new world record 1174mhz DDR2

[cupra]

nice clocks but like "jamo" say the bandwidth is to low in the ddr2 and intel...

[zakelwe]

Call me an old fuddy duddy but to be honest

I HAVE GOT NO FECKING CLUE

what 587Mhz DDR2 RAM at 5-5-5-7 on a P4 actually equates to in real life.


I'm afraid my real life equates to new stepping 90nm AMD and new on the market bh-5 with tight latencies I'm afraid. As does most people I would guess right now, at least in the overclocking community.


Regards

Andy

[Tony]

The best way of running ddr2 is 1:1 so you copy ddr1, I have seen just shy of ddr800 with 4-2-2-5 timings which gave some real sick memory writes, the reads are not as impressive if the ram clock speed is lower.

I have a feeling the 5400EB will do this also but im a 4-2-2 fanboy over 5-5-5 anyday.

[saaya]

Call me an old fuddy duddy but to be honest

I HAVE GOT NO FECKING CLUE

what 587Mhz DDR2 RAM at 5-5-5-7 on a P4 actually equates to in real life.


I'm afraid my real life equates to new stepping 90nm AMD and new on the market bh-5 with tight latencies I'm afraid. As does most people I would guess right now, at least in the overclocking community.


Regards

Andy

IF he could run that in 1:1 he would probably get 15gb/s+ buffered i think, its time intel unlocks all multis down to 5x on their extreme edition chips

[Qkjhfhaiguihfma]

IF he could run that in 1:1 he would probably get 15gb/s+ buffered i think, its time intel unlocks all multis down to 5x on their extreme edition chips

I dunno if they still have ES chips that are fully unlocked, but if someone has one, those two need to meet up

[Waxman]

Intel should do what amd did:

- multis unlocked downward on regular models;

- multis completely unlocked on the EE's

That way we could squeeze more performance from our cpu's

[brandinb]

yeah his bandwidth scores are limited by his fsb not his memory speed!
i wonder what the bandwidth will be 1:1 with a high clocked ee chip????
________
Wellbutrin class action

[[XC] moddolicous]

wow, thats pretty cool. It would really be something if u broke 1.2gig mark, that would be something.

[[XC] leviathan18]

congrats insane ram you have there.... hope to see something like that with ddr1

[Xassius]

unfortunately, unless intel REALLY wants to and make some real architectural changes to the Prescott/Smithfield...we can't see any multies lower than 14x

As I have constantly said for the last half a year, 3:4 was bad, now 1:2 is next to trash. Sure, you can run your memory at HIGH SPEEDS, but it can't translate to bandwidth because of the FSB bottleneck.

For example, at ONLY DDR2-640 with 3-2-2-1 timings, a configuration that ALL good MICRON memory is nearly "Guaranteed" to do [e.g. These Wintec modules are capable of DDR2-720 at 3-2-2, but at the inefficient 3:4 ratio - I cannot run 1:1 as CPU limited] I get

[cpulloverclock]

unfortunately, unless intel REALLY wants to and make some real architectural changes to the Prescott/Smithfield...we can't see any multies lower than 14x

As I have constantly said for the last half a year, 3:4 was bad, now 1:2 is next to trash. Sure, you can run your memory at HIGH SPEEDS, but it can't translate to bandwidth because of the FSB bottleneck.

For example, at ONLY DDR2-640 with 3-2-2-1 timings, a configuration that ALL good MICRON memory is nearly "Guaranteed" to do [e.g. These Wintec modules are capable of DDR2-720 at 3-2-2, but at the inefficient 3:4 ratio - I cannot run 1:1 as CPU limited] I get

which mb?

[saaya]

hmmm true, wasnt there something about the multis with the latest prescotts? are lower multis still possible at all? i havent seen a chip with a multi lower than 14 from intel for a year i think... hmmm multipliers are inside the cpu, right? could it be that intel chips are incapable of lower multis actually?

[zA.Gosu]

Whats the architectural reasons for high multi. Its a 31 stage pipe, isnt it? A64 is like 12. 5x multi is lowest, right? So is it like lowest multi has to be = or > than x/2-2, where x is the pipline stages.

