I think I'm stable at 465 fsb and 3958 ghz.I will use these settings 24/7 and I try to up the fsb 480..I hate C1 stepping q9550....
http://img128.imageshack.us/img128/5673/47810764ee6.jpg
Printable View
I think I'm stable at 465 fsb and 3958 ghz.I will use these settings 24/7 and I try to up the fsb 480..I hate C1 stepping q9550....
http://img128.imageshack.us/img128/5673/47810764ee6.jpg
I wouldn't be worried the 2 are reading different things, sure it would be nice to see them both the same but the Core temp vid is a variable value where as the CPU-Z vid or vcore is what you are actually running (or near as 4 my money)
The vcore value set in BIOS is out of whack with reality I think, I haven't measured actual vcore on the UT but my E8500 needs the same CPU-Z voltage on my DK which I have tested so I think I'm right in saying its the UT thats at fault.
Having said that the DK & UT set vcore differently UT sets a straight voltage like 1.25v & DK adds voltage to vid so perhaps if I left vcore auto & added voltage via special add the UT's voltages would run closer together....... will test later.
Good luck & don't keep us guessing ;)
You are very welcome :up:
480~485fsb is doable but I had to bump vdimm up a lot to prevent the system locking up!!!
I also had to change my 4870X2 out for my 3870x2 as the 4870x2 kept crashing the video driver after a few hours of Prime!!!!!... very strange perhaps my new 4870x2 is bad or perhaps there is a small incompatibility here :shrug:
http://i106.photobucket.com/albums/m.../435095HGo.jpg
Now I have been priming for the night, this board is harder to tweak than I thought because im not used to intel. What I discovered so far is that it does not boot at 412 fsb no matter what settings. I have seen a test of an early QX9650 that stopped at 413, so u can see where the problem is. Getting an E8500 or something soon. Anyway, here is a screenshot of things, multi keeps going down to 6 when idling, under priming multi was 11 (364x11=4004GHz). Now in Windows at same settings but at 4.1GHz, one of the cores failed right away :( Anything I could tweak to get it stable with no increase of vcore?
Have you considered there may be a fsb hole at 412fsb? I would try jumping the fsb up a bit like 420, 425, 430, 435, 440, 445 etc. I personally have not seen anything worthy of calling a fsb hole on the UT but my DK has a funky spot between 400 & 450fsb depending on the divider & a lot of other settings.
Maybe try upping the fsb from within windows using set fsb to see how far the CPU will go.
The multi drops because you have EIST & or C1E enabled in BIOS, its not a bad thing I know XS members that swear by using it but you got to have a nice stable rig in the 1st place.
As for tweaks your C0 stepping Will likely run different than my E0 but I would try bumping the clock amplitude, CPU Skews, GTL's, try setting GTL Buffer strength to weak etc.
P.S You got to love DFI tweaking :D
I considered the fsb hole, but I only tried going 420, never thought that the hole could be that big. I will try 450+ on fsb, see if there turns out to be an fsb Grand Canyon on this board :rolleyes:
I managed to get vcore down about 0.02 and still maintain prime stability at 4GHz btw.
Right now, sitting in a jacket with open windows and heaters off, will try to push it to some high MHz, what is the safe voltage when running water at about 5*C+/-?
And where is that tea? Im freezing here!
I tried like 450 on fsb, its a no go. Later on i will put out some screenies of cpuz etc, right now i will try to get sub 10sec spi times. Hit a wall at 10.100s now. Would be so much easier with dual core
I did most of the things you said.
When it comes to GTL, i tried that at 0.67 but I gett better results with them arond 0.63 :)
Right now, <<<<<<cpu VTT is at 1.4, is that to high or still to low?
Getting 10.047 in spi now
Edit: 10.015
Edit2: I did it in 9.999 ....... could not be clolser to my goal, lol
If he is only bench running that's super low. Probably above 1.6vtt for bench runs would be considered a bit high.
