Associate
Can't answer all of your questions, Fat Rakoon, but I can explain some of the differences you are seeing with Disk Mark, SuperPi, 3DMark, etc. When you are changeing the bus speed this is having more of an effect on the whole system than just changing the multiplier. This is exactly why using multiple benchmark programs lets you get a much better picture of your Overclock than just checking a 3DMArk score. 
250x9.5 is going to make the overall system faster than a 230x10. Now I am going to go and get all the terms wrong here, so I can't be too specific. But basically the faster you are running the system bus, the faster everything else on the motherboard is getting it's data to process. Whereas the multiplier will more directly effect the CPU speed.
I don't see anything saying you have locked your memory at 200, so this means everything that is needing some form of memory access is going to be sped up at 250 instead of 230. Faster the memory can be run, the faster the data can be churned through it.
Though there is a point when you can run memory too fast and start knocking out errors. This will be at the edge of the memory overclock. This is where tweaking all those memory timings will gain most benefit. Increasing the CAS/tRCD/tRAS/rRP as you are allow the high FSB to produce less errors in the memory. Kinda like slowing it down to handle your faster requests.
Now 3DMark coming out with a lower score, but higher FPS at 230x10 instead of 250x9.5 is an odd one. Though this is going to be because of the combination of tests within 3DMark. Compare all the different tests and you will see areas where your higher FSB is a boost, and other places you will see areas where the overall CPU speed has more of an effect.
I remember tuning my old Socket A T-Bird. I used SisSoft Sandra to check what was happening as I tighten the memory timings. Even though 3DMark wasn't always showing a difference, Sandra whould show in detail when I had pushed my memory too hard and caused it all to slow down again.
Another thing to reassure you of is the 32-bit Socket A Barton vs 64-bit A64 Winchester. You are now comparing two different types of processor design running on two different chipsets with different memory. Also remember that this 64-bit processor was designed first for 64-bit calculations, then they would have made sure it could also handle 32-bit. Get yourself a 64-bit copy of XP and a 64-bit copy of SuperPI. Then I would expect you will see a big advantage of the A64.
As I said, I ain't no expert. But you are comparing two very similar systems with Barton vs Winchester. Though the Socket A has been around long enough to be tuned completely to the mature 32-bit OS. Wheras that A64 is still a relitvly new platform and hardly any of us are using the OS yet.
Overall, all of your benchmarks are pretty close, percentage wise. What does the overall system feel like? Don't give up on your A64 yet. Especially not until you get your hands on a 64-bit OS, and all the drivers etc have matured on that 64-bit platform. Then I believe we will see more benefit. I remember the 16-bit to 32-bit leap using a Pentium Pro. Comparing the mainly 16-bit Win95 to 32-bit NT 4.0 on the same multi-boot PC was very noticable. (Though that will also be due to the crappily writen Win95...)
Whoops... quite a waffle. Hopefully some sense can be made out of it. Main thing is to start looking into how your computer actually works. Research the internals a bit to understand better how your overclocking affects things. Find out what those numbers really mean.
250x9.5 is going to make the overall system faster than a 230x10. Now I am going to go and get all the terms wrong here, so I can't be too specific.
But basically the faster you are running the system bus, the faster everything else on the motherboard is getting it's data to process. Whereas the multiplier will more directly effect the CPU speed.I don't see anything saying you have locked your memory at 200, so this means everything that is needing some form of memory access is going to be sped up at 250 instead of 230. Faster the memory can be run, the faster the data can be churned through it.
Though there is a point when you can run memory too fast and start knocking out errors. This will be at the edge of the memory overclock. This is where tweaking all those memory timings will gain most benefit. Increasing the CAS/tRCD/tRAS/rRP as you are allow the high FSB to produce less errors in the memory. Kinda like slowing it down to handle your faster requests.
Now 3DMark coming out with a lower score, but higher FPS at 230x10 instead of 250x9.5 is an odd one. Though this is going to be because of the combination of tests within 3DMark. Compare all the different tests and you will see areas where your higher FSB is a boost, and other places you will see areas where the overall CPU speed has more of an effect.
I remember tuning my old Socket A T-Bird. I used SisSoft Sandra to check what was happening as I tighten the memory timings. Even though 3DMark wasn't always showing a difference, Sandra whould show in detail when I had pushed my memory too hard and caused it all to slow down again.
Another thing to reassure you of is the 32-bit Socket A Barton vs 64-bit A64 Winchester. You are now comparing two different types of processor design running on two different chipsets with different memory. Also remember that this 64-bit processor was designed first for 64-bit calculations, then they would have made sure it could also handle 32-bit. Get yourself a 64-bit copy of XP and a 64-bit copy of SuperPI. Then I would expect you will see a big advantage of the A64.
As I said, I ain't no expert. But you are comparing two very similar systems with Barton vs Winchester. Though the Socket A has been around long enough to be tuned completely to the mature 32-bit OS. Wheras that A64 is still a relitvly new platform and hardly any of us are using the OS yet.
Overall, all of your benchmarks are pretty close, percentage wise. What does the overall system feel like? Don't give up on your A64 yet. Especially not until you get your hands on a 64-bit OS, and all the drivers etc have matured on that 64-bit platform. Then I believe we will see more benefit.
I remember the 16-bit to 32-bit leap using a Pentium Pro. Comparing the mainly 16-bit Win95 to 32-bit NT 4.0 on the same multi-boot PC was very noticable. (Though that will also be due to the crappily writen Win95...)Whoops... quite a waffle.
Hopefully some sense can be made out of it. Main thing is to start looking into how your computer actually works. Research the internals a bit to understand better how your overclocking affects things. Find out what those numbers really mean.
