Prime 95 Blend Test

[cjkart20]

Hey guys, so I am running a prime 95 blend test to make sure that my RAM is stable. When I look at task manager and then performance it shows that Im only using like half of my RAM. Is this normal? Nothing has failed in prime 95. see pics. Thanks!!!

 

[EarthDog]

Yep, normal. You can see in the blend test settings how much ram its going to use.

 

[ATMINSIDE]

You can also set the RAM amount higher in blend

 

[cjkart20]

So is using just the blend test typically enough to check for stability? Or do I need to raise the RAM amount higher?

 

[EarthDog]

I have never raised it for blend and was fine.

 

[cjkart20]

Okay thanks! I have passed and am stable with a small fft for 3.5 hours. But if I fail in blend test do I raise the DRAM voltage or what steps do I take to correct a fail?

 

[ATMINSIDE]

Okay thanks! I have passed and am stable with a small fft for 3.5 hours. But if I fail in blend test do I raise the DRAM voltage or what steps do I take to correct a fail?

You could need to change the NB, CPU-NB, or DRAM.
Change only one at a time so that you know which fixed it.

Either slow down the speed or raise the voltage (within safe voltage limits).

 

[cjkart20]

Right now I am at 2200MHz with the CPU/NB frequency, CPU/NB Voltage 1.231250, DRAM Voltage 1.500, I also changed some of the DRAM timing to match up with my RAM because it wasn't the same in CPU-Z.
I am running 8Gb Viper patriot 1866MHz DDR3.

- - - Updated - - -

this was the timing stuff I changed to make sure it matched the 1866.

 

[ATMINSIDE]

Definitely sounds like your RAM was the issue if you're stable now.

 

[cjkart20]

I just left the blend at default so its only using about 45-50% of my RAM, but so far its been stable for 1.5 hours.

 

[Mandrake4565]

I just left the blend at default so its only using about 45-50% of my RAM, but so far its been stable for 1.5 hours.

If you can pass 2 hours blend without issue in most cases you can throw anything at the Pc without an issue.

 

[johan851]

So is using just the blend test typically enough to check for stability? Or do I need to raise the RAM amount higher?

I like to raise it a bit, leaving a couple GB available so that the OS doesn't freak out. I blend with 6GB of RAM on my box with 8GB. It's probably not really necessary, but I figure I might as well increase the hardware footprint of the test a bit to make sure everything's really getting stressed.

Generally if a system can run blend without errors overnight it'll be quite stable for daily use. I usually throw a couple of hours of LinX at it too to verify that the CPU is solid.

 

[cjkart20]

Alright so failed a blend test after 3 hours. Went to bios and raised CPU/nb voltage and raised the dram voltage. Trying again now. Should I mess with the CPU/nb frequency at all?

 

[Mandrake4565]

Alright so failed a blend test after 3 hours. Went to bios and raised CPU/nb voltage and raised the dram voltage. Trying again now. Should I mess with the CPU/nb frequency at all?

Some people will have different opinions and to let you know where I'm coming from I'll say this. I use my rig for work, gaming and just normal browsing. My work entails having 4 screens, with multiple windows open, news services and such. When things are busy, my rig is running at or near 100% for extended periods of time and if it crashed it could be very costly $$$$$$$ wise for me. I have also run F@H for 2 weeks 24/7 during the chimp challenge we used to do which also runs the Cpu @ 100%. If I can pass 2 hours of Prime Blend with my Oc I have yet to have an issue during my usage. In my book if you didn't have an issue until 3 hours you're stable.

 

[RJARRRPCGP]

So is using just the blend test typically enough to check for stability? Or do I need to raise the RAM amount higher?

Yes. And I would try at least 20 MB FFTs or around there.

 

[EarthDog]

Yes. And I would try at least 20 MB FFTs or around there.

Why?

Most people never budge past the default settings. Is there are reason you recommend this method over the default (which appears to work just fine?)?

 

[RJARRRPCGP]

Why?

Most people never budge past the default settings. Is there are reason you recommend this method over the default (which appears to work just fine?)?

For blend, I usually have the system allocate most of the RAM and use FFTs larger than the processor caches.

This should catch bus and RAM instability.

 

[cjkart20]

So some how I got all jacked up and can't even pass a blend test for 25 minutes now lol. Went back to defaults and am just increasing the CPU/nb voltage in addition to watching temps everytime I fail. If I can find a setting that allows blend to run all night with no errors should I call it good? Or should I then do a small fft to check CPU?

 

[RJARRRPCGP]

If I can find a setting that allows blend to run all night with no errors should I call it good?

The method I mentioned is recommended if you have a lot of RAM and OC'ing the bus to OC. (especially with multiple RAM brands and all DIMM slots occupied.)

Small FFTs only good if you suspect a core instability issue, usually because of needing more Vcore or require lower core temps.

 

[Mandrake4565]

So some how I got all jacked up and can't even pass a blend test for 25 minutes now lol. Went back to defaults and am just increasing the CPU/nb voltage in addition to watching temps everytime I fail. If I can find a setting that allows blend to run all night with no errors should I call it good? Or should I then do a small fft to check CPU?

cjkart20, do yourself a favor and start taking notes of what you're doing it will help greatly when one loses their way.