[saaya]

no, afaik it has nothing to do with the pipeline stages directly

[Xassius]

which mb?

P5AD2-E

[Apoptosis]

Just wanted to give everyone the heads up that the 1.2Ghz memory mark has been reached and i lowered the timings down to 5-4-4 from 5-5-5. Everything else has been kept the same on the system. Still running 2.08V on the memory.



The memory is running 614Mhz or 1228MHz DDR2!

[Pomme]

that's my fsb, but tripled, lol. congratz, insane.

[madgamer]

nuts ram speed, and just wanted to mention that I had read that on the dual core EE intel chips they will be unlocked FX style on the multis.. so could be nice for monster 1:1 bandwidth.

[Apoptosis]

nuts ram speed, and just wanted to mention that I had read that on the dual core EE intel chips they will be unlocked FX style on the multis.. so could be nice for monster 1:1 bandwidth.

Yeah I have the Intel 840 here, and it is unlocked from 14-60.

[[XC] moddolicous]

wow, nice job. Can we see some benchmarks and superpi screenies?

[Apoptosis]

turns out something might be up with these numbers... some are saying it's not a true 1:2 divider (or 1228MHz), but that its a 2:3 divider (or 960MHz) and nothing detects it right.

There is a new version of CPU-Z 1.28.6 that is due out shortly that correctly detects the memory dividers. I tried out the new version and it does correctly detect the memory ratios and frequencies.

Here is the new CPU-Z and a 1066FSB Processor Validation at the DDR800 BIOS setting. Before it showed 1:2 and 1174MHz, now it shows 2:3 and 881MHz.

1GHz Memory Validation

[elec999]

Is it possible to get a intel cpu to run at 500mhz fsb. Seeing the ram at over 500mhz seems sweet. Does this mean ddram 2 is good.
Thanks

[The Stilt]

I think these are the highest speed achieved with tight latencies.

OCZ 4200EB Rev.2 (Micron Technology BT-37E "Texas" IC)
DDR800 3-2-2-1

Corsair 5400UL (Micron Technology BT-37E "Texas" IC)
DDR830 4-2-2-0 (Verified)

Corsair result is only single channel, because I´m currently using DFI 925XE-T2V board which happens to be quite a bit faster than P5AD2-E that I used with OCZ sticks (board can´t run as high in dual channel).

Even I was able to run OCZ sticks at DDR800 3-2-2-1 speeds in P5AD2-E, for some reason they can´t stable run over DDR720 with those settings on DFI

nForce4 IE & Glenwood boards will unleash the capabilities of these sticks, when memory ratio becomes adjustable

[Xassius]

I think these are the highest speed achieved with tight latencies.

OCZ 4200EB Rev.2 (Micron Technology BT-37E "Texas" IC)
DDR800 3-2-2-1

Corsair 5400UL (Micron Technology BT-37E "Texas" IC)
DDR830 4-2-2-0 (Verified)

Corsair result is only single channel, because I´m currently using DFI 925XE-T2V board which happens to be quite a bit faster than P5AD2-E that I used with OCZ sticks (board can´t run as high in dual channel).

Even I was able to run OCZ sticks at DDR800 3-2-2-1 speeds in P5AD2-E, for some reason they can´t stable run over DDR720 with those settings on DFI

nForce4 IE & Glenwood boards will unleash the capabilities of these sticks, when memory ratio becomes adjustable

Stilt - do you have any sandra/everest benchies on that DDR2-800 3-2-2-1 on the P5AD2-E?

[iddqd]

Thanks for all the nice comments...



Finally broke the 1.1GHz mark and almost the 1.2GHz memory mark! You know what the sad part is? This is 100% stable at 1174MHz! The bad news is that the Intel D955XDK reference board only allows the 3.73GHz EE to be overclocked 10% and I'm at 10% already. ONLY the new Smithfields are allowed to be overclocked 30% in the bios.

If anyone knows how to raise the FSB on the desktop let me know and I'm sure I'd be able to top 1200Mhz on this board. Heck, It's sitting here folding right now at 4.1Ghz and DDR2- 1.17GHz and can run superpi without errors...

It has more headroom!

I thought Intel finally decided to leave the EE multiplier unlocked?