What's the opposite of a divider?

[Angry_Games]

there's actually some hidden stuff in the new revE cpu's that does allow for the memory to run faster than the cpu...just like the new Intel DDR2 boards that allow the cpu to run @ 400FSB and the memory @ 533FSB

 

i think the 7/2 alpha bios has this bit of code in it

 

not recommended though unless you got all sorts of backup plans in place

[pershoot]

ahh yea, i completely forgot. he has got a winnie so the >200 divider option will most likely not show/work properly.

[THunDA]

THunDA posted above that there's no way to use a divider to make RAM run faster than the cpu.

 

Thats before I heard about the 7-2 bios.. :cool:

 

 

ahh yea, i completely forgot. he has got a winnie so the >200 divider option will most likely not show/work properly.

 

Yup, it only works for rev E cpu's..

[Surrept]

his cpu limits him, so therefore his throughput is limited by max htt allowed. by cranking the memory higher then the HTT, he gets the added throughput, but keeps his max CPU MHZ. this way he can max out his ram mhz/throughput as well as his max cpu mhz.

 

I understand that he's getting a problem on max speed for his cpu, I just don't understand how faster ram would improve anything. I'm seeing three speeds when I think of this situation, cpu/cache fsb and mem. Mem runs much slower then cpu obviously and fsb is the speed that the cpu/cache communicates with the rest of the system. I just don't see the theory behind any improvement if the fsb isn't moveing the data just as fast as the mem.

 

Correct me if i'm wrong, I prefer experiencing and theorizeing rather then researching so being wrong is something I often deal with.

[pershoot]

I understand that he's getting a problem on max speed for his cpu, I just don't understand how faster ram would improve anything. I'm seeing three speeds when I think of this situation, cpu/cache fsb and mem. Mem runs much slower then cpu obviously and fsb is the speed that the cpu/cache communicates with the rest of the system. I just don't see the theory behind any improvement if the fsb isn't moveing the data just as fast as the mem.

 

Correct me if i'm wrong, I prefer experiencing and theorizeing rather then researching so being wrong is something I often deal with.

 

the memory controller is ondie, so therefore, the memory controller will run the ram as fast as you want, and take this straight to the cpu. the cpu communicates with the rest of the system via HT/LDT Bus. thats why when you ramp up the HTT, even if the memory is divided to run lower then the actual HTT (which is a artificial clamp), throughput is increased (via CPU MHZ). the same holds true if one were to do the opposite. there is no bottleneck, since communication of the cpu/chipset/ram doesn't need to filter through a 'northbridge' of sorts. thats why you see gains if the memory is run faster then the actual HTT. there is more then one way to gain throughput on A64. latency, cpu MHZ, and now with the addition of 702, memory freq > cpu MHZ. a good combo of all 3 makes for a nice balance in those systems where the ram scales past the cpu's limits, w./out keeping a low multiplier to only satisfy the ram and maintain a lower then max cpu limit.

 

hope that made sense (still drinkin my mornin coffee), hehe.

[Surrept]

Thanks that made perfect sense, explained away the bottleneck I had imagined. Ohh I hate you a64 guys even more now!