Build Your Own Arcade Controls Forum
Main => Main Forum => Topic started by: jelwell on August 27, 2005, 09:50:27 pm
-
My cpu/motherboard finally gave out. It had been on it's last throws for a while. So I need to buy a new motherboard & CPU combo. I'm not terribly concerned about price. I'm more concerned about MAME compatibility.
I recognize that Dual processor and Dual Core CPUs (http://www.mameworld.info/ubbthreads/showthreaded.php?Cat=&Board=mamechat&Number=22980&Forum=All_Forums&Words=AMD&Match=Entire%20Phrase&Searchpage=0&Limit=25&Old=allposts&Main=22711&Search=true#Post22980) aren't going to help performance in MAME.
Tom's Hardware has a great benchmarking (http://www23.tomshardware.com/index.html) page for CPUs. What I want to know is which Benchmark is most similar to MAME - with it's single core/cpu restriction?
Are the AMD FX processors dual core?
Lastly, I know that AGP Fastwrite (http://forum.arcadecontrols.com/index.php?topic=35234.0) makes a large difference. But newer motherboards are PCI-Express based with some supporting SLI. Do either of these features help MAME performance?
Joseph Elwell.
P.S. I already know that most games will run on a low end PC. I'm looking to try to support future releases/games. As well as wanting to play Simpon's Bowling which I read (http://www.mameworld.info/ubbthreads/showthreaded.php?Cat=&Board=mamechat&Number=24474&Forum=All_Forums&Words=simpson&Match=Entire%20Phrase&Searchpage=0&Limit=25&Old=allposts&Main=24471&Search=true#Post24474) requires at least an Athlon 3400+.
-
SLI is meaningless for MAME.
-
The FX processors are still single core. I don't think any of the benchmark test truely relate to what will happen in Mame. I just finished building a computer and truthfully I wouldn't jump on the Dual core wagon or the SLI one yet. If you have the money an FX-57 is your best bet. If not look at the 4000+ or 3700+ San Diego cores. No system is capable of full speed in all games.
I'm still going through finding out what games were unplayable on my xp3200+ that are now playable on my 64 4000+ @ 2.7ghz.
I always use the "pacman" test to see the speed difference between each new computer the 3200+ runs 2,000% while the 4000+ runs %2700. While the POS Dell I use at work an intel 2.8 only gets 300%. Hope that helps
-
Lastly, I know that AGP Fastwrite (http://forum.arcadecontrols.com/index.php?topic=35234.0) makes a large difference. But newer motherboards are PCI-Express based with some supporting SLI. Do either of these features help MAME performance?
The other questions were answered, so I'll tackle this one:
PCI-E has equal or faster bandwidth than AGP. Fastwrites are a feature only some cards support well. By that I mean there are plenty of cards that OFFER support, but turn it on and you get system instability.
PCI-E is fine for MAME. You definitely won't be going backwards in performance. Plus with all of these media libraries supoprting fairly direct writing these days (DirectX, SDL, SVGALib, VBE), performance with any modern video card will definitely not slow MAME down.
The biggest bottleneck in MAME still is the CPU. My own benchmarking site is grossly out of date, but during the AthlonXP vs P4 days, the P4 "more MHz" won every time. John V (MAME32 author) has more up to date benchmarking on his site:
http://www.classicgaming.com/mame32qa/bench.htm
http://www.classicgaming.com/mame32qa/old_bench.htm
Performance varies quite a bit from game to game. There are noticable performance differences in various drivers under various hardware configs.
At the end of the day, you just have to buy the best you can afford, and be happy until the next upgrade. :)
-
Tom's Hardware has a great benchmarking (http://www23.tomshardware.com/index.html) page for CPUs. What I want to know is which Benchmark is most similar to MAME - with it's single core/cpu restriction?
Hrm.. missed that bit...
The best benchmark for MAME is... MAME! That doesn't help if you don't have the CPU to test, of course.
Otherwise I find SiSoft Sandra's ALU performance mark seems to accurately indicate a performance scale similar to the needs of MAME. That's probably the closest I can find of any synthetic benchmark to MAME performance.
And one day I will get off my arse and update my benchmark website to be user-contributable. I haven't upgraded my own hardware in some time, and it's getting harder to find people willing to let you borrow their PCs for half a day to benchmark MAME. :)