Asus/NCIX Competiton Entry - dejanh

[dejanh]

I think these high vdroops are becoming larger with each step because the electrical circuit gets less capacity to deliver the set voltage. Maybe you could try a v-mod?!?



Maybe you should try to post with only one stick of memory if you didn't try yet.



I'm totally clueless at this point as well. I would expect that lower bclk would put less stress on every component connected to it and therefore providing more stability.

Did you already do a google search for extreme OC results on a R2E, performed with nearly the same cooling solution as you are using?

I would think that if certain voltages are offered the manufacturer would have made sure to use adequate components to avoid circuit limitation issues. It would appear that is not the case

On that note, I did try posting with a single stick of DRAM with the same results. I do not think any of the sticks from the triple channel kit like higher voltages. Insted of more stability I end up with significantly less stability. Furthermore, I also noticed that if I loosen the timings past 9-9-9-24-2T even a single stick will no longer boot and the board just hangs at DET DRAM. Upper and lower ceiling on timings it seems...or???

In terms of the BCLK that is exactly what I would think as well. Lower BCLK == less stress overall but appears not necessarily. I will follow up on your suggestion and do a bit of a grind through Google to see if I can find any further details on R2E OCing other than what I know to date.

While I was doing this I fiddled with Uncore on my everday overclock and now I have Uncore set to 4.32GHz for everyday use but that requires roughly 1.6V at load on part of QPI/DRAM to be stable. Maybe the issue is really that for Uncore as high as 4.6GHz I need in excess of 1.8V on QPI/DRAM? That seems awful high...

Could it also be that the system would just behave better if I actually had some DRAM that was designed to run between 1.9V and 2.1V that's clocked high like for example the Corsair Dominator DDR3-2133C9DF sticks?

[funk32]

GOOD work looks nice.

[SteveRo]

Edit: Updated BCLK result. Only 0.02MHz higher but hey, you never know, it may make a difference between winning or losing

Edit 2: Pics! I finally got some up on the second page of this thread. Head over there now to view Shows the components I used to build my chilled water setup, very simple but very effective. Don't hesitate to drop me a line but of course do it here in the thread so I get more replies!

Well, let's raise the bar a bit

Here are the default settings for the system. All reported voltages are what is set in the BIOS unless explicitly stated otherwise.

CPU: Intel Core i7 940, 3836A405
Motherboard: Asus Rampage II Extreme, 1201 BIOS (Beta)
DRAM: 6GB Corsair Dominator 1600MHz Triple Channel Kit, 8-8-8-24-2T
Graphics: BFG GeForce GTX 260 (216) 718/1223/1538 (core/shader/memory in MHz) w/ Custom VGA BIOS
HD Drive(s): Generic Serial ATA
Optical Drive(s): LG LightScribe DVD+/- RW
Display: NEC MultiSync LCD225WXM
Cooling: Custom Chilled Water Cooling w/ Coolit Freezone Elite TEC Cooler, Swiftech MCR220-QP Rad, Swiftec Micro Res V2, Stock Coolit FZE Block

Please note that for the BCLK test the BCLK is set to 222MHz in the BIOS but it does not show up any higher in Windows than 220.8MHz no matter what.

Anyway, enough chatter

CPU Clock (HyperThreading On, 4 cores/8 threads): 4.73393 GHz
CPU Voltage: 1.68125V
QPI/DRAM Voltage: 1.56875V
DRAM Voltage: 1.65681V

http://valid.canardpc.com/show_oc.php?id=526588



CPU Clock (HyperThreading Off, 4 cores/4 threads): 4.75672 GHz
CPU Voltage: 1.66875V
QPI/DRAM Voltage: 1.43750V
DRAM Voltage: 1.65681V

http://valid.canardpc.com/show_oc.php?id=526581



BCLK: 220.82MHz Reported, 222MHz BIOS
CPU Voltage: 1.10000V
QPI/DRAM Voltage: 1.35000V
DRAM Voltage: 1.65681V

http://valid.canardpc.com/show_oc.php?id=529086 (see first screenshot below)



DRAM: 1800MHz, 8-8-8-18-1T (from 1600MHz, 8-8-8-24-2T stock), 4320MHz Uncore
CPU Voltage: 1.36875V
QPI/DRAM Voltage: 1.561V Actual (X.M.P. defaults to 1.35000V in BIOS but automatically increases with Uncore clock)
DRAM Voltage: 1.65681V

http://valid.canardpc.com/show_oc.php?id=526532 (see second screenshot below)



I will be posting images of the BIOS towards the competition end

221 bus speed - I don't think I can come anywhere close to that!

[sakelover]

some great overclocks u got there

[dejanh]

221 bus speed - I don't think I can come anywhere close to that!

