Cell-SHOCKS ag. DDR3-2000 7-6-5-18 1T Prime_Pi Stable

[saaya]

yeah and in this case in the end the pig (ddr3) turns out to be "more equal" and not the race hors or fox like youd expect. Didnt have time to look at the other blue kits yet, too much to do, damn

[Lastviking]

I wonder how much vdimm they will take before they die , to bad the ddr3 sticks are not so cheap.

[saaya]

Lastviking! Hey man! how are you doing?

Well using more volts than the sticks like doesnt make sense, you only get worse results. The more tight the timings the higher the sweet spot, but even for 665 the highest sweet spots are 2.25v. Well im using a p5k3 so i cant use more than 2.25v...
So i guess with cas665 the max that makes sense for some sticks is probably 2.35v, for cas5 maybe up to 2.45v but i dont think cas5 is worth it and cas6 and 7 make most sense. so the max voltage that makes sense to use is 2.3v and i doubt it will die at that voltage, unless it runs hot of course...

You have a lot of experience with mem, have you ever experienced something like this before? I never had any mem that had a sweet spot for voltage and didnt like more. if it ran too hot it oced worse or gave errors, sure, but thats not the case here.

[Yoxxy]

I am glad that you have said these were 1800C6 ES sticks. I and I am sure others would have been disappointed to find out the retails did not have the great results of Victors (consistently that is).

Thanks for clearing things up. Also >2.15v seems to lose a lot of gains. I think there might be on-die regulation finally on Micron chips. Just something to look at.

Also 2.45v through a 78nm circuit is going to annihilate it very quickly. The max voltage you want to use is around 2.3v. I predict >4 Hours @ 2.45v is going to kill your sticks.

First one to hook a pot up on the board will be able to tell though.

[saaya]

Thanks Yoxxy

I hope i wasnt too late, Shaminos flight was on Tuesday so me and Fr3ak showed him around in Nuernberg and Frankfurt, was real fun
I arrived in the office on wednesday and was burried in work and emails and then posted on thursday as soon as i realized victor thought this was 1800 mem.

Im really sorry if anybody bought the 1800s thinking itll clock like this.
I did some quick testing and some of the 1800 kits actually get very close to this on a p5k3, so they might clock the same on a blitz.
but the average 1800s clock worse, especially with cas6 and 5.
Kam and booblemagnifique received random 1800s, so their results are more representing.
The blue kits are just a preview of our (hopefully soon, looks at colleague ) upcoming 665 mem.

[dinos22]

excellent clocks

surprised a bit with seemingly slow Pi times but i assume you didn't really give it a hard bench

cellshocks did well with DDR2, hopefully retail chips are like this...looks promising for sure

[saaya]

the retail 665 should clock like this

[dinos22]

the retail 665 should clock like this

nice
you work for cellshock??

[saaya]

Yupp
Very nice people, if you ever come to germany lmk and ill show you around
The fabs are really cool.

[Aerou]

Yupp

good to hear ... always been a CellSHOCK fan
and these 666 are nasty - CSH must have some godlike ICs on them
wow ... the CAS5 performance is outrageous!

I really am curious about the higher VDIMM survival of the DDR3 Microns

[dinos22]

Yupp
Very nice people, if you ever come to germany lmk and ill show you around
The fabs are really cool.

nice man gotta hook us up with some of that lovely ram :p j/k

hey what city is the fab in??


you really should have called those ram sticks 666

[stone_cold_Jimi]

The ones I just received are the 876 then. Still, I see elsewhere that they do:

"CellShock DDR3 @ 1928 with 7-6-5-16 (Micron D9GTR) and 2,2 V"

Which is worth playing with.

[indiecom]

@Saaya : If i buy from memoryC, can i find chip like this..??

Thanks

[STEvil]

Lastviking! Hey man! how are you doing?

