Printable View
Guys,
Here my current spec. Asus A8N-SLi Premium with 3700 San Diego ( stock cooling )
My OCing is as follows 260 x 10 with Vcore 1.525
Temps are 36 idle 44 under load.
Its failed after around 6 mins. Will raising the vcore make it stable or do i need to do something else? I have not touch the ram settings yet. My ram is Kingston Pc3200 DDR 400. (nothing special)
Cheers chaps
Darren
Is it memtet86 stable? Do a few passes of test 5 (c, 1, 3, 5, enter, 0) to be sure.
Are you running a divider on your ram? which one?
On my dfi nf4 and many other nf4 boards, the option called "ram speed" is often just the divider setting (and the options will be a ratio, not a number), and the acctual speed is the fsb if you havent changed the divider (meaning your ram would now be 260mhz and probably not stable). When booting it also says "DDR 200" if running at 1:1 ratio, for whatever fsb speed you are at, which could be confusing.
2.6ghz is quite good you can probably get 2.7 maybe 2.8 depending on temps.
Stick core voltage at 1.6 and see if it makes it more stable. If it is, overclock it a bit further. 1.6v is probably going to be the max you should use for your cooling - I would expect the temps to stay below 55-60°c
Memtest86+ test #5 should (let it do about 20 passes) find any obvious memory errors.
Prime95 is far more sensitive to errors with the CPU itself, if your temps are good, more voltage should fix the problem (unless your at the limit of your CPU).
Cheers guys,
I'll download memtet86 and try that tonight. Its been running "memtest" over night with no errors so im happy with that so far. It's still running now so when i get home tonight at 7pm it would have been running for 22 hours......
Does memtest 86 test the cpu though? I know prime 95 does a bit of everything.
i would guess it need a tiny bit more of vcore
I havnt had a crash yet during windows. Ive been playing games and benchmarking and so far (touch wood) its been ok.
I happy with 2.6 on stock for the moment, i'll get a good heat skins fan ( xp-120) and then i'll crank her up!
I for one, think memtest is useless. It found 15,000 errors on 1 pass of TCCD i tryed 2hand,, yet the memory was stable.
Id use 32mb Super PI as a stabilty first, and once you get errors in that, play around with voltages and stuff, and then try Prime95. If prime fails after 6 mins, it's definatly not stable.
will i need to play with the voltages on the RAM?
No it wasn't. Memtest errors = unstable memory subsystem. It may have been stable enough for you, but that doesn't mean it's stable.Quote:
Originally Posted by Smokey21
Now that's fine, but don't make the mistake of thinking that just because a program isn't any use to you, it isn't any use to anyone;).
Quote:
Originally Posted by Smokey21
15,000 errors do not happen for no reson.
The main issue with your idea of superpi is that you would have to boot windows first, which could easly get corrupted if its not stable. Not really a good idea unless you already know your memory is stable and in that case whats the point?
I only use test5, as the lower number tests dont usually find anything and the higher numbers take far too long to compleat although probably more effective.
I useually do 2-3 passes of test5 which should rule out overheating, which one pass will miss.
how do you select it just to do test 5?
See the first line of SilentDeath's first post;).
Quote:
Originally Posted by Rave
sorry about that......you can tell its friday!! :Oops: again!
Good news so far.....
past memtest 86 16 times with no errors!!
past superpi up for 16mb (not had time to do rest) i'll run rest later.
Will try prime95 tonnight.
memtest means dick all.
Y bother with it. Just use prime95, it stresses it a lot more.
I've used Memtest on a number occasions to identify memory problems. It is the first thing I try when a PC becomes unstable - each time Memtest has found errors, replacing the RAM has solved the problem for me. Therefore, I would say that Memtest is very useful, and means slightly more than dick all.Quote:
Originally Posted by Smokey21
Prime95 is only worth while for stressing the CPU. It will find obvious memory errors, but there are so many things that can cause serious instability that nither Prime95, nor memtest86, nor any other single program, will ever find.
I've rarely had to replace memory that got errors with memtest, becasue most of the time the memory is not defective, just set improperly. Memtest86+ test #5 also gets memory hotter than anything else I've seen. This is important because memory can produce errors if it gets too hot and I regularly run my memory in excess of 3 volts.
Typically, I'll run 3 instances of Prime 95 simultaniously, each one doing a separate test, for at least 24 hours. Then I will run about 50 passes of the standard memtest86+ test, then at least 2,000 passes of test 5, then one full pass of all tests. Then I will stager as many instances of SUPER PI as memory allows, so I keep the CPU at a constant 100% load and make sure the memory is stressed as much as possible (both in ammount used and keeping that under constant use). I will usually keep Super Pi running for about 8 hours constantly this way. Then I play many games which are said to be inherently unstable, but which really are not (BF2 comes to mind, I've had it in one game for more than 16 hours straight, playing on and off for about 8, without a crash). Any crashes, bluescreens, or errors of any kind tell me the system has a problem somewhere.
Ok so far the story goes......
past Everything apart from prime95 which ran for 1hr 30 then failed on "Hardware Failure"
Cpu volts are now 1.552. Idle temps are 33 full load 44
What should i do now guys, play around with memory settings?
Machine is solid in windows though.....
The hardware failure errors, in my experience, are almost always CPU related. Rounding errors are often memory.
Stil playing with the memory settings might help. What speed, timings, and voltage are you using now?
Multi is X10 HTT 260 = 2.6ghz
I have had rounding errors as well, this was the first time it said "Hardware"
I am struggling to find my Memory settings in the BIOS on this asus mobo....will take a look now.
