Dolphin, the GameCube and Wii emulator - Forums

So much for just a gamecube emulator =/. Guess i will just play it on my desktop computer, who would think 400 more mghz would make a difference between unplayable to playable.

Will be there any future work to make it use all 4 cores? The cpu might be weak but the emulator isnt using all its power neither...

(01-26-2012, 12:47 AM)hharry Wrote: [ -> ]Will be there any future work to make it use all 4 cores? The cpu might be weak but the emulator isnt using all its power neither...

no, next time use the search function for another question
http://forums.dolphin-emu.org/showthread.php?tid=20573

(01-24-2012, 01:22 PM)hharry Wrote: [ -> ]Currently testing with the game D.O.N. FPS at menus is 60, but in game while fighting it goes down to 25-26 FPS and the game runs in slow motion making it unplayable.


CPU: Amd phenom II 2.0ghz N930 quad core.

Also i have tried it on my old desktop computer, which has a Q8200 and Ati radeon HD3450 video card(which is a horrible one, 64 bits and 512mb of memory i cant run nearly any game with it), yet the same game runs with 10 more FPS in that one.

The CPU is pretty much the only thing that matters.

Go to http://www.cpubenchmark.net/high_end_cpus.html and check what points your CPU gets, you need around 5000 points in order to get full speed in the majority of games.

The N930 doesn't even get half that, 2400 points. As such, it will never be able to get full speed with the current client.

Your Q8200 is a bit better, it gets 3270 points and as a result it also gets better fps, but still not good enough for full speed.

Quote:Go to http://www.cpubenchmark.net/high_end_cpus.html and check what points your CPU gets, you need around 5000 points in order to get full speed in the majority of games.

1. Passmark results rarely correlate well with performance in dolphin.

2. Passmark is a terrible benchmark to use in general (in fact you should stay away from synthetic benchmarks in general). Results are often very inconsistent with real world performance.

And let's not forget that passmark is heavily multithreaded, dolphin is not.

(01-26-2012, 09:55 AM)NaturalViolence Wrote: [ -> ]

Quote:Go to http://www.cpubenchmark.net/high_end_cpus.html and check what points your CPU gets, you need around 5000 points in order to get full speed in the majority of games.

1. Passmark results rarely correlate well with performance in dolphin.

2. Passmark is a terrible benchmark to use in general (in fact you should stay away from synthetic benchmarks in general). Results are often very inconsistent with real world performance.

And let's not forget that passmark is heavily multithreaded, dolphin is not.

Do you have any dolphin "non-synthetic" benchmarks with every cpu on the market?

Because if you don't, the synthetic cpu benchmarks are what we got. It is much better than the alternative, hearsay from people who has never used the cpu, which is what information you are coming with. Also, would you show some sources for your first claim? That there is next to no correlation between cpu benchmark scores and Dolphin performance? I can make it easy for you, since it rarely correlates, according to you. Take the top ten benchmark scorers and give some examples of how poorly they perform in Dolphin. As the "results rarely correlates", almost none of them should be any good in Dolphin and thus it should be a very easy task for you.

Quote:Do you have any dolphin "non-synthetic" benchmarks with every cpu on the market?

Because if you don't, the synthetic cpu benchmarks are what we got.

So unreliable results are better than no results?

Quote:It is much better than the alternative, hearsay from people who has never used the cpu, which is what information you are coming with.

That sentence makes no sense. If someone says that they are getting say 24 fps running the 1st level of SMG on a 64 bit windows build and another person gets 32 fps running the 1st level of SMG on a 64 bit windows build that is a lot more reliable than using a completely different application, a heavily multithreaded synthetic application, to benchmark performance.

Quote:Also, would you show some sources for your first claim? That there is next to no correlation between cpu benchmark scores and Dolphin performance? I can make it easy for you, since it rarely correlates, according to you.

Search the forums. I've been an active member for almost 3 years on these forums. Nearly half of all threads are about performance issues. Stay here long enough and you'll build up a pretty good idea of how well each type of cpu architecture performs with dolphin.

Quote:Take the top ten benchmark scorers and give some examples of how poorly they perform in Dolphin. As the "results rarely correlates", almost none of them should be any good in Dolphin and thus it should be a very easy task for you.

That's poor logic.
Your logic is the equivalent of "the fastest cpus perform well in both applications therefore all of the results correlate well". The fastest cpus on the market usually perform well in all software because they don't cut corners anywhere (lots of cores, big cache, high clock rates, high IPC, etc.), but this has nothing to do with the point at hand. Many cpus perform far better in certain types of applications that play to there strengths well.

