Printable View
Ah I see, I wasn't aware that with the Quad CPUs you could lower the multiplier but not increase it, I thought one could simply not change it all. Good idea, although I've already got my Kingston HyperX DDR3 1625 modules! I don't think Kingston will let me return them, outside the 30 day window.Quote:
Originally Posted by Ghostleader
thanks pyro glad to see 901 seems ok , on a side note do you know what these do
PULL-IN OF CHA PH1==========================DISABLED
PULL-IN OF CHA PH2==========================ENABLED
PULL-IN OF CHA PH3==========================DISABLED
PULL-IN OF CHA PH4==========================DISABLED
PULL-IN OF CHA PH5==========================ENABLED
PULL-IN OF CHB PH1==========================DISABLED
PULL-IN OF CHB PH2==========================DISABLED
PULL-IN OF CHB PH3==========================DISABLED
PULL-IN OF CHB PH4==========================DISABLED
PULL-IN OF CHB PH5==========================DISABLED
also does anywhere know where theres a post/review explaining the rampage various bios settings,, thanks
they "pull-in" the phases. :p:
essentially, they tighten the performance level for whichever phase you adjust
ie. i have set a PL of 8, but the channel A phase's, 2 and 5 have been "pulled-in" to level 7
cheers pyro appreciated :), i have the rampage coming tomorrow but already have maximus extreme and the rampage has a lot more bios settings so a bit of a learning curve
Hi Pyr0 :)Quote:
Originally Posted by Pyr0
are you running the two 2x1 kits in your sig as 4x1GB ? Impressive compared to my weak Transcend/OCZ 4x1GB effort.
Any unexpected non-POST behaviour around 400 and 500FSB ? I found different POSTs allowable with 2x1 versus 4x1. Could just be down to badly mismatched kits for 4x1GB though in my case :D
hope it treats you well m8 :up:
i moved to the RE after killing my old p5e3 prem. and i deffo think this one feels like a more solid, stronger board
[edit]
hi cheapseats
yup, i'm running the cellshock and crucial kits in my sig
i've had no problems with POST'ing with these kits since setting skews
before i set the skews, sometimes the kit would lock up in windows, and sometimes i would not be able to POST, let alone get into windows
if i were to change the memory speeds/fsb, the skews would likely need adjusting again.
:shrug: s'all just trial n error and diff kits seem to like diff skews :shrug:
Interesting Pyr0, cheers. Had success here with NB and DRAM skews before with 2x1Gb, but 4x1 was just for fun really.
Using CTL skews I got 86+ stability at 792mhz (but CL8 only) for mine as 4x1GB. so I wasn't that motivated to carry on....:D
WoW nice ... ;) KensekQuote:
Originally Posted by Kensek
Merlin
will try thatQuote:
Originally Posted by Pyr0
but i found out whats the issue
its the dram vtt... that causes the hang
so eventhough 86 says ure mem is stable at that vdimm
u need to balance the mem vtt/nb dram vtt with the vdimm.
i guess thats why this mobo is very fussy about overvolting ure vdimm.
also this is really a good mobo to clock stablely...
its hard and has a lot of options
this would make the transition to i7 easier
one thing though..
nobody has posted linpack stable 2000mhz 2x1gb .. atleast 50 runs around 50 mins..
even for i7...
u'd be surprise on ure gtl's setting how wrong it is..
leave them for now..Quote:
Originally Posted by GLENBOY
its more beneficial if ure running 2x2gb kits..
as often they cannot reach exact tRDs as the 2x1gb kits...
also its more beneficial on high bus rates above 550 for 2x1gb kits
Hi MnM, you're right :) Performance Level = tRD (= memory controller Read Delay).Quote:
Originally Posted by MnM
I wasn't aware of matching tRD values to specific ClockTwister settings, just been working on the basis that both are timing based, and can have similar effects on Read and Latency. Haven't noticed non-allowable POST behaviour created by using any CT (if the CL/tRD/FSB/Strap combination is already POST allowable).
Lowest (allowable) tRD will give best performance, but Clocktwister seems to have diminishing returns beyond Lighter > Moderate.
I've done some more tests to compare Everest results between different Strap/ClockTwister/tRD combinations. Sorting screenshots and will post later.
bro try the nb dram vtt to dram vtt..Quote:
Originally Posted by cheapseats
can lower the ram voltage
to confirm teh dram vtt use memtest hsi...
if bsod instead of error then change dram vtt
the use ibt to determine the nb dram vtt
all assuming ure other voltages are already correct..
interesting stuff... this mobo..
learning a new thing everyday
also been doing this for 2 sticks only
use test 5 without any skewing to determine vNB and vdimm.. run it in loops a few times more than 2
and then restart from default test 0 to confirm test 5...
then only do the stuff above..
memtest86 doesnt prove as a good initial test for 4x1gb... usually use test 5 but it will only show the vnb /vdimm thats accurate for 2 dimms only..
the test above will determine the rest and then only rerun memtest86 a few loops
this solves the weird errors u will get later with test 7/8
two most nb dram vtt and dram vtt A/B that i have been using is
1. nb dram vtt - -5 dram vTT A/B 12.5
2. nb dram vtt - 12.5 dram vTT A/B 25
interesting stuff..
