Printable View
Thank you. I figured it out a few minutes after I posted :D I was able to pass the 5 Ghz mark indeed. However I need a bit more Vcore to get the cpu stable past ~4.95 Ghz.Quote:
Originally Posted by abo333
Did you do the programming on this? If so nice work, it works like a charm, however I'm getting som odd sisandra results. They don't increase at all, but 3dmark and superpi increases almost linearly with speed.
Hey guy's this is Apop over at Legit Reviews... Been testing on the same exact setup and found some weird things going on with the 1:2 or 1066Mhz memory divider. It doesn't seem to be working correctly. I have a thread going on here: http://forums.legitreviews.com/about1931.html that goes into further detail on the issue. I'm looking forward to what you guys have seen and can if you can confirm what I have been seeing during testing.
Awesome overclocking with the 3:5 divider... some major bandwidth! :toast:
Hi Apoptosis,Quote:
Originally Posted by Apoptosis
Unfortunately I have to disagree with you :D
Atleast with retail boards it works perfectly.
Sandra Unbuffered with 3:5 divider
Sandra Unbuffered with 1:2 divider
Asus P5WD2 Premium "54MG0" with 0408 bios.
So it is too early to say if those CPU-Z H.O.F scores are invalid.
Well I can't run 4-4-4 timings at 1066 with my memory and unmodded board so I ran 265FSB @ 1:2 and 3:5 memory ratios with Sandra 2005 SR1 on buffered memory testing and got the following results.
Results with the 1:2 divider (6700mb/sec)
Results with the 3:5 divider (6600mb/sec)
I'm also on a retail ASUS P5WD2 Premium Motherboard with the 0408 BIOS.
What are your thoughts? I only ran the tests once, so these are well within the margin of error to be identical results.
Can someone please send me the cpuz 1.28.7 please? :)
Thanks
Can you run the unbuffered test please?Quote:
Originally Posted by Apoptosis
400MB per sec difference in unbuffered certainly doesn´t fit in the error margin.
Like The Stilt already mentioned you really should try sandra unbuffered and/or some benchmarks that actually measure something that has to do with actual performance. Sandra Buffered is one of the worst benchmarks there is.Quote:
Been testing on the same exact setup and found some weird things going on with the 1:2 or 1066Mhz memory divider. It doesn't seem to be working correctly.
Its well known that Sandra buffered is very dependant on system FSB on P4 rigs and thats the reason you dont see much of a difference between different dividers - Your FSB is the same!
Try unbuffered or superpi or 3dmark2001 lobby hi and you should see a difference between the dividers.
The real world performance gain might not be too big thou as the system FSB might be the limiting factor there.
Let me run a few more CPU's & benchmarks and see what happens... either way it's a good problem solver...
well i ran some more benchmarks and it looks like the difference between buffered and read scores are pretty much nill between 900 and 1066 memory clocks. It also looks like the dividers are working right and that everest is showing the wrong values. Sad that the only time you'll see the improvment on these overclocks is on certain benchmarks or when testing unbuffered and write tests.
This corsair 5400UL is some pretty amazing stuff. I'm happy at 5-3-3-8 timings at 1051MHz as it's getting Sandra buffered scores of 8,000MB/s and unbuffered scores of 5,000MB/s
Stilt you got some crazy fast Corsair modules on your hands! Congrats and keep those results coming! :woot:
The Stilt, i've looked my bios on my IC7 but there is no option for running the mem higher than the cpu... is there any special option m8?
Cheers,
B
Anyone tried Memtest at 320FSB with 3:5 ratio (1.06GHz 4-3-3-4 for ram)?
Memtest v1.55 reports following:
L1 Cache: 31375MB/s
L2 Cache: 24383MB/s
Memory: 5007MB/s :ROTF:
EDIT:
330FSB with 3:5 (1.1GHz 4-3-3-4 ram)
L1 Cache: 32353MB/s
L2 Cache: 25144MB/s
Memory: 5163MB/s
...
Select 533PSB NB Strap and it will appear.Quote:
Originally Posted by Byron
Thou there is no way that unmodded IC7 would handle it.
Damn, that's like A LOT af bandwidth there.... :p:
:D
Corsair 5400UL @ DDR1142 4-3-3-1-2 - SuperPI 1M (Verified)
Apoptosis:
I think that 5086MB/s unbuffered score in your article is a bit weird.
Did you deselect all eight boxes in the option menu of Sandra?
I can hit only ~ 4900MB/s with identical setup. I have higher clocks and tighter latencies than you did (320FSB, DDR1070 4-3-3-1).
Damn, this DDR2 "war" with Amin Soltani is getting tight :D
Currently we are neck by neck in frequency, but I can´t do any better from this: Corsair 5400UL @ DDR1170 5-3-3-1-2
Getting that far took so high VDD, that I don´t even dare to say it loud ;)
Maybe it is time to move for Corsair 6400 stuff, and see how it does.