You know what's annoying is that my BCLK is actually set to 222MHz but I cannot get it to stay set at 222MHz. It is only showing up correctly as 222MHz at POST but once I boot anywhere or even just start Memtest86+ 2.11 I am not longer running 222MHz BCLK but more like 221MHz. The higest validation I managed to get it 220.82MHz which is what you have seen on the front page. I wonder why this is the case...

Also, I tried everything imaginable to hit past 222MHz but no dice. 1MHz higher and no POST.

some great overclocks u got there

Thanks!

[Infa]

That's some nice mem speed you got there mate 21m/s im at 18.4 at cl7 1600, keep pushing!

[dejanh]

That's some nice mem speed you got there mate 21m/s im at 18.4 at cl7 1600, keep pushing!

Thanks It's all in the Uncore

Gonna go for max memory OC that I can get with my sticks in the next few days. I think I can get them up to 1975MHz or so. Not enough for first place but I think still pretty good considering I am dealing with 1600MHz sticks

[Rodriguez]

Nice cooling you've got there, How low can you go with that & does it hold the temp during long benchmarking periods (Prime small FTT, IntelBurn)?


Regarding your post on using 1.9 or 2.1v mem, I'm using 12Gb "old school" Diablo 2000 GTR at 1600 with 1.56v QPI, but it can clock 6Gb 2300Mhz no prob with 1.65v QPI & 2.1vDIMM although I haven't tested for stability

[dejanh]

Nice cooling you've got there, How low can you go with that & does it hold the temp during long benchmarking periods (Prime small FTT, IntelBurn)?


Regarding your post on using 1.9 or 2.1v mem, I'm using 12Gb "old school" Diablo 2000 GTR at 1600 with 1.56v QPI, but it can clock 6Gb 2300Mhz no prob with 1.65v QPI & 2.1vDIMM although I haven't tested for stability

Hey, nice to see you around

Take a look at the screenshot below It is a short LinX run with my cooling but it holds up just as well under extended testing. LinX gets the temps the hottest though. This was done with 17C ambients and a source of fresh air for the fans. During my daily use the temps never exceed 56C on the hottest core with fans on silent. That's quite good for an i7 running at 4.2GHz

Now, a question re: memory. Did you clock the GTR sticks on the i7 to 2.3GHz? What timings did you use? Stability is not an issue really, just pure performance in MHz That seems very impressive indeed. I wonder if their new HCF0s clock the same when you juice them up a bit past 1.65V or if they fizzle like some other out there...HCF0s should do well with higher voltages. Did you get a chance to test these? Also, where did you buy your GTR sticks? I don't know any retailers near me that carry these...

[Pisklink]

And did you already find something usefull with a google search?

Btw I was wondering how you can run a stable OC with such a high uncore? Because you are the one who was telling in a thread that the QPI link / uncore ratio needs to be 9:8. So an uncore of over 4Ghz would imply an ever higher QPI link. How do you get that stable?

[dejanh]

And did you already find something usefull with a google search?

Btw I was wondering how you can run a stable OC with such a high uncore? Because you are the one who was telling in a thread that the QPI link / uncore ratio needs to be 9:8. So an uncore of over 4Ghz would imply an ever higher QPI link. How do you get that stable?

So far I did not yield much more than what I already knew. I am inclined to think that it is one of the two...

1. The memory that I am using has problems when voltages are being pushed past 1.65V on DRAM directly affecting the ability to clock a high Uncore.
2. The board and/or the chip simply cannot handle such a high Uncore and still be bootable.

I will try one more thing soon and that is to do a fresh reinstall of Windows with stripped down services and then see if I can actually boot into Windows with a high Uncore.

In terms of running a "stable" high Uncore, well, the ratio I identified is still the safer choice and is basically guaranteed to be stable, where what I am running is not. I am using the term stable here a bit loosely in that I did not run days of tests on it, but rather a few hours only. Overall, to run a high Uncore day-to-day you need to really increase the QPI/DRAM (VTT) voltage and I am not sure what the reprecussions of that will be for the long run. Right now I am booting and using 4.32GHz Uncore with OCd DRAM at 1804MHz using 8-8-8-18-1T timings but to do that I need to have my QPI/DRAM at 1.607V which is 0.257V over the Intel's specified safe maximum of 1.35V. In addition, unless your chip has a very decent IMC and cache you won't be able to stably do this in most cases. There are a lot of people that simply cannot get their Uncore so high regardless of how much voltage they pump through. Uncore was really not generally designed with the intention of running this high. I also think this is where a lot of my limitation comes in. If I need 1.607V to be stable at 4.32GHz Uncore, then I quite likely need as much as 1.85V or maybe even 1.95V for 4.6GHz Uncore. Way to much voltage for VTT. I would probably kill the chip with this extremely fast.