Been more then proven over this year that no 45nm Quad died because of high VTT's that's an old myth like many others out there, especially with such a good board filtering all the noise before it hits the cpu.
Re: Safe voltages for 45nm CPUs
XS member leefgoofd suggests (somewhere here) that voltage limits for 45nm CPUs, for 24/7 use are: Vcore <= 1.40, Vfsb <= 1.40, Vpll <= 1.60, Vnb <= 1.60.
I think that Vnb = 1.60 is OK for a P35 NB, but I don't know what the max is for a P45, which uses smaller technology and seems to run lower. Suggestions, anyone?
I'm posting here because I just had a PM from fallwind re. the "QX9650 @ 420x10" in his sig. He now has it at 400x10, because after running for a few months at 420x10 with Vcore=1.45, one of the cores degraded. I thought that I'd warn you guys!
I'm very wary of high FSB voltages too. This setting controls power to the transistors that drive the external signal lines from the CPU to the NB, and they have to handle much bigger currents than the transistors inside the CPU. That means that incresing Vfsb generates much more heat. Leefghoofd posted some figures on temperatures of his QX running at the same CPU speed but with different (multilpier x fsb) settings, and the temps were much higher with higher FSB speed. That shows that a big fraction of a CPU's heat is generated by those FSB drivers. Take care of them!
Thank you to all for your advice on safe voltages. I found myself most of the time saying "wow, I am way below that" and on rare occasions saying "ooooops.... "
Well, I promised to put out some screenies of my results I had yesterday, and here they are.
Attachment 89975
CPU-Z screenie shows my max overclock so far, I had about 1.55 on the vcore, remember this was in a cold room. My C0 is so much behind all you guys with E0 quads :down:
Attachment 89977
My final SPi score was below 10seconds, so I reached the goal I set to my self.
And probably one of the most exciting overclocks of them all, breakthrew in my rather low FSB wall!
Attachment 89979
After the room temps have gone up to be comfortable levels and I was about to quit, I tried booting up at fsb 455, and it did. I tried 465, 475, 485, 495 and botted no problem every time. Finally I managed to boot in at 500 as you can see on the screenie, but my memory is not working right and has to go back. It craps the pants anywhere near 1000MHz, so while no problem at fsb 495, I had to slack the timings for 500. I think with better memory (should arrive in a few days :) ), I wil continue climing the fsb wall.
I left it priming at 10x390 for the night, but the stupid Vista is newly installed and not configured yet, rebooted for some updates (says in the log) and interupted my priming session :shakes:
Enjoy!
Now thats what I'm talking about :clap:
FSB Grand canyon :rofl::ROTF:Quote:
After the room temps have gone up to be comfortable levels and I was about to quit, I tried booting up at fsb 455, and it did. I tried 465, 475, 485, 495 and botted no problem every time. Finally I managed to boot in at 500 as you can see on the screenie
Haha, I almost forgot to thank you for the idea of a possible fsb hole there C-N :)
I later tried to boot at 430 and 440 on fsb, and it did. It just was a bit weired, like when I hit F10 to save and reboot, it turned off, but booted in to windows when I turned it on again.
Getting 2x2GB OCZ 8500 tomorrow :) Looking forward to it. So far, I had a lot of bad luck with memory.
2 sets got bad in time, and 2 sets arrived but never performed at the rated speeds. All in one year one after the other. High hopes for tomorrow.
Ye its a bit strange to try to explain thats why I said my DK gets a bit "funky" sometimes it works sometimes it doesn't & the "FSB Grand Canyon" to coin your great phrase moves depending on the way the board is setup, voltages, divider etc all make a difference.
The only 2 fsb's that don't run on my UT are 467 & 487 so I just jump them.
486fsb looks to be the max I can hold prime blend stable before my board/setup locks up. The q9650 chip will small prime 500fsb, it would be interesting to know how you get on.
