|
![]() |
Overclock.net - Overclocking.net > Intel > Intel Memory | |
So FSB Straps Affect Bandwidth?
|
||
![]() |
|
|
LinkBack | Thread Tools |
|
|
#1 (permalink) | |||||||||||||
|
Wannabe Photographer
![]() |
Well I just did some benchmarks and discovered my brother's Kingston crap RAM runs faster at 900MHz 5-5-5-15 than 960MHz 5-5-5-15.
The values I found were: 900MHz 5-5-5-15 1.9V read: 7009MB/s write: 6341MB/s copy: 6760MB/s latency: 71.3ns 960MHz 5-5-5-15 1.94V read: 6773MB/s write: 6336MB/s copy: 6719MB/s latency: 77.9ns This was using Everest benchmark. Am I right about the FSB straps or is this something else?
__________________
|
|||||||||||||
|
|
|
|
#2 (permalink) | |||||||||||
|
Overclocker
![]() |
Your mobo is the Asus Rampage Formula X48?
__________________Then, unlikely. Way more info needed. CPU info (multi x FSB was what?) RAM ratios for the above. Subtimings info (tRD was @ what?).
|
|||||||||||
|
|
|
|
|
#3 (permalink) | |||||||||||||
|
4.0ghz
![]() |
OP, what speed is your FSB?
__________________
Quote:
16hours Blend Stable 20 pass IBT Stable
|
|||||||||||||
|
|
|
|
#4 (permalink) | |||||||||||||
|
Wannabe Photographer
![]() |
12.5x multiplier, 300MHz FSB
Ram ratios were 16:8 for 900MHz and 16:10 for 960MHz. All timings were the same, although I gave the RAM an extra 0.04V in between those settings.
__________________
|
|||||||||||||
|
|
|
|
#5 (permalink) | ||||||||||||
|
Overclocker
![]() |
Quote:
If FSB is 300 + DDR2-900, that's 2:3 ratio, which is on the 266 strap by default. If FSB is 300 + DDR2-960, that's 5:8, which is 333 strap by default. Anyway, 266 will apply a tighter tRD (& possibly other subtimings) if it's left on auto, which is precisely why you are getting better scores @ DDR2-900. You can see tRD (Performance Level) with MemSet. If you set tRD manually to the same, DDR2-960 will be faster.
|
||||||||||||
|
|
|
|
|
#6 (permalink) | |||||||||||||
|
Wannabe Photographer
![]() |
I'm sorry, I meant 2:3, not 16:8. Can't give that a try at the moment since the ram sticks are on my brother's computer, but I'll try tomorrow.
Thanks for the response!
__________________
|
|||||||||||||
|
|
|
|
#7 (permalink) | |||||||||||||
|
4.2GHz
![]() |
FSB straps often set the default tRD. Slower tRD at the faster strap is what's causing the performance discrepancy.
__________________
|
|||||||||||||
|
|
|
|
#8 (permalink) | |||||||||||||
|
4.0 GHz
![]() |
Well also higher rate fsb itself allows for higher bandwith.
See http://www.overclock.net/7724692-post1436.html same memory clocks/timings (except for trfc but it doesnt do much) and even lower cpu speed but because fsb/rated fsb is higher everest bench numbers are also higher across the board for memory.
|
|||||||||||||
|
|
|
|
|
#9 (permalink) | |||||||||||||
|
Wannabe Photographer
![]() |
FSB was left at 300MHz, I didn't change it at all. Too much work for tomorrow so I might not be able to bench these until Tuesday.
__________________
|
|||||||||||||
|
|
|
|
#10 (permalink) | |||||||||||||
|
what the pho
![]() |
OP, does the mobo offer a selection of the same mem multiplier but using different straps? For example, on my mobo bios, I have a 4.0, 4.0+, and 4.0~ settings--each denominator corresponds to a different strap--800, 1033, and 1333, IIRC. Is there an option for that in your bios? That might be a better indicator of strap effects.
Also, for your RAM, were ALL of the RAM timings/latencies manually set? Sometimes they are hidden in the main memory timing/latencies menu, and some are set to auto, which might also be the cause of the discrepancy.
__________________
59 6f 75 20 61 72 65 20 61 20 6e 65 72 64 2c 20 61 6e 64 20 61 70 70 61 72 65 6e 74 6c 79 20 73 6f 20 61 6d 20 49
|
|||||||||||||
|
|
![]() |
| Currently Active Users Viewing This Thread: 1 (0 members and 1 guests) | |
| Thread Tools | |
|
|