290 Mhz 2.5-3-3-6 or 310 Mhz 3-4-4-6?

[ivanvt]

Ive been trying to find out how the memory bandwidth interacts with the latency but i still cant find the answer i want.

 

do you have any guide where i can find that info?

 

Iam asking this because as the topic says i dont know which one is better (290 Mhz 2.5-3-3-6 or 310 Mhz 3-4-4-6)

 

I pretty sure that my memory can give me at least 320 considering the timmings but i was to tired to keep trying.

 

As for my CPU it tops at 2766 Mhz with 1.66 Volts, considering it is a 3800+ venice, i think it has more potential do you think is safe to increase de CPU VID Special Control further? (im using 123%) on stock air cooling? im getting 59 celcuis fully loaded with prime95.

 

Another fast question, my HTT (mobo) tops at 310 Mhz, how can i increase this value? which changes do i need to make? ive seen DFI Ultra-D reaching 330 Mhz.

[(Rockafella)]

The best way to find out what settings give you the best performance is to run some tests and find out. There are numerous test that will tell you the performance of your memory. My personal forvorite is everest home edition. It tells you your memory's read/write speeds ect. Do a search for it and download it and run the tests on your memory with different sets.

[loc.o]

SPI likes tight timings. Have a look at 290 2,5-3-3- and see what it does. Get better cooling for cpu, its quite hot.

 

edit 2,5-3-3 will not make up for the 180mhz loss if you must use 9 multi.

[Surrept]

My guess would be 2.5 over 3. On my previous system it took at least 10-15mhz before benchable gains were made up for when going from 2 to 2.5 At the same time the crisper feel of lower cas isn't shown very well through bench's, really the only number I can think of that shows this is speed factor in sandra. The way I look at it is that tighter timeings offer improvement when quick changes are required and it can be felt on windows boot and surfing. Looser timeings probably offer improvements when dealing with larger more predictable tasks, encodeing perhaps?

[madpete]

My experience of this has been that the benchmarks will come out better with the higher mhz, however in the real world your programs and apps prefer the lower timings.

 

I would probably say the 290mhz @2.5-3-3 but like the lads say run some tests, try something like 3dmark 2001se too as this will give you a better feel than sandra etc.

 

If I remember correctly, test 2 is sensitive to memory performance.

[xen]

does anyone still use pcmark as a valid benchmark anymore?

[Surrept]

I've never considered pcmark an overly usefull benchmark, so far as the final number is concerned at least. The only use I've found with it is watching each seperate benchmark from clock to clock to see how each is effected.

 

Again about speed factor, from what I remember multi effects speed factor much more then timeings which makes it not a very usefull measurement of timeings on memory. Higher fsb can make up for looser timeings in theory, you just need enough of a gain. Tight timeings take fewer cycles to make a transaction, however if theres more cycles in the same amount of time you can make a read/write transaction in less time. Still theres alot to be said in feel of a system that isn't translated in bench numbers.