***Asus Rampage III Extreme Owners Thread***

[XII]

Looks like I can get my ram to 1866 Mhz Cas 7 stable, I just needed to increase my QPI:



Gogo 4.65 Ghz CPU without HT next, please work.

Nice run.
Have the same sticks, same scores, but hit a hard wall passing 185fsb. 24/7 ~180 mem@ 7-8-7-24 1800+ Rex3(First release)
Can copy the results with 2 kits, for 24Gb`s, same limitations, though.
How old`s you're Rex3?

[Mungri]

You have 6x4 Gb Geil Ultras according to your sig?

If you can reach 1800 Mhz Cas 7 with 6x4 Gb sticks I wouldnt worry at all, thats impressive for 24 Gb.

I've used these sticks in both my UD5 and R3E with 2 CPUs. 1800 Mhz Cas 7 is no problem, but pushing any higher on X58 is a major PITA. I can boot at 1900 Mhz Cas 7 with all 12 Gb detected, but there is nothing I can do to get it stable, even 1.65v and 10-10-10-30 2T timings are unstable when going over 1866 mhz.

Any better ram than this stuff is major rare in the UK, and even if we do get it, it costs too much for ram.

4.66 Ghz @ 1.425v on the CPU with HT off is somewhat stable with 1866 Mhz Cas 7 ram:



It only made it to 22 mins in Linx, but isnt having any problems so far in gaming.

I was also crazy enough to order a 512 Gb Crucial M4 because they were on a daily offer at £499. Thats a lot of money on an SSD, but I will ditch my Raid 0 F3s for the 512 Gb M4 for my gaming drive, and keep the two 2 Tb F4s for backups and non game files.

Pretty much most of my money gone on the I7 980 + Phanteks cooler plus a 512 Gb Crucial M4. Yes I am crazy.

Oh, and my R3E was bought on release date, so its 14 months old now. I'm going to keep it and my new CPU + SSD for at least 3 more years, but hopefully more.

[Mungri]

Does anyone know if using a 4 pin to 8 pin ATX converter on my second 8 Pin CPU power socket will make my overclock more stable?

http://www.amazon.co.uk/dp/B0051Y0B0...SIN=B0051Y0B0Y

My PSU only has an 8 pin + 4 Pin CPU power connectors, both of which are currently plugged in. Will using this cable to convert the 4 pin to an 8 pin output help at all?

[XII]

Does anyone know if using a 4 pin to 8 pin ATX converter on my second 8 Pin CPU power socket will make my overclock more stable?

http://www.amazon.co.uk/dp/B0051Y0B0...SIN=B0051Y0B0Y

My PSU only has an 8 pin + 4 Pin CPU power connectors, both of which are currently plugged in. Will using this cable to convert the 4 pin to an 8 pin output help at all?

It didn`t give any advantage for me, but core-speed is limited to ~4,2 @ 1,3V running on air.
My main focus is getting 1866 7-8-7 @ 24Gb
As for core, anything +4 is sufficient for me, pref +HT.

Have you pulled the spreaders for a chip id?
E952V8AD941

PS. I`m from across the channel, were weed`s legal.

[Mungri]

No I dont want to risk damaging the chips.

There was a thread in Xtreme bandwidth on my modules where someone did that, and they couldnt be identified. Most people thought they are either Elipda Hyper or BBSE (whichever one is the slower one of the two is what they perform like).

By the way have you got a link to your 4 Gb geil modules so I can check them out?

[XII]

No I dont want to risk damaging the chips.

There was a thread in Xtreme bandwidth on my modules where someone did that, and they couldnt be identified. Most people thought they are either Elipda Hyper or BBSE (whichever one is the slower one of the two is what they perform like).

By the way have you got a link to your 4 Gb geil modules so I can check them out?

Tested now 5 boards, and I seem to run into again, the hard wall @1814 using 24Gb`s
Rex3, Rex3Gene, and Rex3BE all have the ~same limit. Core`s don`t seem to be the problem, 2x W3680, X980 and X990 all give the same results, and speed only varies marginal.

Time to raise the multi on the X990 and find the limit

[Mungri]

Ram isnt limited by the mobo on X58, its limited by the CPU as thats where the memory controller is.

If you try 5 different boards on the same CPU, your ram overclocks wont change.

[XII]

Ram isnt limited by the mobo on X58, its limited by the CPU as thats where the memory controller is.

If you try 5 different boards on the same CPU, your ram overclocks wont change.

