Dolphin, the GameCube and Wii emulator - Forums

How demanding a game is on dolphin has almost nothing to do with the graphics. It's a combination of many things, the most important are simply what instructions the game uses and how often it uses them.

Exactly, Mega Man 9-10 will run slower than many games with 3D graphics which pushes GC/Wii to the limits.

Uhh... there are several things with the list that I find very confusing...

• * Athlon II and A4 series are two tiers lower than the Phenom II?
  * Phenom, Athlon 64 x2, and Core Duo are two tiers below lower-end Core 2 Duos?
  * The Pentium 4 isn't in an "uber-slow" tier?
  * The AMD E-series (slow), AMD C-series (uber-slow), and Intel Atom (uber-slow) are not listed?

One thing in particular is that I feel there should be a tier between "moderate" and "slow". Moderate says CPUs can reach full speed with heavy overclocking, but slow are said to simply be too slow even with overclocking for most games. I feel that these "Moderate/slow" CPUs are more of a "Usually below fullspeed even with overclocking but is still enough to be decently playable".

Quote:Uhh... there are several things with the list that I find very confusing...

Alright, then let's address your criticisms one by one.

Quote: * Athlon II and A4 series are two tiers lower than the Phenom II?

I see no problem with this. Athlon IIs are significantly slower than their phenom II counterparts in applications that require a lot of cache, like dolphin. Dolphins users report better performance on average with core 2 duo cpus than with athlon IIs. A4 cpus have basically the same specs and architecture as an athlon II so they should perform the same.

Quote: * Phenom, Athlon 64 x2, and Core Duo are two tiers below lower-end Core 2 Duos?

.......what exactly is your concern here?

Both Phenom and Athlon 64 X2 are MUCH slower than core 2 duo, athlon II, and Pentium Dual core. Core duo performs similarly to phenom and athlon 64 X2.

Quote: * The Pentium 4 isn't in an "uber-slow" tier?

While I could break the slow tier further apart into two subtiers there would be no point. The slow tier refers to any cpu that can't run the majority of games well regardless of whether it runs them at 15 fps or 5 fps.

Quote: * The AMD E-series (slow), AMD C-series (uber-slow), and Intel Atom (uber-slow) are not listed?

They are listed (scroll down to netbooks) and ranked appropriately. See above for an explanation of why I didn't include a "very slow" tier.

Quote:One thing in particular is that I feel there should be a tier between "moderate" and "slow". Moderate says CPUs can reach full speed with heavy overclocking, but slow are said to simply be too slow even with overclocking for most games. I feel that these "Moderate/slow" CPUs are more of a "Usually below fullspeed even with overclocking but is still enough to be decently playable".

There are several ways that I could have chosen to design the list. I could have even eliminated tiers completely and just made one big list ranking them all in order. The problem with having too many tiers or having a completely tierless list is that things get very complicated very fast and you have to account for differences in models.

For example a 2.0GHz core 2 duo will perform similarly to most pentium Ds while a 3.0GHz core 2 duo will perform similarly to phenom II. So where do I rank core 2 duo when the low end and high end models are so far apart in performance? The answer is you can't. So instead I build an arbitrary list based on various "user experience" levels. In other words the general user experience that you would expect to get from a cpu rather than it's specific performance numbers. This results in few tiers which makes the list simple and easy to use.

If you want to have an idea of the challenges I faced when trying to put this together you can take a look at the old thread here: http://forums.dolphin-emu.org/showthread.php?tid=24678

Just look at all that data I had to sort through. Does that look readable to you? My goal was to make a clear and concise list based on that data.

I'll admit, I did not realize Dolphin was so cache-heavy. That does explain some of the ranking choices.

As for the Athlon x2, Core Duo, and Phenom thing, it's just that I personally feel they are considerably faster than anything based on Northwood or Prescott and can even present a decently playable environment at higher clockrates (such as with the 3GHz Athlon 64 x2 6000+). I mean, the performance jump between a Pentium 4 and an Athlon 64 is one of the largest performance jumps in CPU history, and a Core Duo performs near identically to the Athlon 64 on a per-GHz basis. In fact the per-GHz performance gap between the Athlon 64 and the Pentium 4 is around the same as the difference between an Athlon 64 and Sandy Bridge:

Quote:I'll admit, I did not realize Dolphin was so cache-heavy. That does explain some of the ranking choices.

