Printable View
I would encourage everyone who thinks that their setup is "Prime Stable" to run a blend test for 24 hours and not just 12. I can pass 12 hours many times but 24 only rarely (many times failing after 16, 17, or 18 hours). Also, as others have reported, Memtest has not been giving me any problems either.
-OmegaSys
I think even a stock system is going to give small errors after continuous usage. Prime is probably the best test for stability, but I think there should be a limit on it. Most people are not going to run their system @ 100% for over 8 hours, without letting it cool. Unless you fold or do media encoding, I'd say it's fair to claim stability if it can run overnight. That's just my opinion.
Again Omega, blend is for your RAM. It might be heating up too much after 16-18 hours and spitting out errors. People have had their memory run memtest stable, but fail Prime and blamed Prime. People have Prime run stable and fail memtest and blamed Prime. Bottom line, if one program fails - it's not stable and it's not the program at fault, it's your system.
durring my OCZ VX testing:Quote:
Originally Posted by saaya
Quote:
testing with memtest #5, i can run 260 x9 1:1 dual (3.4V 222,8,10,12)
25xpass
but i havent been able to find a P95 stable setting yet...
For all the people who are so certain this couldn't be a Prime95 software issue, cast your memory back to the recent past, I can think of more than one occasion where new hardware was released, everyone got into a tizzy about Prime95 not running, all shipped their hardware back, only to find a patch for Prime95 released a few days later due to a software problem on Prime95's part.
So if this does turn out to be a software problem, don't be surprised. ;)
What if this is a hardware problem and AMD releases the new stepping admitting that the winnie's had a problem, do we get to trade our old ones in ;) ???
I certainly hope you are right! did anyone bother emailing the author notifying him of this?Quote:
Originally Posted by st0nedpenguin
Whether on a newcastle or a winchester chip Prime95 runs the same way making long floating point calculation before comparing it with known correct values, why people would compare it to the likes of Superpi for "stability" testing is beyond me.
I see people running their winchester 3200+ @8*250MHz (2GHz) and calling it stock speeds when clearly the memory controller is working well beyond it's 200MHz rating. :stick:
Has anyone try to run prime95 with fastwrite disabled? (from bios, chipset settings or something...)
Please disable it and report back, some guys said that that was the problem with some winchesters/mobo incompatibilities.
Cheers
The latest version of Prime 95 works fine.
http://www.houseofhelp.com/forums/sh...ad.php?t=22249
Hi
Well I have had P95 running now for over 24 hours on my Asus A8V with a 3200+ @ 2600 mhz 260x10 1.530 vcore, so I doubt the program has any problems or am I just lucky :).
Menty
Yeah, also you can Prime for say 16 hours, but if you run 3DMark and Prime in the background, you'll get an error in a few hours :) However, separately both Prime and 3D are stable.
after hours of priming i'm thinking that the case will warm up, hence the mobo will build up enough heat enough to make one component unstable (if it is a mobo issue from then and on).
So sometimes it all comes down to overall system temperature.
Hi
Well I have played HL2 for hours while prime ran in the background, and I have also been editing movies, and for the fun of it, I have been compressing and extracting some big files while priming cause that usually makes errors, ooh and did I forget to say prime is still running fine. Im watercooled on the cpu, nb, gpu, Casetemp never goes over 25, cpu temp after 24 hours is stable @ 40 degrees and the room temp is 21. :banana4:
Menty
That version of Prime95 was released February 17, 2004, I think it's quite likely that's the version people are using now, also that was an update to fix a problem with machines running 1Gb RAM, not machines running Athlon64 Winchesters...Quote:
Originally Posted by Beenthere
Well here's my suspected faulty 3200 winnie almost 18 hrs with no errors at stock. THis isn't Blend this is the "small fft's" torture test. Maybe I should've let it run a whole 24 hrs or more. But this chip won't overclock anymore. 2.7- stock speeds in 2 days. Any ideas why a stable overclock would diminish in 2 days?
http://xtremesystems.org/forums/atta...chmentid=20722
all right... as promised i'm coming back with some more Prime95 testing ;)
after countless voltage configs i gave up on 265x10 :( couldn't run Blend for more than 3 hrs... so i lowered the OC to 260x10 but cranked up vcore and vdimm just to be sure and hoping that my Winnie takes the heat :)
and here's my "over 15hrs" of Prime95 Blend :banana: i think it's pointless to run it over 24hrs. i stopped at 15th hour because i need to see how low on voltages i can go...
260x10=2605MHz @ 1.82v REAL LOAD, DDR520 6-2-2-2.0 @ 3.88v REAL LOAD;OCZ DDR Booster
MSI K8N Neo2 Platinum - Q47353.1 0440A2 S Taiwan
A64 3200+ 90nm "Winchester" - CBBFD 0441UPCW
Thermalright XP-90 w/ YTech 120mm
CLICK for Full Size Screenshot - 2048x1536 302KB
http://www.xtremesystems.org/forums/...chmentid=20725
[QUOTE=bachus_anonym]all right... as promised i'm coming back with some more Prime95 testing ;)
after countless voltage configs i gave up on 265x10 :( couldn't run Blend for more than 3 hrs... so i lowered the OC to 260x10 but cranked up vcore and vdimm just to be sure and hoping that my Winnie takes the heat :)
and here's my "over 15hrs" of Prime95 Blend :banana: i think it's pointless to run it over 24hrs. i stopped at 15th hour because i need to see how low on voltages i can go...