strap 200 fsb 400 1600 CL7 tRD 5 now possible stable.. at vNB at 1.33
i've had no probs with dram/nb vtt. it was def. the skews causing hangs for me. i may have a play l8r tho to see if i can lower any voltages :up:Quote:
Originally Posted by cstkl1
i always use hci memtest, memtest86 takes far too long to find any errors
nah dude that solved the hanging issueQuote:
Originally Posted by Pyr0
now trying to solve the gtl issue
try linpack 50 run for 2x1gb and 15 runs for 4gb
u will get what i mean..
it was a issue that i had running 2x1gb at 2000mhz CL8
memtest HCI pass... Vcore and GTL confirmed..
bun linpack can even run 50 runs
but at the 13th run the results once in a while shows faults...
ure cpu skew/nb skew shows promise + the dram vtt ...
oh yeah even had some weird results with sb 1.1v
but that one still inconclusive..
again a crazy board..
I've tested and thought some more on the 200 strap....
The main issue with tRD and cL combinations in certain FSB ranges, is non-allowable combinations for POST, resulting from MCH minimum timing limitations.
Earlier tests showed that the 200 strap allowed a lower tRD value to be used for a given CL etc, compared to what 266 and 333 straps would allow for POST.
eg: 200 strap allows POST/stability with CL7/tRD5, but 266 and 333 straps only allow minimum tRD6 with CL7.
Regardless of whether it works for DDR3, the existing (DDR2) tRD equation doesn't include a discrete reference to the Strap in use (other than including the divider in use). There would seem no allowance in the equation to predict the differences in POST behaviour, when using 1:2 divider on different straps.
Yet on this board there are differences in Strap/tRD POST allowability, which also seem independent of FSB (in 400-500FSB range I tested).
Despite the 200 strap allowing lower tRD values to POST, I would guess that the minimum allowable MCH timing (for POST) may not actually be lower on the 200 strap.
If lower allowable MCH latency resulted from using the 200 strap, then 200strap/CL8/tRD6 would be faster than 266 or 333 strap with CL8/tRD7.
Instead (I suspect), some other latency is introduced/forced by the 200 strap, which even with a lower tRD value, does not fall below the net minimum time for allowable POST.
If it's something along those lines, a forced increase in overall MCH latency (as suspected of the 200 strap) may allow POST/stability with lower tRD value (than 266 and 333 straps allow), but may not actually be better than neutral in terms of net MCH 'bandwidth' and/or stability, compared to other 'tighter' straps with higher minimum allowable tRD.
According to Everest, the 200 strap as tested gives lower performance, tRD for tRD, than 266 or 333 straps, with similar low performance apparent on the 400 strap too.
This doesn't mean that the 200 strap isn't worth using - it may be more stable - just that any expected benefit of allowing lower tRD values, may not come with it.
slowly piecing it together I hope, but just from incomplete information/knowledge :)
475x9 / 950 / 1:2 / CL8 / CT Strong / tRD6 / 200 strap
http://i447.photobucket.com/albums/q...24_PL6s200.jpg
475x9 / 950 / 1:2 / CL8 / CT Strong / tRD7 / 200 strap
http://i447.photobucket.com/albums/q...24_PL7s200.jpg
475x9 / 950 / 1:2 / CL8 / CT Strong / tRD7 / 266 strap
http://i447.photobucket.com/albums/q...24_PL7s266.jpg
475x9 / 950 / 1:2 / CL8 / CT Strong / tRD7 / 333 strap
http://i447.photobucket.com/albums/q...24_PL7s333.jpg
475x9 / 950 / 1:2 / CL8 / CT Strong / tRD7 / 400 strap
http://i447.photobucket.com/albums/q...24_PL7s400.jpg
All results with 475x9 / 950 / 1:2 / CL8
CT AUTO / tRD6 / 200 strap
http://i447.photobucket.com/albums/q...L6s200_CTA.jpg
CT Lighter / tRD6 / 200 strap
http://i447.photobucket.com/albums/q...6s200_CTLr.jpg
CT Light / tRD6 / 200 strap
http://i447.photobucket.com/albums/q...L6s200_CTL.jpg
CT MODERATE / tRD6 / 200 strap
http://i447.photobucket.com/albums/q...L6s200_CTM.jpg
CT STRONG / tRD6 / 200 strap
http://i447.photobucket.com/albums/q...L6s200_CTS.jpg
CT STRONGER / tRD6 / 200 strap
no result - POST ok but no Windows startup
CT LIGHTER / tRD7 / 200 strap
http://i447.photobucket.com/albums/q...7s200_CTLr.jpg
A big thanks Cheapseats for the informative posts :up:
I had no idea the staps had that effect on performance. I'll try some different combos on mine later.
Does anyone know if the Zalman 9500 LED will fit on this board? I've had a difficult time trying to research if it will or not:
http://zalman.co.kr/DataFile/product...0ALED_m(0).jpg
Also, as a measure to reduce the weight hanging off my board, I'm considering changing my Noctua U12P for a Noctua U9B, it's a whole 200 grams lighter and yes the cooling performance isn't as good, but I'm thinking of upgrading the fan to something slightly faster.
