FSB Strap for noobs.

Here it goes.

Your north bridge has an internal clock speed and latencies just like your CPU and memory. The FSB of your north bridge can be found by dividing your original CPU multiplier by your set CPU multiplier and then multiplying by your FSB.

So if you are running a E6600 (266 * 9) at 400Mhz x 8 your NB FSB is:

(9 / 8) x 400 = 450Mhz FSB (1800Mhz Total)

Just like your memory may be able to run at 4-4-4-12 at 1000Mhz but needs to run at 5-5-5-15 at 1200Mhz, your north bridge has a series of latencies which it must adjust in order to maintain stability at its FSB. These latencies seem to play a far more significant role in system performance than memory latencies.

Intel has predefined specific latencies at specific NB FSB speeds. They are referred to as straps. There is a strap for when the NB FSB is 1066Mhz and under, 1333Mhz FSB and under, 1600Mhz FSB and under, ect. When you go from the 1066Mhz FSB strap to the 1333Mhz FSB strap, the north bridge's internal latencies loosen to allow for greater stability.

ASUS has redefined the NB strap so that the 1333Mhz FSB strap does not come into effect until 401Mhz FSB (1604Mhz). Other perimeters of straps are someone unknown.

Tony, the guy who pretty much figured all this out and he works for OCZ. He is now a lobbyist trying to get ASUS and other major motherboard manufacturers to give the end user the control of when straps start.

There are 2 ways to beat the NB strap:

1. Boot to windows in the 1066Mhz strap and then use Clockgen to increase your CPU's FSB. You can then get to a much higher FSB while maintaining the 1066Mhz strap simply because the BIOS does not adjust the north bridge's latencies in real time.
2. Get a X6800 or QX6700 (or even a ES chip). To the north bridge, you are always at a default multiplier with a Extreme Edition processor. This allows you to set a much lower or higher multiplier without the NB FSB being effected.

Answered my own question :)

Quote:

---

Originally Posted by pauldovi

Answered my own question :)

---

TBH Im not sure about NB strap. Any info on 965 would be of interest

thank

dave

http://www.thetechrepository.com/showthread.php?t=30

Here it goes.

Your north bridge has an internal clock speed and latencies just like your CPU and memory. The FSB of your north bridge can be found by dividing your original CPU multiplier by your set CPU multiplier and then multiplying by your FSB.

So if you are running a E6600 (266 * 9) at 400Mhz x 8 your NB FSB is:

(9 / 8) x 400 = 450Mhz FSB (1800Mhz Total)

Just like your memory may be able to run at 4-4-4-12 at 1000Mhz but needs to run at 5-5-5-15 at 1200Mhz, your north bridge has a series of latencies which it must adjust in order to maintain stability at its FSB. These latencies seem to play a far more significant role in system performance than memory latencies.

Intel has predefined specific latencies at specific NB FSB speeds. They are referred to as straps. There is a strap for when the NB FSB is 1066Mhz and under, 1333Mhz FSB and under, 1600Mhz FSB and under, ect. When you go from the 1066Mhz FSB strap to the 1333Mhz FSB strap, the north bridge's internal latencies loosen to allow for greater stability.

ASUS has redefined the NB strap so that the 1333Mhz FSB strap does not come into effect until 401Mhz FSB (1604Mhz). Other perimeters of straps are someone unknown.

Tony, the guy who pretty much figured all this out and he works for OCZ. He is now a lobbyist trying to get ASUS and other major motherboard manufacturers to give the end user the control of when straps start.

There are 2 ways to beat the NB strap:

1. Boot to windows in the 1066Mhz strap and then use Clockgen to increase your CPU's FSB. You can then get to a much higher FSB while maintaining the 1066Mhz strap simply because the BIOS does not adjust the north bridge's latencies in real time.
2. Get a X6800 or QX6700 (or even a ES chip). To the north bridge, you are always at a default multiplier with a Extreme Edition processor. This allows you to set a much lower or higher multiplier without the NB FSB being effected.

Quote:

---

Originally Posted by pauldovi

Here it goes.

Your north bridge has an internal clock speed and latencies just like your CPU and memory. The FSB of your north bridge can be found by dividing your original CPU multiplier by your set CPU multiplier and then multiplying by your FSB.

So if you are running a E6600 (266 * 9) at 400Mhz x 8 your NB FSB is:

(9 / 8) x 400 = 450Mhz FSB (1800Mhz Total)

Just like your memory may be able to run at 4-4-4-12 at 1000Mhz but needs to run at 5-5-5-15 at 1200Mhz, your north bridge has a series of latencies which it must adjust in order to maintain stability at its FSB. These latencies seem to play a far more significant role in system performance than memory latencies.

