| Azzanation said: Erm.. PS3 says otherwise. |
The GPU didn't have access to a Tessellation unit. The Xbox 360 did via it's Truform based geometry unit, but that relied on N-Patches and could not even hold a candle to today.
Particles are bigger and more plentifull today as well. HBAO+ was usually mimicked and "baked" into the texture work last gen, which was free from a performance perspective but also required more development time. Even this generation of consoles struggle to use the best quality Occlusion in games.
| eva01beserk said: I dont play multy player ever, so I would like to know how big of a diference could a game in 30 and 60 be competitevly? is it big enough for devs to say hey, we need to separate x1 and scorpio players cuz its unfair? If so, would they then need to make separate servers? |
I mean. I would be telling lies if I said there wasn't a difference between 30 and 60fps competitively. Because there is a difference.
How big exactly? That depends on the game and even what role you are playing.
For instance in Overwatch it's fine to play a character that doesn't rely on firing precision like Mercy, Symmetra or Torbjorn. But a character like widowmaker, soldier Mc'cree really does perform better at 60fps.
But if you are playing competitively or professionally, then your approach to the game should be more than just a casual one, so you will tend to gravitate to the platform that performs best anyway. Like Scorpio.
| SvennoJ said: Yeah I was just thinking that. Original XBox had a good cpu too, as well as XBox 360, heck even the ps2 was no slouch. It's not easy to emulate XBox 360 games even though the architecture isn't all that different. It's this gen that introduced rather weak cpus pared with much more powerful gpus. And true, ps3 orginally wasn't even meant to have a gpu, 2 cell processors instead. |
Using flops within that context doesn't tell us the performance of those chips.
Think about it: If the Cell had 230+ Gflop of real-world performance and the GPU had 400+ Gflop...
Then why does the Switch which clearly has a CPU with less flops and a GPU with less flops, have it's games look significantly superior? Think about it. Please. And stop using flops in the context you are using it in.
********
The original Xbox's CPU was trending towards the low-end.
It had a coppermine based core with half the cache.
Intel had already moved onto Tualatin by the time the Xbox had released and had CPU's with almost twice the clock speed. (1.4Ghz verses 733mhz)
4x the amount of cache, prefetching etc'.
Intel also had Willamatte Pentium 4 chips operating at 2Ghz+
And AMD had it's Palomino based Athlon XP's on the market with PR ratings of 2100+ based on the amazing K7 core, this was the fastest chip at the time.
So whilst the original Xbox was beastly for a console, it paled in comparison to what the PC was offering in 2001, let-alone in 2004.
******
As for Cell. The only time that CPU showed any kind of decent performance is when using iterative refinement floating point. Otherwise it was terrible. It performed terrible. It was terrible to program for.
The Cell was never going to be a GPU replacement either. It doesn't even have hints of the appropriate silicon to replace the task of a GPU, nor does it even have a fraction of the performance for it either.
The Cell processor was 234~ million transistors at 221mm2 at 90nm SOI.
Roughly comparable to intels Pentium D which released before the Cell.
However, Sony had to disable a SPE for yield reasons. It would have likely been cheaper to manufacture than the Pentium D for that very reason.
Can't forget that Core 2 released around the time of the Playstation 3 either... And that was a superior CPU uArch.
The Cell was also shit at integers. Games and game engines don't just using floating point you know.
*******
The Xbox 360's CPU isn't anything special either. Sure, 3x cores operating at 3.2Ghz all those years ago might have sounded flashy to the uneducated, but it's an in-order design like the original Intel Atom in phones and tablets.
It had a long pipeline, not much in the way of prefetching or branch prediction and only 8-way associative cache, with that cache running at only half the CPU clock rate where-as most x86 CPU's since the Pentium 3 era, the cache runs at the same speed as the CPU clock.
PowerPC hasn't really been a contender against x86 for a very very long time. IBM tends to go for extremely wide core designs which appeals to professional markets, for everyone else, it's not ideal.
... And I could go on. But I think I have made my point.
www.youtube.com/@Pemalite
