CPU Benchmarking - Quad vs. Dual in various RPGs

[Jlagreen]

Since I recently changed my CPU because of PCSX2 I decided to make a small comparison between my old CPU and my new one. And of course I want to share my results with you

1. Testsystem

Compared CPUs:

• Intel C2Q Q6600 G0
  
• Intel C2D E8400 E0
  

Compared CPU clocks:

- Intel C2Q Q6600 G0

• 2000 Mhz
  
• 2600 Mhz
  
• 3200 Mhz
  

- Intel C2D E8400 E0

• 2000 Mhz
  
• 2600 Mhz
  
• 3200 Mhz
  
• 3800 Mhz
  
• 4400 Mhz
  

Other Hardware components:

• Giagabyte GA-P35-DS3
  
• GeiL 2x2GB DDR2-1000 RAM
  
• Leadtek GF260GTX+ (216 shader) @ default clock
  
• WD Raptor 150GB (OS, PCSX2 and ISO on it)
  
• Windows Vista x64 Home Premium
  

PCSX2 Version and Settings:

- PCSX2 0.9.5 Beta SVN 394

• EERec with VU0rec+VU1rec ON
  
• MTGS ON
  
• Dual Core Mode ON
  
• Frame Limiting Normal
  
• Disable Forced ABS ON
  
• EE/IOP Sync Hack ON
  
• Tighter SPU2 Sync ON
  

PCSX2 Plugins and Settings

- GSdx 0.1.10 852m MSVC15.00 SSE3 (for both CPUs)

• Screen Resolution: 1920x1200 60Hz
  
• Renderer: Direct3D 10 (Hardware)
  
• Shader: Pixel Shader 3.0
  
• Interlacing: None
  
• Aspect Ratio: Stretch
  
• D3D internal res: 1920x1200
  
• NLOOP hack: grey
  
• Texture filtering ON
  
• Wait vsync OFF
  
• Logarithmic Z ON
  
• Alpha Correction ON
  

- ZeroSPU2 0.4.6

• Time Scaling ON
  
• other options OFF
  

- LilyPad 0.9.4 for Controllers

- P.E.Op.S CDVD 1.3.0

• Caching mode: Thread read
  
• Use addtitional 4 MB ON
  
• other options OFF
  

- Other plugins the Null driver is used


2. Tested Games

I tested only some NTSC RPGs as these are the only Games I use PCSX2 for. For my PAL Games I can use my PS3 but for the NTSC Games I prefer PCSX2 over my old PS2.

For Benchmarking I used FRAPs set to 30 seconds benchmarking time. So I load a save state in the game and try to do the same thing every time. Like running around or battling some enemies. Because a one-time run is not enough I loaded the save state 3x for every setting and took the average as a result.

Dark Chronicle
   
Since I don't have any saves for this one on PCSX2 I decided to use the first battle right at the beginning for benchmarking (see attached image). Usually that battle finishes before the 30s end so some of the following cutscene is also benchmarked. But it is no problem as this only happens on higher clocks and there the speed is higher overall.

Dragon Quest VIII
   
Here I used also a scene right at the beginning of the game. It is after you take your first quest and leave the town. The green landscape takes a lot of performance from the CPU so running through here is perfect for benchmarking. Here you can see by the results that having more frequency can actually make a game much more playable. Yet I have to admit that fortunately the game performs much better in battles and there it counts .

Final Fantasy X
   
I don't think I have to say anything about this game as it was one of the first to perform nicely on PCSX2. As such I also included it but of course you can play it fine with less powerful CPUs.
For benchmarking I used a battle with the sinspawn when the crusaders attack Sin. If you haven't played the game yet, please do not look at the screenshot as it might be a spoiler for you.

Final Fantasy X-2
   
I love-hate this game. Really it is just strange. I love it because it has characteres and places from FF X but I hate most of the minigames and some missions are very annoying. Also it has no story at all but the perfect ending is just what I needed after the ending of FF X .
Anyway for benchmarking I use the final boss fight as I recently played that game on PCSX2 so if you don't know this game yet, it might be a spoiler for you to see the screenshot. Besides I disabled the default patches as it runs fine without them, I even found a way to see the FMVs quite well .

