Running at 1806 and being fast, 2 totally seperate things..........;)Quote:
Originally Posted by BeepBeep2
Printable View
Running at 1806 and being fast, 2 totally seperate things..........;)Quote:
Originally Posted by BeepBeep2
trfc was 110ns, I've got trc at 41 atm though ;)Quote:
Originally Posted by chew*
Gonna switch to a different install bout an hour from now.
I still dont know most tweaks though :doh:
Gonna give you a small piece of advice that to be quite honest someone has needed to give you as of late.Quote:
Originally Posted by BeepBeep2
Less chat more splat, nicest way I can put it.
This is try four (going on right now), I crash loop 8 to loop 10. 2.33 to 2.4v, makes no difference, but I think CPUNB is too high (this CPU doesnt like to go over 2800). Core speed doesn't make a difference on if it crashes or not.Quote:
Originally Posted by chew*
Timings are 6-6-6-18-11 1T 110
Points taken, from both threads.
It's just friendly advice, trying to help you out :)Quote:
Originally Posted by BeepBeep2
sorry my bad pre-fatbodies :pQuote:
Originally Posted by Don_Dan
Nomatter what I do superpi will fail at loop 8. Tried lower core speed, lower cpunb speed, even backed ram down to 1750 ish cl6. It's always loop 8 nomatter what combination I try...
Not as easy as it looks is it :p:Quote:
Originally Posted by BeepBeep2
Add in the fact that the CH III is probably the most effecient board ever made for deneb, which as a side effec is very stressfull on ram and it's that much harder.
i still love mine :p i would put the M4A78T-E right next to it thoughQuote:
Originally Posted by chew*
I booted 1820 lol. If it can't get past loop 8 at 1600 C6 then I call shenanigans. :p:Quote:
Originally Posted by chew*
man not that had to get 1600 cas 6 stable did that with D9's and HCH9Quote:
Originally Posted by BeepBeep2
M4A78T_E and M4A79_T clocked ram easy, usually thats a sign that the board is much less effecient.Quote:
Originally Posted by cdawall
Even the CH4E is a very very loose board, it can't hold a candle to CH3F when the cpu in question is deneb.
To let the CH4E fly you must use thuban and have ram clocked to the moon.
This paints the story very well. The deneb on CH3F has a slight NB disadvantage.
2 totally different paths to acchieve the same goals.
Deneb
http://www.hwbot.org/result.do?resul...&title=Highest score of chew*
Thuban
http://www.hwbot.org/result.do?resul...&title=Highest score of chew*
1680 failed loop 6 :d
oh i know it did doesn't mean i did not like the board ;) :rofl:Quote:
Originally Posted by chew*
2.02v was enough for 1600MHz 6-6-6-18 with my 965, however I don't want to do it again now as a birdy has told me that Thuban doesn't like that kind of voltage :p:
got to turn extreme overvolt on for 2.8v
EOV on 2.8v
http://img.techpowerup.org/110226/IMG_3503.jpg
Apparently the newer denebs are fragile to, its got to be the low K dielectic.Quote:
Originally Posted by [XC] Oj101
I killed a newer week 49 560 with d9 volts, quick very quick while testing.
The 965 I used is a very old chip.
Yes, I haven't tested it with high volts myself, but I noted that posibility, due to the Thuban-like behaviour (except the DDR-2000 capability, at least mine chips).
Wonder how is the efficiency, compared to old Deneb.
Same effec, all it changed was scaling characteristics.Quote:
Originally Posted by I.nfraR.ed
I would not expect effeciency to change or clock walls unless they had a microcode update, it's probably a good thing they didn't.
On average NB clocks are 200-300 higher, on average cpu clocks are 200-300 higher.
Mine was an old C2, a very nice clocker (4GHz 1m on stock voltage) and handled as much voltage as I could throw at the RAM (I think I stopped at around 2.45v, it was still scaling past 2.4v but only slightly, is that GTR or GTS?). I never manager 1800MHz 32m stable though, nice RAM :up:Quote:
Originally Posted by chew*
Now, I've never been very good at 32M pi. I popped my hypers in my UD5P today to give it a shot... Holy hell. Chew*, you literally are the master. The chip I've got is pretty useless for this competition on air though. Cpu-nb tops at around 3000 on air, but gimmeh some DIce and up to 3600 she goes. ;) On these hypers, 1800+ cas 6-6-6-16-18-1T is easysauce. Just couldn't get anywhere on the cpu-nb. :shrug: Pretty bad efficiency since I was in diagnostic mode on my 24/7 OS, but my best was still 30s+ too slow. :-/
I'm no master chance, I'm just a regular guy that puts alot of time in to figure out how to get the most performance out of a platform.
Doing the homework is the hard and tedious part and many times means pushing ultimate limits just to find out if something is indeed better or worse.
Once the homework is done it's just a matter of applying it.
I believe chew has realized I'm not a guy of patience but I also never get to work :p:Quote:
Originally Posted by chew*
I realize valids are nothing and stability is everything, but nonetheless these sticks have a lot in them:
DDR3-1900 6-6-6-15 on 890FX (Actual is 1886, ths board has HTT fluctuation)
http://valid.canardpc.com/show_oc.php?id=1680718
http://valid.canardpc.com/cache/banner/1680718.png
Voltages werent anything crazy, I'd almost feel comfortable at these volts over long term.
I would like to believe I'm at troubles with the IMC as far as 32M, hopefully chew can elaborate. Maybe it's just this copy of windows, some bug, I dunno. I couldn't get 32M to pass at 1660 last night.
Like I said in orginal post, suicide is easy, many guys have gotten ram very high, stable and fast however is a totally diff ballgame, I can boot and run a few loops at 1900+, I can get to about loop 23 at same speed as hypers also, Once you go for a "tuned run" you need to be ultra stable however.
Also the method your using to get there, aka 1/4 isn't very efec, much easier to get high clocks that way. 3/10 is a :banana::banana::banana::banana::banana: and hell on ram.
Ok, I think I'm gonna give up...
I regret the decision to skip the dry ice this time. If I had bought some, now I'd had CPU-NB @ 3.6GHz.