P5B deluxe FSB 401-460?

There are new BIOSes for P5B series and Commando from Asus.
The new BIOS versions are alowing native support of FSB-1333.
http://www.vr-zone.com/?i=4547

um, ok so the new BIOS will allow for FSB1333, but what does this mean for FSB set between 401-460? apparently 400FSB will provide 500mb/s MORE bandwidth than 401...

I have no idea, someone who has P5B will find out after testing it with the new BIOS

Correct me if I'm wrong but Dual channel only works up to DDR2-800 on that board.

The problem has to do with chipset straps. The straps are 266-333-400-467and 533 the bandwidth ramps up and down in between these settings and is most noticeable in the 401 to 460ish range.
The decrease in memory bandwidth affects overall performance
The 1333 bios upgrade is for CPU support.
The chipset strap is already in place.

this board will do well over 1200Mhz :wink:

so what would you suggest? only try to hit 266, 333, 400, 467 or 533 on the FSB and nothing else?

If you can hit 430 to 450 fsb the drop is only memory bandwidth and you can compensate with good ddr2 800 cas4 ram.
The CPU performance still climbs respectfully.
The best performance of course is right at the straps.
Depending on your CPU find the combination that is best suited for the multiplier.

In answer to your question 400 is the best bandwidth and anything higher you need a little more skill and a very good cooling solution.
When you start hitting 500 FSB your Voltages NB, Memory, Vcore all need to increase to stabilize the system and this is not as easy as it sounds.
Very achievable but to maintain this OC you need to open your wallet  
My suggestion is as long as you’re stable and within the thermal limits hit whatever your overclock permits :wink:

hmm i see

well ive been trying out multipliers and finding the maximum overclock and this is what i can hit...

394 x9 ... 3.54ghz

435 x8 ... 3.52ghz

465 x7 ... 3.25ghz

theyre the absolutely highest i can go, all running 1:1 ratio, on this setup:

c2d e6600
p5b deluxe
2gb corsair xms2 pc2-6400C4
tuniq tower 120
corsair 620W

which of those combinations do you think gives me best overall performance?

In my experience no matter the speed, the higher the Multiplyer (x) the better atleast in synthetic benchmarks (thats why we pay more for 6600's compared to a 6300), also you should be able to get much more out of your OC.

I also have a p5b and e6600 but with kingston value ram ddr2-667, and i have had it upwards of 450 x 9

Keep the 9X multiplier and stay within 333 or the 400 FSB straps for 1:1
When you lower the multiplier the additional increase in the FSB will not give you an increase in performance.
At the base frequency with the stock multiplier (lets use your 9X) the chipset is at its optimum latency.
When you drop the multiplier and raise the FSB the higher frequency forces you to raise the Northbridge volts to compensate. You actually loose ground
Some people say the latency changes (loosens) and they may be right.
The 9 X multiplier gives you a higher frequency base that requires less increase in fsb and in your case a better chance of maintaining 3.6GHz

but doesnt increasing the fsb give me better memory performance at 1:1 as youre increase the memory frequency as well that way?



wow, what voltages are you using? at a 9 multiplier my computer wont even run stable at 400fsb 8O

Yes
8)

1.55 vCore on a zalman 9700
30c @idle
67c @max TAT test

lol so i dont get it, how does running a 7 multiplier with a 465 frequency not improve performance? because my memory is running faster that way

or are you saying the overall clock speed is more important than having a high clock and memory frequency?

Youre overthinking  
If your at ddr2 800 at 3.6 Ghz you can loosen the latency and increase the memory to DDR2 1000.
Theres your boost.
at the lower multiplier your chipset is doing all the work or front side bus if you may.
Let the cpu strut its stuff with the 9X multiplier and finish off the overclock with an increase in ram frequncy or lower timings

this board will do well over 1200Mhz :wink:

In dual channel mode? man did I have some wrong info :?

I think I'm in love with the P5B Deluxe by the way... :? :twisted:

Give me a minute
Ill get some screen shots

Sorry it took so long
lost a hard drive  
3.6 Ghz 400 fsb E6600 X9 multiplier
DDR 1000 speeds


DDR 800 speeds 3.6 Ghz 400 fsb E6600 X9 multiplier




3960 MHz ddr 1100 440fsb E6600 X 9 multiplier



DDR 1200

I remember when I was OCing a CPU a while ago, I noticed that certain lower FSB OCs, would cause the PC to crash, where'as a slightly higher one would run perfectly  

Someone gave a reason for this in a post.