Final Fantasy XII
   
The best FF ever in my opinion as I hate random encouters. The battle system has improved so much with this game and I pray that it will never change back. Unfortunately the story isn't that great but Basch and Balthier are excellent characters.
As for benchmarking I used the second battle against Gilgamesh since it eats a lot of CPU power as Enkidu is rendered by the CPU. Of course I made sure the view was always on the enemy and thanks to Gambits I didn't have to do anything else.

Persona 3 FES
   
I bought this game recently by suggestion of one of my friend and reading some reviews. I only played for a few minutes but I already know that I won't regret it as I am a big Anime Fan and this RPG is just right for me.
For benchmarking I use the opening ceremony in the high school at the beginning of the game. This game performs quite bad even with high CPU clocks but since I didn't use any patches and special configuration it might change with proper settings.

Persona 4
   
I bought it together with Persona 3 as I am the kind of guy who always buys a successor. I have no idea about the compability of this game but I already read some threads and it seem quite good with certain patches.
Anyway for testing I used the cutscene when the main character drives into town by car at the beginning of the game. As the cutscene is rendered by CPU and GPU it can be used for benchmarking quite well.

Shadow Hearts 3
   
Shadow Hearts has a unique battle system and I liked the first game of the series very well. I didn't play the second one yet as it doesn't run on neither my PS3 nor on PCSX2 but fortunately this one runs quite well in PCSX2.
The Benchmark start right at the beginning in front of the house where you need to gather information. So I used the 30s to walk around and ask people. It was a little annoying as you have to repeat the same pattern all the time.

Xenosaga I
   
The Xenosaga saga is the reason I acutally decided to change my CPU as I really want to play these games.
In Xenosaga I the point of benchmarking is a battle during the tutorial. I always attacked in the same pattern and therefore the benchmarking results should be accurate enough.

Xenosaga II
I had this one as PAL for a long time but I never played it because I wanted to play EP. I first. I wasn't satisified with just watching cutscenes from the first game. Now I bought the complete saga and I look forward to play it on PCSX2
As for benchmarking I use a point at the beginning of the game where you are walking in a robot suit and fight some other mechs. The results you find to this game are not really accurate in my opinion. The reason is simple because the performance when you walk around in the world is much less than during battle. It's like you have 45-50 FPS in world mode and up to 80 FPS in battle mode. Still the game is playable that way as at least the battles have no slowdowns.
What I am trying to say is that it looks the game is playable with an average of 45FPS on a C2D @ 2Ghz but actually in world mode you walk around with 10-20 FPS which is very frustrating.

Xenosaga III
The third installment of the saga is all I can say about it for now.
For testing I use the point where you have to breach a security check by destroying some colored objects in a certain pattern. During benchmarking I didn't care about the pattern I just destroyed them all directly to not start any cutscene.
The performance is better than in EP. II but it is indoor and I haven't played anymore so I can't say if the game is fully playable or not. I didn't test any battles as I don't want to spoil to much for myself.


3. Results

Attached you will find pictures of the benchmarking results for all Games with the different CPU clocks. Of course the average FPS are displayed as a result.

Since I compared the Quad- and Dualcore also on the same clock speed with same FSB and RAM speed one thing is clear. There is no point in getting a Quadcore for PCSX2 at all. I didn't turn off any background process so my antivirus program was still running and also Vista's indexing service. Still the speed is almost the same, actually the Dualcore outperforms the Quadcore a little on same clock speeds. The reason for that is probably that the 45nm Dualcore has a higher L2 Cache (6MB/2Cores vs. 4MB/2Cores on the Quad).
Also in my daily use I don't feel any difference between a Dualcore or Quadcore. So instead of buying an expensive Quadcore it is better to get a Dualcore with higher clocks for PCSX2. Of course if you have PC applications or games which profit from Quadcores (Video encoding or GTA4) then a Quadcore is a better choice. But for PCSX2 only a Dualcore is a better choice.