Intel has predefined specific latencies at specific NB FSB speeds. They are referred to as straps. There is a strap for when the NB FSB is 1066Mhz and under, 1333Mhz FSB and under, 1600Mhz FSB and under, ect. When you go from the 1066Mhz FSB strap to the 1333Mhz FSB strap, the north bridge's internal latencies loosen to allow for greater stability.

ASUS has redefined the NB strap so that the 1333Mhz FSB strap does not come into effect until 401Mhz FSB (1604Mhz). Other perimeters of straps are someone unknown.

Tony, the guy who pretty much figured all this out has since been hired by OCZ (they were impressed). He is now a lobbyist trying to get ASUS and other major motherboard manufacturers to give the end user the control of when straps start.

There are 2 ways to beat the NB strap:

1. Boot to windows in the 1066Mhz strap and then use Clockgen to increase your CPU's FSB. You can then get to a much higher FSB while maintaining the 1066Mhz strap simply because the BIOS does not adjust the north bridge's latencies in real time.
2. Get a X6800 or QX6700 (or even a ES chip). To the north bridge, you are always at a default multiplier with a Extreme Edition processor. This allows you to set a much lower or higher multiplier without the NB FSB being effected.

---

Nicely stated.

*EDIT*

pauldovi fixed it ;)

Quote:

---

Originally Posted by alexio

Very nice explanation, but Tony has been an OCZ employee for a little longer than that ;)

That's the only tiny remark on this very nice write-up :)

---

Opps. I wrote this up for another forum where the majority of people would not know who Tony is.

I moved my explanation to the top for anyone to read, and I fixed the mistake!

Nice job pauldovi. Easy to understand.

Quote:

---

Originally Posted by pauldovi

ASUS has redefined the NB strap so that the 1333Mhz FSB strap does not come into effect until 401Mhz FSB (1604Mhz).

---

Which ASUS motherboards allow FSB up to 400 before switching to the 1333 strap?

This is another way to "beat the strap" for most people who can't or don't need to go above 400FSB.

Quote:

---

Originally Posted by pauldovi

Here it goes.
...
There are 2 ways to beat the NB strap:

1. Boot to windows in the 1066Mhz strap and then use Clockgen to increase your CPU's FSB. You can then get to a much higher FSB while maintaining the 1066Mhz strap simply because the BIOS does not adjust the north bridge's latencies in real time.
2. Get a X6800 or QX6700 (or even a ES chip). To the north bridge, you are always at a default multiplier with a Extreme Edition processor. This allows you to set a much lower or higher multiplier without the NB FSB being effected.

---

Good write up thanks,

One question in regards to "beating the NB strap". Are you in effect "controlling the Strap" on the Intel BX2 when you adjust the "reference frequency" setting ?

I always assumed that was the case, but Im not sure.

Quote:

---

Originally Posted by Yreka

Good write up thanks,

One question in regards to "beating the NB strap". Are you in effect "controlling the Strap" on the Intel BX2 when you adjust the "reference frequency" setting ?

I always assumed that was the case, but Im not sure.

---

I believe so, although I am not 100% certain.

Can we get the actual outcomes in terms of performance, stability, etc., of what you just explained?

Quote:

---

Originally Posted by darcon

Can we get the actual outcomes in terms of performance, stability, etc., of what you just explained?

---

http://www.xtremesystems.org/forums/...d.php?t=114998

Quote:

---

Originally Posted by pauldovi

ASUS has redefined the NB strap so that the 1333Mhz FSB strap does not come into effect until 401Mhz FSB (1604Mhz).

---

I ask again... what ASUS Mobo's do this?

Quote:

---

Originally Posted by Yreka

Good write up thanks,

One question in regards to "beating the NB strap". Are you in effect "controlling the Strap" on the Intel BX2 when you adjust the "reference frequency" setting ?

I always assumed that was the case, but Im not sure.

---

yes, this is direct control over the strap.

here's some work i was doing for the other BX2 thread
http://www.xtremesystems.org/forums/...1&d=1167934741

Quote:

---

Originally Posted by virtualrain

I ask again... what ASUS Mobo's do this?

---

all the asus non integrated video 965 boards do this to some extent. depends on the BIOS revision as to where they draw the line. Basically any board with an intel chipset save the current big 3 in 975 (infinity abit and BX2 that allow strap control) have to manipulate the strap based on the current FSB speed. You can test the basics out for yourself by plotting memory latency as a function of FSB speed, or just open memset at varrious FSB speeds and compair the extended timings.

Quote:

---

Originally Posted by Blauhung

yes, this is direct control over the strap.

here's some work i was doing for the other BX2 thread
http://www.xtremesystems.org/forums/...1&d=1167934741



all the asus non integrated video 965 boards do this to some extent. depends on the BIOS revision as to where they draw the line. Basically any board with an intel chipset save the current big 3 in 975 (infinity abit and BX2 that allow strap control) have to manipulate the strap based on the current FSB speed. You can test the basics out for yourself by plotting memory latency as a function of FSB speed, or just open memset at varrious FSB speeds and compair the extended timings.

