The problem with this analysis is it's purely theoretical. It shows PS3's Cell as ~3X faster than Xbox 360 and yet almost all cross-platform title ran better on Xbox 360. One of the major reasons Xbox 360 was so popular last gen is because outside of 1st party titles, Xbox 360's games ran faster and/or had higher IQ/textures. It wasn't until very late in that console's cycle that PS3 started to catch up with cross-platform parity.
Per Wiki on the Cell, "At 3.2 GHz, each SPE gives a theoretical 25.6 GFLOPS of single precision performance."
There are 3 other big points:
#1 per IBM: "Tests by IBM show that the SPEs can reach 98% of their theoretical peak performance running
optimized parallel matrix multiplication". That means 98% using a niche use case. In the real world? WAY less than 98% of its theoretical throughput. Therefore, simply taking the maximum theoretical throughput of the PowerPC main CPU core (PPE) + 8x SPEs is pure marketing BS for this reason alone.
#2. "This Cell configuration has one PPE on the core, with eight physical SPEs in silicon.[19] In the PlayStation 3, one SPE is locked-out during the test process, a practice which helps to improve manufacturing yields, and another one is reserved for the OS, leaving
6 free SPEs to be used by games' code"
I believe later on in PS3's cycle, they may have unlocked 7 SPEs but it was never full 8. This has to be accounted for in the real world as the Cell on PS4 could never access all of the theoretical SPEs.
#3. Real world IPC of the Cell was and is horrible for games. Why does this matter? It matters A LOT when comparing the Cell to x8 Jaguar because it means the Cell's theoretical GFLOPs throughput can not be taken at face value in the real world, even with full on optimization. Xbox 360's 3-core PowerPCs were not even as fast a single core in a typical Core i7 920:
Metro 2033 developer stated the the entire Xbox 360 CPU was only as good as 70-85% of a
single Nehalem Core i7 (say i7 920):
http://www.eurogamer.net/articles/digitalfoundry-tech-interview-metro-2033?page=4
This means the single PPE core of the Cell (without the 6 SPEs) was garbage, basically 1/3 the horsepower of the Xbox 360's CPU, and that Xbox 360 CPU was a joke compared to a 1st generation i7. Not looking like a powerful gaming CPU that Sony claimed.
In other words, I don't even know how the person in your thread derived 230 GFLOPs for the PS4 but even if you take the throughput of the main PowerPC core + 6-7 SPEs, the theoretical throughput needs to be adjusted dramatically for real world standard gaming code against a Jaguar CPU core.
Since MS was able to emulate PowerPC code on x86 Jaguar, if the PS3 had just PowerPC cores, PS4 could also easily emulate most PS3 games. Since the coding for games on PS3 involves a combination of PowerPC + specialized SPE units, it would be very expensive to convert all the code. Could PS4 have the power to run emulated PS3 games though? Possibly. Why? Because most PS3 games look worse than Xbox 360 games, which means to achieve the same level of graphics + FPS as a PS3 game would be a walk in the park for PS4. The horsepower is easily there but the problem is converting the code would be prohibitively expensive. If you threw 1000s of software engineers on the problem, could you get PS4 to run PS3 games? I think you could since no PS3 game is graphically impressive compared to the CPU+GPU horsepower available on PS4 and comparing the best ever PS3 game to the best Xbox 360 game.
Sony -
"We are just taking two different approaches. Unfortunately there are just not sufficient enough software engineers in the world for everyone to do everything. Each platform holder has to make their choices, we made one and they made another. Their choice is entirely legitimate, and I think our choice is legitimate, too. In some ways it is quite nice to have points of difference between the two platforms, and people will decide which approach suits them best."
http://www.mcvuk.com/news/read/play...t-s-good-we-re-doing-different-things/0151317
Facebook
Twitter