U12P on the left, U9B on the right:
http://icrontic.com/draco/images/art...b_compare1.jpg
[ 0901 ]Quote:
Originally Posted by Kensek
----------------------
Fix if use 2 pieces of HD4870X2 to build CrossfireX it may get no display from the VGA card which is inserted in the PCI-E X16_1 slot
Yes the 9500 will fit - I had one before fitting the TRUE120. Will this mean that your Noctua U6 NB sink will fit using this combo ?.Quote:
Originally Posted by aln688
Thank you cstkl1 and cheapseats for this thing with strap 200.:clap:
I did some tests with 400x10 1600 and I can confirm that with 1600 CAS7 tRD 5 give the same performance as 1600 CAS6 tRD6 but with lower DIMMv and NBv , quite awesome I think.
1600 777 21 tRD5 Strong Strap 200 DIMMv ~1,80 NBv ~1,33
http://i282.photobucket.com/albums/k...RD5StrongS.png
1600 666 18 tRD6 Strong Strap 333 DIMMv ~2,00 NBv ~1,51
http://i282.photobucket.com/albums/k...RD6StrongS.png
Lower the timing, 1600 666 18 tRD5 Strong Strap 200 DIMMv ~2,00 NBv ~1,35
http://i282.photobucket.com/albums/k...RD5StrongS.png
Not much of a different, conclusion, using strap 200 at 1600 777 21 make it possible to use tRD5 with low DIMMv and NBv.
Lower the tRD, 1600 666 18 tRD4 Strong Strap 200 DIMMv ~2,00 NBv ~1,70
http://i282.photobucket.com/albums/k...RD4StrongS.png
Back to Strap 333, 1600 666 tRD5 Strong Strap 333 DIMMv ~2,00 NBv ~1,70
http://i282.photobucket.com/albums/k...RD5StrongS.png
Not much of a different in bandwidth but in latency and with much higher DIMMv and NBv.
And once again, using strap 200 at 1600 777 21 make it possible to use tRD5 with low DIMMv and NBv and it gives a pretty good performance, quite awesome in my opinion.
Credit to cstkl1 and cheapseats for this.:up:
Absolutely, because the width of the U12P was hitting my U6 Northbridge heatsink, the U9 is significantly narrower, they would both definitely fit. Good point. It's just I was adding up the "extras" being put on this board:Quote:
Originally Posted by rob2k
* 1 x HR-05 125g (Southbridge)
* 1 x HR-05 125g (Northbridge)
* Noctua U12P 760g (CPU)
* Enzotech copper heatsinks 20g (MOSFETs & Chokes)
That's over 1kg of metal hanging off the board. I need to ways to reduce the weight and think the U9 will be perfect, although coupled with a higher CFM fan.
How did your Zalman 9500 perform? I'm considering one of these instead of the Noctua U12P, mainly due to its reduced weight. I also see the Xigmatek S-1283 is around 600g with fan.
The Zalman is something I would consider, except it hangs off the top part of the board. On the compatibility chart they specifically ask users to make sure there is a gap between the top of the board and the power supply.
This was in the early days of my RE ownership. I was struggling to get more than 420FSB and getting very frustrated at my OCZ memory not running at its rated speed so I went for the TRUE120 for a no compromise CPU cooler. From memory the Zalman ran 10 degrees hotter under load. The TRUE120 is HUGE and I had to remove a side fan from my case at the time which was a C/M Stacker 832 to fit it inside. Of course there were other issues, early bios, the huge number of bios options and N/B temps that also hampered me. When I moved to water I found the board to have warped 3-4mm when sat in its free state on the table although when fitted in my new case it 'bolted up' flat and there are no long term ill effects. Of course I never measured the mobo flatness when it was new so it may not have been perfect then and may not have warped at all - how is your new board ?.Quote:
Originally Posted by aln688
Here is a useful comparison review of the Noctua U12 Vs the Noctua U9 :-
http://www.silentpcreview.com/article295-page1.html
and a more general H/S summary table :-
http://www.silentpcreview.com/article30-page1.html
HTH
I decided on this combination, which is a good balance of performance to weight:Quote:
Originally Posted by rob2k
Southbridge heatsink, Thermalright HR-05 SLI/IFX:
http://www.thermalright.com/new_a_pa...05-IFX-SLI.gif
Northbridge heatsink, Thermalright HR-05 IFX:
http://www.thermalright.com/new_a_pa.../HR-05-IFX.gif
Northbridge heatsink fan, Sanyo Denki 80x80x25mm (Part #109R0812M4011):
http://store.thermalfx.com/merchant2...R0812M4011.jpg
CPU heatsink, Noctua NH-U9B:
http://www.xbitlabs.com/images/coole...an/17_n9_p.jpg
CPU heatsink fan, Sanyo Denki 92x92x25mm (Part #109P0912F4011):
http://store.thermalfx.com/merchant2...P0912F4011.JPG
This combination is just over 200g lighter than keeping the Noctua U12P. I already returned the Noctua U6 but can get my hands on a HR-05 IFX instead.