A short story behind Corsair Dominator @ DDR3-2462 CL9

[Yellowbeard]

This is the EXTREME BANDWIDTH section

[bachus_anonym]

Alright, this will have to do for now, guys. I don't have much time today, unfortunately.

1. Below you see 2 different systems, cooled by Corsair Nautilus 500.
2. One running with BIOS 0301 that I mentioned is essentially what I used for 2462MHz, and the other is flashed with BIOS 0401, freshly DL'ed from Asus website. Look at CPU-Z Motherboard Tab for BIOS info.
3. Each has a 2GB kit of Corsair Dominator that you're looking at in this thread. Windows Vista 64-bit on the 1st, and Windows XP on the 2nd.
4. Running ORTHOS on both, at DDR3-2133 CL9.
5. One system at 533FSB x 6, and the other at 533FSB x 7.


5hrs ORTHOS Blend at DDR3-2133 CL9, BIOS 0301

--- Click image for larger version ---



Over 3hrs ORTHOS Blend at DDR3-2133 CL9, BIOS 0401

--- Click image for larger version ---



Boot up @ 612FSB - BIOS 0301



Boot up @ 610FSB - BIOS 0401




Let's put the speculation about the "developers" BIOS to bed now...

[metro.cl]

From my POV i like to see what manufacturers can do in their lab, so thanks Bachus for the post and info.

For me BIOS 401 is much better than 402.

[kiwi]

So what are these non micron chips? Do they like volts? Do they like tight timings/sub timings? Because if they don't they are not any better. I'd like to see at least 13K read speed @ 4Ghz

[M.Beier]

Interesting Tony & Bachus...
Bachus, the BW, frustrating low :| But chipset and other settings must be your problem.. I mean... sub 800mhz 11GB/s is possible on P35 as far as I remember...

From my POV i like to see what manufacturers can do in their lab, so thanks Bachus for the post and info.

For me BIOS 401 is much better than 402.

The kings of lab testing is.....

PATRIOT....
Sooo cool results that they've acheived.... =)
Sadly, none is published to my knowledge!

EDIT:
Ohh crap, entered message too fast, didnt write my opening comment

[saaya]

thanks michal!
makes me wonder why you used the special bios to begin with as 0401 seems to give you better scores?
read is still very low though... is 2000 cas7 bencheable or stable? what about cas8?
I think when your pushing the clockspeed up too high perf actually degrades...
unless you keep the timings tight like jodies 2200 cas8

[dinos22]

lol guys

i think ppl are forgetting we are in the Xtreme Bandwidth section here and getting their knickers in a twist of some new ICs which are still in ES stage.....

i spent a couple of hours with Michal in the labs a few days ago and there was no funny business going on here. These results are truthful. The setup was as described and CPU watercooled....These motherboards are pretty good though. Asus has done some improvements over the X38 chipset which is testament to the fact Corsair had many many setups with these boards doing great FSB and clocks fairly easy...........

FSB however is hard to tweak if the board ain't all too happy which is what my X38 boards are like and i am pushing limits on Intel chipsets with DDR3 gear on air/water cooling

i'll have to buy one of these Premium boards though after seeing what Michal can do with them.....

i'll do some testing with my 2133MHz kit to see what they will do....first i have to get some rest/sleep after such a short trip/no sleep and these long ass 14 hour flights between sydney and San Francisco...........man oh man i'm just buzzing with all this new tech coming out and all



but i'll do some benching on 790i which kicks the pants off a lot of Intel setups these days as people can do that kind of FSB very easy >> i've got a new P05 bios from nvidia for 790i so let's see what that does in terms of some known bug fixes as well

keep pushing it as Oliver would say

[M.Beier]

Its in my nature to argue, so Dinos, here goes :
Since its ES, it doesnt mean anything....

Unless its one of them fancy "ES spree" but hopefully, Corsair doesnt lowern themselves to such level and writes ES just for making it look "cool"... If its truely ES, then the final specs sure aint decided, aint certain, can be changed....
- And if thats the case, then, wuhuu lets wait for the final product...

However, it does look promising...
I will ask for what killer IC's are on the market the upcoming monday, and try to get some ram with such IC's shipped...

[Oliver]

Looks really good and quite interesting...

Lol Dino ...

[eva2000]

interesting dinos 2133 ES ram that's already at v4.1 ????

[gundamit]

but i'll do some benching on 790i which kicks the pants off a lot of Intel setups these days ...

Hersey! x48 vs. 790i and jumping in (late) to DDR3 ... so hard to choose after so much quick and dirt(y) cheap DDR2. Oh well ...

BYW - everyone keep encouraging company guys like bachus_anonym to stay in the mix at XS. They bring a lot to the table.

[saaya]

i think ppl are forgetting we are in the Xtreme Bandwidth section here and getting their knickers in a twist of some new ICs which are still in ES stage.....

i think we are all spoiled from michal
he posted a lot of really really impressive threads with nice results and pics and scores, and this time its a suicide screenshot and then 2133 999 with average performance results...