Ok, Currently its all set to auto........
the timing for the ram are 2.5 3 3 7.........also should i put it on 1T or 2T?
Your memory is probably substanitally overclocked. I would not expect that memory to be able to do 260Mhz stable. It's probably running at a ratio already, but you will need to know for sure.
Run CPU-Z (available at www.cpuid.com) and tell me exactly what speed your memory is running at.
The memory should be at T1.
The requency is 217.3m mhz
FSB DRAM cpu/12
CAS LATENCY 2.5
RAS TO CAS DELAY 3
RAS PRECHARGE 3
CYCLE TIME 7
BANK CYCLE 22
DRAM IDLE TIMER 16
I think it is set to run at 1t as defualt??
let me know if you need any other info.......thanks very much for your help so far!
Yeah you are running at a ratio already, the memory is slightly overclocked, should bebale to handle it with some more voltage though, try 2.8 vdimm. I would also raise tRAS from 7 to 8.
Quote:
Originally Posted by oralpain
will try that now then run prime 95
Just seen that my memory was at 2T. Have now put it to 1T and have not changed any settings......will try prime now.
Setting T1 without doing aything else will not help stability.
I have to disagree with you on this one. Ive seen many systems that unstable by using 1T over 2T.Quote:
Originally Posted by oralpain
I would certainly try changing it to 2T and rerunning the test/s. The only thing you have to loose by doing it is the time it takes :)
He said he was changing from T2 to T1. He had it at T2 and it was unstable.
I said that changing to T1 would NOT help stability.
You disagree with me by saying nearly exactly what I just said, in more words...
Anyway, I think he should use T1 because of the significant improvement in memory performance, then stabalize it by increasing voltage or reducing memory speed. Going to T2 should be a last resort type of thing.
Ok heres the mornings update.........
I have changed the vdimm voltage to 2.8 like you suggested and change the tRas from 7 to 8...........it failed again after about an hour. It said Hardware Failure.....do you think it might be my cpu that has a fault?
When you say reducing the memory speed.....the greater the figure the "looser" the timings? Just to confirm ive got that right and i'm not going things the wrong way around.
What should i do next oralpain?
Thanks again!
Only skimmed over the thread, sorry about that :)Quote:
Originally Posted by oralpain
Yeah, you've got that right, higher number = looser timings. As for what to do now- you need to isolate what's causing the problem. Set the RAM to a lower multiplier so that it's running at under 200MHz. If it still fails then you know the processor is at fault and needs more volts/better cooling/reduced overclock. If it's stable then you know the RAM is at fault and you need to tweak the settings some more.Quote:
Originally Posted by darrensen
To be honest though, a computer that is P95 stable for an hour or more is pretty unlikely to crash during normal usage. The only reason I insist on total stability is that I run Find-A-Drug on my computer and I don't want to return duff results, and I encode some video, so I don't want to potentially introduce errors. If you don't run a DC program or do video encoding, you could probably leave things how they are now.
Ok i'll try that and ge back to you.Quote:
Originally Posted by Rave
:rockon:
You are running at 260x10 and you want to keep the memory as closr to 200MHz as possible. See if you have a 150MHz (CPU/13) or 133MHz (CPU/14) memory ratio.
I doubt loosening timings any further will be of much help.
If reducing the memory speed does not help, your cpu could very well be at it's limit.
One other thing. Have you reduced the HTT(LDT) multiplier? Stock is 5, at 260Mhz I would be using 3.
HTT is stock.
That could be the problem.
It failed again after i reduced the HTT to 3 and ran the ram at 200mhz....
what im doing right now is i have reset the bios back to defaults....so the cpu is stock and ram (well everything is how it was out of the box) and now im running prime to see if it fails. If it does fail then i have problem with the cpu or ram, right?
cheers all for the great response, i knew i could could count on you hexus boys!!
Yeah, if it fails at stock, chances are that the CPU, ram, or motherbaord is screwy.
if it passes stock fine, just bring things up a bit at a time. Or start at 250x10 and work your way down as it didnt seem super flaky at 260.
ok its been fine on stock!!
I have now reduced the clock speed to 2.5ghz. I have sneaky feeling that my ram is holding me back. I'll buy some corsair stuff, some of the low latency stuff from scan.
You should test the processor and the RAM seperately- if it is the processor then new RAM isn't going to help. Like I said, run the RAM really slow, like set to 133MHz in the bios, and then see if it's stable at 2.6GHz.
Have you tried raising the CPU core voltage to 1.6?
Hi all,
Todays update is as follows.........it ran for approx 11hrs 30 yesterday at 2.5 (250 x 10) on 1.512V without any problems! So thats good news. I tell you what i did..........
Reset all BIOS back to defaults and increased the CPU to the above settings. Left everything else as stock.
Now im pretty sure that my RAM is holding me back, so for the moment im happy sitting at 2.5......(errrrr no im not, what was i thinking!!!) I'll be going home tonight and i'll try 237 X 11 and failing that i'll work down.
I reckon i need,
http://www.scan.co.uk/Products/Produ...oductID=118549
and a XP-120 heatsink cooler!
On the Asus mobo the max is 1.575vQuote:
Originally Posted by joshwa
Morning all,
I have been running prime with the below settings
255 x 10 @ 1.512 Volts
It went for 12 hrs without a crash and i have now increased it to 257 X 10 @ 1.512
We'll see how it goes.....