Do we know how many cores PS2 / Forgelight can use?

[kaffis]

For people considering upgrading -- lots of people make the mistake of ignoring their CPU -- modern gaming uses a lot of CPU, simply because the rate of advancement in CPU design has been slower than the leaps and bounds made in GPU design over the last 10 years.

Furthermore, display sizes have stagnated since the introduction of HD TV's, so again, the requirements for the GPU haven't scaled in proportion to the needs for CPU power.

[Coreldan]

While not the best architechture and all that, I believe I should manage pretty long with my AMD Phenom II x6 1090T. Currently only mildly overclocked at 3,5ghz (stock 3,2ghz), should be able to get it to 4ghz with default cooler, even higher with a new cooler.

I'm just a bit noob with computers, so I'm afraid I break something

[Was_Ash_Emerald]

Originally Posted by Rbstr

This is definitely not true. Clock-for-clock Core is much more powerful...Hell in some cases Atoms can beat the P4.

Yes, it is more powerful, but consider that the Pentium 4 3.0ghz (with HT) existed in 2003 and had a pipeline that was mostly starved due to slow memory.

One can run Windows 7 32 bit on such a machine just fine even today as an acceptable desktop computer for normal stuff.

Speed improvements, based on single cpu core, have slowed drawn dramatically in the last 10 years, as the clock rates cannot get much above 3Ghz. Almost all of the "speed" improvements have been more cores and lots of cleverness.

If you use the benchmarks of 1993 vs 2003 computers were radically faster, almost incomparably so. No 1993 computer could run Windows XP Circa 2003.

The top computer in 1993 was a 486 or Pentium 1 at a blazing 66MHZ. In 2003, the fastest cpu was 3000Mhz!

To bring the discussion back to Planetside- making use of more cores seems paramount to getting the huge scale of battles to work well.

[Landtank]

Originally Posted by Was_Ash_Emerald

To bring the discussion back to Planetside- making use of more cores seems paramount to getting the huge scale of battles to work well.

I'll say it again, the amount of cores for a game like this will not be critical to the games performance. A good dual core will likely run the game just as well as a good quad core depending on the support given to the quad core.

More cores=/=better performance for all games. It depends on how much of the game can be parallelized, and in a game like this, a huge majority of the calculations can be parallelized, especially with hyperthreading.

More cores improves performance for games that have multiple applications, such as strategy games which require huge amounts of processing power to calculate the AI and massive amounts of troop movements.

[CyclesMcHurtz]

Originally Posted by Landtank

I'll say it again, the amount of cores for a game like this will not be critical to the games performance. A good dual core will likely run the game just as well as a good quad core depending on the support given to the quad core.

Depending upon how settings are implemented in a game, this can be true. It's not the kind of statement you can make on broad terms, however.

Originally Posted by Landtank

More cores=/=better performance for all games. It depends on how much of the game can be parallelized, and in a game like this, a huge majority of the calculations can be parallelized, especially with hyperthreading.

This is not true. There are significant dependencies on calculations and at many points you need to wait for certain operations to complete and other operations are not possible to spread across threads. The complexity of coordinating data structures across threads can be more expensive than the gains you get from spreading the work across the cores.

Linked lists are the devil on multi-threaded systems and also for cache coherency but they are also the most efficient type to manage, while arrays are the best for multi-threaded systems and cache coherency but are a terror to manage efficiently.

Originally Posted by Landtank

More cores improves performance for games that have multiple applications, such as strategy games which require huge amounts of processing power to calculate the AI and massive amounts of troop movements.

Physics and animations are typically much more expensive than AI, when done correctly. Many games try to make AI that operates at CPU speeds which causes significant performance problems - reducing the rate at which the AI makes decisions results is better AI behavior and lower CPU cost. One of the most expensive operations in AI is path-finding, and most of that is pre-calculated in modern games to some extent (which can dramatically increase memory requirements).

Animations and Physics typically need to be done each frame of the game and there is typically no way around this at the macro level.

[Whalenator]

Originally Posted by LtStaley

Do you really think PS2 isn't going to be too CPU intensive? I've been debating on whether I should build a new gaming desktop or not. The one I currently have might still be strong enough to handle PS2. Here are my specs, let me know what you think...

