Hyperthreading and FPS

[NaturalViolence]

Gamespeed is usually determined by fps usually so no that is not fullspeed. Nowhere near it.

Quote:vps sounds like vertical interrupts.

Yup.

[tuanming]

I highly doubt that overclocking a P4 CPU would yield much fps or performance...

[NaturalViolence]

Why not? It's a weak cpu so you're guaranteed to have a linear speedup (30% OC = 30% faster). You still won't be able to achieve playable speed in most games if that's what you meant.

[Squall Leonhart]

Quote:(30% OC = 30% faster).

Thats not how overclocking works. not on P4's anyway.

[NaturalViolence]

Why? Because of stalls in the execution pipeline? Just wondering since I would think a cpu as weak as P4 by modern standards would have great OC scaling.

[Squall Leonhart]

clock rates never scale 1:1 with performance. there are decreasing gains the higher you go.

[Kodiack]

(01-06-2011, 12:24 PM)Squall Leonhart Wrote: clock rates never scale 1:1 with performance. there are decreasing gains the higher you go.

Definitely this. You may see semi-close to a 1:1 ratio, but the higher you go, the more diminishing returns you'll note. 2.4 GHz to 2.6 GHz, for example, may yield 5% more performance, while 3.6 to 3.8 GHz might only yield 2% more performance.

You can get a rough estimate of diminishing returns by looking at various CPU overclocking benchmarks, like the ones here at Tom's Hardware. Also note that these benchmarks are very CPU-bound, and don't really rely on other components (i.e. GPUs) nearly as much as something like Dolphin or a video game would.

Hyperthreading may help your FPS, but NetBurst's implementation was really poor, showing, if memory serves, about 20% increased performance in synthetic benchmarks. In real-world application, the results wouldn't even be that promising. If you're still running a Pentium IV, you're going to have to compromise some performance or quality. You'll definitely want to look into an upgrade if you're still on a single core CPU.

[NaturalViolence]

That is mainly due to stalls, bottlenecks by other hardware, or bottlenecks within other parts of the cpu. Considering a pentium IV execution pipeline isn't bottlenecked by anything else in nearly any circumstance I would think it would scale really well, better than a modern cpu, unless it's an issue with pipeline stalls?

Anyways I have been looking at tomshardwares annual cpu benchmarks for years now and they usually seem to show that the law of diminishing returns does not kick in on cpu bound applications until very high clock rates (usually 3.8GHz or higher) are achieved. And usually having faster memory read/write fixes that.

[Squall Leonhart]

lol, well my cpu is overclocked by around 43% and the performance gains aren't 43%

[NaturalViolence]

What type of cpu and memory do you have?

And are you running synthetic benchmarks to make sure nothing else is the bottleneck?