To me this thread has a strong hype factor for average users but its not impressive/interesting for ocers when you look at the actual stable/benchable speeds... at least from what ive seen so far.

we have all seen a lot of 1800+ cas6, 2000+ cas7 and recently 2200 cas8...
so 2133 cas9 isnt exactly amazing compared to those results
its been about a year since michal posted everest@13K, and now we see 11K...

having said that, im sure you and michal will up the ante and raise some everest hell with theese sticks so bring it on!

interesting dinos 2133 ES ram that's already at v4.1 ????

hahah yeah, i wondered about that too and pmed michal, he confirmed its supposed to show 4.1... i dont get why tho
maybe all sticks that say 4.1 use the same chips, and all sticks that say 4.0 use the same chips etc, accross the board. that would be cool, still you cant see what your buying until you actually have it in hand, but then you can be sure what chips it is.

[George_o/c]

Interesting info dinos, waiting for the results in 790i ... They should be kick-ass

i've got a new P05 bios from nvidia for 790i so let's see what that does in terms of some known bug fixes as well

Any other improvements with this new bios ? ...

keep pushing it as Oliver would say

He has now updated to " keep tearing it up ", which I like even more (Nice one Oliver)

But I'm sure he will come with more updates

[saaya]

a couple of people already tried the pr05, its been out for a week or two i think.
some people still get corruption with it, even at low clockspeeds, but most say its very good.

[bachus_anonym]

interesting dinos 2133 ES ram that's already at v4.1 ????

hahah yeah, i wondered about that too and pmed michal, he confirmed its supposed to show 4.1... i dont get why tho
maybe all sticks that say 4.1 use the same chips, and all sticks that say 4.0 use the same chips etc, accross the board. that would be cool, still you cant see what your buying until you actually have it in hand, but then you can be sure what chips it is.

On a label, ver4.x indicates ICs used, not stages of development of the part.

To me this thread has a strong hype factor for average users but its not impressive/interesting for ocers when you look at the actual stable/benchable speeds... at least from what ive seen so far. we have all seen a lot of 1800+ cas6, 2000+ cas7 and recently 2200 cas8...
so 2133 cas9 isnt exactly amazing compared to those results
its been about a year since michal posted everest@13K, and now we see 11K...

There are few main points of this thread, If I may:

- show my take on what may be the upper ceiling for DDR3 speed in a foreseeable future, hence DDR3-2462 CL9 1-up, and DDR3-2268 CL9 2-up;

- inject some fresh ingredient into DDR3 overclocking by showing very good, high-speed alternative to Micron DRAM components;

- present 100% solid, fully stable DDR3-2133 configurations based on Asus P5E3 Premium motherboard - not only SPi32M stable as we all know, 32M at max speed does not equal 100% stability; correct me if I'm wrong but DDR3-2133 ORTHOS runs shown here are the only to date, regardless whether it's on X38/X48 or C73;

- introduce what could be our new product that has been kept on hold for a while due to rather high requirements for X48 chipset based setups, and C73 being still in early stages, with its share of "intermittent or random corruption" issues once memory frequency is pushed above DDR3-2100. We have an expertize to produce DDR3-2133 product for both platforms, but we decided to take a wait-and-see approach


As for the bandwidth results, I respectfully disagree with the assertion that they are sub-par. If you look at DDR3-2228 @ nearly 3.9GHz shot in the 1st post, bandwidth scores are actually very good for 9-9-9, and even for some 8-8-8, if we keep the same CPU frequency - 11.9k Read, 11.7 Write, 11.7 Copy. I'm not talking about uber tweaked out timings either, as results shown here can still be improved for efficiency. Additionally, please do not compare it to C73's Read score as BIOS on this chipset has performance paths enabled that give substantial boost in bandwidth, yet e.g. clock-for-clock C73 is still a bit slower in SPi32M than X48.

i'll do some testing with my 2133MHz kit to see what they will do....first i have to get some rest/sleep after such a short trip/no sleep and these long ass 14 hour flights between sydney and San Francisco...........man oh man i'm just buzzing with all this new tech coming out and all

I'm looking forward to seeing what you manage to do with those on your C73 board, Dino.

[saaya]

On a label, ver4.x indicates ICs used, not stages of development of the part.

so my guess was right, cool

- show my take on what may be the upper ceiling for DDR3 speed in a foreseeable future, hence DDR3-2462 CL9 1-up, and DDR3-2268 CL9 2-up;

why do you think suicide screenshots have any relevance on what will be possible stable in future? yes, speeds that we can only reach suicide stable today might be possible stable in future, but its not like a suicide screenshot at any speed proves anything. its not like it shows this speed WILL definately be stable in future. ive seen 2000+ cas6 and 2200+ cas7 suicide shots and videos... but they are pretty meaningless imo. yet to most people who dont know a lot about memory this sounds amazing, and from what you posted and what corsair PR emailed to the press, its sounds like 2460 was stable. and even though i doubt it was worded to sound like this on purpose, it wasnt worded to make clear that it was just a suicide screenshot and not stable either.

thats all im saying, imo you should focuss on the "real" results of 2133 999 and 2200 999 and add the suicide shots as a goodie, and not draw all attention on a suicide shot that doesnt really mean that much.

im looking forward to some more real results now
dinos has shown that he has a very good clocking 790 board, so im looking forward to see his results.
after he caught some sleep

[bachus_anonym]

It's me aaaagaaaain! I hope you haven't had enough of it yet...