One can always hope
And G58A-OC showed, board does matter

PS. The R3BE is much cooler, and needs a little less voltage compaired to R3`s

[Mungri]

And G58A-OC showed, board does matter

PS. The R3BE is much cooler, and needs a little less voltage compaired to R3`s

I will agree with that in regards to CPU and QPI voltage, my R3E needs a lot less of both for stability than my UD5 needed.

But changing X58 mobos wont affect the maximum attainable ram frequencies, only changing the CPU will.

On X58, the mobo affects voltages and stability, CPU determines max CPU + Ram frequencies. I really hate CPU integrated memory controllers with a passion now.

[XII]

Second that

[Mungri]

Pushing past 186 BCLK - 188 BCLK and 1884 Mhz ram:



At 188 BCLK, I can choose 4.5, or 4.7 Ghz CPU clocks. I havnt tested 4.7 Ghz yet, but I dont think it will fare too well as I get minor instability at 4.66 Ghz.

And for some reason, 190 BCLK and 1900 Mhz ram has never been stable, but I will try again, 189 and 190 BCLK next. The ram is overclocking a bit better than it did on my I7 920 at least, I couldnt get it this high before.

It does around 1700 Mhz @ 6-8-6, and just under 1900 Mhz @ 7-8-7 (maybe more if I had a luckier CPU).

Nice, 189 BCLK and 1894 Mhz Cas 7 ram passed:



And at 190 BCLK I get 'Not exact in round' error right away.

[Cookiesowns]

What's your voltage?

I'm pushing 4.3ghz, 205x21 HT off, 1.39v in bios 1.41v cpu-z. 24GB , 3500 QPI/IMC, 1.30v. It's running pretty stable. I had a validation of I think 230bclk on my chip before. Can't seem to find it. I'll be benching my system a crapload before I sell it though

http://valid.canardpc.com/show_oc.php?id=2133260

[XII]

@bhavv - How high can you take Uncore?(x2+1)
Seems to be my limit with 24Gb ~1815 7-8-7
Going to X1,5 doesn`t kick it for me, but 2000 8-9-8 12Gb was stable enough for benching

[Cookiesowns]

@bhavv - How high can you take Uncore?(x2+1)
Seems to be my limit with 24Gb ~1815 7-8-7
Going to X1,5 doesn`t kick it for me, but 2000 8-9-8 12Gb was stable enough for benching

The uncore really depends on your IMC. my chip always likes voltage! I think I broke my own personal record with less than 1.42V with even HT off. I was doing 1.34v ( in bios with QPI LLC ) at 3600mhz with DDR3 1700 at C9 with 24GB.

[Kosmic71]

I need help guys.

I read threads left and right about the R3E sata3 capabilities and saw conflicting results/opinions. With an OCZ vertex 3 in single drive, what speed can it achieve? I'm still debating if I should buy a sata2 or sata3 ssd.

Thanks

[Cookiesowns]

I need help guys.

I read threads left and right about the R3E sata3 capabilities and saw conflicting results/opinions. With an OCZ vertex 3 in single drive, what speed can it achieve? I'm still debating if I should buy a sata2 or sata3 ssd.

Thanks

If they are around the same price get the Vertex 3. The SATAIII controller on the R3E pretty much sucks, in raid it sucks even more. Hence why I'm upgrading to SB-E

[Kosmic71]

Yeah that's pretty much what I've read. Do you have a ballpark bandwidth number that the R3E marvell 9128 can achieve in sata3 with only one high performance sata3 ssd drive?

[spoolindsm127]

New King of efficiency

80.9820 GFlops / 4.01984 GHz = 20.15 GFlops/GHz

4.02GHz (167x24) 2010MHz 7-7-7-20 (12x) 3850MHz Uncore (23x)
HT On, CIE & EIST Off. CPU 50% LLC, QPI LLC Off

VCore 1.193v (load_real)
VTT 1.289v (load_real)
DRAM 1.629v (load_real)



Was actually just trying to stabilize a 4GHz daily 24/7 OC and realized efficiency was really good. BTW how's the high perf club idea coming along Zoson?