It's not super cache heavy, but cache performance/size does make a difference, more than most applications.

Quote:As for the Athlon x2, Core Duo, and Phenom thing, it's just that I personally feel they are considerably faster than anything based on Northwood or Prescott and can even present a decently playable environment at higher clockrates (such as with the 3GHz Athlon 64 x2 6000+).

Pentium D, athlon X2, core duo, and phenom all perform almost identically in most single and dual threaded applications. They are all too slow to run most games well. Search the forums for athlon x2 or athlon 64 x2 and you'll find mountains of threads with people complaining about poor performance on their 3.2GHz athlon X2.

Pentium 4 would perform much worse in dolphin mainly because it's a single core cpu. So it's even worse.

Quote:I mean, the performance jump between a Pentium 4 and an Athlon 64 is one of the largest performance jumps in CPU history,

You must be joking. Clock rate is not performance. IPC is not performance. Performance is performance. IPC * clock rate = performance

Pentium 4 may have had lower IPC (performance per clock) but the overall performance was comparable to athlon 64 (depending on the application they traded blows with each other). I was there in 2003, I saw it happen, I don't know what alternate universe you were in where athlon 64 was a "huge" leap in performance over pentium 4 but it wasn't my universe.

Quote:and a Core Duo performs near identically to the Athlon 64 on a per-GHz basis.

Once again, you must be joking. You just posted a graph which disproves this assertion. The jump from pentium D/athlon X2 to core 2 duo was one of the biggest jumps in IPC that occured for x86 cpus over the last 10 years. The same is true for performance.

Quote:In fact the per-GHz performance gap between the Athlon 64 and the Pentium 4 is around the same as the difference between an Athlon 64 and Sandy Bridge:

True. But IPC is not the full story now is it?

Quote:

If you're going to post a graph you should always provide the page source so that I know exactly what I'm looking at.

(10-10-2012, 12:42 PM)NaturalViolence Wrote: [ -> ]Pentium D, athlon X2, core duo, and phenom all perform almost identically in most single and dual threaded applications. They are all too slow to run most games well. Search the forums for athlon x2 or athlon 64 x2 and you'll find mountains of threads with people complaining about poor performance on their 3.2GHz athlon X2.

Bwah? A Pentium D and an Athlon 64 x2 perform similar? I know a Core Duo, Athlon 64 x2, and Phenom do, but a Pentium D as well? I mean, I've gotten up to 40fps in F-Zero GX of all things with my Brisbane - I really don't think a Pentium D could pull that off.

(10-10-2012, 12:42 PM)NaturalViolence Wrote: [ -> ]You must be joking. Clock rate is not performance. IPC is not performance. Performance is performance. IPC * clock rate = performance

Pentium 4 may have had lower IPC (performance per clock) but the overall performance was comparable to athlon 64 (depending on the application they traded blows with each other). I was there in 2003, I saw it happen, I don't know what alternate universe you were in where athlon 64 was a "huge" leap in performance over pentium 4 but it wasn't my universe.

Yes, maybe back in 2003 when the Athlon 64 was only around 2GHz and single core. However, a Pentium D and an Athlon 64 x2 both max out around the same clockrates - a little over 3GHz. This puts the Athlon 64 x2 ahead by quite a bit.

(10-10-2012, 12:42 PM)NaturalViolence Wrote: [ -> ]Once again, you must be joking. You just posted a graph which disproves this assertion. The jump from pentium D/athlon X2 to core 2 duo was one of the biggest jumps in IPC that occured for x86 cpus over the last 10 years. The same is true for performance.

I'm not disagreeing that the Core 2 Duo wasn't a large jump seeing how it matches a Phenom II x2. Perhaps you're confusing my mention of the Core Duo (essentually a dual-core Pentium M) as a typo for the Core 2 Duo?

(10-10-2012, 12:42 PM)NaturalViolence Wrote: [ -> ]True. But IPC is not the full story now is it?

When both architectures max out at around the same clockrate, then IPC is a very important piece of that story.

(10-10-2012, 12:42 PM)NaturalViolence Wrote: [ -> ]If you're going to post a graph you should always provide the page source so that I know exactly what I'm looking at.

