By using this site, you agree to our Privacy Policy and our Terms of Use. Close
sc94597 on 27 January 2018
Edit Thread
bonzobanana said:
sc94597 said:

L.A. Noire runs at dynamic 1080p on Switch when docked with better textures, shading, and shadows. Where it falls behind (draw distance and framerate) it seems to be a matter of optimization and polish. For example, oddly they tied the frame-rate with the rate of gameplay on Switch whereas on PS3 if the framerate dropped they'd drop frames while maintaining the rate of gameplay. This shows a somewhat lack of polish, because there is no performance reason for why they'd choose to do one over the other. Conversely, when the game was first developed there was heavy optimization in taking advantage of the PS3's SPE's and it was actually something they bragged about. I suspect the PS4 and XBO don't have this problems because their CPU's just have more raw power than the Switch's (roughly 1.8 - 2x more at least.) 

Unless this has changed with a later firmware the Switch still caps its CPU's to 1ghz and only 3 are used for games and this applies to both docked and undocked modes and while these are more capable cpu's than in the 360 and PS3 they easily surpass Switch by the sheer speed they run at, 3.2ghz. While many have said the reason the Switch can't run LA Noire well is its optimised to utilise the PS3's cell processors the 360 did in fact run the game well too and I don't think its unfair to say both 360 and PS3 easily surpass Switch CPU performance.

https://www.youtube.com/watch?v=SZT8lB0icC8

Isn't it something like 9,000 mips for wii u, 13,000 mips for Switch (due to the 1ghz limit) but something like 20,000 mips for 360 and maybe 28,000-40,000 mips for ps3. PS4 and Xbone are up to 34,000 - 38,000 mips. If the Tegra CPU's were run at full speed of course it would be different but they aren't they are only run at about half speed in Switch but you can imagine if Nintendo released more cpu performance with a later firmware it would comfortably surpass 360 and be closer to the other consoles. Both PS4 and Xbone never pushed cpu performance in their consoles being only a mild jump from the last gen.

There are cheap octacore android tablets that exceed 30,000 mips for cpu performance but of course have much, much weaker gpu performance than Switch. For comparison the current AMD Ryzen CPU can exceed 300,000 mips. Cisc chipsets tend to get more work done per cycle as they have a larger instruction set (generalisation). 

However mobile chipsets tend to utilise the main cpu for secondary tasks too and don't have as many support processors as non mobile chipsets. So a comparison of mobile vs non mobile without factoring that in would not be fair. So a mobile chipset 10,000 mips is weaker than a non mobile 10,000mips chipset which again is weaker than a cisc 10,000 mips non mobile chipset. I'm just making the point the issues of LA Noire on Switch are extremely likely based on the weak cpu performance especially as the issue effects both docked and undocked.

A while back there was a thread regarding this topic where I calculated this based on estimates scoured on the internet. 

http://gamrconnect.vgchartz.com/post.php?id=8432244

sc94597 said: 

It is really hard to compare real-world performance for CPU's with such drastically different architectures without benchmarks (and even with benchmarks it is difficult.) 

One way to measure theoretical CPU performance is in DMIPS though (basically how many million instructions per second can the processor perform after considering differences in instruction sets by a generalized benchmark called Dhrystone.) An instruction set is the set of all instructions that the CPU's machine language provides for. 

So for the A57 the recorded statistic is 4.1-4.5 DMIPS/MHZ. Let's just take it to be 4.3 DMIPS/MHZ. Mutiply that by a clock speed of 1020 MHZ, and we get 4182 DMIPS/core. 

Scouring the web it looks like Expresso is 2877.32 DMIPS/core  

The ratio of performance is then 4 cores*(4182) DMIPS/core/ 3 cores*(2877.32)DMIPS/core =  1.93 times more instructions/sec for Switch's cpu than Expresso. 

So more or less twice as many instructions per second, on a basic comparison. 

The Xbox 360's CPU provides 5638.90 DMIPS @ 3.2 MHZ for all cores

Which gives a ratio of (4*4182)DMIPS/5638.90 DMIPS or about 3 times the Xenon. 

Not even going to bother comparing to the CELL because the architecture is so odd. 

Jaguar has about 3.6 DMIPS/MHZ , so 1,750 MHZ * 3.6 DMIPS *8 cores = 50,400 DMIPS (for XBOX ONE)

So the Switch's CPU is about 33% the Jaguar @1,750 MHZ (assuming both use all of their cores.) 

Jaguar in the PS4/XBO ~ 3* A57 in Switch;  A57 in Switch ~ 2*Expresso in Wii U ~ 3*Xenon in Xbox 360 (theoretically; assuming all cores can be used to the max.) 

Performance is of course different, because we know Microsoft and probably Nintendo don't use all of their cores at max. 

Also note that developers have commented that the Xenon has better real-world performance than the Expresso, but that could just be a matter of not bothering with optimization for the Wii U's advantages in ports. 

 

I don't know where you got your MIPS estimates for Xbox 360/PS3. Even if we assume only three cores can be used at max, that still gives us much more performance than the Xenon. 

I got my Xbox 360 estimate from here. 

https://www.neogaf.com/threads/wii-u-cpu-espresso-die-photo-courtesy-of-chipworks.513471/page-15#post-58036908

We also have to recall that in gaming there are diminishing returns the more cores you have, just because not everything is parallelizable.