---

Nice chart!:scope:
I think 2 things make it even better:
1. please add a fsb/memory ratio
2. please specify the software you used to measure latency

If running a x6800 isn't affected by the strap, then what other considerations to consider whe choosing for instance (on phase cooling) 400x11 or 440-450x10 to start with, assuming mem timings are set the same?

So has anyone tested the actual difference of leaving your bord @ 266 into windows, then using clockgen to ramp it way up? My Biostar TForce 965PT hits a wall @ 380 and is 100% prime stable, but crashes in 50 mins @ 381. I will have to try this when I get home but thought I would see if anyone had done it yet. Otherwise I will try it and post my results...

Quote:

---

Originally Posted by pauldovi

The FSB of your north bridge can be found by dividing your original CPU multiplier by your set CPU multiplier and then multiplying by your FSB.

So if you are running a E6600 (266 * 9) at 400Mhz x 8 your NB FSB is:

(9 / 8) x 400 = 450Mhz FSB (1800Mhz Total)

---

Sorry if I'm being a tard, but can someone please explain why this works? I'm sure there's some crucial information I don't have which I need in order for the math to make sense.

Quote:

---

Originally Posted by GPSeek

Nice chart!:scope:
I think 2 things make it even better:
1. please add a fsb/memory ratio
2. please specify the software you used to measure latency

---

whoops sorry

1. on the bad axe 2, memory frequecy is found by (FSB)*(mem freq)/(ref freq) = DDR2 speed
first 3 are 1:1 last 2 are 3:4 and 3:5

2. used sandra for benches memset to read default timings

Quote:

---

Originally Posted by xgman

If running a x6800 isn't affected by the strap, then what other considerations to consider whe choosing for instance (on phase cooling) 400x11 or 440-450x10 to start with, assuming mem timings are set the same?

---

basically, as your FSB goes up, you start to run into looser strap settings. While the theoretical bandwidth of data along the FSB goes up, you are sacrificing latency. I am still playing around so i don't have current #'s, but you will get slightly better performance if you were to max out your FSB on the lowest stable strap using ram dividers and CPU multipliers to achieve your goal for speed on each.

Quote:

---

Originally Posted by Blauhung

whoops sorry

1. on the bad axe 2, memory frequecy is found by (FSB)*(mem freq)/(ref freq) = DDR2 speed
first 3 are 1:1 last 2 are 3:4 and 3:5

2. used sandra for benches memset to read default timings

---

1. Not a real problem but a matter of convenience for a lazy guy like me:D
2. Let's say my aim is to have a lower latency without pushing fsb into higher MHz. It can be achieved by 2 ways:
A - tighter memory timing
B - higher memory frequecy (without running into lousy higher strap)

Usually you cannot do both A and B at the same time. Your table shows the effect of B clearly because A is almost the same. :toast:

However, in an actual overclock situation, it is going to be a comprimise between those factors. Not so easy to find out which factor is more important.

Another issue is that I think there are probablly better programs there for measuring ram latency than the bloat sandra package. :slapass:

yeah, uses sandra because everest is doing the voltage reset thing on the board. That and i still am pretty new to the overclocking thing. This is my first build and I'm mostly just experimenting. I have a huge spread sheet with a metric butt ton of settings that i've run at. If you knew of a better latency program i would be all ears :p:

but on the plus side this makes me feel a lot better about buying a X6800. Due to strap settings and looser timings at high FSB, the unlocked chip really comes into its area to shine. You can do both A and B at the same time. I think i just maxed out my CPU where i wanted to go as far as temp on air cooling at 3.6 (only about 60C load at 1.4v, i'm a pansy but i spent a lot on it) and i did all this on the 800 strap. My plan now is to work my way up slowly and max out on the 1066 strap up to the same CPU and Memory speeds and see how much better you can do with the lower strap.

Hehe, and I've also started bugging the hell out of intel tech support for someone i could talk to who works with the BIOS on their extreme series boards. I picked up all I know about this crap from reading lots and lots of stuff and I needs more!

Quote:

---

Originally Posted by Blauhung

yeah, uses sandra because everest is doing the voltage reset thing on the board. That and i still am pretty new to the overclocking thing. This is my first build and I'm mostly just experimenting. I have a huge spread sheet with a metric butt ton of settings that i've run at. If you knew of a better latency program i would be all ears :p:

but on the plus side this makes me feel a lot better about buying a X6800. Due to strap settings and looser timings at high FSB, the unlocked chip really comes into its area to shine. You can do both A and B at the same time. I think i just maxed out my CPU where i wanted to go as far as temp on air cooling at 3.6 (only about 60C load at 1.4v, i'm a pansy but i spent a lot on it) and i did all this on the 800 strap. My plan now is to work my way up slowly and max out on the 1066 strap up to the same CPU and Memory speeds and see how much better you can do with the lower strap.

