Ddr2 @ 1-1-1-1 by kyosen

[sierra_bound]

Kyosen has been testing the Asus P5NSLI (nForce4 SLI Intel Edition rev A3). An interesting board to say the least. Can run Corsair 5400UL or 6400C3 at CL2 or even CL1.



VDimm mod





http://valid.x86-secret.com/show_oc.php?id=115776


http://valid.x86-secret.com/show_oc.php?id=115782

[ante_ante]

That´s awesome
Any 3dmark with that timings ?

[Gautam]

Even considering its kyosen, I have a lot of difficulty in believing that. It seems to me like a BIOS glitch of some sort.

[sierra_bound]

Don't know about 3D. I think he indicated maximum FSB is not that good on the board.

Even considering its kyosen, I have a lot of difficulty in believing that. It seems to me like a BIOS glitch of some sort.

Yes, that could be. Still, it's fun to see screenshots like that. He has sent in CPU-Z validations for the 5400UL's and 6400C3's.

[SoF]

kyosen really pushes ddr2 to the max...insane, really insane timings

[ZX7891]

to be honest, im pretty apprenhensive to believe its really running 1-1-1-1

[guess2098]

大著 keysen !! my hero hehehehe

but...... what is the voltage for that?

[sierra_bound]

He was using 2.0v for the 5400UL's and 2.87v for the 6400C3's.

Here is CL2. There wasn't much difference with CL3, so it may indeed be a BIOS bug.

[Vassili]

It also says Hyperthreading is enabled @ x6800.

Could you try to test 1-1-1-1 vs 3-3-3-3?

[guess2098]

looks like CL1.5 and CL2 bug on old NF4 chip

He was using 2.0v for the 5400UL's and 2.87v for the 6400C3's.

Here is CL2. There wasn't much difference with CL3, so it may indeed be a BIOS bug.

[Kunaak]

I did this a few weeks ago, with my Corsair 5400UL on a Gigabyte DS1.

I could never verify that it was running ras to cas 1, cause the differences were so hard to spot.

like there was only a 5 second improvement going from ras to cas 2 to 1, with super pi 32M.

I can do this over and over with this board... and I do think it's definatly running ras to cas 2, and when switching to ras to cas 1... it is doing something, cause the overclock definatly drops fast, when trying it...

however, since intel boards are limited to 2T, I hear, that while is may be ras to cas 1... it will always act like ras to cas 2....

http://www.xtremesystems.org/forums/...d.php?t=108116





unfortunatly, there appears to be no support for it on 975X boards, so I cant even try it on this DFI infinity... not even with memset.

[sierra_bound]

The P5NSLI is a fairly low-end board. Retails for around $115-120. But it does support C2D.

Anyway, maybe kyosen will stop by and post his thoughts in this thread.

[Revv23]

wow, cool, too bad it doesnt look like any performance gain, wonder if its really setting that way

[Vincentvega18]

Yeah, i'd like to see some comparison benches. Still looks cool though.

[R@imizA]

Wery nice... but i was thinking that performance diference will be bigger... Anyway grats

[Turrican]

now, that's waht i call tight timings
anyway congratz

[EZClock]

I'm a little skeptical on 1-1-1-1 w/ DDR2, agreeing with what Gautam said.

[sierra_bound]

I think it's worth pointing out that kyosen's intention was not to deceive. The Japanese are very methodical in their testing. He's simply reporting what he found. Take it for what you will.

Many people don't believe the results because they've been told CL2 and CL1 are not possible, and not because their actual testing proved it's not possible. I think it's interesting the BIOS would even have those options.

It's very possible this is a BIOS bug. Anyway, I'll be interested in kyosen's thoughts.

[zbogorgon]

i would like to see that superpi 8m at normal 3-3-3 or even 4-4-4, so that we could see if that 1-2-1-1, that looks so far to be the fastest "is" or it isn't.

[Tony]

Bios may have cas2 etc but DDR2 does not support it.

It took Oskar a year to actually get performance increases from the cas1.5 setting on the expert etc...this will just be a bios option that does nothing i fear.

[sierra_bound]

Kyosen has a lot of screenshots at his c-board. I didn't feel like posting all of them. Here's a comparison of 6400C3 at 3-3-3 and 1-2-1.

[ozzimark]

Bios may have cas2 etc but DDR2 does not support it.

indeed, but 3-1-1 or 3-0-0 should still theoretically be possible

thanks for sharing this sierra, i look forward to kyosen's comments on the matter

[Kunaak]

interesting...

the differences on his board are much more noticable, then on my 965 board at cas 3-1-1.

does the Nvidia board offer 1T at that setting? is he running 1T?

[sierra_bound]

I'm not sure if he was running 1T or 2T. The NForce Intel SLI chipset does allow 1T.

Now it could be that setting RAS# to CAS# Delay to 2 instead of 3 is mainly responsible for the faster time. That's interesting in itself because I didn't know 6400C3's could do that. He also said setting it to 1 was a no-go. But X-1-X-X was possible with the 5400UL's.

Technology sometimes makes what was once considered impossible possible. For many years, people thought the multiplier lock on Intel CPU's was unbreakable. Then along came SpeedStep and suddenly everything changed.

[Eldonko]

Interesting. I dont know if it is really 1-1-1 because that would not be supported but something is obviously making the time faster.