Well using more volts than the sticks like doesnt make sense, you only get worse results. The more tight the timings the higher the sweet spot, but even for 665 the highest sweet spots are 2.25v. Well im using a p5k3 so i cant use more than 2.25v...
So i guess with cas665 the max that makes sense for some sticks is probably 2.35v, for cas5 maybe up to 2.45v but i dont think cas5 is worth it and cas6 and 7 make most sense. so the max voltage that makes sense to use is 2.3v and i doubt it will die at that voltage, unless it runs hot of course...

You have a lot of experience with mem, have you ever experienced something like this before? I never had any mem that had a sweet spot for voltage and didnt like more. if it ran too hot it oced worse or gave errors, sure, but thats not the case here.

Quite likely voltage/current regulation of the board (vtt circuit in particular, the regulator chips are often running horribly hot to begin with).

Yoxxy - On-chip voltage/current regulation isnt really feasible, dont think we'll be seeing it in the dumbed down form that gets kicked around forums.

[eva2000]

Lastviking! Hey man! how are you doing?

Well using more volts than the sticks like doesnt make sense, you only get worse results. The more tight the timings the higher the sweet spot, but even for 665 the highest sweet spots are 2.25v. Well im using a p5k3 so i cant use more than 2.25v...
So i guess with cas665 the max that makes sense for some sticks is probably 2.35v, for cas5 maybe up to 2.45v but i dont think cas5 is worth it and cas6 and 7 make most sense. so the max voltage that makes sense to use is 2.3v and i doubt it will die at that voltage, unless it runs hot of course...

You have a lot of experience with mem, have you ever experienced something like this before? I never had any mem that had a sweet spot for voltage and didnt like more. if it ran too hot it oced worse or gave errors, sure, but thats not the case here.

yeah some ram don't like too much vdimm... some of my kingston and mushkin BH-5 babies liked specific max vdimm voltages.. could be board/memory controller related too.

[saaya]

nice man gotta hook us up with some of that lovely ram :p j/k

hey what city is the fab in??


you really should have called those ram sticks 666

That was the plan, 666 and red heatspreaders, but victor kept asking for blue hehehe then i asked around and everybody liked blue...
Maybe we will use red ones for the final 666

The fab is close to Karlsruhe, south to Frankfurt

The ones I just received are the 876 then. Still, I see elsewhere that they do:

"CellShock DDR3 @ 1928 with 7-6-5-16 (Micron D9GTR) and 2,2 V"

Which is worth playing with.

I checked around 10kits and found 2 1800 kits that do 2ghz 765 with 2.05v
orthos stable for 10mins, so it might not be 100% stable but pretty stable.

@Saaya : If i buy from memoryC, can i find chip like this..??
Thanks

You mean chips as in D9GTR? Our Sticks come with D9GTR and Z9HWQ.
Dont know why micron still labels some chips as Z9...
I checked and D9GTR and Z9HWQ clock about the same, so it seems to be the same chips just with different marking.

If you mean if you will get sticks like victors when buying the 1800 from memoryc... most likely not. there is a small chance... but if you really want 665 sticks like victors then wait a few weeks

Quite likely voltage/current regulation of the board (vtt circuit in particular, the regulator chips are often running horribly hot to begin with).

hmmmm thats what i would have thought if there wasnt such a big variety between the chips. i think its the chips, not the board, i think...

Yoxxy - On-chip voltage/current regulation isnt really feasible, dont think we'll be seeing it in the dumbed down form that gets kicked around forums.

i was also thinking that micron might have added a hard fuse that as soon as the current reaches a certain value starts to leak current to ground before it reaches the logic to prevent the chip from dieing.
But then again, why would micron even care... they dont get any RMAs with dead chips on them, do they?
Unless crucial and large partners ask for an overcurrent protection i dont see why micron would even care... and microns biggest partners are oems, and their customers surely wont overvolt their mem to the point that it fries...

yeah some ram don't like too much vdimm... some of my kingston and mushkin BH-5 babies liked specific max vdimm voltages.. could be board/memory controller related too.

