Dolphin, the GameCube and Wii emulator - Forums

Full Version: Hardware Discussion Thread
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
@wizardsmagic007

This really should be in its own thread. But what the hell, I'll answer it here. List time!

First off you have background tasks running in both instances such as firefox. If you want accurate measurements you need to close them.

Second your assumption that the Intel cpu should be faster is wrong. You're comparing a laptop Intel cpu (35w TDP) against a desktop AMD cpu (125w TDP) with 4 times as much power (wattage). They should be about equal for dolphin if I'm not mistake. The AMD cpu has a 40% higher clock rate (4.4GHz vs. 3.2GHz) but the Intel cpu has a 40% higher IPC in dolphin which should cancel it out.

Third your assumption that GPU load in software is an accurate way to detect bottlenecks is wrong.

And fourth your assumption that GPUs can only reduce performance if they are the bottleneck is wrong. Although I wouldn't expect a discrete GPU to make that much of a performance difference if you are using minimal settings on both setups it is certainly possible.

Also you should list your settings and make sure both are running at 1x IR. Also try both openGL and d3d. Some GPUs respond better/worse to different APIs.
(11-10-2014, 06:49 AM)NaturalViolence Wrote: [ -> ]@wizardsmagic007

This really should be in its own thread. But what the hell, I'll answer it here. List time!

First off you have background tasks running in both instances such as firefox. If you want accurate measurements you need to close them.

Second your assumption that the Intel cpu should be faster is wrong. You're comparing a laptop Intel cpu (35w TDP) against a desktop AMD cpu (125w TDP) with 4 times as much power (wattage). They should be about equal for dolphin if I'm not mistake. The AMD cpu has a 40% higher clock rate (4.4GHz vs. 3.2GHz) but the Intel cpu has a 40% higher IPC in dolphin which should cancel it out.

Third your assumption that GPU load in software is an accurate way to detect bottlenecks is wrong.

And fourth your assumption that GPUs can only reduce performance if they are the bottleneck is wrong. Although I wouldn't expect a discrete GPU to make that much of a performance difference if you are using minimal settings on both setups it is certainly possible.

Also you should list your settings and make sure both are running at 1x IR. Also try both openGL and d3d. Some GPUs respond better/worse to different APIs.

I was running them at the lowest graphical settings (same build and settings). Direct3D has an improvement over openGL for Intel HD 4000. I was just under the impression that if the i7-3630QM has better single thread performance then it would be better than the FX-8120 as long as GPU usage wasn't 100%. Even overclocked the 8120 scores 1.10 in cinebench single core benchmark but the 3630QM scores 1.41. The results are still the same without background programs running.
It's also weird for the games that run slower with the FX-8120 run faster with the i7-3630QM. That is more or so why I am confused regarding the GPU usage.

(11-09-2014, 08:58 PM)Link_to_the_past Wrote: [ -> ]
(11-09-2014, 07:44 PM)wizardsmagic007 Wrote: [ -> ]Intel i7-3630QM = 44fps
AMD FX-8120 = 66fps
Attached Screenshots of Super Mario Galaxy 2

Neither are GPU bound so the i7-3630QM in theory should have more FPS? Unless Intel HD graphics drivers are so bad they degraded CPU performance?

Of course intel graphics are bad, you should use low ir to not have any gpu bottlenecks like 1x (frankly i remember cases when a dedicated graphics card improved fps even at 1x ir by a bit). You can't compare intel hd 4000 with a gtx 470.

I ran them both at absolute lowest graphical settings. I was just under the impression that the i7-3630QM would be faster if the GPU load wasn't max. However it seems I am wrong.
I've posted this at two places now and no one seems to know what is going on and is as flustered as I am trying to diagnose and possibly resolve. I'm gonna try my luck here hoping someone might have read about or experienced a similar issue.




So the gist of it is with the AR GTX 460 in the top pci-e slot and the TR GTX 460 in the bottom pci-e slot I seem to get worse performance in multiple games. Worse than single card performance. There's also graphic anomalies that I'm not sure are artifacts but reading up some anomalies seem to be caused by a loose sli bridge. However switching the cards around so the TR is on top and the AR on the bottom those issues disappear and I experience much better performance.

If I remember correctly this started happening when I received a TR model after I sold my second AR GTX 460 back around the time I built this system. The screws for the heatsink on the TR model are inferior and not spring loaded, the bios on the TR model is a different version than the AR model.
Oh yay, I get to update from the 344.60 nvidia driver to the 344.75 one! I bet a whole lot of exciting new features and optimizations have been added; it's a whole 0.15 new versions, after all!
Just bought a G3258 and 750ti from Micro center. Time to install this and run some tests on it.
That was fast! Guess you're going to test Dolphin and PC games right?
Anywhere I post everyone seems to be as stumped as me about my gpu and sli issues. I tell everyone since swapping the TR GTX 460 to the top pci-e slot and AR GTX 460 to the bottom pci-e slot and upgrading to the latest Nvidia drivers I'm getting awesome performance. I should jump into a full TF2 server to see if anything has improved since I lost my lte net. I jumped into an empty Dustbowl server with my satellite connection and was getting 300fps to 400fps. I would hate to jump into a filled server then leave without playing, ping/latency is just so high I can't really do anything.
(11-23-2014, 09:26 AM)Shonumi Wrote: [ -> ]That was fast! Guess you're going to test Dolphin and PC games right?

Of course. The only thing is I have to wait until tomorrow to finish it because I don't have the right screwdriver (The screwdriver I have won't fit into a tight space and the one I have that will fit won't turn the screw). Also at Microcenter a Pentium G3258 + MSI PC Mate Z97 is only $100.
Yeah, not having the right screwdriver is a PITA (almost happened to me last year when building this). Thankfully my dad's a real car-buff, so we have a dozen of every kind of screwdriver.

Looking forward to pics and accounts of how things go. Have fun now.
(11-23-2014, 02:38 PM)Shonumi Wrote: [ -> ]Yeah, not having the right screwdriver is a PITA.
True. I "had" to buy a new Blu Ray drive because I damaged the screws (screwdriver was too small) when I tried to unscrew the previous BD drive that is in my old PC. My brother also wanted to keep the drive, so I had a second excuse to buy a new one.