Which setting is better if:

[JuanTutors]

My ram is faulty. (Tried cleaning off dust, didn't work.)

Because of my ram, setting allocated memory above around 1000MB causes prime95 to fail mere minutes into stage 2 P-1 testing.

Is it better to reduce the amount of ram I make available to prime95 until it stops failing, or is it better to allocate a tiny amount of memory to Prime95 and skip stage 2 all together?

[retina]

I think it would be better to remove and/or replace your faulty RAM. It will affect more than just Prime95.

[Mr. P-1]

Originally Posted by retina

I think it would be better to remove and/or replace your faulty RAM. It will affect more than just Prime95.

I agree. If you have more than 1 stick, (or more than one pair, if that's how your hardware requires them), then you might be able to identify the faulty one and remove it, leaving yourself with a reliable box albeit with reduced memory. Even then, do plenty of torture testing to verify that it really is reliable.

If you insist on using an unreliable machine for GIMPS, do DCs. These have all had their P-1s done, so you won't harm the project by missing factors that would otherwise have been found, nor will you slow it down, because bad results will be detected immediately. Plus you will be able to keep tabs on your own reliability.

If you, (or anyone else, for that matter) wish to do first-time tests, but can't do P-1 stage 2, I have a large supply of exponents already P-1ed.

[Freightyard]

Oh my gosh--RAM is sooo dirt cheap nowadays--how could you not just replace it? If it's an old machine, a computer shop probably has used RAM just laying around... I know I do. Anyone need a pair of DDR2 667 MHz ECC 512MB? Brand new! I probably paid hundreds for it... who knows...

[JuanTutors]

OK everyone, I was really asking without the need to buy new hardware or change projects. I'd like to know which of the options I gave is better. I appreciate that RAM is cheap now, and I might upgrade because of that. But I am in the middle of testing a 100M digit mersenne, and that's the kind of work I want to do. I don't want to switch to DC or factoring. All my results have been turned in without error and none of my results have been bad. I'd like an answer that doesn't involve spending money or switching projects.

[petrw1]

Originally Posted by dominicanpapi82

My ram is faulty. (Tried cleaning off dust, didn't work.)

Because of my ram, setting allocated memory above around 1000MB causes prime95 to fail mere minutes into stage 2 P-1 testing.

Is it better to reduce the amount of ram I make available to prime95 until it stops failing, or is it better to allocate a tiny amount of memory to Prime95 and skip stage 2 all together?

Per George on another thread you can reduce Stage 2 to as low as 500MB

[James Heinrich]

Download Memtest and try your RAM one stick at a time until you find the bad one(s), and leave it(them) out of your machine. You really don't want to run a computer with faulty RAM, bad things happen. And really, RAM is ~$10/GB now so it's an entirely reasonable suggestion to replace the faulty RAM.

[Mr. P-1]

Originally Posted by petrw1

Per George on another thread you can reduce Stage 2 to as low as 500MB

On current first-time test exponents and those being now assigned for P-1 testing, 500MB should give you a reasonable stage 2. You could probably go down to 200MB for a minimal Stage 2. Going past 500MB gives very little benefit. It's worth doing if the memory's available.