Dolphin, the GameCube and Wii emulator - Forums

(06-09-2012, 10:06 PM)AnyOldName3 Wrote: [ -> ]Meaning weak system.

A system with a single weakness is not a weak system. That's analogous to calling an Olympic gold medal winning triathlete diabetic weak because he needs insulin.

Nope, it isn't. If you put him in a prison with no insulin, he'd be the weakest inmate, as he is completely unsuited to his environment.

Also, I am highly doubtful that someone with type 1 diabetes could complete in a triathlon successfully, as that amount of exercise, without careful monitoring of blood sugar levels, and and stabilising of these with glucose and insulin (which would cause disqualification against non-diabetics, as they don't get haribo breaks) could result in a life-threateningly serious hypoglycaemic episode, so your analogy would have been poor, even if correct.

Well,I am pretty sure phenom II is not 5 times slower then i5 2400,if you said you had pentium 4 that would make sense.
Either is something wrong somewhere with your settings or computer,or games get much more demanding as you play,which you didnt clarify.
I tried about an hour of each game with copy to ram and got 50 fps.
You should be getting maybe around 40 fps.
NOT 8 !!!

yeah, Phenom 2 is only 2.5x slower than a 2400

(06-09-2012, 10:26 PM)AnyOldName3 Wrote: [ -> ]Nope, it isn't. If you put him in a prison with no insulin, he'd be the weakest inmate, as he is completely unsuited to his environment.

No, he'd be dead. Diabetes has nothing to do with a man's strength. Take away his insulin and you take away his ability to balance his blood sugar level. Take away that, and eventually his levels will become too high, causing him to go into seizure and die.

But until he died of sugar poisoning, he'd still be in fine physical shape.

---

(06-09-2012, 10:26 PM)AnyOldName3 Wrote: [ -> ]Also, I am highly doubtful that someone with type 1 diabetes could complete in a triathlon successfully, as that amount of exercise, without careful monitoring of blood sugar levels, and and stabilising of these with glucose and insulin (which would cause disqualification against non-diabetics, as they don't get haribo breaks) could result in a life-threateningly serious hypoglycaemic episode, so your analogy would have been poor, even if correct.

I had to research this first, as I honestly didn't know. Here's what I found:

Diabetic Triathelete David Weingard

Diabetic Triathlete Shannon Standridge

Diabetic Olympic Gold Medalist Gary Hall, Jr.

So, apparently, no diabetic triathlet has won an Olympic Gold yet, but there are people who are diabetic and triathletes, and there are people who are Olympic Gold winners and diabetic.

---

(06-09-2012, 11:03 PM)rpglord Wrote: [ -> ]Well,I am pretty sure phenom II is not 5 times slower then i5 2400,if you said you had pentium 4 that would make sense.
Either is something wrong somewhere with your settings or computer,or games get much more demanding as you play,which you didnt clarify.
I tried about an hour of each game with copy to ram and got 50 fps.
You should be getting maybe around 40 fps.
NOT 8 !!!

Well, I am getting 8. I have no reason to lie about that. And no, my CPU is not 0.2x the speed of an i5 2400. The problem is that the Phenom is apparently ill-suited for the type of operations being performed during EFB->RAM and as such suffers greatly in performance, while the i5 is not ill-suited and as such performs decently.

As far as /why/ this is, I have no idea, but seeing as how my computer has no problems in every other application I run it through, I would hazard that it the fault of Dolphin, and not AMD.

eyeonus Wrote:A system with a single weakness is not a weak system.

If you have a weak cpu your system is weak no matter how fast everything else is. Unless you're running something GPU bound like a video game the cpu is practically the only component that affects application performance (memory or I/O speed might have a small impact in some applications). Most applications function like a pipeline and single weak link slows the entire thing down. Typically that weak link is the cpu.

eyeonus Wrote:The problem is that the Phenom is apparently ill-suited for the type of operations being performed during EFB->RAM and as such suffers greatly in performance, while the i5 is not ill-suited and as such performs decently.