So, I took a week off from torturing this 2133C9 set and decided to come back today, to push it even more, looking for another milestone in ORTHOS / Prime95 stability on DDR3.

In a mid-time, Asus sent the latest beta BIOS last week. I want to add however, that 0401 BIOS did just as well for this new result. For the curious ones, new BIOS creates a separate sub-menu for Skew Control, with per DIMM skew tuning of for command, clock and control (CMD/CLK/CLK) signals in 25ps steps. Also, one can now play with skew of CMD signal at Command Rate 1N, which was not possible before - option was greyed out. But we enthusiasts want to play with all the knobs, right?

I would like to emphasize, that Asus engineers have been great in listening to suggestions for and implementing some of them so far. There are still a couple more things I would like to see in the next official BIOS that I have seen earlier, though. All that effort, makes Asus P5E3 Premium hands down the best to date motherboard for serious DDR3 overclocking on Intel chipset, period!

OK, enough! I really wanted to share what I woul;d consider quite a milestone

4hrs ORTHOS Blend at DDR3-2200 CL9 1T, BIOS 0503
This time, I also went with C2D E8500. Previously I used C2D E8400

vPLL = 1.80v   
vFSB = 1.30v
vDD = 2.02v     
vNB = 1.71v       
vSB = Auto

--- Click image for larger version ---



Dual SuperPi32M & Everest Ultimate @ as above, BIOS 0503

--- Click image for larger version ---

[dinos22]

great work there man

[Movieman]

Thanks guys!

Man, never enough? We're constantly looking into the future, in an effort to bring fastest, and most exciting memory for extreme overclockers, but my humble prognosis is that you will not see DDR3-3000 kits making its way anytime soon, hehe. Even if DDR3-3000 becomes reality at a reasonable voltage on a component level, you will still need a platform that could support DIMMs built with it. I'd wouldn't hold your breath until DDR4 for that...
Always, always As you can see, we know how to make DDR3-2133 happen, and we'll go from there

I admit to knowing very little about memory but when you post I read.
It's not all about the huge numbers to me but reading a thread from someone who does understand how it all works.
I'll add my name to the others here that say we miss your threads and the education we get from them.

[Tony]

In a mid-time, Asus sent the latest beta BIOS last week. I want to add however, that 0401 BIOS did just as well for this new result. For the curious ones, new BIOS creates a separate sub-menu for Skew Control, with per DIMM skew tuning of for command, clock and control (CMD/CLK/CLK) signals in 25ps steps. Also, one can now play with skew of CMD signal at Command Rate 1N, which was not possible before - option was greyed out. But we enthusiasts want to play with all the knobs, right?

Make sure you thank Raja from Anandtech as he was the one who asked for the options... also the 25ps skew was my idea

great work there man

And we are the ones who don't support the enthusiast on the forum

tooting ones own horn LOL

[Harshal]

Nice numbers guys!!!

[bachus_anonym]

I admit to knowing very little about memory but when you post I read.
It's not all about the huge numbers to me but reading a thread from someone who does understand how it all works. I'll add my name to the others here that say we miss your threads and the education we get from them.

Thank for the kind words, Dave!

Make sure you thank Raja from Anandtech as he was the one who asked for the options... also the 25ps skew was my idea

Although not taking anything away from Anandtech crew, they are not the only ones that deserve the credit. There are quite a few other people involved in the process as well, whose feedback, extensive testing and suggestions are being heard. Raja, Gary and Kris @ Anadtech sure are part of it, however not exclusive. Recognizing that is important as well

[Tony]

I just needed the world to know where certain options came from, I know others work to help, but this time that bios was the product of mainly Raja and a little my request also.

So corsair do well with a bios OCZ requested LOL ironic.

have a nice day man

[bachus_anonym]

I just needed the world to know where certain options came from, I know others work to help, but this time that bios was the product of mainly Raja and a little my request also.

So corsair do well with a bios OCZ requested LOL ironic.

have a nice day man

Again, you're not correct, just when you were when you mentioned "developer's BIOS" earlier in this post. Those types of results can be obtained with 0401 BIOS, which doesn't have per DIMM skew tuning, and without modifying the board the way Anandtech crew did. Do not make me prove you wrong again, Tony

On a side note, let's not go any further in this, please. We both know we may not agree, and I don't want this thread to turn into a fight. Thank you, Tony.

[Tony]

you are using the 503???

0401 was earlier...or so I thought, and before that was 0001 which Asus would not release to me.

Im done, I will keep my promise about keeping the bios private.

Have a nice day man