No SiSandra gains on my 1600+ after setting fsb to 200?

[Tipycol]

SiSandra Scores:

CPU 1666MHz @ 133x12.5:
Dhrystone = 6335 MIPS
Whetstone = 2614 MFLOPS
Int ALU = 1969 MB/s
Float FPU = 1849 MB/s

CPU 1700MHz @ 200x8.5:
Dhrystone = 6368 MIPS
Whetstone = 2657 MFLOPS
Int ALU = 2858 MB/s
Float FPU = 2576 MB/s

(ALUs/FPUs included to show mem bandwidth is increasing)

I always thought that even if you kept around same clockspeed, an fsb increase by 50% would have had a larger effect than just .5% increase. Anyone have a suggestion to what's holding it back? Is it usual to get very little to no gain if you keep the MHz around the same speed? Or maybe the architecture of the chip is a factor?

Thanks
Tipycol

 

[deathstar13]

i dont think cpu benchmarks are affected by bandwith as its just testing the cpu itself.
applications are the way to do it. 3dmark or a gaming bench would work.
your thinking "inside the box" not "outside the box" ie:cpu where the increases will show is outside the cpu.
the fsb bandwith allows the cpu to get and send info faster as it opens the pathway larger. even tho the engine remains the same.

one small indication of this will be some people will need xtra vcore to run same mhz but higher fsb as it does process the same speed but just more of it.

 

[Tipycol]

Hmm....I didn't know cpu benches didn't actually increase when you raise the fsb but clockspeed stays the same. I was wondering why I the fsb wasn't having an effect on them (I didn't understand why 3dmark01 increased but they didn't).

Thanks
Tipycol

 

[brennan77]

Raising the FSB raises the bandwidth between the CPU and memory. It affects the system communication as a whole and thus your 3dmark scores go up.

On a benchmark that isolates the CPU, nothing should change at all.

 

[chasingapple]

Sisoft is synthetic benching, test a real world app

 

[dryars]

brennan77 is correct. The only gains from upping the fsb while maintaining the same cpu speed will affect memory performance the most. Run the Sandra Memory Bandwidth benchmark and see for yourself.