Hehe, and I've also started bugging the hell out of intel tech support for someone i could talk to who works with the BIOS on their extreme series boards. I picked up all I know about this crap from reading lots and lots of stuff and I needs more!

---

How about Science Mark 2.0?
I saw anandtech used it as a benchmark:
http://www.anandtech.com/cpuchipsets...spx?i=2795&p=5
I've never used it before.

http://forums.techpowerup.com/showthread.php?t=14736

Looks good to me at least:toast:

hmm, that actually seems to hint that the bulk of the performance there is from the pre-fetching on the CPU. I think i'll try CPU-z, even though I don't use it all that often cause if makes intel desktop util's not work.

Please help me to understand/confirm couple of concepts:

In my mind there are 5 different settings need to be optimized:

i) chipset (NB) latencies (straps)
ii) chipset speed (800mhz, 950mhz, 1125mhz, etc)
iii) cpu speed (8x400, 9x400, etc)
iv) memory speed (bandwidth)
v) memory latencies (3-3-3, 4-4-4, etc.)

We know that chipset latencies (item i) are constant between intervals. For instance, chipset have tighter latencies up to 401 MHz than it is between 401-467 MHz. These intervals are called as straps. And these strap boundaries are different between boards. For P5b, upper boundary of 1066 strap is 400 MHz, as soon as you pass 400MHz (in bios) you are in 1333 strap. As far as I know, new RD600 will give us the ability to set our own strap boundaries – so that we can set the 1066 strap valid until 455 MHz (or whatever we want).

edit: {{{ I think terminology wise more correct version: 1066 strap latencies are valid between 1066-1600 mhz; 1333 strap latencies are valid between 1601-2000 mhz, etc. }}}


We assumed that as chipset speed (item ii) increases we will have a faster system overall. However, as soon as we move into next strap overall system speed actually deteriorates until a threshold fsb is achieved. That’s why 9x400 with tight chipset latencies is faster than 8x450 with loose chipset latencies. The question is where can we see the real performance increase in the next strap (what is that threshold value)? Is it at 450mhz or 500mhz, or better, can we measure it by percentages? Let me explain what I am trying to say, if 1066 strap’s boundary is 400 fsb, then what percentage we need to pass to see a real performance increase? If the answer is 15% then we will have a faster system by running 8x450 (using some type of async settings to set fsb = 920mhz - 15% increase then 800mhz) then 9x400 (fsb = 800mhz). By this way, we can use our magic formula in every board that doesn’t allow us to change the strap boundaries.

Then we have the cpu speed (item i). Can there be a practical guide which will tell us which one is better up to what point: higher cpu speed with loose chipset latencies or lower cpu speed with tight chipset latencies, like 8x475 (3.8ghz) or 9x400 (3.6ghz)? This is slightly different than above paragraph because above we were comparing 8x450 (3.6ghz) against 9x400 (3.6ghz).

Memory bandwidth and latencies (item iv and v): as we know, the faster the better - in benching or psychologically in daily tasks :) But really, in order to achieve the fastest (best balanced) rig which one should we more focus – memory bandwidth or memory latencies? Is it a good idea to run 5-5-5 at 1300mhz or 4-3-2 1000mhz (like old bh5 versus tcdd comparison)? Which is more important in LGA 775 platform > latencies or bandwidth?

I believe this is the fun part of overclocking > testing and trying to find best (fastest) combination. And probably it is unique for each rig… But I would like to hear comments from people that already found their combination :)

Thanks…

Quote:

---

Originally Posted by pauldovi

So if you are running a E6600 (266 * 9) at 400Mhz x 8 your NB FSB is:

(9 / 8) x 400 = 450Mhz FSB (1800Mhz Total)

---

Is it just me or is that calculation wrong? I have no clue where you got 450 from. At (400 x 8) your NB FSB should be 1600. I think you made the calculation more complex than it really is. The NB FSB has nothing to do with the CPU multiplier it is just the FSB x 4. (400 x 4 = 1600)

Quote:

---

Originally Posted by pauldovi

To the north bridge, you are always at a default multiplier with a Extreme Edition processor. This allows you to set a much lower or higher multiplier without the NB FSB being effected.

---

Noob question alert!

Does this mean that Intel is fu*king us up the as* with cheap chips? Or can you change the Strap manually if you have a good motherboard?

What happens if my motherboard does not have this strap settings?