Compare a hexacore sandy bridge like the 3930K against a quad core sandy bridge like the 2600K in passmark. The hexacore chip performs 50% better in passmark because it has 50% more cores and passmark is a heavily multithreaded application. But do you honestly think that the hexacore chip will perform 50% better than the quad core cpu in dolphin, an application that only uses 2 cores?
Hint: The answer is no.

The benchmark is fail because dolphin only uses 2 cores
AMD 6 cores and Faildozer pass 5000 points but they ain't good enough to run dolphin
PC games : the more core the more speed you get
Dolphin : CPU architecture (SB, Ivy Bridge,haswell...)and clock speed (3.0ghz) are the most important things

Quote:So unreliable results are better than no results?

Of course, and it is you who claim they are not relevant.

Quote:That sentence makes no sense. If someone says that they are getting say 24 fps running the 1st level of SMG on a 64 bit windows build and another person gets 32 fps running the 1st level of SMG on a 64 bit windows build that is a lot more reliable than using a completely different application, a heavily multithreaded synthetic application, to benchmark performance.

The fault with your logic is that with hearsay there could be any number of reasons for the fps. Such as lots of programs running, viruses, faulty or slow ram, wrong settings, graphics card, misinformation and so on, there are too many external (non-CPU related) factors to consider.

Quote:Search the forums. I've been an active member for almost 3 years on these forums. Nearly half of all threads are about performance issues. Stay here long enough and you'll build up a pretty good idea of how well each type of cpu architecture performs with dolphin.

So, no sources then, as I expected. This happens all the time unfortunately, someone claims something is true but has no way to back it up other than saying "I have been here for almost three years!!1, and because of that I must be right!!" and on a side note, it is much closer to 2 years than 3 years, and shorter than me.

Quote:That's poor logic.
Your logic is the equivalent of "the fastest cpus perform well in both applications therefore all of the results correlate well". The fastest cpus on the market usually perform well in all software because they don't cut corners anywhere (lots of cores, big cache, high clock rates, high IPC, etc.), but this has nothing to do with the point at hand. Many cpus perform far better in certain types of applications that play to there strengths well.

It is your statement that is wrong to being with. You claim that cpu benchmark performance "rarely" correlates with dolphin performance. Obviously, if all the top ones perform good in Dolphin, that is an extremely good correlation. For if there was no correlation, picking a random cpu would have the same chance of being good in Dolphin as picking the top one. Also, picking the bottom ones would have an equal chance of being good as a random one. So, to be extremely generous, please pick any of the bottom 300 CPU's and name one that runs dolphin well. If not, then there is indeed a correlation, and a sound one at that. For then it does mean that CPU benchmarks do give a very strong indicator of dolphin performance, and, that you are wrong.

Of course there is a correlation between benchmarks and dolphin performance. The question is if the benchmark information is actually valuable. I mean, everbody can tell that a CPU clocked at 3.5GHz will perform better than one clocked at 2.0GHz in dolphin. You don't need a benchmark for that.

Here's a real-world example: I have two CPUs here, an Core i5 450M @ 2x2.40 GHz, and a Core2Duo E6420 @ 2x2.7GHz. The former has about twice as many points in almost any benchmark, still it performs only marginally better (if not worse) in dolphin.

So, to put it short: I'd trust experienced users' opinion on what CPU will work more than those benchmarks. They seem rather random to me, unless they're exactly for the task you want to do (in which case they're of course very accurate). You can't assign a CPU a number and say "that's how good it is and the higher the number the better". It depends on what code is actually executed on the CPU.

@Paladia

Please try to answer this:

Quote:Compare a hexacore sandy bridge like the 3930K against a quad core sandy bridge like the 2600K in passmark. The hexacore chip performs 50% better in passmark because it has 50% more cores and passmark is a heavily multithreaded application. But do you honestly think that the hexacore chip will perform 50% better than the quad core cpu in dolphin, an application that only uses 2 cores?
Hint: The answer is no.

