In Place large FTTs vs Custom

[Oceanborn]

Hey overclockers out there I'm having a weird "issue" with prime 95, right now my system is running as it says my sig and I ran prime95 on both cores to make sure it was stable...trouble is when I did it back then it completed the test for 8 hours without any hiccups.

 

The trouble is I was running the "In Place large FTTs Test" and it wasn't until recently I read somewhere here at the forum that the right way to test the system is using a "custom" method in which you select the amount of memory you want to allocate to the test. I allocate 900mb for each core since this leaves me with 200mb for the system. And after running this test one of the cores fails after 3:45 Hours whereas the In place large FTTs run fine for 8 !!

Anybody can tell me what's wrong?? or how can I correct this?

[NEODARK]

Large FFT's test RAM only... Thats why it didnt stress your CPU, and it didnt fail for a while

Small FFT's test CPU..

Blend test both.

 

custom (default) and changing memory to whatever you need, is like BLEND, but with only the allocated ram.

 

you should run BLEND or custom to test full system stability.

 

This is how I run my custom:

 

 

 

I allocate 800MB per core, since I have 2gigs. with 1 gig, do 400MB per core. And finally with Single core 1 gig, do 800.. and so on, get it?

[Capt Proton]

I think he already 'had it". Looks like test was set up correctly, and there is a problem with one of the cores. A little more VCore perhaps?

[Oceanborn]

Large FFT's test RAM only... Thats why it didnt stress your CPU' date=' and it didnt fail for a while

Small FFT's test CPU..

 

If large FTT's stress memory only, why does it say next to the test "Maximum Power and Heat Comsumption" ?? and both cores work at 100% if it doesn't stress the CPU??

[ramik]

What's the max ammount of RAM that can be used?

 

blend mode uses nearly 1GB, but as i have 2GB of ram, i tried to use custom with more ram (1.5, 1.7gb) but in process explorer i see that it uses only 1GB.

[Oceanborn]

According to what I saw in a guide here somewhere the rule is to leave 200mb for the system no matter what amout of ram you have

[Oceanborn]

I think he already 'had it". Looks like test was set up correctly, and there is a problem with one of the cores. A little more VCore perhaps?

 

That's just it...running the "maximum heat/power consumption" that seems to stress only the CPU with 1.4 vcore I'm stable for 8 hours....running the custom one that also stresses memory(judging by the fact you are specifying the amount of memory tested) with 1.425 volts one of the cores craps out at 4 hours....that's the reason I'm amazed....

[RLM]

Just had similar results with my overclock, albeit with small FFTs. Could prime no problems with small FFT's. I changed that to custom min 8, max 4096 700mb, and one core failed after 16 minutes.

 

Have just done 8 hours 22 minutes, but needed to up the volts by 0.025.

 

RLM

[Oceanborn]

But so far I haven't had a conclusive answer....I mean it appears as though In place large FTT's only the CPU is tested...and in my case it's 8 hours stable....

When I use "CUSTOM" one of the core fails...."custom" it seems RAM is tested a lot, since you have to specify the amount of memory you wanna test....

so if large FTT's (only CPU) primes fine, custom( CPU and RAM) fails....why do I have to increase Vcore ??????

[RLM]

....I mean it appears as though In place large FTT's only the CPU is tested...and in my case it's 8 hours stable....

 

Large in place FFT's is more Ram stress, not cpu Oceanborn. Or atleast that's how it reads. Small FFt's is more cpu stress.

 

I know what you mean though. Your question applies more to my scenario, where I've been using small FFT's without a problem. You would think for the custom test that upping something like the vdimm voltage would be what's called for, not vcore.

 

That said more ram stress, I'm guessing would still bring the cpu into the equation. Another guess is that the memory controller is on the cpu. Haven't a clue what a memory controller does mind :confused: (off to read up, handles the communication between cpu and ram I think)

 

Have been wondering whether tinkering a bit more with the memory settings, drive strength etc, might help with the custom test. I've tried loosening things a bit, but that hasn't helped. Would like to get the voltage back down a tad.

 

RLM

[twivel]

Yea, oceanborn probably needs a tad more voltage - or perhaps a slight drop in CPU OC. I've found that when I have close but unstable RAM, it fails in the large FFT - but if I am pushing the onboard CPU mem controller a tad too far, it fails in small FFT.