DDR1300 - DDR1400 would be quite nice even with CL5 latencies :D
Holly JESUS! 5K memory... drool... that is some serious data transfer m8 :DQuote:
Originally Posted by The Stilt
Me with 875 chipset and turbo at 261fsb's is 3.8K exactly with DDR1 though.
Nice overclock, with nice timings! :toast:Quote:
Originally Posted by The Stilt
I see You got the 1. place again!
I won't try to clock my Centon any further. Ill wait untill I receive my new PC5300 modules from Centon, hope they can do better, with better timings also.
Heya stilt, I've been watching this thread so closely =)...any chance you can post the result for the MSI P4N as well?
I am a bit worried about a post I read before that C19 doesn't fully support dual core though (the system will run fine, just that only one core will be recognized, or something like that).
And btw, nice OC. If you can have the chance to test the 955x vs C19, do advise which one to get =). I am leaning over to C19 but Mesce's experience somewhat dishearten me but I'd really like SLI. And last two questions, what is the highest vcore, vchipset, vdimm on unmodded board does that mobo is capable of? does it need vdroop mod too?
I just cancelled my MSI P4N board and ordered P5ND2 SLI-DLX instead ;)Quote:
Originally Posted by death metal
Too bad that Asus still use 3 stage cpu power cirquit in P5ND2 :slapass:
No way it will handle dual core if it´s having hard time with single 5GHz Prescott...
I´ll try to arrange something little, concerning that NF4 IE vs. Glenwood thing ;)
EDIT:
Maximum voltages for P4N in v1.0 bios:
Vcore: VID - 1.55V
VDDR: 1.8V - 2.4V
NB Voltage: 1.xx - 2.xx (?)
Thanks for the quick reply =)....I guess I'm going for the Asus 955x then =)...and I hope you don't mind my queries again: can you post the stock vCore, vAGP, vDIMM for the Asus i955 mobo? Do I have to vdroop mod it to stabilize the vcore or can it hold its own on a simple/basic Swiftech Rev3a cooler kit without any droop mod?
Have you tried the Corsair 6400 RAM? I have ordered a Gb of it to go in my P5WD2 board and was wondering how it went and the best timings for it.Quote:
Originally Posted by The Stilt
P5ND2:Quote:
Originally Posted by death metal
Vcore: 0.9375V - 1.700V (+ 0.01V overvoltage option available)
DDR VDD: 1.8V - 2.3V
Chipset Core Voltage: 1.4V - 1.5V
South bridge SATA Voltage: 1.5V - 1.6V
CPU Termination Voltage: 1.25V - 1.4V
For basic use, there is no need for Vdroop mod because Asus has used different regulator chip than in their previous boards. P5WD2 board uses Analog Devices 3181 regulator chip instead of older 3180 chip which have almost removed the droop "problem". I think that P5ND2 uses the same chip as P5WD2.
Wow, thanks for the info The_Stilt =)....I didn't know that revision on the vregulator chip can do such enough wonder, at least for non extreme OCing =)...I'm a bit worried about the Chipset Core Voltage though, 1.5v seems a bit low. I am used to pushing about 1.65 to 1.8v during bench run for older chipsets, even with a 1066Mhz 925xe (I use about 1.7v for DDR750+).
I guess vNB modding is called for this time, I am targetting 285FSB to 295FSB with NB1C cooling on the chipset. I haven't used the CPU termination voltage though (I haven't used any P5AD2-E Premium or Fatal1ty board yet). I wonder why it's such a low setting though, vCore, and vDD both seems "within" my expectations.
And hey, am looking forward for that Asus C19 bench run soon :)....(please, don't confuse me which should I get this time: Asus i955 or Asus C19 *lol*, if you came up with great score on C19, I'll be resetting my mindset again hahaha).
Im watching this thread carefully as well. I have p5nd2-sli and have tested 2 different 5400ul's and the same thing happens. I cannot manage 2:3 or even 1:1 at 4ghz on my 660. The timings of 3-2-2-8 even dont work. I will watch to see how the stilt manages to fix this problem bceause I am all alone in searching for people with p5nd2's
The 0408 bios really did solve a LOT of problems for me. I couldn't get the memory to post at anything above DDR730 before, or the motherboard above 290 Mhz fsb. Now I'm running 300 Mhz fsb and DDR1000 stably.... :D Only at 4-4-4-12 timings but still, no mods and default chipset voltages
I tested the NF4 asus a while back and its not to good, i also hit 5007MB/s in memtest with the new 800EB which uses the same IC's as the corsair.
DDR2 is actually getting quite nice now, the only issue is the cpu's it uses are poor.