::WORLD RECORD:: DDR2 @ 1GHz

markodude · Post by **markodude** » Mon Apr 18, 2005 10:36 am

I cannot overclock the FSB at all with this setting, and at stock 200FSB on the processor the memory bandwidth is under 5000Mb/s in Sandra.
I moved back to DDR2-720 using the 3:5 divider and 216FSB, gives me 3.67ghz on the processor. Memory bandwidth is 5313Mb/s, not bad for an 800FSB proc.
I am going to try PCMARK and 3DMark to see what it is like. Ideally I need a better processor in there

markodude · Post by **markodude** » Mon Apr 18, 2005 11:08 am

Sorry to be a post whore, but I am working on this as we speak.....Intel have the new BIOS and drivers for this board on their website - do you think it is worth flashing the production bios on this board or may we lose some of the features ?????
http://intel.com/design/motherbd/bk/bk_drive.htm

EDIT : BTW Intel Desktop Control Center 2.0 is out as well

Post by **Apoptosis** » Mon Apr 18, 2005 12:06 pm

give it a try and let me know if anything is better on the new BIOS and all!

Today I am playing with some DDR1 memory on the DFI SLI-D board. Hitting DDR600 at 2-2-2, so so far so good. The thread on that is here.

markodude · Post by **markodude** » Tue Apr 19, 2005 5:59 am

Flashed new BIOS and you can now monitor temps with Intel Active monitor. You can also adjust FSB in windows with the new IDCC

I was thinking I bet it is not that hard to get an 800fsb proc to be recognized as a 1066FSB one, just some conductive paint on the BSEL pads. I wonder if that would that help overclocking the 840XE?

Marc · Post by **Marc** » Thu Apr 21, 2005 6:10 pm

http://valid.x86-secret.com/show_oc.php?id=13543

Owned ;)

But i think in both case it's a bug from CPUZ when dealing with FSB1066 CPU and FSB:DRAM i955X ratio detection ...

infinitevalence · Post by **infinitevalence** » Thu Apr 21, 2005 10:36 pm

Nice work man, its only a few MHZ but good going

edit/ i have to say its really cool that the top two memory scores in the world came from this forum right here.

Post by **Apoptosis** » Fri Apr 22, 2005 9:20 am

hum, wonder what board he did that on!

Doesn't matter cause my setup has a bunch more overhead... here is a teaser for him and everyone else!

A NEW world Record!

http://valid.x86-secret.com/show_oc.php?id=13657

I also tightened up my timings and am also running 5-4-4 timings

- Nate

Marc · Post by **Marc** » Sat Apr 23, 2005 5:20 am

Apoptosis wrote:hum, wonder what board he did that on!

Doesn't matter cause my setup has a bunch more overhead... here is a teaser for him and everyone else!

A NEW world Record!

http://valid.x86-secret.com/show_oc.php?id=13657

I also tightened up my timings and am also running 5-4-4 timings

- Nate

You still use the CPUZ bug ... your real ratio FSB:DRAM is not 1:2 but 2:3 (the one for DDR2 at 400 MHz with FSB at 266 MHz), it's a bug from CPUZ with 955X, confirmed by Franck !

So your DRAM speed is ... 460 MHz (DDR2-920)

Post by **Apoptosis** » Sat Apr 23, 2005 11:15 am

hum...i shot intel an e-mail, but Intel saw my results and said "good work", I'm suprised their engineers didn't tell me the numbers were off.

markodude · Post by **markodude** » Tue Apr 26, 2005 9:09 am

Why do we think this is a bug? I cant understand the issue, the DDR2-667 and DDR2-800 dividers work OK on an 800FSB proc. What is the problem?

Post by **Apoptosis** » Thu Apr 28, 2005 12:09 pm

Intel has confirmed that the memory ratios are correct on the Intel D955XBK motherboard.

The ratios should be as follows:

FSB1066:
4:3 DDR2-400
1:1 DDR2-533
4:5 DDR2-667
2:3 DDR2-800

FSB800:
1:1 DDR2-400
3:4 DDR2-533
3:5 DDR2-667
1:2 DDR2-800

CPU-Z showed that the DDR2-800 option on a FSB1066 processor was 1:2, but it was really 2:3. This means that we were running only 882MHz and not 1176MHz. This is a huge difference in numbers. When dealing with new hardware and not having all the data to plug into software programs errors like this can happen. CPU-Z has fixed the big in versions 1.28.6 and higher. I also believe the only on 1.28.6 and higher will i955X be VERIFIED by CPU-Z as authentic.

