:eek::shocked:Quote:
Originally Posted by Solarfall
I've been testing it for a day and no fuss so far. Might flash back to 0601 since I know it works and I found ZERO improvement.
Printable View
:eek::shocked:Quote:
Originally Posted by Solarfall
I've been testing it for a day and no fuss so far. Might flash back to 0601 since I know it works and I found ZERO improvement.
like i said im not sure if it was the beta bios.. i was benching my E8600 at 635fsb and the system crashed. i tryed to boot back to windows with 590 and i kept on getting a bluescreen.. temps and all were fine on the board. its pretty mystical what caused it.Quote:
Originally Posted by Think
Damn...well it could of been worse. Keep us posted.Quote:
Originally Posted by Solarfall
Quote:
Originally Posted by Think
I linked to these at OCZ before and maybe you've seen them anyway. There's some useful general info on the various skews.
OCZTony's "Intel FSB Over Clocking, clock skew and the effects it has on system memory"
http://www.ocztechnologyforum.com/fo...ad.php?t=40746
and a bit more here
"Taming the Asus P5E3 Premium" also by Tony.
http://www.ocztechnologyforum.com/fo...ad.php?t=41042
I notice that Tony has an RE now, and has worked through some (OCZ) 4GB kits.
"Rampage Extreme tweaking and tips"
http://www.ocztechnologyforum.com/fo...ad.php?t=44088
Not that much extra info yet but could be worth checking out for general tips as time goes on.
As Tony says in the "Intel FSB Over Clocking.." guide, unless you have the kit "to read the clocks and dial them in", any skews you might need are just down to you to find, as they'll be specific to your individual kit/clocks etc.
If your really stuck you could give this a go -
http://www.xtremesystems.org/FORUMS/...postcount=1128
still works well for me (CTL skews), though there might be more elegant way.
Just a word on the "Memtest86+ method": I only used Test 3 to speed up the trial and error process - the other Tests may or may not give better/more meaningful results in finding the 'best' skews, or for CMD/CLK skews which I haven't used.
Using Memtest86+ like this isn't about testing memory for final stability - as others have said, HCI should be better for that. However, it might take forever to find DRAM skews with HCI, and you can't really use low VDIMM in the same way.
Having said that, I've found that 86+ can find plenty it doesn't like even when HCI passes 1000+%. This could be down to failures in memory that HCI doesn't cover, or some other system instability which the HCI loading doesn't expose. I'm still using 86+ to double check after successful Blend and HCI testing.
:)
-----------------------------------
Finally, I do have some fairly unexciting results to post shortly for the Transcends, which I think confirm Stelios' earlier advice of Samsung IC.
Sorry to hear that...as now I might wait a bit to flash 701 myself:(. Still thinking of backing down to 501. What kind of disk and were you in RAID or Straight up?Quote:
Originally Posted by Solarfall
Hopefully you have a couple spares lying around.;)
yep thankfully i have another raptor.. i had the one that died in a straight up set up.. :(Quote:
Originally Posted by Expat GriZ
Any word on the changelog for 0701? Is it better even for dualcore cpu's than 0403?
Thank you very much!Quote:
Originally Posted by cheapseats
:toast:
If you have mounted a GTZ on the RE, what did you do for the backplate? I modded my TRUE backplate to miss the chip but the GTZ plate looks a bit flimsy to be cutting a hole in the middle of it. I thought about using the TRUE's plate but the standoffs a much higher and I am afraid you won't get enough pressure on the block before the bolts bottom out.
I have a GTZ fitted and used one of these :-Quote:
Originally Posted by Pntgrd
http://www.scan.co.uk/Product.aspx?W...ing+the+cooler
The threaded portion is a very similar design to what comes with the GTZ.
I also changed to EK barbs by swaping the O ring from the Swiftech barbs.
My Common Performance Level is set to 6 for A1/A2/B1/B2 by having it set to AUTO.
If I want to set this manually to improve stability (reduce performance) would this become a 5 or 7 ?.
Apologies if I have missed this in the bios help. Cheers.
Lower PL = Tighter internal chipset timings = Less system stability at a given VMCH/VDIMM voltageQuote:
Originally Posted by rob2k
Reverse is also true. :up:
Usually with a FSB of 450-480 and Memory CL 7 a PL of 6 gives the best performance and stability. With a CL of 8 you need to raise PL to 7 or system might not boot.
Thanks EF - I understand how the scale works now :up: :up:
Froze last night as HyperPi was calculating the summary - I will leave PL alone and continue to tweak CPU & NB GTLs. Might even try the PCIE 101 trick then look at some voltages if i'm not happy.
Well, it's taken longer than the OCZ to get anywhere stably above rated with these sticks, though this seems more down to some board/BIOS/settings problem than the memory itself.