[claydough]

I apologize for cross posting this query from other forums ( perhaps I should have jes started here in the first place! )
However I am finding it hard to believe no one has tried to use a 4-way bridge on the REIII considering the ASUS manual states that one should span the second slot for tri sli!
( before one jumps the gun an advises jes to use tri sli or ribbons please read the following: ( and thanks in advance! )
is 4-way sli bridge possible on a 3-way sli?
crossposting from the evga boards:

Originally my manual for my ROG Rampage III Extreme states that correct configuration for 3 way sli ( on the REIII @ least ) was supposed to be:

x16
skip
X8
x8

In which case a 3 way sli bridge will not span that distance.
The new directions given by ASUS on their support forums is..
To use the following configuration instead:

skip
x16( supposedly )
x8
x8

Which is then supposedly fine and does not result in slower speeds.
Not that I do not trust getting x16 from the second slot... however For purely aesthetic reasons:
( My motherboard's EK cooling Block looks better exposed with the first configuration. In fact it's :banana::banana::banana::banana: win with the first configuration )
I would just use the single band sli bridges to span the cards but I have more stable luck with my 3-way sli ( used for dual sli at the moment )
What I would prefer is to span that distance with a 4-way sli bridge if it's possible to use one for 3-way instead.
I am paranoid because the pcb is not quite the same as a 3-way bridges pcb ( while there is solder for every connection on a 3-way bridge the 4-way bridge alternates from one to two connection solder )
Hope that make sense? If not check out the pics:




vs.





I am hoping someone has used the 4-way sli bridge for a 3-way sli that spans the second slot successfully.


come to think of it...
Does anyone know if all 4-way sli bridges are designed this way? Perhaps there is one with a pcb configured for every connection?

would really prefer not to daisey chain ribbons. but if I have to what would be the right way to configure the ribbons to span the second slot?


I would have thought that in enthusiast forums that a succint "no i tried that and it won't work or yes I tried that and you are good to go!" would be simple enuff.
I think the forums I tried were not as enthusiast centric as I had first thought ( neither am I cuz the price of a quad bridge is a little pricey to simply leave to chance! )

[zoson]

4ghz runs will not be considered for efficiency in the hpc thread.
4.5ghz is the cutoff as previously discussed.

I'm super busy with work right now, but expecting it to calm down once our new release matures a little bit.

[donmarkoni]

New King of efficiency

80.9820 GFlops / 4.01984 GHz = 20.15 GFlops/GHz

4.02GHz (167x24) 2010MHz 7-7-7-20 (12x) 3850MHz Uncore (23x)
HT On, CIE & EIST Off. CPU 50% LLC, QPI LLC Off

VCore 1.193v (load_real)
VTT 1.289v (load_real)
DRAM 1.629v (load_real)

http://img545.imageshack.us/img545/8...5402ghzcl7.jpg

Was actually just trying to stabilize a 4GHz daily 24/7 OC and realized efficiency was really good. BTW how's the high perf club idea coming along Zoson?

Congratulation! The crown is yours. But don't get too comfortable...

4ghz runs will not be considered for efficiency in the hpc thread.
4.5ghz is the cutoff as previously discussed.

I'm super busy with work right now, but expecting it to calm down once our new release matures a little bit.

Until you open up HPC, we need to stay in shape, but no need for taking it to the max while training. Waiting for the competition.

Take your time, work must come first. Well, that's what pays for toys.

[Cookiesowns]

Yeah that's pretty much what I've read. Do you have a ballpark bandwidth number that the R3E marvell 9128 can achieve in sata3 with only one high performance sata3 ssd drive?

Honestly, if you don't need more than 6 SATAIII ports i'd just put them on the ICH10r because of Intel RST, very useful and neat utility. The marvell will have higher sequential, but the random read writes will be a tad slower than the ICH10R.

It's totally up to you, for me I'm running two Agility 3's on the ICH10r. But just because of USB3, efficiency, and SATAIII I'll be switching over to Z68 / X79.

[Kosmic71]

Ok thanks man.

[Mungri]

Ok, this is the most 'efficient' I can get mine running:



I didnt test the gigaflops yet, just the stability. It will do 4 Ghz on stock voltage of just 1.21v, but just going up to 4.05 Ghz requires around 2.625v, so the perf / volts cuts off and becomes very inefficient over 4 Ghz on my chip.

The voltages might go slightly lower, but the ram is pushed to its absolute limit and wont be stable at less than 1.65v @ 1900 Mhz, just the CPU and QPI might work a bit lower.

I was only pushing it to the extreme when I first got it to test its capability, but this will be my 24/7 setting.

My GTX 560 tis run at 880 / 4600 at the minimum voltage of 0.950v, so everything is nice and cool for daily use, and then I go mad when running benchies :p

[zoson]

bhavv,
you haven't tested any stability there.
run linx for 20 loops with all mem, and then you can call something stable...
Prime doesn't prove anything any more and superpi mod is a performance test, not stability.