The image said Tom's Hardware so I thought that was good enough in itself. Here is the article for reference:
http://www.tomshardware.com/reviews/processor-architecture-benchmark,2974.html

Quote:I'm not disagreeing that the Core 2 Duo wasn't a large jump seeing how it matches a Phenom II x2. Perhaps you're confusing my mention of the Core Duo (essentually a dual-core Pentium M) as a typo for the Core 2 Duo?

Yes. That's my mistake. I read it as core 2 duo for some reason.

Quote:When both architectures max out at around the same clockrate, then IPC is a very important piece of that story.

But they didn't max out at the same clockrate. Neither Athlon 64 nor athlon X2 got close to 3.8GHz even after years of development and slow but steady increases to the clock rate every year.

Quote:Yes, maybe back in 2003 when the Athlon 64 was only around 2GHz and single core. However, a Pentium D and an Athlon 64 x2 both max out around the same clockrates - a little over 3GHz. This puts the Athlon 64 x2 ahead by quite a bit.

They leapfrogged each other for years. I remember this quite well because this was around the time when I started to take an active interest in computers. Both architectures clocked higher in 2004. Both architecture switched to dual core in 2005. Pentium D was replaced by core 2 duo in 2006 as athlon X2 continued to gain clock rate. At no point was AMD ahead by a huge margin.

I'm not going to compare athlon 64s and athlon X2s from 2007 (which almost nobody bought) against pentium 4s from 2004/2005, that's just not fair. The performance in the chart doesn't refer to the highest model from each family, it refers to the average of all models from all years for any given family.

Quote:Bwah? A Pentium D and an Athlon 64 x2 perform similar? I know a Core Duo, Athlon 64 x2, and Phenom do, but a Pentium D as well?

When they were released yes. I bought a 3.6GHz Pentium D. It was the first cpu that I bought myself for my first home built PC. At the time the fastest Athlon X2 was 2.8GHz if I recall. The athlon X2 was faster on average, but not by a huge margin (depending on the application, netburst does much better with predictable workloads like video encoders).

(10-10-2012, 01:12 PM)NaturalViolence Wrote: [ -> ]But they didn't max out at the same clockrate. Neither Athlon 64 nor athlon X2 got close to 3.8GHz even after years of development and slow but steady increases to the clock rate every year.

D'oh, I completely forgot about the 65nm Pentium Ds.

(10-10-2012, 01:12 PM)NaturalViolence Wrote: [ -> ]They leapfrogged each other for years. I remember this quite well because this was around the time when I started to take an active interest in computers. Both architectures clocked higher in 2004. Both architecture switched to dual core in 2005. Pentium D was replaced by core 2 duo in 2006 as athlon X2 continued to gain clock rate. At no point was AMD ahead by a huge margin.

I'm not going to compare athlon 64s and athlon X2s from 2007 (which almost nobody bought) against pentium 4s from 2004/2005, that's just not fair. The performance in the chart doesn't refer to the highest model from each family, it refers to the average of all models from all years for any given family.

Ahh, see, that's where I'm seeing the difference. It now being 2012 I'm including the 3GHz Athlon 64 x2 models that came out after the Core 2 Duo.

(10-10-2012, 01:12 PM)NaturalViolence Wrote: [ -> ]When they were released yes. I bought a 3.6GHz Pentium D. It was the first cpu that I bought myself for my first home built PC. At the time the fastest Athlon X2 was 2.8GHz if I recall. The athlon X2 was faster on average, but not by a huge margin (depending on the application, netburst does much better with predictable workloads like video encoders).

But what about in Dolphin? It was my impression that Netburst was pretty bad in it. Maybe I'm wrong, but this still makes it a bit strange to have the single core Pentium 4 and Athlon 64 in the same tier as their dual-core versions, especially since the dual core models had a higher average clockrate.

Quote:But what about in Dolphin? It was my impression that Netburst was pretty bad in it. Maybe I'm wrong, but this still makes it a bit strange to have the single core Pentium 4 and Athlon 64 in the same tier as their dual-core versions, especially since the dual core models had a higher average clockrate.

They're both too slow for most games. Therefore they both belong in the slow category.