Printable View
I can't seem to keep my rig stable at the same settings now that I switched to Windows 7 RC1 64 bit. Do you guys think the OS puts more stress on the components being 64 bit, or that it is just not stable?
I was running 191x21@1.35 vcore, but now the voltage has to go up a few notches (along with the temps)
now I've got peak temps of 82 deg C, so I backed off the clocks for now.
well... from my experience when hardware starts to come of age then it can be a big factor thats what it sounds like.Quote:
Originally Posted by mikesrex
There are various threads in the AMD section where people reported lower max oc's with 64bit os vs. 32 bit, but I have not seen anyone reporting the same yet for Intel cpu's. However, considering how similar AMD and Intel Cpus are it is certainly possible that it is a issue on Intel cpus as well.
Most definitely.
so basically I need to try out a D0-stepping i7 or perhaps find a better cooler than the TRUE120 lol
Threads like this come up every once in awhile (tho usually not in a 32 vs 64-bit context) and the general consensus has always seemed to be that hardware stability due to overclocking is totally independent of the OS being run. According to the first few posts in this thread perhaps that is not the case as I've always somewhat suspected (though it's hardly illogical to think OS should not matter).
Further thoughts?
I'm still a noob at this whole thing, but it was pretty apparent that XP32bit would crunch for days on end with no problems whatsoever at 191x21 with 1.35 vcore. With W7 RC1 64 bit, though, I had to make changes. Maybe it is just bugs in the actual OS as opposed to a difference in inherent stability?Quote:
Originally Posted by WrigleyVillain
64 bit is more demanding from the hardware.Quote:
Originally Posted by mikesrex
If you are on the edge of stability with your CPU (on 32 bit OS) , it's expected to have a slightly lower maximum clock with 64 bit OS.
I have experienced the same thing
ah this makes sense. I don't have time to tune this thing right now, so I feel kind of disappointed for letting it crunch at stock clocks. I have to make a haircut appointment and then go to work. I do have the whole weekend off work (for the first time this year) but I am graduating from college tomorrow.Quote:
Originally Posted by Jaco
better cooler than the TRUE 120 LMFAO you cant get better other than watercooling .......:rofl:Quote:
Originally Posted by mikesrex
i was hoping to get a laugh out of that one.Quote:
Originally Posted by dan7777
i noticed someone tested the prolimatech megalems against the TRUE Copper and found it to outcool the TRUE Copper. I wonder if I should give it a try. I need a new cooler for one of my rigs, so I may try it out.
That'd definitely be a waste of money IF you even get 4c lower temps, which you won't.Quote:
Originally Posted by mikesrex
yeah , you even need good watercooling to beat the TRUE.
Won't happen without a triple rad.
This totally reminded me of Spock in the new Star Trek. have you seen it lately? :rofl::rofl:Quote:
Originally Posted by WrigleyVillain
This is Xtreme, we use car radiators now:rofl::rofl:Quote:
Originally Posted by Jaco
i've always had lower clocks in 64bit than 32bit at the same voltage, this is on 2 amd x2's and 1 intel quad (x3330)
Well, depends on how you define stability. If you checked your OC using prime95 on XP (the standard 32bit 25.6 prime), and it was stable, then it should and will be stable on any x64 OS as well.
However, if you only tried crunching on XP 32 vs. a 64bit OS, you might experience instabilities because your setting wasnt really any good in the first place, it just didn't show because the CPU wasn't loaded that hard with 32bit BOINC.
ouch.Quote:
Originally Posted by jcool
there are a lot of boinc projects that are more demanding than prime95 and therfore 'produce' 'more' errors. fight aids at home is the most demanding i know, but it changed recently.
on the other hand i agree that a cpu that does no errors @ 32bit but does @ 64bit wasn't ever stable after all. but it just happened that the 32bit software did its job slightly different, putting load on a different - more stable - part of the cpu.
but from the standpoint of the user this would seem as if 64bit software was more fault provoking, magically reducing the oc-ability. :rolleyes:
64 bit OS and apps require more stability and more power.
In 32-bit mode parts of CPU are not used, and though power and clocks are distributed to them, they do not use as much power as in 64-bit mode. For example, all general purpose registers are used only in half, all ALUs work at half width etc.
Ouch.Quote:
Originally Posted by enteon
Do you even have an i7? :rolleyes:
Prime blend is more demanding than ever on the i7 arch.. only to be surpassed by LinX which is beyond any and all realistic load anyways.
There is definitely not a single WCG project harder on an i7 than prime, sorry.