hmmmm well but think about this, if you have a few sticks that can do 1900 with a certain voltage in sc, some of them will do 1900 dc with the same voltage as well. i dont know what you think of this but to me this looks like the memory controller isnt limiting. otherwise the difference between sc and dc would be around 50mhz. then again, there are sticks that reach amazing speeds in sc, but in dc theres a wall no matter what timings, no matter what voltage.
one of the p5k3 boards i have here, one from the very first batch that was shipped, only does ~1825 dc max...

like victor said, its hard to come to conclusions about why ddr3 works like it does
well its still very early and jedec still hasnt really finalized ddr3 specs.
But the fun ddr3 brings will surely be enough motivation for us to figure out why it works like it does

[dinos22]

That was the plan, 666 and red heatspreaders, but victor kept asking for blue hehehe then i asked around and everybody liked blue...
Maybe we will use red ones for the final 666

The fab is close to Karlsruhe, south to Frankfurt

666 red FTW

[STEvil]

ship with heatspreaders that change color depending on temps.

[dinos22]

ship with heatspreaders that change color depending on temps.

ROFL nah i think Crucial should get into that blingy stuff ROFL

seriously who really cares what colour they are

i mean WTF do you guys actually look inside your PC and thing wow that looks nice

[saaya]

well if id mod my case with lights n all then yeah, id want the colors to match... but besides that... i personally dont care about the color.
Still i wouldnt buy pink memory for example, even if its locked in my case and i never see it

[stone_cold_Jimi]

Still i wouldnt buy pink memory for example, even if its locked in my case and i never see it

Can I have the pink 666 sticks you don't want?

[Yoxxy]

Quite likely voltage/current regulation of the board (vtt circuit in particular, the regulator chips are often running horribly hot to begin with).

Yoxxy - On-chip voltage/current regulation isnt really feasible, dont think we'll be seeing it in the dumbed down form that gets kicked around forums.

On die voltage regulation is already on Elpida, Infeon/Anfeon, and Hynix DDR2. It is also on Elpida DDR3. It is pretty easy to add.

Maybe someone from Micron Texas will enlighten us.

[noobzed]

in daily use, do you see something change in using 2 ghz ddr ?

[saaya]

Can I have the pink 666 sticks you don't want?

hehehe, nice try

On die voltage regulation is already on Elpida, Infeon/Anfeon, and Hynix DDR2. It is also on Elpida DDR3. It is pretty easy to add.

Maybe someone from Micron Texas will enlighten us.

Are you sure?
Well the Micron D9HNL chips sure look like they have an overvolt protection built in... but it seems to be broken as they stop scaling above 1.8v already.

But I dont think DDR3 has overvolt protections... Elpida DDR2 and DDR3 stop scaling with higher volts, but they dont produce errors. Micron DDR3 errors and even refuses to boot if the voltage is too high... so if they have an overvolt protection then its not working properly id say.
Plus Micron can take 2.3v+ and Elpida 2.1v+...

So IF elpida and Micron DDR3 ics feature an overvolt protection then its not working properly. If it would work dont you think Micron and Elpida would want to have it kicking in way before 2.1v for chips that have a default voltage of 1.5?

If Elpida has an overvolt protection for DDR2, then it kicks in at 2.1v.
So for DDR3 they would definately want to lower it to less than 2.1v...

Also dont forget, when lowering the latencies on Micron and Elpida DDR3 the sticks suddenly like higher volts than they do at higher latencies.
How would you want to build an overvolt protection that works like this?
Possible? sure! but not easy to implement, and if you screw it up your chips only run half the speed they would if that screwed up overvolt protection wouldnt be there.

in daily use, do you see something change in using 2 ghz ddr ?

Yeah, the reflection of a big grin on the monitor

[erwinz]

hey saaya..

whats a safe voltage for DDR3 for 24/7 use?? in overclock mode of course..