XMP
----
Due to my unfounded concern over first time POSTability, I reset (0501) BIOS to default, so all on AUTO except VPLL, VTT and VSB which were set to min.
900mhz / cl8 is rated at 1.80V in the XMP. I didn't select XMP just AUTO but it POSTed at 450 x 7.5 with 1:2 straight from BIOS defaults. This gave rated speed for the memory straight away, with the nearest approximation of E8600 at stock. Memtested OK, so not bad for first POST with the kit. Never realised that XMP set the whole board up!
Obviously to be 'guaranteed' at 1.80V the XMP would need to include some extra VDIMM allowance so I worked down the VDIMM and found they only need 1.617V set in BIOS for 3000% HCI stability.
http://i447.photobucket.com/albums/q...3000_AXE_1.jpg
and with the same settings, more of a workout for the CPU -
http://i447.photobucket.com/albums/q...OB12_AXE_1.jpg
BIOS
----
Quite quickly ran into instability above 450fsb on 0501 and 0601. Went back to 0403 which seemed a fair bit better for memory at 500fsb. All results here are with 0403.
Clocking and Latency
--------------------
I was hoping for a kit that would do 950mhz / CL7 with reasonble VDIMM. 475FSB with CL7 / PL6, gives better memory performance than 500 / CL8 / PL7, and at lower board voltages.
CL7 is a non-starter with this kit, which is not stable at 900mhz/CL7 even up to 1.95V + skews. With no decent CL7 clocks, pushing upwards from 900mhz at CL8 is the only option.
Highest clock I've reached with 32M stability is 1030mhz / CL8 with 515FSB and 1.97V. As you can see from the 1030mhz screenshot, Windows can't find the GFX, so nasty things are happening during boot. Still passed 32M though.
http://i447.photobucket.com/albums/q...SP32_AXE_1.jpg
VDIMM
------
Although the very low VDIMM is nice for the rated speed, it doesn't translate into much in the way of clocking/voltage headroom.
So far the max VDIMM these modules can handle without giving uncontrollable errors is about 1.97V (set). This should still give 0.35V of usable voltage response from 900mhz/1.62V.
It gets used up quite quickly above 900mhz, so that by 950mhz about 1.85-1.90V is needed. The voltage demand then seems to ease off so that 1.95V allows 1000mhz/CL8 with 2000% HCI stability.
http://i447.photobucket.com/albums/q...2000_AXE_1.jpg
At 'higher' clock speeds (950+) VDIMM also needs to be kept within a fairly narrow margin. Overvolting, even a couple of notches past what's needed for a given clock, will give errors in HCI and 86+.
For now 1010mhz/CL8 @ 1.97V is probably about the limit for HCI stability, though this may improve.
VNB and NBGTL
---------------
Quite pleased with 1.37V for VNB at 900mhz/CL8/1N/PL7, since 1.41 was needed for the same (but at 2N) with the OCZ3P18002GK.
As with VDIMM, the VNB requirement shoots up above 450FSB/900mhz, reaching 1.56 - 1.60V for 500FSB/1000mhz/CL8/PL7. The HCI 2000% pass required 1.56V at 500fsb, though this went up to 1.59-1.60V for 2-4 hour Blend. This is the same with any CPU multiplier. Unsurprisingly, OCCT 2hour mix passes at VNB 1.56V.
Was hoping that raising VTT and testing for a VTT/NBGTL sweetspot, might allow a drop in VNB. So far used up to 1.38V VTT and tested dozens of VTT/GTL/VNB/VCORE combinations.
No chance yet for lowering VNB stably, though I found NBGTL at -5mV for up to 1.30VTT and -10mV for 1.30-1.38V, worked best for me across the VNB range I've used. For example, the difference here between NBGTL on AUTO or -10mV with 1.38VTT/1.60VNB was no Windows versus 2+ hour Prime blend. Positive values for NBGTL also worked up to a point, but Large FFT failures during Blend became less predictable timewise.
As a result, my 500x9/1000/CL8/PL7 settings are the best compromise I've found so far that is 2-4hour Blend 'stable', but no more.
http://i447.photobucket.com/albums/q...SP32_AXE_1.jpg
Unless my high VNB for these clocks is just an ugly and unavoidable side effect of the Transcend kit with the available BIOSes, I think that a more stable combination of VNB/VTT/VCORE/PLL/GTLs, with some lower voltage levels, may be achievable with work.
Overall
------
It doesn't appear that the memory itself is causing trouble at 1000mhz but considering the higher voltages needed and limited Blend stability, I'm not happy with 500x9/1000 as a usable clock yet. I would rather be at 475fsb / CL7 / PL6, but it's going to mean another memory kit which is fairly disappointing.
I've tried BIOS 0701 but found that no settings stick with the BIOS, even full AUTO, and always revert to downclocking upon booting, tried both my 2x2GB A-Data and 2x1GB CellShock.
Back to 0403...
