Predicting the gaming future involves estimating the processing power required for upcoming advancements.
Predicting the gaming future involves estimating the processing power required for upcoming advancements.
The system operated with two dedicated x86 cores per module. The term SMT isn't a standard label; Intel's Hyper Threading is actually a related concept. SMT refers to a single x86 core that can handle two threads at the same time, distinct from the FPU.
The level of detail seems unnecessary since the main issue is the FPU bottleneck similar to SMT. Even though SMT aimed to use both threads and cores efficiently, sharing resources would limit performance. It’s hard to capture all factors affecting CPU speed. A clearer name might be more appropriate instead of relying on misleading numbers like 8 cores.
Also just for clarity, I made that name because it’s simpler. Windows shows it as a 4 Core 8 Thread, which makes it much easier to explain SMT instead of getting stuck on the design details. It’s more common to understand how an SMT CPU works, so it fit well in this discussion.
The Floating Point Unit handles floating-point operations. It wasn't part of the original X86 design, like the 8086 processor. This was an SMT feature, meaning a single core without a floating point unit. Multi-threading was the focus, not an interpretation of the CPU itself. The FPU is built into the CPU, while SMT is a capability specific to certain X86 cores. It seems there might be some confusion about how these components relate. I’ll try to agree to disagree and let the discussion continue as planned.
The concept revolves around how processors are designed and perceived. Four modules equate to a quad-core setup due to only four FPUs, even though each core actually has eight separate X86 cores. This misunderstanding has persisted for years. Many assumed every core needed an FPU to qualify as an X86 core, but that’s not the case by design. The confusion stemmed from outdated standards where each module was treated as single-threaded. In reality, with modern architectures supporting simultaneous processing, having eight cores per module could deliver better performance. The debate often circles around whether these numbers are exaggerated or if they reflect real capabilities. Ultimately, the focus shifted to speed through advanced pipelining rather than raw core count.
Avoid referring to it as having 8 cores or quad-core; that term doesn't accurately reflect its configuration. It actually contains four FPU units despite having eight physical cores. Calling it an 8-core processor is misleading since the architecture differs from standard interpretations.
I wasn't familiar with the term CMT, so I'll begin using that word as it more accurately describes the FX CPUs.
BF4 and BF1 are titles that can benefit from more CPU cores or threads. My FX-8350 running at 4.9 GHz performed quite well, which is why I kept using it for a long time (just playing BF4, Doom 2016, StaCraft2, and r00nscape). The issue lies with the FX series CPUs being underperforming per core. When comparing a Phenom II CPU to an FX-series CPU, the Phenom II would actually be faster—especially with the first-gen FX 'Bulldozer'. For example, a Phenom II X6 1090T at 4GHz outperformed an FX-6300 at 4.0 GHz. It seemed the FX chips were slightly better in certain scenarios, like with refresh FX 'Vishera' chips. In short, you’re still fine with 4-core / 8-thread or 6-core / 12-thread setups for now, though leaning toward the latter. You’ll likely reach limits with just four cores. The Ryzen 3300X is hitting similar issues to the Ryzen 3600 in terms of high clock speeds out of the box. The core modules and CCX setup matter a lot for games that prioritize CPU speed over core count. The 3300X uses one 4c/8t module, while the 3600 has a 3+3 layout with 3c/6t per module. The 3600 also has two 4c/8t modules doubled up with one disabled in each, and there’s latency in the communication bus between the chiplet groups. Ryzen 3300X vs Ryzen 3600: