It really is starting to look like Newcastles are the way to go at the mo.
Printable View
It really is starting to look like Newcastles are the way to go at the mo.
I'm going to put my winnie on LS this evening with neo2. Pray to the overclock Gods for me. I am going to Jinu's apt. tonight to try my cpu out on his neo2 and LS, just to make sure my board is the reason I am having so many problems. Hopefully the gods will be in a good mood and the overclock will be prosperous.
if only OcUK were on time id have a cpu right now as ebuyer came to pick up the other chip alreadyQuote:
Originally Posted by Brum Man
cpu less :stick:
10x258MHz on the CPU side running memory 1:1 at 2-2-2-10 with 3.51VDIMM seems to be my absolute maximum with this 3500+ processor. I still plan to experiment with all BIOS files from 1.3 Official to 1.51 Beta. If I do not gain a single MHz with any of the other BIOS releases, I'll conclude that the above configuration is final and will then commence testing to discover the lowest voltages I can run for CPU and memory while remaining stable and error-free.
The Bottom Line: I'm a happy camper with my Newcastle 3500+. :D
http://webpages.charter.net/conrad-d...258pitests.png
Conrad, are you online bro?
Let me know within 30 mins if you are and I will email a screeny to ya if that's cool... you'll love this one ;)
PS, that 1MB SuperPI is for that clock... I havent kept up with all this but seems the DualChannel really makes a diff in SuperPI. With a loose 10x260 I get 33s and with a tight 10x256 I sometimes get 33s-34s.
Note to ALL:
I have found that when I had my Trc-Trfc @ 15/17 I could clock higher with just 3.2v... eg, 256-257 was the limit... now that I have tightened those timings to 11-15 and also reduced a few others I am having to run 3.3v at 10x256. Still great memory, but keep that in mind when "pushing for max"... Trc & Trfc are biggies ;)
Very nice oc there conrad :thumbsup:Quote:
Originally Posted by conrad.maranan
Is that prime stable? And what bios was that oc achieved with?
Im going to try the new beta bioses aswell
Very good observation on the TRC/TRFC. You just beat me to posting the same thing. On My DFI NF2 board The default is 13-15. I usually lower it to 12-14 as that played nice with my BH-6 and also gained me slightly higher 3dmark01 scores. Earlier today in desperation I figured I would raise those two values back to default (13-15) as it was about the only thing left to try before I gave up trying to get any higher on this memory. After raising those I was able to go one mhz higher on my fsb with fewer errors than before. Not a massive difference as you have also noticed but still something to keep in mind.Quote:
Originally Posted by Rabbi_NZ
Orker - I posted a screenie earlier. Go back a few posts and you'll see 9 hours 14 minutes 13 seconds of Blend Torture Test Priority 10. :D BIOS is 1.36 Beta.
Well. I seemed to have isolated the problem and I just might be able to find my missing megahertz...all 265 of them...and hopefully more. My problem is heat - a lot of heat. I removed the 92mm Sunon that was cooling my DDR Booster and VX modules and replaced it with a 120mm Sanyo Denki DC San Ace (Model 109E1212H101). I also cut some heavy duty cardboard and made a wall to channel all of the air solely on the DIMM region. What did I gain? So far, 3MHz. :D
http://webpages.charter.net/conrad-d.../261sandra.png
http://webpages.charter.net/conrad-d...261everest.png
http://webpages.charter.net/conrad-d...261pitests.png
I plugged my CompuNurse Digital Thermometer to the edge of my CPU. The temperature while running Memtest and Super PI was a killer 68C!!! If I still had my Thermalright SLK-948U to play with, I'm positive I can keep the CPU and memory controller cool enough to push for more.
Time to test 10x263MHz.
:toast:
Congrats Conrad... 263 is a REAL nice clock dude. Try setting Trc=15 Trfc=17... you will lose a little bit of performance but you may gain 2MHz-4MHz ;)
Take care of the CPU too... looks like a keeper :toast:
And biggy thanks for hosting too bro, I will post my screenies in a sec
The above shots are 10x261MHz. I'll test 10x263MHz after I grab a bite to eat. I'm starving from all of this testing. :slobber:
Awesome news! I suspect heat is an issue with lots of people running this memory. Even at 3.3v, with a strong enough fan this stuff gets quite warm to the touch. Even just putting your hand in the general area of the dimms has you feeling the heat that is being radiated from the sticks. I can only imagine 3.6v with a booster that is known to get quite hot to begin with.
These are with 3.4v set in BIOS but as soon as ANY sort of memory load comes into play it quickly drops to 3.31v-3.35v and jumps round there... so effectively it's only 3.31v seeing as the weakest point is what counts.
9x270 2-2-2-10-1T 1x512MB OCZ VX - Sisoft Sandra Memory Bandwidth
9x270 2-2-2-10-1T 1x512MB OCZ VX - 3Dmark01
This one can sit at windows all day but benchies die as soon as load is put on the memory... I NEEEEED a 3.3v rail mod but no way in heill do I want to kill this PSU... we'll see... hehehehe
3.4v BIOS (3.31v under load)
http://webpages.charter.net/conrad-d...-2-2-10-1T.JPG
Good job, Alex. What's your maximum Super PI 32M stable with just one stick? I think I'll start playing with single channel after I get a result with Hexus Pifast and complete a full Super PI marathon.
http://webpages.charter.net/conrad-d.../263sandra.png
http://webpages.charter.net/conrad-d...263everest.png
If I can get through Super PI 16M and 32M at 263MHz, it'll be a miracle. Memtest was at 70C while I was testing memory at this speed. Even though 28 loops ran without error, I think I've reached the absolute maximum heat my CPU will tolerate and an increase of just 1C will probably cause Super PI to error out. There's only one way to find out. ;)
CPU temperature fluctuated between 71C and 72C. Amazingly, I pulled off the Super PI and Hexus Pifast runs, but I'm not going to be running with these temperatures on a daily basis. I'll resume dual channel testing again when I receive my phase change creation from chilly1.
http://webpages.charter.net/conrad-d...263pitests.png
fook me your doing well conrad :)
Off topic mode on but still related to A64's------
one thing i have noticed with A64's tho is a wierd 30 second delay in windows for all the drivers to laod, some open straight away but others lag, after 30 seconds it will be fine
it did this on the 3200winchester and 3500 newcastle i had before (tried reinsatlling XP, same problem), it appears the chip is fine when you first stick it in but after a day or this delay kicks in (cpu is 0% for this 30seconds btw)
The chip i stuck in today, my new 3500 turned up (NC) doesnt have this pause at all and is super silky, im guessing i may have damaged the mem controllers with too much voltage on the others, as i have a tendancy to chuck volts in them within the first few mins, not letting them burn in long enough to get a rough idea of how they clock??
With this cpu im not going over 1.65v for a few days, to see if this silky performance remains and i dont get the dreaded 30 second wait!!
Off topic mode off ------
edit: this has only been on 939 with the same msi neo 2 btw
guys y havent none of u who have the vx mem tried 2.5,3,3,6 timings to see how far they ll go???? i d interested to see the max fsb at these timings! Cheers :D
i think its been said that it prob wont scale much higher with slacker timings but i guess the only way to find out is to test themQuote:
Originally Posted by Mtemel
let the testing begin! ;)
Noone on intel??? :(
10x263MHz is Prime Torture Test stable!!! :D
http://webpages.charter.net/conrad-d...X/263prime.png
conrad, what temps are you getting and what are you using to cool your system?
71C on the CPU side under Prime Torture Test. I'm using the AMD retail-boxed HSF. The VX modules and DDR Booster should be pretty cool since I employed a 120mm fan along with a wind tunnel to isolate the air in the DIMM region.
I'm not going to run my CPU at 10x263MHz because the temps are so high. I just wanted to see if 2630MHz would still be stable with insane heat.
EDIT:
I just noticed that Memtest recognizes my Newcastle as an .09 process chip.
very well done. I think you have one of the best pairs of those sticks that I've seen. What volts did you need at 263 and have you been doing a lot of burning in to get that stable?
3.51VDIMM measured with a digital multimeter. Any more voltage than this will cause errors. I'm pretty sure there's a little bit more to be gained from the VX modules, but at this point, the CPU is running way too hot. I've seen 265MHz 2-2-2-10 on my 3000+ Winchester, so I know that the VX definitely has potential. I've been burning these modules with high voltage since the day I received them. And I was among the first batch of people to order them from excaliberPC on November 30th.
I'm now tempted to purchase a Thermalright SLK-948U to mate with a 92mm Vantec Tornado that I have in my storage bin. But since I'll be moving up to phase change, it might be a better idea to just wait. I'm fairly unpredictable and react on impulse when it comes to purchases and acquisitions so you never know. ;)
I have a feeling my RAM will do 270MHz 2-2-2-10 in dual channel with phase change. We'll wait another month and see.
EDIT:
Mtemel and BAcon - I have tested with slacker timings. I get errors the moment I change a value of 2 to 2.5, 3, or 4. I only tested with high voltage but I'm fairly confident that if I were to drop the VDIMM a few notches, I can run 2.5-3-3-10. I'll test later when I have a chance. I'm also going to test the following to see what performs better at DDR400 using 2-2-2-10 in 1:1 dual channel.
11x200MHz | 2-2-2-10 | OCZ VX
11x200MHz | 2-2-2-10 | OCZ 3200 Platinum Deuce