Strap Question (1066 vs. 1333)

I was reading on about the Bad Axe 2 on extremesystems.org. People were complaining about the recent bios release how they couldnt get over 400FSB on the 1066 strap like they could on a previous version.

The consensus there from folks smarter than me was to use the 1333 strap for FSB frequencies over 400FSB. Instead of registering there I'll ask the question here:

Don't alot of other boards automatically switch you to the 1333 strap when you go over a certain FSB? I could of sworn I read that somewhere, maybe anandtech about the DS3 beta bios or something. Just curious how the Bad Axe 2 is any different than other boards in regards to the strap.

Ha I found it. It was about the DS3 beta Bios. Apparently the DS3 automatically switches to 1333 at 401FSB and the Bad Axe 2 does not. So if you have to manually do so in the Bad Axe 2 that doesnt make it less of an overclocker than the DS3. For either board you are going with the 1333 strap over 400FSB so I'm not getting the critique of the Bad Axe 2???

I can't believe none of the resident experts have anything to say on this. Anybody???


While increasing our overclock capability by 30FSB is substantial we will have to see in our benchmarks if there is an actual performance difference. The switch of the memory controller strap from 1067 to 1333 at the 401 FSB setting creates a memory performance penalty due to the relaxed MCH timings. On the average P965 motherboard and BIOS setup, the MCH is actually being overclocked from the 365 FSB to 400 FSB level.

http://www.anandtech.com/mb/showdoc.aspx?i=2854&p=2

OK, come on. With all the peep claiming they have over 400FSB nobody has an opinion???

ToddMcF2002 said:

Ha I found it. It was about the DS3 beta Bios. Apparently the DS3 automatically switches to 1333 at 401FSB and the Bad Axe 2 does not.

Click to expand...

not just the DS3 but pretty much all 965 mobos whereas only a couple of 975s can actually use the 1333 strap & afaik it's not automatic.
Thing is that a 975 running the 1067 strap & doing e.g. 410 fsb will outperform a 965 using the 1333 @ 410 fsb.

All I can tell you so far, is I couldn't boot over 400 with even the 1333 strap using the 2395 BIOS. Im back to 2333, but haven't tried 1066 strap yet.

Ok after a bit more tinkering, I think I have the 2395 bios working.

My old settings on the 1333 strap were 4:5 on 425 FSB so ram was running about 1066 at 4-3-3-7

Going to the 1066 strap, I could no longer run 4:5 regardless of my timings, so Im now at 425 1:1 so 850 5-4-4-10 but believe it or not, its benching noticeably faster so far.

So to answer your question *as I understand it* , yes the DS3 and AFAIK all the other boards automatically switch the strap depending on your FSB. The BX2 allows you manipulate it via the "reference frequency" setting.


And here: some more info on FSB straps and how they are manipulated ( still waiting for clarification on my statement above in this thread)
http://www.xtremesystems.org/forums/showthread.php?t=128913

Edit:: Some data on the performance difference (I am much to lazy to do anything like this )
http://www.xtremesystems.org/forums/showpost.php?p=1926313&postcount=679

This is great info Yreka. Thanks for posting your results.

Just for clarification - are you saying that you are back on the 1066 strap at 425 running 1:1? So the faster strap at 1:1 beats the slower strap at faster 4:5 RAM?

Since this is a 975 chip vs. a 965 chip - I'm the big OC difference is in fact the automatic switch to 1333 on the 965? (DS3)

I'd really like clarification too!

Yreka said:

Going to the 1066 strap, I could no longer run 4:5 regardless of my timings, so Im now at 425 1:1 so 850 5-4-4-10 but believe it or not, its benching noticeably faster so far.

Click to expand...

I believe it. the FSB already gets saturated with memory (& system) bandwidth at 1:1, going to a faster memory ratio doesn't usually help.
The 1:1 ratio will perform better most of the time. Theres a slight advantage of being at 1:1, plus you can usually run at tighter timings.

I read that Core 2 Duo don't see performance benefits from memory timing tighter than 6-6-6-6-18 (compared to AMD procs running off the same RAM frequency but tighter timings, the Intel Core 2 Duo still kept up.) I cannot find the article but it was discussed on these boards.

So wouldn't this mean that using the 1333 strap (which means looser timings but higher FSB) would definitely increase performance?

Not sure if you guys have heard of that timing thing. If anyone knows a link to the article that would be great.

I'm a total noob so I also think I am completely wrong in thinking this has anything relation to memory latency. That link you all posted talks about latency in the NB not sure how this would relate to memory latency and what I mentioned above. Someone help

ToddMcF2002 said:

This is great info Yreka. Thanks for posting your results.

Just for clarification - are you saying that you are back on the 1066 strap at 425 running 1:1? So the faster strap at 1:1 beats the slower strap at faster 4:5 RAM?

I'd really like clarification too!

Click to expand...

Yep, looks like thats the case. I did some quick benches tonight before I went back to the rev ..33 bios to play around a bit more.

I can't answer definitively on the 965 board, I dont have too much time with them. The only one I have is an ECS, Fry's electronics special. Not a whole lotta options to play with there.. According to Blauhung on XS, all intel chipets have "straps", the only ones with manual control over them are the infinity abit and BX2 975 boards. As for the rest of the 965/975, and I believe even the Nvidia boards, the only way to manipulate the "strap" is as a function of the FSB.

In both data sets, I am running 423FSB X 9 (3.8G) and Memory at 1:1 5-4-4-10. The only difference between the data sets was changing the reference frequency from 333(1333strap) to 266(1066 strap).

First Data Set (1333 Strap)

Super PI - 13.485

Sandra Memory Bench
Int buffered - 7052 (bandwidth)
Float buff - 7064 (bandwidth)
cmb-index - 18,810mb/sec (latency)

Sciencemark--






Second Data Set (1066 Strap)

Super PI - 13.312

Sandra Memory Bench
Int buffered - 8252(bandwidth)
Float buff - 8270 (bandwidth)
cmb-index - 20,445mb/sec (latency)

Sciencemark--

Nice stuff.

No question now I'm getting the BadAxe 2. Thanks for the test data!