard of Hybrid SLI being in the works for ages, but today at CES, NVIDIA officially announced Hybrid SLI technology and a slew of products for this new ecosystem. What a way to start 2008 with a bright upswing and we've got the full details for you.
By Vijay Anand -
Hybrid SLI - The Much Awaited Advancement
2007 has been a busy year for NVIDIA in both the graphics and desktop solutions department and to prove that they've the momentum going strong, no sooner than a week in to the brand new year, NVIDIA announced several new updates concerning their 'secretive' Hybrid SLI technology. We've actually mentioned that the big green has been working on Hybrid SLI technology for a long time and we ourselves have been predicting such a concept even before that. The idea of SLI (Scalable Link Interface) itself was a major step in the graphics performance evolution back in 2004 and today, they are at it again to further progress this concept.
NVIDIA's Hybrid SLI is addresses two key concerns of cranking up graphics performance ever higher (as well as maximizing graphics silicon) and reducing power consumption. To deal with them, Hybrid SLI is a combination of two sub technologies - GeForce Boost and HybridPower.
GeForce Boost - The First Enabler of Hybrid SLI
GeForce Boost is a technology that provides discrete GPUs (dGPU) more graphics processing prowess when used on an NVIDIA motherboard GPU (mGPU) platform. This is the main concept behind Hybrid SLI where users were hoping to leverage performance of both a motherboard or notebook's integrated GPU capabilities in addition to the discrete GPU for a combined performance boost than any one solution alone. The closest implementation of this to date is a manual switch that toggles between either a discrete or integrated GPU option on a few select notebooks such as the Sony SZ series. With the newer NVIDIA platforms supporting Hybrid SLI, this will be a thing of the past. Our more technical readers might be wondering how two very unlike GPUs can be made to work together to boost performance as there's bound to be syncing or latency issues. For the moment, NVIDIA hasn't delved into the details, but they did attribute the bulk of this capability is thanks to the driver development team who've incorporated advanced intelligence and balancing techniques to split work load appropriately between the mGPU and dGPU respectively. Naturally, the discrete GPU is going to be doing the bulk of the actual rendering while the motherboard integrated GPU tags along to boost performance to its best ability.
How GeForce Boost works is that the driver intelligently figures out what portion of the workload is to be handled by each GPU and dispatches the respective graphics commands to both GPUs. Each GPU will work to render a separate set of frames, and the result of which is stored in their respective frame buffers. Block transfers (BLT) then occur to transfer the discrete GPU's rendered image to the mGPU's frame buffer because the primary output display will be from the integrated GPU.
After receiving workload instructions from the graphics drivers, each GPU will work to render a separate set of frames, and the result of which is stored in their respective frame buffers.
Block transfers (BLT) then occur to transfer the discrete GPU's rendered image to the mGPU's frame buffer.
mGPU then outputs the display to the screen.
HybridPower - The Second Equation of Hybrid SLI
HybridPower is perhaps the most enticing part of going the route of Hybrid SLI as you can finally choose between high performance and low power computing. Given an SLI setup on a Hybrid SLI ready platform, many standard applications require no more processing power than the motherboard integrated GPU. As such, when HybridPower is enabled, the drivers automatically detect the running applications and operating state to shut off the discrete GPU(s). This is done via the tried and trusted SMBUS interface that can initiate shutting down the dGPUs, as well as power them up. The driver technology signals the SMBUS to shut down the dGPU when not required. When a 3D application is invoked which requires more performance, SMBUS can re-enable the dGPU, which will then work in union to boost performance (GeForce Boost). The beauty of this is that switching occurs on-the-fly and is seamless to end-users.
HybridPower is perhaps the most enticing part of going the route of Hybrid SLI as you can finally choose between high performance and low power computing - either via profiles or just letting the drivers figure it out intelligently.
And when NVIDIA means shutting down of the discrete GPUs, they literally mean it. It would be as though the discrete GPUs were never plugged in at all. With zero power consumption of the dGPUs, you can also expect far lower operating noise levels - perfect for movie watching utilizing the onboard integrated GPU.