So, after clearing up the multiplier issues the highest we have been able to hit is 1GHz memory clock, which still hold the record for highest CPU-Z validated memory.

I'd like to thank Intel's PR Department, Intel's BIOS Development Team, and Franck over at CPU-Z for helping determine the dividers and know 100% sure what was going on. It took a number of weeks and many people to come together to get the right data out there for everyone to use. Now Franck finally has the data he needs to make CPU-Z current!

infinitevalence · Post by **infinitevalence** » Thu Apr 28, 2005 1:26 pm

thats still no less than amazing, if DDR2 and DDR3 keep scaling then we could be entering a era where memory is just as fast as the processor and then speeds really will run 1:1. I can just imagen the performance that would give, it would be like having 2gb of L2 cache.

markodude · Post by **markodude** » Fri Apr 29, 2005 5:19 am

Cool, well done for finding that out Apoptosis

Seems my 800FSB processor is accurate on the memory front.
On another related subject, I have tried 2 different 3.46EE's in this board and neither will POST, and Intel site says it is supported......my theory is that the 955 boards may not have support for older 130nm processors (this is the only LGA775 130nm processor with the higher 1.5vCore) - I am going to test the 3.46 in a D925XECV2 this weekend to confirm if it is the board or the CPU.

infinitevalence · Post by **infinitevalence** » Fri Apr 29, 2005 11:30 am

i want to know where you get 2 other 1066fsb EE chips from, do you have your own shop or something???

deadly-app · Post by **deadly-app** » Mon May 02, 2005 10:50 am

Corsair has specifically mentioned legitreviews in their newsletter, "Corsair Currents"

CM2X 512A-5400UL
Sets the World Record!
LegitReviews Uses It to Break the 1GHz world record!

"Using a 30% overclock on our Intel 840 dual core processor and 5-5-5 timings I was able to hit 1228MHz on the memory. This would of course beat our last world record of 1000.6GHz. At 1GHz the system is stable enough to be benchmarked."

freakgeek · Post by **freakgeek** » Sat May 21, 2005 10:54 pm

I do not see any difference in the 5400 & the 5400UL DIMMs 'cause I have my 4x 'CM2X1024-5400' modules running at 800MHz on the Intel 955XBK ref board at 4-3-3-4 timings (2.08V)! The SPD on the DIMMs is set at 4-5-5-15.
Also, I do not understand why Corsair refuses to publish this part number on the products page but is willing to supply (Notice the 1024. These are 4 modules of 1GB EACH.)

Apoptosis wrote:It seems a bunch of other sites are linking directly to this forum post. The full article can be seen here with all the benchmarks used posted.

I also ran everest and got scores of 5800MB/Sec on average. I'd run more testing, but with retail boards coming in I'm going to hold off as 90% of our readers buy boards by Tier 1 and Tier 2 motherboard makers. (ASUS, MSI, ABIT, EPOX, and so on)

I'm able to post 5-4-4 timings at 1GHz, but the system gets the BSOD when I run any applications. I can hit 4-4-4 well above 800MHz, but just not hit the limit. I'm working on the C19 article tonight.

-mogwai · Post by **-mogwai** » Sun May 22, 2005 1:24 am

why does it keep saying "this id is invalid" whenever i click on the link?

Post by **Apoptosis** » Sun May 22, 2005 11:56 am

because Franck removed some CPU-Z scores on the 955X chipsets till he could modify his software... Our results were valid

-mogwai · Post by **-mogwai** » Sun May 22, 2005 12:38 pm

aaah ok, cool

zolkorn · Post by **zolkorn** » Tue May 31, 2005 11:16 am

With A-DATA ram buy in thailand

http://www.overclockzone.com/zolkorn/05 ... ndex3.html

Legit Reviews

::WORLD RECORD:: DDR2 @ 1GHz

Corsair 5400 & 5400UL