Which set of RAM timings is faster? (specifics inside)

[avi85]

Which set of RAM timings is faster?
2-3-2-6-1T @ 200MHz or 2.5-3-3-7-1T @ 220MHZ
These timings are on an X2-4200+
TIA



EDIT: Also, is there some sort of performance penalty when using a divider?

 

[cmdrdredd]

go with the little higher frequency.

 

[Ricemarine]

I'd go with 2.5-3-3-7 because of higher bandwidth.
Run Everest Home Edition and test.

And, I don't believe there is a performance penalty by using a divider.

 

[avi85]

Originally posted by: cmdrdredd
go with the little higher frequency.

What is the reason that it would be better? after all, it's a higher latency. Do A64's need higher bandwidth more than lower latency?
I'm just trying to understand this thing.

 

[BitByBit]

This is the last time I am going to post this.

Latency = CAS / Frequency

 

[avi85]

Originally posted by: BitByBit
This is the last time I am going to post this.

Latency = CAS / Frequency

Yeah, I saw that you posted that in another thread but my question is what do A64's need more, lower latencies or more bandwidth?

 

[cmdrdredd]

Go with the higher frequency.

 

[myocardia]

Originally posted by: avi85
my question is what do A64's need more, lower latencies or more bandwidth?

Optimum performance comes from having both. If you have to give up one of the two, go for bandwidth, at the expense of latencies.

 

[Idontcare]

Originally posted by: BitByBit
This is the last time I am going to post this.

Latency = CAS / Frequency

AT needs a Memory Forum section and a round-up of your latency-related responses to the populace ought to be stickied at the top. Please don't give up the good fight.

 

[Duvie]

Zebo did this for you guys al ong time ago...The extra 40mhz of ram speed witll not overcome the cas 2.5 penalty...beyond the synthetic world some of you naively live in the reality is the cas 2 timings will likely win most real world benches...

The AMD64 systems are not bandwidth limited but instead the on-die memory controller likes the low cas latency....

This has been gone thru back in late 2004 early 2005....

 

[Praxis1452]

Originally posted by: Duvie
Zebo did this for you guys al ong time ago...The extra 40mhz of ram speed witll not overcome the cas 2.5 penalty...beyond the synthetic world some of you naively live in the reality is the cas 2 timings will likely win most real world benches...

The AMD64 systems are not bandwidth limited but instead the on-die memory controller likes the low cas latency....

This has been gone thru back in late 2004 early 2005....

exactly

 

[cmdrdredd]

Originally posted by: Duvie
Zebo did this for you guys al ong time ago...The extra 40mhz of ram speed witll not overcome the cas 2.5 penalty...beyond the synthetic world some of you naively live in the reality is the cas 2 timings will likely win most real world benches...

The AMD64 systems are not bandwidth limited but instead the on-die memory controller likes the low cas latency....

This has been gone thru back in late 2004 early 2005....

Depends on the system...a C2D on P965 benefits tons from bandwidth. Very little from latency. I've tested this many times.

 

[Duvie]

Originally posted by: cmdrdredd

Originally posted by: Duvie
Zebo did this for you guys al ong time ago...The extra 40mhz of ram speed witll not overcome the cas 2.5 penalty...beyond the synthetic world some of you naively live in the reality is the cas 2 timings will likely win most real world benches...

The AMD64 systems are not bandwidth limited but instead the on-die memory controller likes the low cas latency....

This has been gone thru back in late 2004 early 2005....

Depends on the system...a C2D on P965 benefits tons from bandwidth. Very little from latency. I've tested this many times.

Fine...But the OP was talking about his system with the x2 4200+...I know the INtel is a different beast.....It likes cas timings to and generally a bit more balanced then the netburst days when all it wanted was bandwidth

 

[avi85]

Originally posted by: Duvie
Zebo did this for you guys al ong time ago...The extra 40mhz of ram speed witll not overcome the cas 2.5 penalty...beyond the synthetic world some of you naively live in the reality is the cas 2 timings will likely win most real world benches...

The AMD64 systems are not bandwidth limited but instead the on-die memory controller likes the low cas latency....

This has been gone thru back in late 2004 early 2005....

Given what you're saying, would it be worth it to go down to ~180MHz to do 2-2-2-5-1T timings (if possible, I haven't actually tested if it is)?

 

[Duvie]

Originally posted by: avi85

Originally posted by: Duvie
Zebo did this for you guys al ong time ago...The extra 40mhz of ram speed witll not overcome the cas 2.5 penalty...beyond the synthetic world some of you naively live in the reality is the cas 2 timings will likely win most real world benches...

The AMD64 systems are not bandwidth limited but instead the on-die memory controller likes the low cas latency....

This has been gone thru back in late 2004 early 2005....

Given what you're saying, would it be worth it to go down to ~180MHz to do 2-2-2-5-1T timings (if possible, I haven't actually tested if it is)?

Possibilty....You could test it more thoroughly but I dont know if it is enough of a change to offset the lower bandwidth....cas 2-3-2-6 vs cas 2-2-2-5....I dont think it will make enough...L:eave it where you have it...

 

[cmdrdredd]

In reality it's what gain? .2fps? .02sec in semperpi?

 

[Duvie]

Originally posted by: cmdrdredd
In reality it's what gain? .2fps? .02sec in semperpi?

I tested vitually the same thing the op mentioned and it made as much as a 4-5 seconds in a 2 minute test file for video encoding...take that for a 2 hour actual file it added up to like 4-5 minutes...

Nothing phenomenal....

Zebo saw more the .x fps...it was m,ore then that but not likely more then 1-2% I thik in order of importance when it came to A64 memory was...1) dual core over single core 2) 2t over 1t 3) lower cas timings