260x10=2605MHz @ 1.82v REAL LOAD, DDR520 6-2-2-2.0 @ 3.88v REAL LOAD;OCZ DDR Booster
MSI K8N Neo2 Platinum - Q47353.1 0440A2 S Taiwan
A64 3200+ 90nm "Winchester" - CBBFD 0441UPCW
Thermalright XP-90 w/ YTech 120mm
1.8vcore??!! wtf?
oh well, like i said i raised it to have this "little" margin :D
i just started Prim95 again at Blend @ 1.71v and 3.71 LOAD... so far so good :)
my system won't even post if I ran that. how much does your vcore fluctuate? I beginning to think my problem is due to power. Or because I moved my raid stripe from nf3 250gb 754 to nf3 ultra 939. I just get these random reboots all the time!!Quote:
Originally Posted by bachus_anonym
i measure all voltages with multimeter (vdimm on the 7th pin in DIMM socket and vcore on the mosfet's leg). so....
@ 260x10 (CoreCenter @ 1.80vcore), BIOS 1.4B1
IDLE ---> 1.77v and 3.88v
LOAD ---> 1.82v and 3.82v
as a matter of fact i'm not using SATA, just PATA. as for SATA it's advised to use 3 and 4 ports when overclcoking and locking AGP at 67 (if you didn't know that already).
i'm using Fortron/Source 530W with one HDD and CDRW, GF4Ti4200.
rails fly little bit higher at 12.20v/5.20v/3.40v LOAD
i don't have any SATA drive so can't check if it might be the problem :(
I'm on bios 1.51 and ocz powerstream rails are like 12.20v on the 12v 3.35 on the 3.3 line and 5v rail is even 5v. I don't have a multimeter to test, but this is what I get from the bios, I'm not sure how acurate it really is. my cpu temps are around 35-39c load. 3 sets of tccd all work fine in memtest. but as soon as I overclock past stock and go into windows, problems occur. For example I tried 2.6ghz and lowered ht link to 3x. agp/pci lock is set at 66, but it seems to lock fine in clockgen. I am using sata 3and4 in raid0, and that seems ok. I ran prime 95 at stock almost 18hrs on small fft torture test, didn't get a single error, but I ran the same test after setting to 2.6 and passed up to test 3 and then system reboot and windows xp screen just hangs until I hold power button. Normally you would get an error and prime will automatically stop, but in this case system just randomly reboots itself. I am really out of ideas here and I'm begging for advice. neo2 is week 41 on nf3 chip. could this be a bad bios? or is board just shot from too much ddr booster? I never ran booster over 3.5v mostly it was between 3v and 3.3v. cooling is not really an issue here, so it's either bios, power flux., or faulty board. I thought it might be the cpu at first, but how do you explain 18hrs of prime with no errors f the cpu was at fault? i tried low vcore and vcore up to 1.72, fastwrites is off, everything else like nv/ati is disabled, I read the msi 939 thread and I'm following all the advice. I also lowered oc to 2.4 and 2.5 and adjusted ram voltage and vcore acordingly.Quote:
Originally Posted by bachus_anonym
here is my 3500+ winny
http://anarchb.free.fr/primestable.200412933212.jpg
EDIT: don't know why CPU-Z 1.26 say 1.45v....it was at 1.6v
How much voltage did you run through the chip @ 2.7ghz?Quote:
Originally Posted by trans am
@Edit:
I just completed 27hours of prime on maximum heat torture, 267x9 stock vcore, Memory @ 218.5 3/3/3/6 1T timings. I say this is long enough to say the rig is stable.
What is troubling is the fact at 1:1 the system would only pass 14 hours of prime and always bum out at the same spot, 3/4/4/8 2t timings when in memtest the memory will pass up to 293 at 3/4/4/8 timings. Maybe it is memtest86 not detecting the errors properly on the account of the ondie memory controller?
1.475+3.3%=1.52v I noticed in my screen shot I had tras set to 7. I changed it later to 10 and that was when I started noticing problems. But tras 10 is still perfect in memtest86 test5,6 and 8, up to 280mhz 2.5-3-3-10. I always heard tras 10 was the best for A64. I don't know if that would really cause these issues. ANd I was on the 1.51 bios. is this the best bios yet? SHould I flash back to 1.41 modded? Should I call Monarch and have them send a new board? Oh, off topic, how do you know if board is 6 phase power? Is it 6 phase if board has 6 mosfets?Quote:
Originally Posted by phobix
Is this board 6 phase?
http://xtremesystems.org/forums/atta...id=20567&stc=1
The board employs a three phase power regulator. There are 2 FETs per phase, 4 phase would of been better.
Like I said, I am noticing funny things with memtest86, it'll pass all kinds at high speeds with my TCCDs but what is memtest passible is far from stable when it comes to prime. I had to run my memory async to get some stability at 267 but I am happy I found a comfortable stability point.
The more problems I see, the more discouraged I am to mount this rig in my prommy. I know the CPUs have plenty of life in them, it is the memory controller I don't trust. Sure, some people seem to be getting excellent results, but these "super" specimens are few and far between. One thing I can say rather conclusively is that a 130nm is less of a crapshoot than these 90nm chips.