AMD 官方 Blog 对独立 GPU 存废的回答: H.E.L.L No.

Edison

August 6, 2010 by Godfrey Cheng


Share



One of the questions I get asked frequently is:  “When AMD rolls out Fusion APUs, will discrete GPUs die out, or become ‘mostly dead’ like Wesley in Princess Bride?”  The answer is not a simple “yes” or “no,” but I believe that the best days of the GPU are ahead of it. In fact, according to Mercury Research, AMD is now the market share leader for all discrete GPUs, and we intend to expand that lead.

Allow me to set the stage for my explanation with some brief definitions:

Discrete graphics processing unit (GPU):  In simple terms, it is a graphics processor that is separate from both the chipset and CPU; it has its own memory, connects to the system through PCI Express, and is usually found on an add-in graphics card.
Accelerated porciessing unit (APU):  Also in simple terms, it’s an x86 CPU combined with the memory controller and graphics processor.
CPU & GPU: The traditional CPU is primarily a scalar processor, while the GPU is primarily a vector processor.
The eventual creation of the APU was one of the key reasons for AMD’s acquisition of ATI in 2006.  This acquisition brought to AMD industry-leading GPU and chipset technologies, and a software team to support them.  As we sit on the precipice of launching the first generation of APUs, people are wondering if discrete GPUs will disappear altogether.

To best answer that question, you have to know why discrete graphics exist in the first place. One of the main reasons people want a discrete GPU is  to improve gaming performance.  Gamers have remained loyal to AMD & ATI for more than a decade – a decade that has seen some pretty incredible changes in the industry.  This segment shows no signs of disappearing anytime soon and we intend to continue to provide enthusiasts with leading discrete graphics solutions.  In fact, we have several generations of APUs and GPUs in development right now! Sigh of relief No.1.

When designing silicon products, one must make tradeoffs in performance, power and size (cost).  When we designed our initial APUs, we chose to place our emphasis on reducing power consumption and cost.  In so doing we chose a level of graphics performance lower than that provided by our highest end discrete GPUs.  Here is another way to look at it. An APU must run the whole system, which requires a mix of vector processors, scalar processors and cache memory. A discrete GPU is dedicated to graphics and parallel compute, so almost all the silicon can be dedicated to vector processing. So area for area, a discrete GPU will pack more punch for the gamer or video enthusiast.  Sigh of relief No.2.

AMD’s APUs are designed to deliver exceptional performance – and we believe this will be readily apparent when compared to integrated graphics products available from the competition. Some of our APUs, by sheer performance alone, are expected to be faster than some discrete GPUs on the market. And as our GPU cores improve, you can expect our APU graphics performance to similarly improve.  AMD’s strategy is to leverage these world-class graphics cores and continue to raise the bar, as we have always done.  OK, you get it.  AMD’s APUs are designed to be as fast as some discrete GPUs but not as fast as the higher end discrete GPUs.

But this is just the start of where AMD plans to innovate…

To really answer the question “will discrete GPUs die out?” we need to look at the quantum level.  Despite having the power budget for nearly unbounded performance, one of the bottlenecks for discrete GPUs is PCI Express, the interface to the system.  In the case of graphics workloads, normally PCI Express does not present a constant bottleneck, however, everyone and their dog has heard of parallel computing.  This is the case where there is a lot of traffic between the discrete GPU, CPU and main memory, and PCI Express becomes a liability in terms of bandwidth and latency.

Many people do not realize this, but the “life” of a CPU is dreadful and boring. It exists primarily waiting for data.  The use of discrete GPUs for parallel computing through PCI Express will not improve the “quality of life” of the CPU.  While some discrete GPUs will offer graphics and pure FLOP performance over an APU, the performance will be limited, in some cases, by the PCI Express interconnect.

This is where AMD Fusion APUs will shine.  AMD Fusion APUs have not only been designed to offer great graphics performance, they also have been designed to offer great parallel compute performance.  The fact that the CPU core resides next to the GPU core connected by a bus of mere nanometers, helps diminish the bandwidth and latency issues presented to parallel computing on a PCIE bus.

The design plan for successive generation of AMD APUs includes architectural innovation, as well as tighter and faster interconnects between the CPU cores and the GPU cores. One goal is to advance the parallel compute capabilities without sacrificing x86 and graphics performance.

So, to finally answer the question whether discrete GPUs will die, the answer is: H.e.l.l No.

AMD relies on our ability to generate world-class discrete graphics cores and then leverages them in our APU product lines.   You should also expect AMD to innovate when an AMD discrete graphics is plugged into an AMD APU, the mythical word of “synergy” will be realized.  We are expecting to improve both graphics and compute performance.

Besides the technical reasons why discrete GPUs will not disappear anytime soon, AMD likes the taste of success.  To stay the market leader in discrete graphics solutions, AMD plans to continue to develop better and better discrete GPU products.   And to make the point, we’ve scheduled a whole new generation of discrete GPUs for release later this year.  Beware ye of little faith.

Next Question!

gz_easy

对AMD的H.e.l.l No，大家应该喝彩：H.e.l.l Yeah！！

西把拉古

With no exception, seeing Edison is seeing English

齐樂樂

等翻译帝