Not true at all. The i5 is just faster in general. And both architectures will take a similar performance hit from efb copy to ram emulation which is more dependent on memory and pci-e performance. You make it sound as if there is some specific type of operation that efb copy to ram emulation makes use of that the K10 microarchitecture is not good at.

eyeonus Wrote:As far as /why/ this is, I have no idea, but seeing as how my computer has no problems in every other application I run it through, I would hazard that it the fault of Dolphin, and not AMD.

Dolphin is simply more demanding on the cpu than most applications that you're using. And the difference in performance is very noticeable since it can be directly visually observed. It's not like dolphin has some special bias against AMD. Most cpu bound applications will run faster on an i5 regardless of whether you notice the difference. The only time a phenom II X6 will get close to the performance of a sandy bridge i5 is in applications like video encoders that use stream processing and therefore are able to put full load on all 6 cores.

It is indeed odd that you are getting 8 fps anywhere in MP with efb copy to ram when my Q6600 @ 3.2GHz gets 40-60 fps depending on the area. My cpu is about 25% faster than yours as far as dolphin is concerned so it wouldn't make sense for me to be getting 5 times the performance. My results are pretty consistant with everyone else's whereas your results are not (as AnyOldName has pointed out).

Can you post your settings and revision?

(06-10-2012, 09:47 AM)NaturalViolence Wrote: [ -> ]Can you post your settings and revision?

Certainly.

The build is Release x64 3.0-709, compiled using MS VS 2010. The revision is Hashless 16a611bafae7 + ES_LAUNCH 52def7cebd32.

Hashless may be causing issues, try these settings: http://forums.dolphin-emu.org/showthread.php?tid=23618

Try this build: http://www.dolphin-emu.org/download.html

(06-10-2012, 10:58 AM)Starscream Wrote: [ -> ]Hashless may be causing issues, try these settings: http://forums.dolphin-emu.org/showthread.php?tid=23618

Try this build: http://www.dolphin-emu.org/download.html

Neither of those will do any good. I've done a lot of experimenting, with many different builds. I've used the master branch exclusively, I've used the build specifically recommended for MPT on its thread, I've used old builds, builds compiled by other people, so on and so forth, and it all comes down to one setting:

If I have EFB->Texture, I get ~40fps

If I have EFB->RAM, I have ~8fps.

Nothing else has any effect, unless the effect is a worse framerate. What I am using currently has the best performance I've been able to get, which is ~10fps with EFB->RAM. (A marginal 2fps above master, but still something.)

Keeping in mind, of course, that I'm speaking specifically of MPT:MP1/2/3. Other games perform differently- usually faster.

---

I have determined the problem.

It has nothing to do with Dolphin's settings. Nor does it have anything to do with my hardware.

It's because of SETI@Home and poclbm, programs I run continuously in the background that use the graphics card to perform calculations, specifically because the graphics card does it much, much faster than the cpu does.

I didn't consider them as possibly being at fault earlier because they automatically suspend themselves when the card is being used for its actual purpose, so I assumed they they did so when Dolphin was run as well.

poclbm does not suspend its activities when Dolphin is running, so when Dolphin and poclbm are both running, they are fighting for control of the graphics card, and poclbm wins. I do not know why this happens. I also don't know why it's only apparent when EFB is copied to RAM, and no other settings have any effect. What I do know is that this is the only time I have seen where a graphics-intensive program (AKA game) such as Dolphin does not trigger poclbm to suspend. In any case, closing poclbm when running Dolphin, I have a framerate of ~33fps on EFB->RAM playing MPT:MP1 with the settings as in my previous post.

For anyone that doesn't know, poclbm is a bitcoin miner. Beyond that, I suggest Google.

E5300 @ 3.6Ghz (i could oc to 3.9Ghz but the game run more than full speed at 3.6Ghz so i don't need to go any further)
3x IR (1080p) , no AA , 16x AF
Metroid Prime : 35-60FPS (EFB to Ram)
Metroid Prime : 40-60FPS (EFB to texture)
I think you should oc your CPU