I forget what it was now  

Hmm, so according the first and second pictures running asynchronously at DDR2-1000 speed does yield a boost (over DDR2-800) in performance even if a small one. And it definitely runs in dual channel over DDR2-800. Very good to know, thanks for the shots 8)

I’ve run a few tests on your theory that the higher FSB with the lower multiplier could give you better performance.
As you can see below this is just not the case and with an inferior CPU with less level 2 cache and an even lower multiplier the performance can drop is even more noticeably.
You might say the increase in performance at 9X is minimal in comparison but over longer periods of time in number crunching or DVD encoding the higher multiplier would dominate the lower multiplier.
And that’s what it’s all about.


E6600 460FSB 8X multiplier.
The memory bandwidth has dropped over the 9X multiplier settings

CPU performance has little or no gain at 3600 MHz over stock 9 X multiplier

E6600 at 400 FSB 9x multiplier has the higher memory bandwidth and better CPU performance than the 8 X multiplier


470 FSB 8X multiplier 3760 MHz


418 FSB 9X multiplier 3760 MHz

With the 9X multiplier the memory bandwidth has dropped slightly because of the strap change but CPU performance gains slightly.

So in conclusion there is little benefit to increased FSB speeds.
You can even loose ground trying to hurtle the walls and stay within the chipset straps.
Even though 3760 is possible on air these tests were done with water cooling and in no way should be attempted without proper cooling, and Testing for stability staying within reasonable thermal guidelines.
There I added my disclaimer 8)

Excellent info Kwalker. As I'll be overclocking an E6600 as soon as I return from this deployment I find the more settings and info posted a wealth of information for me.

Thanks!

thanks for that, looks like ill stick with 9x multiplier then

i cant quite hit 400FSB with it though, how far away from 400 should I be before I start to notice big drops in memory bandwidth?

Your fine until 401 then its uphill again until around 467

oh sorry i meant the other side of 400...is 395 ok?

395 is fine
You will have very good bandwidth.
the strap change's again at 401

ah ok

i apologise for asking so many questions, but you seem to know what youre talking about   theres just one more thing that i need help with understanding really if you dont mind

there are quite a few settings in the bios for voltages. the manual explains what the all do (kinda) but i dont understand how they affect overclocking

so things like...
cpu vcore
mem voltage
fsb termination
nb vcore
sb vcore
ich chipset

this is what i understand so far...
cpu vcore - the higher it is the higher you can set your overall clock speed
mem voltage - higher gives a higher possible memory frequency
fsb termination - no idea
nb vcore - voltage going to the north bridge? not sure how it affects OC though
sb vcore - no idea how this helps get a higher overclock either
ich chipset - same as above

so basically my overclocking atm is just changing the cpu vcore around to try to make it stable. im sure if i understood the other voltages then my overclocking would work better

FSB termination
It may be used for the memory controller FSB supply or host
This term I am most familiar with in the CPU function when you exceed thermal limits
Two methods used are reduction in multiplier and or Vcore.
If the processor reaches anywhere from 75 to 85 C the Vcore is reduced or pulsed at a lower voltage (duty cycle) until the CPU recovers to normal thermals.
This value would increase the termination voltage used in modulating. (Throttling)

In the FSB circuitry Termination voltage is used to increase current flow from the CPU to the MCH (memory controller hub) At 266 FSB Intel limits the voltage to 1.29 volts
But allows a range of .83 to 1.65 standard.
Increasing the Vtt helps with higher FSB but can increase CPU temps.


NB Vcore
Memory controller voltage.
When you reach limits or hit a wall you may need to increase this to stabilize or boot past the stalled FSB.
This also helps to increase bandwidth between the processor and memory.
Same principle that you would apply to increase ram voltage to reduce latency (tighten timings) and increase overall frequency.

SB Vcore
PCI bus / PCE EX bus
This helps with increased frequency in the FSB. You should increase this value if you set PCI-EX over 100MHz also helps for graphics and other pci bus devices

ICH chipset
SATA PCIX other I/O devices
Again increase when you have issues with higher FSB such as data loss or drive disconnects ….communication errors

ohhh ok i see now

doesnt look like i can do much else in terms of overclocking then. i cant increase the nb voltage without reducing the pci-e slot to 1x speed. and if i increase the termination voltage anymore then the computer might melt

i guess im trying to push the system because its not the best overclock possible. im currently running p95 to see if things are stable, but right now ive managed to hit 390 cpu frequency with a multiplier of 9...which isnt the best. however it gets really hot. idles at about 36C, but with full load ive seen it reach 70C 8O

my fans are already working their hardest, so what else could i do to bring the temp down?i dont want to lower the FSB anymore because its not a great overclock as it is, maybe there are other things that i could turn down to the cpu isnt working so hard?