Hey Solarfall!! How are you making out after the "crash"??? Are you still trying 701?? Anybody else? I've just flashed from 601 back to 501 and none of my previous settings will work!!!:eek: I'm almost back to auto on everything and stable @ 5 ghz with lower volts all around.:D Will start cranking it up and see what happens. :yepp:Quote:
Originally Posted by Solarfall
I flashed back to 0601. Found 0701 to be unstable with my configuration.
@Think Thanks for the reply...might just stay away from it. As I stated above, just flashed back to 501....everything looks great but now Orthos is giving me hell!!! Back to the ol' drawing board!! :shakes:Quote:
Originally Posted by Think
Flashed 0701 and found no problems so far.
Have a slight impression that with the same settings I used on 0601, I'm having better memory performance.
Folding @ Home Deino SMP (project 2665) Min. Time / Frame improved 3-4 seconds.
Everest memory and cache benchmark didn't change in a noticeable way.
My 2x2GB kit (see sig) is running at 1705MHz, 1T, 1.90v
1st set of timings: 7-7-7-18-4-60-11-5
2nd set of timings: auto
3rd set of timings: 23-7-1-8-8
DSRC: Enabled
DDWC: Enabled
Sweks: auto
AI Clock Twister: strongest :D
AI Transaction Booster: Manual
Common Performance Level: 06
Pull-ins: all disabled (If I try to enable a single one, system locks while saving BIOS settings. Already happend on previous versions)
Other probably "important" settings:
CPU PLL: 1.51v
FSB Termination: 1.28v
NB Voltage: 1.28v
NB DDRVref: +12.5v
DDR3 Channel A Vref: auto
DDR3 Channel B Vref: +37.5v
SB 1.5 Voltage: 1.57v
SB 1.05 Voltage: 1.07v
I'm running my QX9650 @ 3833MHz (9x 426)
Has anybody been able to overclock one of these 2x2GB kits to 1800MHz?
I suppose you're all singing "Dream Dream Dream" while reading... :ROTF:
I'm seeking 24/7 stability and no matter how relaxed I set my memory timings and other settings, I can't make those wonderful Samsung IC based DIMM's fly faster than now.
My goal is FSB450, 2:1 divider.
Anyway I'm pretty happy with the results I've achieved, taking into consideration that it is a 2x2GB and not a 2x1GB kit.
I was just curious if there is some juice left in these DIMM, that requires some very nasty settings which I don't master.
I'd really appreciate some feedback on this. :up:
Thanks for all the useful posts in this (and other) thread(s).
I've been learning a lot with you people. :clap:
Quote:
Originally Posted by Expat GriZ
Haven't tried 0701 yet but I've had this a couple of times after flashing BIOS 1 to 0206 and 0403 - couldn't save any settings or AUTO. Had to use BIOS 2 (still as shipped) to restore BIOS 1.Quote:
Originally Posted by eternal_fantasy
Also a couple of wipeouts during Prime with 0403 - random restart and CMOS reset / cleared completely.
Thought it was probably bad flashes or maybe a flaky BIOS chip :shrug:
Hmm have flashed onto the second BIOS chip many times with other BIOS with no such problem.. although it may be a bad flash... anyway I think I'll let the BIOS mature a little with xx03-xx04 versions. :)Quote:
Originally Posted by cheapseats
Well, I'm starting to like this 501.:D Back stable @ 5 ghz, mem 2000 mhz 1N after 1 1/2 hours of Orthos, etc. I don't have to have the skews all over the place as opposed to 601 & volts are lower all around. Will give it a bash tomorrow as 1:15 am here and I'm whooped!
loljust noticed:rofl:Quote:
Lots of very expensive, soon to be obsolete parts!!
Well for a laugh I've just written to Asus requesting better compatibility with OCZ ram in their future bioses. I'm sure if I get response it will be along the lines of "Of course sir, we will get right onto it" joke.
I've referred to http://www.bit-tech.net/hardware/200...mory-roundup/8
and
http://www.ocztechnologyforum.com/fo...ad.php?t=44088
Also taking into account my own experiences and the fact I could consistently bench my ram on the 790i at 1940mhz 8-7-6 1T timings.
Albeit I can't confirm it's a ram issue, I'm really running into a brick wall with board for any decent overclocking and 3d benching. Specifically the likes of 3dmark 06 and 05.
Would be nice to have the sort of bios updates that I believe the Premium had
Real fun'n'games.
RLM
Desperately trying to learn about CPU GTL settings, read lots of information although some is very technical and others conflicting. Read that the data bus values should be higher than the address bus values so going to try these tonight.
The other thing I have read refers to multipliers, 0.63x should be set to AUTO on a quad with the calculated difference to GTL ref V set in mv to the 0.67x multipliers.
Does anyone know which GTLs refer to 0.63 and 0.67 multipliers ?.