I think that's complete bunk, to be honest.
An OS doesn't use a lot of power at all. You may notice something in the range of 1-5% background usage on a 'clean' OS installation compared to nothing at all. But certainly nothing that makes a 733 PIII-derivative run as fast as a 2.8 GHz Pentium 4.
If there is any such difference at all, especially with emulators, then the difference can only be explained by very inefficient application code. Eg using very poorly written graphics handling, or not making any use of extended instructionsets such as MMX/SSE at all.

I switched from DOS to Windows when I had a 486DX-2 at 80 MHz. And even then the overhead of the OS during a fullscreen DirectX graphics application was in the range of 1-5%. When I upgraded to Pentium, it became "don't care".
OSes may have become heavier since, but CPUs have increased in performance faster than that. Even on a single-code Pentium 4, an OS like Windows 7 doesn't take much CPU. As long as you have enough memory, a single graphics application can still make almost full use of the processing power.

Scali wrote:

OSes may have become heavier since, but CPUs have increased in performance faster than that. Even on a single-code Pentium 4, an OS like Windows 7 doesn't take much CPU. As long as you have enough memory, a single graphics application can still make almost full use of the processing power.

Depends on GPU (and definition of "almost"), because if you go all in on GPU side, you get yourself into bottlenecking scenario even in 4k with AA (and that's after CPU overclock) : LINK 😉

Tertz wrote:

Not 2 core, but 4 core CPU >=3 GHz of Core2 family, combined with good video should be still playable (>20 fps) for today games, some maybe after tweaking. That dude could take Q6600 (released in Jan 2007), overclock it to 3 GHz and get much better results. People still are using Sandy Bridge without issues, while they are not much better than Core2 (+10-20% on same clock).
I suspect it's possible to get Q9500 or Xeon E5450, overclock them to ~ 3.5, combine with low end GTX960/GTX680 and get 30 fps in Witcher 3 and far better in "modern" GTA 5, on not max settings certainly.

Did I screw up by going from a Core 2 Duo E8500 to an Athlon x4 860k then? If Sandy Bridge is only marginally faster than a Core 2 Quad, and completely beats the crap out of AMD's offerings, does this mean my "upgrade" was actually a downgrade?

yes according to tertz, because it can't quake in dosbox better and is certainly not the fault of dosbox and its odd max cycles behavior on the faster core processors 🙄

Standard Def Steve wrote:

There are very few apps where Sandy Bridge is only 10-20% faster per clock than Core 2.

For most people, the only thing where practically more performance is interesting is games. With GTX960 or faster, on middle graphical settings or higher, I doubt you'll get significantly more than the mentioned +20% fps on Sandy Bridge, compared to 4-core Core2 on the same clock.

mr_bigmouth_502 wrote:

Did I screw up by going from a Core 2 Duo E8500 to an Athlon x4 860k then? If Sandy Bridge is only marginally faster than a Core 2 Quad, and completely beats the crap out of AMD's offerings, does this mean my "upgrade" was actually a downgrade?

It needs comparision in modern games. For older 1-2 threads games this may to be "downgrade" on close CPU clocks.

Tertz wrote:

For most people, the only thing where practically more performance is interesting is games. With GTX960 or faster, on middle graphical settings or higher, I doubt you'll get significantly more than the mentioned +20% fps on Sandy Bridge, compared to 4-core Core2 on the same clock.

That's not the processor's fault though, that's because you're GPU-limited. The CPU is actually much faster per clk.

Kerr Avon wrote:

Granted this is just one person's assessment, but it's an interesting view on how much CPU power is used by a modern version of Windows and other concurrent software, as compared to another system that's not running multi-tasking software.

its not about multitasking, its about bloat and proper software architecture for a given task; i made sometime in the past a comparison with some program i made, the pc wasnt fast, sluggish in windows, but even just a loop ran visibly faster, then i added more functionality the difference became huge, tens of %;

i surf the web all the time with my Pentium 3-s 1.4ghz 2gb of ram Geforce 6800 ultra, it actually run smooth. I even go on face book (little sluggish at first when loading the page) I also use it for playing DVDs and mp3s, as for YouTube that's a different story.
I do however use a Pentium 4 3.8ghz socket 775, 4gb of ram, hd4850 at work for everything I was even able to watch Netflix with no skips.

Arctic wrote:

E5450 @ 3.6GHz (1.19v) user here: This CPU is awesome!
I have not run into any problems yet. I am still trying to get as much clock for as little current as possible.
But I also don't own The Witcher 3 and GTA 5 (yet) 😀

I have a q6600 at stock and both of those games and they run fine

Scali wrote:

I think that's complete bunk, to be honest.

Maybe it is, I have no idea at all myself. But hcf has ported programs well enough to the XBox, so he is presumably knowledgable about the XBox side of things, though of course he could be missing some important things Windows-side, or perhaps even grossly exaggerating the difference (I'm not saying he is, of course, from his posts and his freely given XBox programs he seems a decent person, but maybe for some reason his contention is totally wrong).

luckybob wrote:

4x ati 480's and I will be set for the next 4-5 years

I see... you're putting your bets on DX12's Multi-GPU!!!!

