Prime 95 failing

[oralpain]

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.

[darrensen]

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.....

[oralpain]

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?

[darrensen]

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.

[darrensen]

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?

[oralpain]

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.

[darrensen]

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!

[oralpain]

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.

[darrensen]

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.

will try that now then run prime 95

[darrensen]

Just seen that my memory was at 2T. Have now put it to 1T and have not changed any settings......will try prime now.

[oralpain]

Setting T1 without doing aything else will not help stability.

[Agent]

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.

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

[oralpain]

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.

[darrensen]

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!

[Agent]

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...

Only skimmed over the thread, sorry about that

[Rave]

When you say reducing the memory speed.....the greater the figure the "looser" the timings?

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.

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.