Low CPU GAME on Shadow Tomb Raider Bench
Low CPU GAME on Shadow Tomb Raider Bench
Do you know the reason behind this setup? --SPECS-- CPU with 32 threads (3.2–3.8GHz), 4 physical processors, supports up to 2000MHz DDR3 RAM, RAM of 1333MHZ (16 sticks, 64GB total) in dual channel (DDR3), PSU of 1000W, GPU is RX 580, Monitor with 144+ Hz, Radeon Thermal Grizzly
This configuration is far too strong for real-world use. It won't run at full capacity across every thread in any game.
They were indeed Opterons. Thought the scaling performance matched what I got from my similarly old Xeon models, though I doubted it would be the same. So that’s the conclusion.
They operate under a Piledriver model, which could mean their actual core count isn't what they believe. It's unclear whether AMD applied similar accounting tricks to the cores involved in the lawsuits over Opteron FX chips from that time.
The term "fast" refers to speed, but the specific metric isn't clear. Ghz is a common benchmark for throughput. You might also consider latency or bandwidth as alternatives.
Improvements in CPU performance across different architectures mean relying only on clock speeds won’t give a full understanding. A modern i3-12100 performs better than an i7-7700 even though they share the same cores, threads, and clock rates. The AMD processors from that time offered a large core/thread count for their era (though not as impressive as what AMD was expected to deliver), could be overclocked significantly but at the expense of high power consumption and heat, making their IPC less relevant in real-world use. Intel’s chips from the same period clearly outperformed them, and even AMD’s earlier Phenom models had strengths. All these points apply mainly to mainstream gaming-focused AMD products, while server processors will be quite different.
IPC refers to how efficiently a CPU performs operations relative to its clock speed. It highlights the number of tasks a processor can complete per clock cycle. Modern CPUs with higher IPC values execute more work in the same time, making them faster even if their base frequency is the same as older models. IPC isn't listed on typical specifications because it's hard to measure accurately and changes depending on the workload. For example, one CPU might handle two additions per cycle while another does four. When dealing with heavy calculations, performance can double or increase by a similar margin, though real gains usually range from a few percent to a couple of decades depending on the application.
Somewhat surprising how a 3.6ghz Pentium 4 can be easily beaten by a 2.1ghz Core 2 Duo just a few months later. Architecture differences make a big difference, not just the clock speed.