Ryzen vs FX?
Ryzen vs FX?
I understand Ryzen offers superior performance compared to the FX lineup. The key distinction lies in how CCX and the FX series were structured and promoted. For instance, the FX 8350 was marketed as an eight-core processor, but each chip actually contained four dual-core modules—something that often confused me given AMD’s use of a chaplet design.
The 8350 was labeled a quad-core because each pair of cores shared cache and it likely included a floating-point unit. This meant that what appeared as separate cores actually relied on shared resources. It wasn't a fully independent core but rather a more compact version. FX processors used a monolithic design, whereas Ryzen chips employ a chiplet architecture to boost manufacturing efficiency and reduce defects.
FX CPUs featured half floating point and full integer cores, meaning an 8-core model was essentially a 4-core design on FP, though it still supported 8 integers. Or vice versa. Back then, it wasn't clear, but guides explained it. Now Ryzen 8 offers 16 threads with chiplets, yet each core retains full floating point and integer capabilities.
The CCX and Bulldozer modules differ significantly in structure despite sharing similar ideas. The key distinction lies in their architecture: for Zen designs, every core functions independently with an L3 cache shared among them. In contrast, Bulldozer uses individual cores that contain most typical CPU components but lack a floating point unit (FPU). The shared cache in Bulldozer is limited to L2, while each core has its own L1 cache. This setup hampers effective caching because the tiny L1 caches lead to frequent starvation, and the small L2 cache makes data sharing between cores difficult. Modern CPUs typically feature L1, L2, and an FPU, making them more efficient. The FPU was a longstanding standard for many years, but today it’s often seen as outdated. While Ryzen offers full cores that can enhance multicore tasks, Bulldozer’s design often falls short in handling parallel workloads effectively.
The FX processor units consist of two 64-bit cores sharing the FPU and cache. Four modules provide eight cores in total. A Ryzen chip on a single CCX offers eight cores, each with its own L1, L2 cache, and dedicated FPUs. And the x3 integration is really impressive!
The 8 cores per CCX claim applies specifically to Zen 3. Earlier versions offered just 4 cores per CCX, while Zen 2 achieved 8 cores per CCD by using two CCXs per CCD. The initial Zen and Zen+ models also provided only 4 cores per CCD, with each chiplet receiving four cores. This explains why the 1000 and 2000 series reached their maximum of 8 cores collectively—limited to a maximum of two CCDs per consumer Ryzen CPU.
Everything is accurate! I thought focusing on the latest would be a good approach. The CCX 5800X and 3D design stand out as my top priorities.
It's worth noting that an FPU doesn't always ensure each core operates independently, especially before floating-point processing became common. In reality, FX chips had eight x86 cores but only half the FPUs were used. The nodes lacked sufficient room for the remaining four FPUs, so FX opted for Netburst with long pipelines to boost frequency. Consequently, AMD introduced the first consumer 5GHz processor, though it suffered from poor IPC performance.
At one point, the FPU operated as an independent unit separate from the CPU. By the time Bulldozer entered the scene, those old days were behind them, which created a challenge. It was confusing for buyers to think of the FX-8150 as an "8 core" processor because it suggested it had double the cores compared to a Core i5 or i7. In reality, this isn't accurate in a broad sense. A CPU core doesn’t require an FPU at all. It doesn’t need cache, modern instruction sets, or even an x86 architecture. If AMD had clearly stated that their cores weren’t equivalent to Intel’s, there would have been less confusion. But they didn’t—by implying Bulldozer cores matched Sandy Bridge cores, they created a misleading impression. This misunderstanding affected performance in many multi-core tasks. Today, discussions continue about how Ryzen compares to Intel in handling AVX versus core series features, and whether certain instruction sets matter for efficiency. It’s worth noting that prioritizing integer operations over floating point was a factor in VIA’s decline, and AMD’s approach with the Bulldozer raised serious concerns. Ultimately, consumers were misled, and the court ruled against AMD for this deceptive marketing. A Bulldozer core isn’t just about speed—it’s fundamentally different from a modern Core series core. The Ryzen 7 was a strong example of what could have been if clarity had been maintained.