This brings us to another point that we've touched on in our tech articles. To date, no integrated motherboard GPU is capable of handling Full HD video due to inadequate local memory bandwidth. However, NVIDIA promised that the mGPU that will be featured in the new Hybrid SLI capable platforms will tackle this smoothly due to advances in compression algorithms and efficient memory utilization. Thus this will answer any queries as to how Hybrid SLI would tackle calls to process Full HD videos, which is all tackled by the new mGPU. In addition to the intelligence packed within the drivers, the control panel will also allow end users to utilize profiles to set options as to which applications being invoked should use Hybrid SLI features/technologies. For the moment, we are certain that the full functionality of Hybrid SLI will function under the Windows Vista OS, but there are no clear indications of Windows XP or other operating systems support.
The Hybrid SLI Ecosystem
So far, Hybrid SLI looks very appealing. However, we hate to break the news that no current motherboard supports the Hybrid SLI initiative and almost most of the discrete GPUs are not qualified for Hybrid SLI. For that, you'll need new hardware.
As such, today's official launch of the Hybrid SLI technology also marks that start of a new ecosystem of products supporting Hybrid SLI. The first group of platforms is targeted at AMD processors based on the new nForce 780a SLI, nForce 750a SLI, nForce 730a and the GeForce 8200 MCPs. Intel processor based equivalents are slated for the second quarter of the year though.
NVIDIA MCPs | Target User Segment | Target Motherboard Form Factor | Hybrid SLI Technology | Other Features | |
HybridPower | GeForce Boost | ||||
nForce 780a SLI | Enthusiast (US$250+) | ATX | Yes | Yes | 3-way SLI, ESA, SLI Memory |
nForce 750a SLI | Performance Gamer (US$120+) | ATX | Yes | Yes | 2-way SLI |
nForce 730 | Mainstream Gamer (US$80+) | ATX | No | Yes | 1 x16 PCIe 2.0 |
GeForce 8200 | Mainstream Vista Premium Users | microATX | Yes | Yes | 1 x16 PCIe 2.0 |
The nForce 780a SLI and its Derivative MCPs
The nForce 780a SLI mostly shares the same capabilities of the nForce 780i SLI launched late last year, while the nForce 750a SLI is a performance oriented board minus the enthusiast touches. Both the nForce 780a and 750a platforms have integrated GPU solutions that incorporate the GPU prowess equivalent to that on the GeForce 8200 MCP - which has the graphics crunching power to tackle Full HD videos with its PureVideo HD engine (though NVIDIA declined to comment on which VP engine it is running at the moment). Take note however that the mainstream nForce 730a does not support HybridPower but will still feature GeForce Boost for the added performance.
A block diagram of the nForce 780a SLI chipset
The nForce 200 companion chip to the nForce 780a SLI and nForce 750a SLI handles the full bandwidth PCIe 2.0 busses on these motherboard platforms just as it has done so for the Intel equivalent platform. However, NVIDIA has recently been awarded patents to some of the technologies incorporated on the nForce 200 chip which were not shared previously disclosed and we'll cover them in brief:-
- PW Shortcut - When one GPU issues a command to other GPUs to perhaps move/transform a particular set of vertices or color a section red, this instruction normally goes all the way to the CPU before it is directed back to the chipset which again directs it to the nForce 200 to dispatch the command to the respective GPUs. Now with PW Shortcut in the nForce 200, it intercepts such command requests and immediately issues to multiple GPUs (or any one GPU) without having to intervene the CPU/chipset as well conserve precious bandwidth and time.
- CPU Broadcast - When a graphics instruction is dispatched from the CPU, this feature in the nForce 200 helps to duplicate and target multiple GPUs, thus conserving PCIe bandwidth and CPU cycles to reissue the instruction multiple times.
The GeForce 8200 MCP
The GeForce 8200 MCP itself is targeted as the new low cost entry level DX10 capable GPU that has potential to set a new baseline for Vista Premium mGPU certified board. Other features of the GeForce 8200 MCP is its PureVideo HD engine that fully offloads processing of all CODECs used on Blu-ray and HD DVD discs (MPEG-2 , VC-1 and H.264) so that low cost CPUs can be used, Hybrid SLI, MediaShield with RAID 5 and Gigabit with FirstPacket technologies. The GeForce 8200 is mighty robust for a single-chip platform solution.