What is also nice is the scaling of PCSX2 with clock speed. By changing from [email protected] to [email protected] the clock speed increases by 37.5% and you can see that speed increase in most of the games. This is good news as for example Persona 3 seems to profit a lot from it. I really was surprised and of course I like it as it improves playability a lot.

I do not regret downgrading from Quad- to Dualcore at all. It was a very good decision as now I can play most of the RPGs without slowdowns on PCSX2. I just can't stand if a Game performs less than 60FPS and for that I am ready to pay. A Q9000 or Nehalem wouldn't perform much better than a Dualcore but cost much more so right now I think an E8400/8500/8600 E0 stepping overclocked is the best choice for PCSX2.

That said, should the day come when PCSX2 supports several Cores (multi-threaded) I will immediately order a Nehalem .


EDIT:
Unfortunately, I reached the limit of attachements so I can't add screens of Xenosaga II + III

[rama]

Very nice thread
We can link people here that ask on what speed to expect.
Thanks for the extensive testing

[Bositman]

Nice Make an extra reply for the rest of the attachments
Also it would be nice if you added both processors at stock speed,for those who don't want to OC but want to see how they perform out of the box

[Zeydlitz]

Yes, and add 60FPS line -- to made a good enough speeds more obvious.

[Thrive]

Uhh..doesn't PCSX2 only thread to two threads at the moment? So technically this is a Dual Core (2 cores utilized) vs. Effective Dual Core (2 cores utilized, 2 idle) comparison.

Also..were people actually buying Quads for PCSX2 specifically, expecting to see superior performance compared with similarly-clocked Dual-Cores?

Nice work nonetheless -- it's really nice to see a controlled benchmark showing how much clock speed affects performance. As a Q6600 owner myself (G0 rev sitting at 3.2GHz), I am a bit envious of the monstrous overclocking the E8400 is capable of. Still, I wouldn't consider downgrading just to run PCSX2 a little faster, as it's hardly the primary task for which I use my system

[Jlagreen]

Nice Make an extra reply for the rest of the attachments
Also it would be nice if you added both processors at stock speed,for those who don't want to OC but want to see how they perform out of the box

nice idea, I will run the tests with default clocks aswell.

I will add a 60FPS line in red to make it obvious later and concerning Thrive's post. Yes, there are people who ask if a Quad is useful because of background processes. It isn't at all, I run several processes in the background and have no speed impacts with the E8400 at all. Of course if I wanted to encode while playing PCSX2 it would be a different matter but then PCSX2 would be slower on Quad as well as the encoding would use HDD and RAM performance. Besides Windows Vista and especially XP aren't that good with multi-threading as I experienced in the past .

Besides another interesting information. I copied the game ISOs on different locations and realized that it does matter where you run it from.
Before I always wondered why I had sometimes slowdowns in FF X during cutscenes. Then I found out it depends on access time of the media. From DVD it was horrible as the FPS sometimes dropped to 30. But also on my USB HDD it was terrible and improved much by copying the ISO on a faster HDD. Then finally I put it directly into my PCSX2 directory on my Raptor and since then I haven't experienced any slowdowns during cutscenes anymore. Turning off processes which access the HDD frequently also helps a lot. One for example is the Windows Indexing service.

[Thrive]

Yes, there are people who ask if a Quad is useful because of background processes.

Jesus, what sort of background processes were these people be running to ask that?

Turning off processes which access the HDD frequently also helps a lot. One for example is the Windows Indexing service.

Ew, yes -- especially on Vista; HDD thrashing from the Search Indexing service can be brutal. Disabling it is also unlikely to affect the average user..I can't remember the last time I used file search >_>

[sephiroth61]

Try Dragon ball z tenkaichi 3 and watch the difference ...

[Jlagreen]

Sorry, but I don't have that game besides I don't like Dragonball at all...

[stpdrgstr]

Too bad I can't get my e8400 to 4 ghz without it going nuts. But nice, this gives me an approach of what games I'm gonna buy . Xenosaga games seem to run really fast. And Dark Chronicle, is it good?