What could go wrong? [/sarcasm]

Kerr Avon wrote:

grossly exaggerating the difference

Pretty sure this is the case.
The CPU in the Xbox is slower than a regular PIII at 733 MHz (it's a special sort of 'Celeron', with less cache, 128K vs 256K. Also it is bottlenecked by having to share the system memory with the GPU, so you get lower memory performance than on a regular PIII system).
Now, the performance difference between a PIII 733 MHz and a P4 2.8 GHz is very substantial. Especially if you also factor in new P4-only features such as support for faster memory, use of SSE2/SSE3 or even 64-bit instructions and the extended register set that comes with it (depending on which exact version of the P4 you take. We can even throw in HT for good measure).

Then we're talking about a CPU that may be roughly 3 to 4 times as fast as the PIII 733, let alone the downgraded one used in the Xbox. So the claim would be that roughly 75-80% of CPU time on a P4 is taken up by the OS?
Really, that's a pretty retarded statement to make.

See this chart for example: http://www.tomshardware.com/charts/cpu-charts … 03-CPU,434.html
The difference in performance between a P4 2.8 GHz and a PIII 600 MHz or 800 MHz huge.

mr_bigmouth_502 wrote:

Did I screw up by going from a Core 2 Duo E8500 to an Athlon x4 860k then?

That really depends on how much multi-tasking you do. In my case I specifically opted for an FX 8120 for the advantage of higher integer IPC throughput per-process without the additional need of a hypervisor or marshal service. Even if you have an "equivalent or better" total IPC, it's very difficult to write a scheduler to try and balance it out; and impossible to do so if someone else is running wild of which you have no control over.

If you only play games then it would have been a bad decision some time ago but there are quite a few games now which do much better with four cores over two.

IMO, you're fine.

Kahenraz wrote:

In my case I specifically opted for an FX 8120 for the advantage of higher integer IPC throughput per-process

Since when do the FX have better IPC than anything?
In terms of IPC they get thrashed by the older Athlons/Phenoms, let alone the Core CPUs.
They compensate this with higher clockspeeds, to a certain degree. But obviously IPC is 'instructions per cycle", so clockspeed does not come into play.
See here, for example: http://www.cpu-world.com/benchmarks/AMD/FX-8120.html
Phenom X6 1065T at 2.9 GHz perform better than the 8120 at 3.1 GHz. That's lower clockspeed AND less cores, but still faster.
Or here, where the Phenoms (at 3.3 GHz) beat the AMD FX (at 3.6 GHz) at various single-threaded tests: http://www.anandtech.com/show/4955/the-bulldo … fx8150-tested/7
And we all know that the Phenoms never matched even the original Core2 in IPC.

Scali wrote:

And we all know that the Phenoms never matched even the original Core2 in IPC.

Nah, Phenom II is at least as fast as 65nm Core 2 in IPC.

I have an X6 1090T and a Q6700, both overclocked to 4GHz on DDR3 motherboards. In most of the benchmarks I've tested, the Ph2 is slightly faster than the Q6700 at single-threaded workloads.* Once multi-threading comes into play, it's really no contest. For example, the Ph2 X6 can handle 4K VP9 youtube streams without any difficulty while the Q6700 drops a few frames here and there.

*So far, the only real exception I've found is that ancient, x87-loving benchmark called SuperPi, where the 4GHz Q6700 just surges past the 4GHz Phenom.

Standard Def Steve wrote:

Nah, Phenom II is at least as fast as 65nm Core 2 in IPC.

Nope: http://www.anandtech.com/show/2702/15
The X6 1090T may skew the results somewhat because it has an L3 cache.

Scali wrote:

Standard Def Steve wrote:

Nah, Phenom II is at least as fast as 65nm Core 2 in IPC.

Nope: http://www.anandtech.com/show/2702/15
The X6 1090T may skew the results somewhat because it has an L3 cache.

OK. So the Phenoms may be slightly slower than the 2nd-gen, 45nm Core 2s, but they appear to be at least as fast as the 65nm variants.

Standard Def Steve wrote:

OK. So the Phenoms may be slightly slower than the 2nd-gen, 45nm Core 2s, but they appear to be at least as fast as the 65nm variants.

Well, sadly the fastest 65 nm variant in that test seems to be the Q6600, but as you can see, it's keeping up quite well between the 2.5 GHz 45 nm variation and the 2.33 GHz one.
There isn't much performance difference between the two. I suppose this again depends on whether the 45 nm Core2 can flex its larger caches.

Scali wrote:

Kahenraz wrote:

In my case I specifically opted for an FX 8120 for the advantage of higher integer IPC throughput per-process

Since when do the FX have better IPC than anything?

I went into detail right after your quote..

... for the advantage of higher integer IPC throughput per-process without the additional need of a hypervisor or marshal service.

etc..