Based on what I read in this thread, 500+ is not common for this board, right? I hope I will get somewhere with my new memory. Otherwise, it was not just my memory that was stopping me :S
I still have to say, this board is so much fun. Even if you hit a wall at some point, there is aaaalways hope, that u can change some setting and it will go further. I love it!
Edit: My ATI4850 just arrived, so my DFI set up will look better from now on. Was going to run 2 nVidia 7900GTO in SLi yesterday until I realized the obvious. Never mind, just a habit :)
I haven't seen/heard of anyone holding 500fsb prime blend stable with a quad & this board & as I only have the one Q9650 that may or may not be good for 500fsb blend stable I couldn't honestly say. I hope for the boards sake I'm wrong & not to put too much pressure on you but you may be just the man to prove it one way or another with your QX :D:D:D
Some run a mile at a challenge some relish it, the boss at my 1st job used to call it character building & now I'm old enough to realise what he meant I have to agree.
just to comment the FSB grand canyon :
I had P5K-E with Q6600 and it had what I call 'FSB black hole' - from 380 to 470 !!!! So, it is possible
My P35 seems to be playing nicely with my Cellshock D9GKX.
Ran into some trouble. After running prime-blend for 12 hours @ 3600 I experienced a bsod and lost all stability( this was the next day after a few hours of gaming ). I could still boot into windows but any stability testing ( linpack-prime blend ) would hard lock the system ( upon system reset I got the C1 error ) or bsod again. I think I may have cooked the ram playing with some of C-N's voltage settings (only bois changes made). Anyway I managed to get linpack stable at 3600 again but prime blend bombs out instantly. Prime small FTT's are running now with no errors. I will try and run memtest in the am before I go to work. Any suggestions?
Update- memtest for windows produces errors and if I let it run it crashes the system. Testing@ stock with back up ram.(same set older though)
Update- Ram died! Old set passed. Oh well at least I had a spare set....
Yup looks nice & I would have been happy but you pretty much said it all in the linked post.Quote:
I know at first glance this doesn't look very impressive (500FSB, 600MHz C5 on the memory, PL6) but this was with my DFI LT P35 and my Cellshock 8000 C4 kit. Supposedly P35 and D9GKX don't get along, but I only needed 2.05V set in the BIOS (don't believe Smart Guardian's reading). I had to tweak the Clock Delay settings, but I don't really see much difference between using these and D9GMH.
Sorry for your loss, The RAM is rated at 2.3v so it should have been alright but personally I would always use active cooling & I would have throttled the vdimm back as far as I could once I got the rig stable before pounding it with Prime95.
Now you've learnt the hard way, you wont do that again will you ;)
So I got my new set of OCZ Reaper 8500 4GB ram, good thing a guy from DHL dropped of a Vista SP1 64bit DVD right after that, so i can use all 4GBs and not just 3 of them :rolleyes:
My previous RAM was at its limit at about 1000MHz, but it looks like so was my board. I am struggling to repeat that fsb500 stunt, but......ah there we go, it crashed 2 minutes after priming at 7x495.
The things I discovered that work for higher FSB on my board are:
IN DRAM TIMING section!
-Set performance LVL (read delay) as low as possible, 14 or 15.
-REF to ACT Delay (tRFC) for my ram at least down to 60.
IN VOLTAGE settings!
SB Core/CPU PLL Voltage at 1.75 (anything higher does not help) + (this voltage bump helps alone with out above mentioned tweaks to at least make Vista boot process start)
Hope this is easy to understand, otherwise ask.
Other than that:
CPU VCore about 1.40-1.42
CPU VTT 1.414
Clockgen voltage 3.60
GTL Values all at 0.63 (works best here)
EDIT: Priming again with same settings but both CPU Clock Scews are set to 500. I googled it and it turnes out the right setting there can give us 10-15% better overclock. The article was a few years old, but 8 minutes and still priming. Coinsidense?