A new class-action lawsuit against AMD is arguing that the company engaged in fraud and deceptive marketing when it claimed that its Bulldozer processors had eight cores. The suit, Dickey v. Advanced Micro Devices Inc, argues that AMD's Bulldozer (and presumably Piledriver, Kaveri, Carrizo, and so on) all effectively contain just half the number of cores AMD claims they do, and that the company's decision to combine certain aspects of the core into a module and share resources constitutes a deceptive practice.
If the writeup at Legalnewsline.com is remotely accurate(Opens in a new window), the lawsuit is utterly without technical merit. The suit supposedly alleges that because Bulldozer shares certain core resources, the cores can no longer work independently and the chip is no longer capable of performing eight instructions simultaneously. If that's the hook Dickey is hoping to hang his lawsuit on, he picked a bad one. While it's true that AMD shared core resources within Bulldozer, the chip doesn't work the way Dickey alleges it does.Diving into 'Dozer
Four years ago, we published an investigation into how Bulldozer handled multi-threading scenarios and what kind of penalty the chip took when running two threads on the same module. Our investigations showed that Bulldozer took roughly a 20% performance hit when scaling up to eight cores compared to the scaling we'd expect to see from a conventionally designed processor.
Cinebench scaling when using four threads. Clock speed was locked to 3.6GHz.
The graph above shows Bulldozer performance when Cinebench has been told to use only four threads in every case. We then controlled where the threads executed. The 4M / 4C test means we locked one thread to each module. The 4M / 8C test means we specified four threads, but let them run on any cores. The 2M/4C test means we locked all four threads to run on a single module. The results show that the 4M / 4C mode is clearly faster than the 2M/4C mode -- about 18% faster.
Bulldozer's total performance with all eight cores enabled, however, was still higher than the 4M / 4C mode, as shown below:
Cinebench performance when all eight threads ran on eight cores. Clock speed was locked to 3.6GHz.
An eight-core Bulldozer was capable of roughly the same performance as a six-core Thuban. The performance hit from combining and sharing cores, in other words, was nowhere near the 50% that Dickey alleges.
Since Bulldozer shipped, AMD has made modest improvements to the CPU's overall efficiency and performance. Kaveri cut the penalty for multi-threading in half, from ~20% to 10% compared with typical core scaling. If AMD hadn't been forced to lower clock speeds to compensate for its 28nm manufacturing process, Kaveri would've outperformed Richland across the board. The 10-20% penalty for multi-threading compared to conventional core designs is nowhere near the halving that Dickey alleges.
Now, it's absolutely true that the Bulldozer family of products has had much lower single-thread performance than either previous AMD CPUs (in many cases) or Intel chips (in virtually all cases). But this lawsuit doesn't appear to argue that AMD mismarketed its CPUs because single-threaded performance was weaker than expected, but because multi-threaded scaling was critically harmed by the decision to share various aspects of the underlying architecture. Weak single-threaded performance and high power consumption created a situation in which BD could neither hit its target clock frequencies nor its IPC targets. Critically, these issues do not disappear when the CPU is run in one-thread per module mode.
Dickey's lawsuit is wrong on other areas of fact as well. Bulldozer does share a single FPU block per work unit, but consumer workloads are rarely FPU-heavy. Each CPU module does contain the eight integer pipelines you'd expect in a typical dual-core conventional chip (4 ALU + 4 AGU per module). Dickey refers to Bulldozer as being unable to "perform eight calculations simultaneously," but this is imprecise, inexact language that does not reflect the complexity of how a CPU executes code. Bulldozer is absolutely capable of executing eight threads simultaneously, and executing eight threads on an eight-core FX-8150 is faster than running that same chip in a four-thread, four-module mode. Bulldozer can decode 16 instructions per clock (not eight) and it can keep far more than eight instructions in flight simultaneously.Courts can't define cores
This lawsuit essentially asks a court to define what a core is and how companies should count them. As annoying as it is to see vendors occasionally abuse core counts in the name of dubious marketing strategies, asking a courtroom to make declarations about relative performance between companies is a cure far worse than the disease. From big iron enterprise markets to mobile devices, companies deploy vastly different architectures to solve different types of problems. An eight-core, Cortex-A7-based, mobile SoC is a very different beast from an eight-core big.Little Cortex-A57 / Cortex-A53 configuration. That chip is very different from an Oracle M7 or the SPARC T5. The T5 doesn't pack the per-core performance of Intel's 18-core Xeons, or IBM's Power8.
As this chart makes clear, CPU IPC and clock speed vary widely, even within Intel products over time.
Bulldozer may have performed more like a quad-core chip from Intel, but that doesn't mean it actually was a quad-core chip. The performance benefits from running the CPU in a quad-core-equivalent configuration weren't nearly large enough to make that claim. The argument might stand if AMD had marketed BD as having great floating-point performance, but the company's disclosures and briefings all clearly stated that BD would have just four floating-point units. Anyone buying the system for FPU work would have known that long before hardware shipped.
AMD has, in a very real sense, been thoroughly punished for the CPU it brought to market in 2011 -- and this lawsuit makes claims that don't hold up to technical scrutiny.Tagged In
FX-8150 Bulldozer Piledriver AMD Eight-coreMore from Extreme