ASUS A8N32 SLI Deluxe Mobo
AMD 64 X2 4800+ @ 2.4 Ghz
Dual BFG 7900GTX GPUs in SLI
3GBs of RAM
Windows XP x86

I would upgrade to 4GB (2x2GB) of RAM at least. It's cheap, so pile it on.
I would also upgrade to Win 7 (x64)
I would get an Intel Processor over an AMD anyday.
The 7900GTX should be able to run PS2 smoothly (35FPS+) on Medium.

[Landtank]

Originally Posted by CyclesMcHurtz

Depending upon how settings are implemented in a game, this can be true. It's not the kind of statement you can make on broad terms, however.




This is not true. There are significant dependencies on calculations and at many points you need to wait for certain operations to complete and other operations are not possible to spread across threads. The complexity of coordinating data structures across threads can be more expensive than the gains you get from spreading the work across the cores.

Linked lists are the devil on multi-threaded systems and also for cache coherency but they are also the most efficient type to manage, while arrays are the best for multi-threaded systems and cache coherency but are a terror to manage efficiently.




Physics and animations are typically much more expensive than AI, when done correctly. Many games try to make AI that operates at CPU speeds which causes significant performance problems - reducing the rate at which the AI makes decisions results is better AI behavior and lower CPU cost. One of the most expensive operations in AI is path-finding, and most of that is pre-calculated in modern games to some extent (which can dramatically increase memory requirements).

Animations and Physics typically need to be done each frame of the game and there is typically no way around this at the macro level.

Thank you very much for the clarifications! My knowledge of processors is limited, I'll be the first to admit that.

[kaffis]

Yeah, it always makes me happy to see CyclesMcHurtz happy to come around and explain some of the neat technical details about the challenges he and his fellow coders face.

[Whalenator]

Originally Posted by elfailo

35 FPS is horrible to play at. I will turn down every single setting until I can hit 120 on average.

The Human eye, even trained and young, cannot discern a difference in framerate after ~60FPS. 45FPS is optimal for twitch shooters and fast-paced games, but I can play slower ones like Arma 2 at 30FPS.

After 65FPS you're just senselessly heating up your processor.
Besides the point that most modern monitors can't display more than 75 FPS...

[Rbstr]

Originally Posted by WildVS

Historically they have always favored lower voltage (and have hinted at that already for the next gen cpus) and more cores. But you really approach a point of diminishing returns in doing that. They claim 3X clock at same power level of todays chips or either run the cores at 0.5 volts and same clock. Hope they venture more into the former.

There are problems with increasing voltage when you go down in size because things like electromigration become more significant the smaller your feature size. (A small movement doesn't mean much to a big thing but can mean everything to a small thing).

I would upgrade to 4GB (2x2GB) of RAM at least. It's cheap, so pile it on.
I would also upgrade to Win 7 (x64)
I would get an Intel Processor over an AMD anyday.
The 7900GTX should be able to run PS2 smoothly (35FPS+) on Medium.

That's a complete conjecture. That card is over 6 years old and doesn't even support DX10.
Everyone needs to stop quoting system specs like they have any idea how the game runs.

[Whalenator]

Originally Posted by elfailo

Thanks for quoting popular science. Unfortunately it's wrong, but there's no way for me to prove it. You have to play a game at 120FPS on a 120Hz monitor and see it for yourself. I can guarantee you that you will see the difference with 60FPS.

Is the difference worth it?
You expect a game, in beta, with a 1km viewdistance running 1,500 other players on a 10x10km map, to run at 120FPS for anyone?

If that's your minimum you sir need to lower your expectations.

[Was_Ash_Emerald]

Originally Posted by Whalenator

The Human eye, even trained and young, cannot discern a difference in framerate after ~60FPS. 45FPS is optimal for twitch shooters and fast-paced games, but I can play slower ones like Arma 2 at 30FPS.

After 65FPS you're just senselessly heating up your processor.
Besides the point that most modern monitors can't display more than 75 FPS...

You can see it because that FPS is calculated as an average over 1000ms.

At 45fps (even 60fps), you will likely have 10ms, 20ms, 30 ms moments where you are actually below 30fps, if that time were conceptually extended out to 1000ms. Some users are very sensitive and can perceive those moments.

In my subjective estimate, you need to be > ~80fps at all times to be above slight but perceptible motion lag.

[Aaron]

I might be installing a Intel Core i3 2100 for my sister. Do you have any opinions on the PS2 performance, McHurtz?