PJ64: Only using 50% of the CPU?

[SeymourOmnis]

Is this right or wrong? Because I want PJ64 to use 99% of my system's CPU so games can run faster.

[Clements]

If you have a dual-core CPU, or a processor that supports hyper-threading, this is normal behavoir for all single-threaded apps. For PJ64 to use more than 50%, then the app would need to be threaded, which is extremely difficult to code for. Also, since a 1.2GHz processor is all that is needed to run 90% of games, there is little motivation to add such a feature.

[Smiff]

or you may just have a fast enough single core cpu and the speed limiter on. which is correct. why would you want PJ64 to use more CPU than it needs? if you want to increase the framerate, i don't want to explain all that out again, maybe someone can find it :/

[SeymourOmnis]

i have a P4 3.2ghz with HT turned on, should I disable it for faster performance?

because now i'm playing goldeneye and in certain areas i notice framedropping, and i hate them. i love a much stable gameplay without noticing framedrops in goldeneye. how can i fix this?

that's why I want pj64 using 100% of my cpu.

If you have a dual-core CPU, or a processor that supports hyper-threading, this is normal behavoir for all single-threaded apps. For PJ64 to use more than 50%, then the app would need to be threaded, which is extremely difficult to code for. Also, since a 1.2GHz processor is all that is needed to run 90% of games, there is little motivation to add such a feature.

But I DO have many applications that use 99% of my CPU when running, and they do not support SMP. By the way, I have an AMD Duron 1.2ghz that can only handle Super Mario 64 at 50 FPS, can you explain this?

[Clements]

- Goldeneye slows down on a real N64 system. If you overclocked your P4 to 5GHz on Cascade cooling, you would still experience slowdowns. Emulation cannot solve frame rate problems inherent on the original hardware.

- Hyperthreading does not decrease performance by 50% like Task Manager implies. Disabling it can only increase performance by a small amount in single-threaded apps (normally 5% according to the reviews I read when HT was first introduced).

- I've had early AthlonXPs that can run SM64 at 150+fps with the limiter off. Your Duron is slower than Athlon processors due to it's tiny 64kB L2 cache and bandwidth etc. Althon or Pentium 3 processors are recommended in the manual.

- Unless I am complete wrong about how HT works, it'll only register 50% of the CPU in single-threaded apps. Run SuperPI and you will see that. System Idle Process is not a process.

[SeymourOmnis]

thanks clements, but I also experience framedrop problems with conker's bad fur day. is this normal?

also, my dad has a Pentium 3 1.0ghz. Is this faster than my Duron 1.2ghz?

[gandalf]

The frame limiter makes the CPU works at half load (or random load), to make it works at 100% you need to disable de frame limiter, but try to play without it xD

I think that P3 it´s faster than the Duron

[Toasty]

- Unless I am complete wrong about how HT works, it'll only register 50% of the CPU in single-threaded apps.

Correct. Task Manager (or any CPU monitor) is a little quirky with HT, because there are two logical processors, but only one physical one. The physical processor's full potential can be used in a single thread, but Task Manager still believes there is another equally powerful processor that is not being utilized - which is not true. Together, the two logical processors can only use as much CPU time as is available from the one physical processor. How this time gets split up varies, but the Task Manager doesn't account for this by weighing one processor as a greater or lesser percentage of the CPU's total ability.

[squall_leonhart]

yeah basically. the task manager is reporting the usage wrong

it is actually using 100% cpu usage (which to gandaf. is how it works on all systems, even if they are way fast, doesn't change the amount of processor usage or lower it, just means it can process more at 100% so theres no slowdowns as much) its just that the system see's 2 cores so the load is split between them oddly, (in reality there is only on core, HTT is just a method of making a virtual core so that the system believes you have 2 cores,... which this is actually slower and worse then a standard single core.)

disabling hyperthreading will allow full usage of the core, and remove the 5% overhead that creating a virtual core causes.

also

CBFD has these slow downs to begin with, its a heavy game on the N64 so its just as heavy on the emulator. theres also the fact that certain effects aren't rendered on the emulator and some times even pj64 is the cause of slow downs.

to clements

mate, in regards to Goldeneye, on the n64 it worked at smooth speeds, only slowing down when there was heavy amounts of smoke being displayed.

on pj64, it can slow down anywhere. even just walking along.
i remember playing it on n64, and there are points in the emulator that it becomes nigh unplayable, where on the console it was perfectly fine.

also.

The Duron being slower is also related to its FSB and how fast it talks to system memory., not just the 64kb cache.

a computer can only go as fast as it can recieve data from the system memory

[Clements]

On my system, it's actually pretty smooth, solid 60 V/Is throughout on CF1, CPU idle average ~30%, max ~56%, min ~8%. I remember it not running very fast on my old AthlonXP 1800+ @ 2000+, so I had to use 1964. I have completed the game fine with PJ64 without a hitch, and noticed no dramatic slowdowns except for the heavy grey smoke that occurred in the original.

The Duron being slower is also related to its FSB and how fast it talks to system memory., not just the 64kb cache.

a computer can only go as fast as it can recieve data from the system memory

Yes, I know about this, I was referring to that in my previous post. Durons with the Morgan core use only a 100MHz FSB, so don't have a lot of bandwidth.