Pentium lacks AVX that might cost you upto 1%.
what performance decrease should I expect?
|
10-18-2014, 06:34 AM
AnyOldName3 Wrote:Then why have we let people say that a G3258@4.7 = a 4690K@4.6 for months? I'm surprised no-ones noticed how often its said if it isn't true. Because I wasn't here and apparently I'm the only one here that would bother to actually test the claim before making it. All you have to do it open dolphin up on a quad core rig run it once normally and then again with two cores disabled in task manager. Then just measure the performance and load differences. As pessimistic as I am it actually surprises me that nobody bothered to do this. I'm also surprised that people have been saying there is no difference considering I and many others here used to point out the difference constantly back when we were active in this subforum. You can search my post history to confirm that. So the information was already out there for anyone willing to google it, they didn't even have to test it. The G3258 comes remarkably close to the upper end haswell cpus when OCed considering its price. Which makes it an excellent bang for your buck cpu for dolphin. But to say that its performance is exactly the same is false. Also don't forget the cache size differences. That has an impact too.
"Normally if given a choice between doing something and nothing, I’d choose to do nothing. But I would do something if it helps someone else do nothing. I’d work all night if it meant nothing got done."
-Ron Swanson "I shall be a good politician, even if it kills me. Or if it kills anyone else for that matter. " -Mark Antony
I have a G3258 and a GTX 750 setup. I'm one of the unlucky one's too with a poor overclock. Won't go over 4.2GHZ. I have a spare one but haven't had the time to install it yet.
I ran the original Delroth benchmark and got 7:18 seconds. My games seem to have a pretty comparable frame rate to those with haswell k cpu's mentioned here. I also ran the Delroth benchmark with the newest dolphin and Window's 10 technical preview. It gave me 4:42 seconds so at this point it seems we're seeing better gains in code then in silicon. A while back I looked up that "20% - 30% greater performance" for quad core quote and traced it back to 2009 or 2010. It seemed to be based on putting LLE on an extra core and possibly some background windows services. But background windows services don't usually use 20%-30% of even one core, so I've been suspicious of it. Edit: Sorry! 10%-20% greater performance 10-18-2014, 11:26 AM
Considering that HLE is quickly approaching LLE's quality, LLE on thread isn't that important any more. Quadcore is still preferrable to dualcore, but it doesn't make as huge of a difference as it would 3 - 5 years ago.
@RepWolf - I'm pretty sure the performance differences that NV mentioned happen regardless if Dolphin is using LLE audio on a separate thread or not. As long as Dolphin is in dual-core mode, it's going to be a resource hog on dual-core CPUs. Background processing for other tasks (e.g. stuff that needs to happen for the OS to run) takes a squeeze, I'd assume, which is reduced when going to a quad-core CPU. I don't know who said 20-30%, but I've seen NV peg the 10%-20% number before, which is much more believable imo.
I'm certain there's a way to limit my i5-2500K to 2 cores, correct? I think I remember seeing something like that when I was messing around with the UEFI. It really shouldn't be that hard to test if that's the case. 10-18-2014, 12:45 PM
@Shonumi - I only have a dual core so I can't test it . I am also more curious about the haswell quad core vs dual core because of the "up to 30%" increase with the new architecture. All these numbers! Who knows?! My games work great thanks to all the developers hard work. Many thanks.
10-18-2014, 07:16 PM
(10-18-2014, 06:34 AM)NaturalViolence Wrote: All you have to do it open dolphin up on a quad core rig run it once normally and then again with two cores disabled in task manager. Then just measure the performance and load differences.No you don't. You need to disable two cores entirely to get the correct measurements. Anyway I've done some testing with 4 cores, 2 cores set with affinity and 2 cores disabled. With two cores disabled I got approximately 77 fps while with all four enabled but two cores only set for Dolphin I got approximately 87 fps. With all four cores enabled and no affinity set I got 92 fps. So the performance benefits of a quad-core is 19.48% (77 vs 92). 10-31-2014, 10:53 AM
Garteal Wrote:No you don't. You need to disable two cores entirely to get the correct measurements. Correct. That's what I meant. Garteal Wrote:So the performance benefits of a quad-core is 19.48% (77 vs 92). Yup. I got the same results when I tried it years ago with dolphin 2.0. I figured not much had changed since then. 15-20% is probably more accurate to say than 10-20%.
"Normally if given a choice between doing something and nothing, I’d choose to do nothing. But I would do something if it helps someone else do nothing. I’d work all night if it meant nothing got done."
-Ron Swanson "I shall be a good politician, even if it kills me. Or if it kills anyone else for that matter. " -Mark Antony |
« Next Oldest | Next Newest »
|
Users browsing this thread: 1 Guest(s)