[dejanh]

Here is a quick update on a memory MHz test that I did...1950MHz from stock 1600MHz. Not too shabby. My replacement sticks from Corsair do not do nearly as well. The new sticks cannot clock past 1800MHz from 1600MHz stock. The original kit of which I have two sticks still could do 1975MHz with a bit of tweaking

[dejanh]

Here is one DRAM that I was looking at to potentially pick up. It uses the Samsung HCF0 ICs though I am not sure how it overclocks. It would definitely allow me to run higher QPI/DRAM volts as the memory is rated for about 2V.

Anyway, check it out - http://ncix.com/products/index.php?s...acture=CORSAIR

[dejanh]

And here is another one from CSX http://www.csx-memory.com/?site=show...article_id=707

I could get them relatively quickly I think but again, I am not sure how they would hold up. According to Rodriguez his older Diablos are great but the triple channel kit is running at 1.65V and supposedly uses the same ICs as above, the HCF0. I'm kind of hoping for him to reply to my question above re: his experience overclocking the older dual channel kit using the D9GTR ICs. Anyone know of any direct comparisons between these?

[dejanh]

I was wondering how you can run a stable OC with such a high uncore? Because you are the one who was telling in a thread that the QPI link / uncore ratio needs to be 9:8. So an uncore of over 4Ghz would imply an ever higher QPI link. How do you get that stable?

Sorry, I had to post about this again...since you asked me how I can run a stable OC with such a high Uncore...

I was doing something today and all of a sudden my system went BSOD. Only change I have made recently is I went from 4.14GHz Uncore at 1.55V QPI/DRAM Voltage to 4.32GHz at 1.6V QPI/DRAM. I guess despite the tests I did something was not stable or deteriorated over a few days...lol

I guess the lesson is like I said before, you cannot be sure that this high Uncore is 100% stable.

[scyzer]

Nice OC. Wish I could break 4.0GHz on mine.

[dejanh]

Nice OC. Wish I could break 4.0GHz on mine.

Did you try the suggestion for finding the max BCLK that I made in the Core i7 Overclocking Thread?

[dejanh]

Did anyone ever try this block on their CPU? I was wondering if it is worth potentially picking this up http://swiftnets.com/products/mcw6500-T1.asp to add to my loop. Kind of a general question I guess...I doubt I'd be able to get it in time to try it out for the competition. What do you guys think?

[Rodriguez]

Hey, nice to see you around

Take a look at the screenshot below It is a short LinX run with my cooling but it holds up just as well under extended testing. LinX gets the temps the hottest though. This was done with 17C ambients and a source of fresh air for the fans. During my daily use the temps never exceed 56C on the hottest core with fans on silent. That's quite good for an i7 running at 4.2GHz

Now, a question re: memory. Did you clock the GTR sticks on the i7 to 2.3GHz? What timings did you use? Stability is not an issue really, just pure performance in MHz That seems very impressive indeed. I wonder if their new HCF0s clock the same when you juice them up a bit past 1.65V or if they fizzle like some other out there...HCF0s should do well with higher voltages. Did you get a chance to test these? Also, where did you buy your GTR sticks? I don't know any retailers near me that carry these...

Hi,

Very good temps, I'm amazed that it can keep up with the heat load, with HT on right?

I use a chiller to cool all my components but I never run it below 12Cº to avoid condensation.

As for the mem speed I MADE A MISTAKE I clocked 6Gb at 2.230 BCLK160 BIOS 1104 - NOT AT 2.300, SORRY this was with timings on Auto 9-9-9-24.
I booted 2 times to Vista64, no stability check. I'm not a dedicated overclocker, I just mess around from time to time. 99.9% of time I run 12Gb without pagefile, only every now & then with a new BIOS release I take out 6gb & try to see how far I can clock it.

My only "goal" at the moment is to reach around 1800 with 12Gb, but now that I've installed an Adaptec 5805 RAID card it takes about 2 minutes after each reboot to start loading OS or to get into BIOS because of the RAID card eternal POST, so my tinckering with BIOS is kept to a minimum.

I bought these sticks at HPM-Computer.de after asking CSX (Mr. Heinrich) for a tailor made 12Gb set D9GTR which they later sent on to HPM under my name. I originally used 8Gb for my previous Striker II Extreme & kept the other 4Gb until X58 came out. I havent tried the HCF0 from CSX, what I really want to try are the ADATA 2133 sticks at 1.50.

In the US CSX parts are distributed by Tankguys.com

EDIT: Sorry again for my previous mistake

[skiman]

I like your 920 results.

[dejanh]

Hi,

Very good temps, I'm amazed that it can keep up with the heat load, with HT on right?