You want some more specific examples?
Bulldozer FX6100, FX8120, and FX8150 perform better than phenom II X6 or phenom II X4 in passmark but worse in dolphin.
Core i7 980 beats the 2600K and 2500K in passmark, but the 2500K and 2600K are a lot faster in dolphin.
AMD llano quad core and phenom II X4 beat desktop core i3 sandy bridge cpus in passmark. But desktop core i3s beat the living shit out of llano or phenom II X4 in dolphin.
And on and on and on.
Just take a look at the cpu in the 4,000-6,000 point range. Half of those cpus perform poorly with dolphin the other half perform well.
Hexacore opterons and core i5 quad cores both score in that range but core i5 quad cores are wwwaaayyyy faster in dolphin.

Dolphin doesn't care how many cores you have, how fast your SSE engine is, how fast the FPU is, or any of these things that clearly make a big difference with passmark. Passmark is a totally different type of application.

Quote:Of course, and it is you who claim they are not relevant.

Read your post again:

Quote:Do you have any dolphin "non-synthetic" benchmarks with every cpu on the market?

Because if you don't, the synthetic cpu benchmarks are what we got.

See. You said we need to use unrelated synthetic results because that is "what we got".

Quote:You claim that cpu benchmark performance "rarely" correlates with dolphin performance.

Read my post again:

Quote:1. Passmark results rarely correlate well with performance in dolphin.

2. Passmark is a terrible benchmark to use in general (in fact you should stay away from synthetic benchmarks in general). Results are often very inconsistent with real world performance.

And let's not forget that passmark is heavily multithreaded, dolphin is not.

Quote:Obviously, if all the top ones perform good in Dolphin, that is an extremely good correlation.

*facepalm*

Please never go into any field that requires engineering or statistics skills. If you ignore all but the top results in any field of data you won't produce a very good statistical analysis.

Quote:So, to be extremely generous, please pick any of the bottom 300 CPU's and name one that runs dolphin well..

Very well.

Pentium G630
Phenom II X3 720
Pentium G840
Pentium G850
Pentium G860
i5-2415M
And so on.
Just to name a few that score under 3,000 points yet perform extremely well with dolphin (faster than a lot of the cpus that score above 5,000 points).

And here are some examples from that same range that perform very poorly with dolphin.
Opteron 1354
Opteron 1356
Opteron 2380
Opteron 1352
Phenom 9650
i5-2557M
A8-3500M
And so on.

And a lot of those cpus perform moderately well.

Good god, look at how innaccurate those results are. G850 outperforms the G860 in passmark. It's the same chip except the G860 is clocked at 3.0Ghz instead of 2.9GHz. How do you explain stuff like that? The results are filled with major anomalies like that. Passmark has a terrible reputation as a reliable benchmarking utility, everybody knows that (except a few people apparently).

Better yet type passmark into the forum search, go on, do it. Then type in "passmark good" into google, or something along those lines and you'll see for yourself how great its reputation is.

Quote:So, no sources then, as I expected. This happens all the time unfortunately, someone claims something is true but has no way to back it up other than saying "I have been here for almost three years!!1, and because of that I must be right!!" and on a side note, it is much closer to 2 years than 3 years, and shorter than me.

The entire forum is my history. I'm not going to spend hours searching through thousands of posts for examples just to point out a fact that should be obvious to nearly everyone who has been here for awhile.

You haven't been very active on the forums, which is kind of obvious given the statements that you're making.

Quote:Such as lots of programs running, viruses, faulty or slow ram, wrong settings, graphics card, misinformation and so on, there are too many external (non-CPU related) factors to consider.

So it's better to use a completely different application to base our expectations on instead of results from dolphin users?

You'll also notice if you browse the forum that we always ask for detailed settings, specs (including graphics card), dolphin revision, and OS. But of course you would have already known that if you actually browsed the forum from time to time.

In Conclusion:
The fastest cpus on the market usually perform well in all software because they don't cut corners anywhere (lots of cores, big cache, high clock rates, high IPC, etc.), but this has nothing to do with the point at hand. Lower end cpus have to cut corners and make tradeoffs. Some tradeoffs will not decrease performance very much if at all in some applications while having a huge performance hit on other applications. Many cpus perform far better in certain types of applications that play to there strengths well. For example dolphin will perform the same with a triple core, quad core, hexacore, or octacore cpu if all the other specs are the same. However a multithreaded video encoder may run twice as fast on the hexacore cpu vs. the triple core cpu.

Therefore it is far more reliable to use the results of actual dolphin users and the knowledge of experienced users and developers than some totally unrelated highly multithreaded unreliable synthetic application.

Don't tell users that they need a certain passmark score to get good performance with dolphin, that is just wrong on so many levels.