What happened to the use of "bits" to describe consoles?

NES - 8 bit
Genesis/SNES - 16 bit
Dreamcast/N64 - 64 bit

PS1 - Disk based, Generation 1 (CD-ROM)

PS2/XBOX/Gamecube - Disk based, Generation 2 (DVD/minidisc)

PS3/XBOX360/Wii - Disk based, Generation 3 (BRDVD/HDDVD)

PS4/XBOXOne/WiiU - Disk based, Generation 4 (4k?)

quote:

---

PS4/XBOXOne/WiiU - Disk based, Generation 4 (4k?)

---

No

I stole this from another forum:

quote:

---

It's worth mentioning that bits as a measurement of console performance fell into disuse after the fifth generation (i.e. the 32/64-bit era). There are 2 reasons for this. Number 1 is that bits barely matter by comparison to other technical specifications. The hertz rate, or number of cycles per second (CPS), is far more important. The DreamCast CPU, for example, is a Hitachi SH7750 (SH-4). It runs at 200 MHz, which, I think, is 200,000,000 cycles per second acccording to International System of Units (SI) specifications. I may be wrong and as with bytes it might increment by i ×= 1024 per prefix, but I don't think it does. The PS2's Emotion Engine is also clocked at 294 MHz. These are listed by you as 128 bit consoles, but in terms of CPS (a more important measurement) they are terrible compared to GameCube's PowerPC microprocessor Gekko (485 or 486 MHz) and the Xbox's Intel Pentium 3 which ran at 733 MHz. With respect to hertz, GameCube and Xbox central processing units (CPUs) lead the way, which is counterintuitive to what you'd expect from your byte readings. GPU is also important. The consoles you claim to have the highest bits (Dreamcast, PS2) also have the worst GPUs. Dreamcast uses a NEC-manufactured PowerVR GPU which runs at 100 MHz, while Graphics Synthesizer (PS2 GPU) runs at 147 MHz. GameCube's GPU, named Flipper in a reference to the Dolphin codename for the system, runs at 162 MHz, while the NVIDIA GPU of the Xbox 360 comes out on top at 233 MHz. Therefore, the two worst consoles (GameCube, Xbox 360) by bit rating would be the best based on GPU, and the best, the worst. It's factors like these that heavily influence console performance, while bits barely play into the big picture anymore. The advantage bits do give in your examples can be explained by the Single Instruction, Multiple Data (SIMD) concept. With 128 bits, as in your example, you can perform the same operation on 4 DWORD values at once. This is only useful when you need to perform the same operation on different data; e.g. in three-dimensional graphical topology. For seventh generation consoles, the multiplex of cores is what matters the most. A single-core is far inferior to a multi-core, and something like the Gulftown 9x0X Extreme versions of the Intel i7 processors, which are hexacore, are extremely powerful. The second reason bits have fallen into disuse as a measure of console performance is because they have no clear referent. Let's use a console you called 128-bit as an example. The Emotion Engine of the PS2 has a 128-bit SIMD ISA, but its floating point unit (FPU) and arithmetic logic units (ALU) are, respectively, 32- and 64-bit. The VPU has some 128-bit registers, but not all its registers handle same-sized operations; and while the integer registers measure 128 bits, their units are only 64 bits, so it's unnecessary for that purpose. With respect to bandwidth, the CPU and IOP are 32-bit, while the DRDRAM has 16-bit channels. Gekko (GCN) is simplified as 32-bit, but has 64-bit components, for example its double-precision FPU.


tl;dr version: After the fifth generation, bits have had a very minor influence on console performance. Other facts matter much more, and to what "bits" refer isn't even clear, though it's usually assumed as word size. With the right specifications, a 16-bit console can output stuff you'd expect from a 128-bit console.

---

quote:

---

What is much more important now is processor speed and the number of cores.

---

Not exactly. The Xbox 360 runs a PowerPC based CPU that has 3 physical cores clocked at 3.2 Ghz. The PS4 and One both have AMD based CPUs with 8 cores running at 1.6 Ghz.

Clock speed, obviously, isn't nearly as big a factor as it was 10 years ago. More cores generally results in faster performance, but there are exceptions. For example, AMD's A8 (4500m) is a quad core processor for notebooks running at 1.9 Ghz, yet it gets surpassed by Intel's i5 (3317u), which is a dual core CPU running at 1.6 Ghz. Half the amount of physical cores, as well as a lower clock speed, yet the 3317u provides similar, if not better performance (depending on the application).

Typically, IPS (instructions per second) is the primary determinant of overall speed, which is calculated by IPC (instructions per cycle) multiplied by the clock speed.

It seems that Intel processors, in general, can execute more instructions per clock cycle, most likely due to the way in in which they're designed (more specifically, their overall architecture). So, even though the clock speed on an AMD processor may be higher, it's getting less done per clock cycle. I'm sure this varies depending on the specific task being performed, but it generally appears to hold true.

Where the hell are you guys in our enthusiasts thread? Seems like something you guys would enjoy...

quote:

---

Where the hell are you guys in our enthusiasts thread? Seems like something you guys would enjoy...

---

I occasionally read in that thread, I just don't always have something to say. A lot of knowledgeable in there though. I'll try to start contributing some though and help build our small community of hardware enthusiasts. Though I am more much knowledgeable on the architecture/emerging technologies front for future processors, which is what my research focuses on. But you guys have taught me a few interesting things to look for when I built my system.

I'm a graduate student pursuing a PhD in computer architecture. Research involves application of new device technologies (e.g., TFETs) at the architectural level. I also do some collaboration with Intel.