I use a chiller to cool all my components but I never run it below 12Cº to avoid condensation.

As for the mem speed I MADE A MISTAKE I clocked 6Gb at 2.230 BCLK160 BIOS 1104 - NOT AT 2.300, SORRY this was with timings on Auto 9-9-9-24.
I booted 2 times to Vista64, no stability check. I'm not a dedicated overclocker, I just mess around from time to time. 99.9% of time I run 12Gb without pagefile, only every now & then with a new BIOS release I take out 6gb & try to see how far I can clock it.

My only "goal" at the moment is to reach around 1800 with 12Gb, but now that I've installed an Adaptec 5805 RAID card it takes about 2 minutes after each reboot to start loading OS or to get into BIOS because of the RAID card eternal POST, so my tinckering with BIOS is kept to a minimum.

I bought these sticks at HPM-Computer.de after asking CSX (Mr. Heinrich) for a tailor made 12Gb set D9GTR which they later sent on to HPM under my name. I originally used 8Gb for my previous Striker II Extreme & kept the other 4Gb until X58 came out. I havent tried the HCF0 from CSX, what I really want to try are the ADATA 2133 sticks at 1.50.

In the US CSX parts are distributed by Tankguys.com

EDIT: Sorry again for my previous mistake

Thanks for the info and don't worry about the error. I do have to admit I was surprised that they clocked that high so easily so I guess I was on the mark. I don't know much about CSX memory since I never used their products but I hear from people that they are quite good. I will probably have to give them a shot at one point in time. My understanding right now is that Elpida ICs are better than the HCF0 that come with the Diablo triple channel kit. I will have to do some more investigation of this. This competition caught me a bit unprepared so I am struggling to keep up now.

I may even give CSX a call and see if they can tailor something for me for future.

In regards to ADATA, I heard that the ADATA are very good, but I cannot seem to find them here. I know that one of our bigger retailers here (London Drugs) carries ADATA but I doubt that they can supply the high end parts. I have only seen them carry 2-3 year old stuff

Are you aware of any US/Canadian retailers for ADATA?

[dejanh]

I like your 920 results.

Thanks, though it's a 940 so maybe not "quite" as impressive hehe

I wish I had some more extreme cooling though considering my setup this is not bad I think.

[Rodriguez]

Thanks for the info and don't worry about the error. I do have to admit I was surprised that they clocked that high so easily so I guess I was on the mark. I don't know much about CSX memory since I never used their products but I hear from people that they are quite good. I will probably have to give them a shot at one point in time. My understanding right now is that Elpida ICs are better than the HCF0 that come with the Diablo triple channel kit. I will have to do some more investigation of this. This competition caught me a bit unprepared so I am struggling to keep up now.

I may even give CSX a call and see if they can tailor something for me for future.

In regards to ADATA, I heard that the ADATA are very good, but I cannot seem to find them here. I know that one of our bigger retailers here (London Drugs) carries ADATA but I doubt that they can supply the high end parts. I have only seen them carry 2-3 year old stuff

Are you aware of any US/Canadian retailers for ADATA?

I've been searching for their X v2.0 kits but they're no where to be found, I only found a japanese page that sells them.

I would definately go for the Elpida chips, I've tried the HCF0 with "old" Patriot Viper 2000 but they were limited to high speed (2100) with very loose timings, maybe the "new" HCF0 chips are better but I doubt it.

[nordien]

With bios 1204 the 21 multi seems to do very well. 4200 @ all cores is very cool. Im glad i stayed with the R2E........Almost got me a Bloodrage.

I hope you succeed with your price Dejanh

[TheKarmakazi]

And here is another one from CSX http://www.csx-memory.com/?site=show...article_id=707

I could get them relatively quickly I think but again, I am not sure how they would hold up. According to Rodriguez his older Diablos are great but the triple channel kit is running at 1.65V and supposedly uses the same ICs as above, the HCF0. I'm kind of hoping for him to reply to my question above re: his experience overclocking the older dual channel kit using the D9GTR ICs. Anyone know of any direct comparisons between these?

If CSX is binning the HCF0 diablos as good as they binned the D9GTR's they will be great

I've been searching for their X v2.0 kits but they're no where to be found, I only found a japanese page that sells them.

I would definately go for the Elpida chips, I've tried the HCF0 with "old" Patriot Viper 2000 but they were limited to high speed (2100) with very loose timings, maybe the "new" HCF0 chips are better but I doubt it.

Fugger's HCF0 im pretty sure holds frequency record for i7. HCF0 is like all other ram IC's some batches are better, and depending on how well it is binned it will OC poorly or wonderfully. Luck of the draw unless you can get a pretested kit. But to say all HCF0 are not high speed chips is misleading