Maybe QMC (another boinc project), but not WCG.
well well well, so I let my computer crunch all day at stock clock and default settings.....
......and upon coming home, my computer has crashed. I think there is something wrong with my windows 7 64 bit OS that causes the computer to crash, not the overclock.
yes, I still consider myself a n00b, but I am not an idiot. I can use my critical thinking and troubleshooting skills to figure things out, even if I don't consider myself an "expert" in a particular field.
So your rig is on stock clock and it still crashes when on load. Have you tried it in a different OS (also in stock clock)? This will verify whether the OS has any fault in this.
Otherwise, seems like a hardware issue considering crunching is just like any other stress test.
Note: There is a remote possibility that it could be unstable drivers.
I crunch 24/7 no problems at 4.0 GHz in XP 32 bit.Quote:
Originally Posted by fungry
I have a feeling this has something to do with the computer going to sleep or something. I'm gonna try to figure it out.
id go with OCCT myself.Quote:
Originally Posted by jcool
but be careful, it has been known to destroy some hardware, namely PSU, CPU, and GPU, and a few mobos.
ive had it increase my temps by over 8C, and thats with boinc 24/7, with gpugird, and rc5-72 in the background. with a DD480 with 8 denki fans on it sitting in my window, sucking cold air in the room.
Does OCCT stress all 8 threads on an i7 if you have HT enabled? Last time I checked it only ran 4 threads... which makes it unfit for i7 testing really.
Like I said, if you wanna go extreme, try LinX...
right now I disabled anything that windows 7 does to put the computer to sleep along with the screen saver and it seems to be working perfect. after a day of so of no problems I'm going to clock it back up to 4.0 GHz and see what happens. I have a feeling that was my problem.
OCCT v3.0.2.b03 has checkboxes for hyperthreading in their occt and linpack test options so I hope that is what it is really testing.Quote:
Originally Posted by jcool
no it streses the 4 real cpus, not the 4 logical ones. the linpak only sees real cores, not HT ones.
@Snow Crash: Good to know. I still prefer prime and LinX tho, but that's just me.
@Skycrane: You are missing the point here.. 4 threaded tools are useless on i7.
btw I put it back up to 4.0 GHz @ 1.35 vcore (which is what I used to run it at in XP32) and it has crunched for 24 hours no problems. The temps are being reported a few degrees higher than what it was with XP, so I'm not sure if it is just a different ambient temp or perhaps the 64bit version of boinc is stressing the CPU more.
Hi ! i confirm that windows 7 64 bits requires more vcore than vista 32.
I test the same OC on both system and it failed on W7. I need to raize the vcore to get stability with Prime95 test.
OCCT v3.1.0 stresses all 8 cores . Works great with i7 . You should try it .Quote:
Originally Posted by jcool
If you don't mind noise put a big fan on it (a big ol' Delta should get you a few degrees)Quote:
Originally Posted by mikesrex
I went from 3.6 to 3.9 on a q6600 for the same temps just with a change to a ffb1212vhe
A big ol' Delta might make the TRUE fly away.. lol
Even my Ultra Kaze 3000's get scared of some of the Delta fans.
Not the really big ones, I reckon you need to mount those on angle iron to stop them torqueing your heatsink of the board :) ...no, the one I used was only 150cfm.....Never did get to trying the 3blade (high pressure) version and I have 16 of those kicking around :eek: ....Yep I was into BIG air
You have 16 of the big Deltas?
http://img261.imageshack.us/img261/7676/housedowant.jpg
Well....you see....it's like this...I was looking at ebay and saw a server case that turned out to be a server (brand new no proc's or ram etc) for $40....When I won I looked again and saw it was for 4 units I was still happy....then i saw the carriage charge...OMG!. 2 days later a pallet arrived with 4 on, weight? 400lbs !!
each unit has: a Delta afb80912vh, 2 Delta afb0812sh, 4 Delta efb1212she...oh and 2 psu's with Delta fans in :)
So I have the case I wanted (kinda stealth looking...not flash at all) and a shed load of fans and stuff (hdd racks spare cases etc)
When I bought my vhe's I think I paid $20 plus for each...plus I had to import from sidewinder....this lot on ebay totalled $160 delivered so not so sad after all
You know how we were talking about box fans and we are talking about deltas... Oddly enough a few months ago a friend and myself were actually talking about taking the fan out of a box fan and filling it with deltas... which turned into talking about making the side of a comp case out of deltas.
I say make a box like that of a box fan but make it more square and not rounded off, make it the size of the Height and Length of the case, and fill it with deltas and mount a psu on top. :D