A block diagram of the GeForce 8200 MCP chip
GPUs need to be Hybrid SLI Compliant too!
Requiring the correct motherboard support is to be expected of a technology like Hybrid SLI, but apparently, even the GPUs need to be designed for this and that sort of put a damper on our high spirits for this new technology. According to NVIDIA's recommendation list, only the 'next generation' GPUs are capable of supporting Hybrid SLI in full swing (GeForce Boost and HybridPower). While NVIDIA was not yet ready to comment on 'next generation' graphics cards, they did hint that these cards will be available in the first quarter of the year. From the grapevines, we hear that these are the new GeForce 9 series of GPUs based on the 55nm process technology. Since we are already in the first quarter of 2008, expect official announcements and launches coming up quite shortly.
Oddly, the GeForce 8500 GT and GeForce 8400 GS are the only currently qualified GPUs to support GeForce Boost to increase graphics rendering power in combination with a qualified NVIDIA mGPU board. However they do not support HybridPower and thus do not support the full capabilities of Hybrid SLI. Stay tuned for more updates in the coming days and weeks as NVIDIA reveals more information.
** Updated on 17th January 2008**
After more email exchanges with respective NVIDIA managers, we obtained more information on why only low-end cards like GeForce 8500 GT and GeForce 8400 GS will officially support GeForce Boost. If you recall, we mentioned that GeForce Boost hedges upon the driver intelligence to adequately support two unlike GPUs for SLI performance boost. However, it seems that further email exchanges with the NVIDIA managers revealed that Hybrid SLI via GeForce Boost will scale best when the mGPU is paired with a performance comparable dGPU. As such, GeForce Boost is able to compensate for some imbalance, but not for a complete imbalance when pairing a high-end dGPU like a GeForce 8800 Ultra with the mGPU. If such a scenario does happen, NVIDIA further commented that the drivers will automatically detect the system platform to disable GeForce Boost.
This certainly explains why NVIDIA has only qualified low-end cards like GeForce 8500 GT, GeForce 8400 GS and equivalent next generation cards. So while the common DIY gamer isn't able to enjoy GeForce Boost for Hybrid SLI, a mainstream desktop PC and quite a bit of the mobile computing audience are likely to benefit from some performance increases when combining the appropriate mGPU with a dGPU of this class (which is quite common these days in several notebooks).
Closing Remarks
Hybrid SLI is truly an extension of the original SLI concept and we're certain that it will only further rally users to support the SLI ecosystem (or more specifically, the Hybrid SLI ecosystem). No doubt that this is yet another blow to the AMD graphics group, you have to realize for NVIDIA's new Hybrid SLI ecosystem to work, you'll need to hop on to a new system altogether with a new motherboard and graphics cards. This is indeed an expensive proposition and not merely a platform upgrade as we initially suspected. Needless to say, it will take a while before a sizeable user group is formed with Hybrid SLI ready platforms and it is anybody's guess if AMD might seize this pie as well. Already, NVIDIA's nForce 780i platforms are slow to penetrate the enthusiast market space and this might be a concern as to the perceived consumer's confidence.
NVIDIA's new Hybrid SLI platform for the performance and Enthusiast folks finally offers more performance as well more power savings. A hard to resist proposition, but it comes at the price of renewing your system.
Still, new technologies must start somewhere and this concern of needing to literally hop on to a new system would diminish in time to come. After all, this requirement is nothing new to the DIY world where embracing new technology to its best would mean disbanding an existing system for a new one. Now if NVIDIA continues to innovate and push out new products in time to keep the pressure on ATI as it has been doing for quite a while now, NVIDIA can certainly stay in the lead. Their execution for the graphics card market has been met with great success, but it hasn't been followed through the same extent in the motherboard market, but we hope NVIDIA does buck up in this new year and better coordinate with motherboard manufacturers. Hybrid SLI's GeForce Boost (though only for low-end, mainstream users) and HybridPower has something for everyone to look forward to and we can't wait to take it to try it out firsthand.
Our articles may contain affiliate links. If you buy through these links, we may earn a small commission.