I have had 2GB of ram in my system, and though it provided increased load capacity, it made my sytem very unreliable. Each gig was purchased from different manufacturers at different times. The computer worked fine in all respects except when using digital audio/giga, which, of course, was the main reason for the upgrade, but alas, I had to remove a gig.
Upgrade season is upon me, and if I go out to buy two matched 1024's, I'd like some suggestions as to which manufacturer is best to go with.
P4 2.93 GHz (HP PAvillion a805n) Not sure about the mobo, but it's an asus.
160 GB system drive (Samsung)
80 GB Sample Drive (Western Digital)
M-Audio Delta 44 (Might add an Audiophile soon)
Sonar 4 SE
Really there's no such thing as 'best RAM'; assuming the RAM is properly spec'd for the motherboard it's either perfect or defective. The performance differences between brands are usually insignificant since performance is usually bounded by disk I/O which is orders of magnitude slower than memory.
When DDR was a new thing some motherboards were sensitive to particular brands but I haven't seen these problems for years and for most motherboard manufacturers the incompatibilities were fixed long ago with bios updates. I've mixed and matched expensive low latency RAM with cheapo (but tested perfect) value RAM in dual channel mode without any problems.
The standard test for basic RAM sanity is memtest86; this free program can be booted from floppy or CD and is also included from the boot menu of some live linux distributions like Kanotix. The only thorough test for total ram perfection is 'prime95'; run the blend torture test and if it's still running 12 hours later the system is solid, otherwise the RAM is to some degree flaky. The system may not even be generally crashy but Gigastudio counts on the integrity of huge data structures like nothing else and goes off into space and crashes in the kernel on bad data.
If you insist on RAM name dropping, I run 2x1GB of Geil value ram (about $100 / Gig at newegg) dual channel, prime95 perfect and crash free. I also run 3x512MB of Corsair Value Ram dual channel & perfect, this is the stuff that sells for about $45 a stick. In the past I had a Corsair fail prime95 and Corsair replaced it no problem. I had a 1G stick of Mushkin fail prime95 but show no errors in memtest86; they would not replace it despite a lifetime warrantee and they claimed that no memory could be expected to be that perfect. I won't buy Mushkin again.
I believe HP wouldn't sell a relatively expensive mobo like an Asus. Finally, I have plenty of perfect RAM from no-name vendors, Micron, Samsung, Geil, Corsair. I've only received bum sticks from respected names (HA!) and it's easy enough to torch any RAM into badness with bad handling (static discharge). RAM is cheap enough now that IMO none of it comes properly tested. That's become the consumer's job, just get it from a place with a good return policy (and don't trust Mushkins warrantee...)
-thanks, sam. I appreciate you thoroughness. All this information will doubtless come in handy during the purchase.
One more question: My original 4x512 setup passed memtest, but I'be never heard of prime 95. Is this a free download somewhere? Or should (can) I buy a copy when I get new memory?
Google for prime95; it's also a freebie. Prime95 is a more thorough test because memtest86 basically confirms that memory holds any pattern briefly but prime95 confirms that a huge memory image never goes bad under stress. I have ditched memory that passed memtest but failed prime95, and my systems never crash, ever. Prime95 lets you rule out hardware problems, so if you crash on good hardware you know it's a software problem.
If you pass prime95 then new RAM isn't going to help you and you have to look at the state of your OS install, driver problems, software conflicts etc. If you fail prime95 I wouldn't expect GS to be super solid regardless of all that.
Yes it will !!! It's increased instrument loading I'm after-not increased stability!
My 2 GB setup ws unstable-my current is not, really.
My system has been passing the test for about 8 hours now on 1 GB of RAM. If it's all good when I get home from rehearsal, then I'll pop GB #2 back in and run it again. We'll see what happens from there.
I ran the torture test foor 12 hours successfully, but then I got curious and loaded a medium-sized gsp and midi file. After tinkering around a bit, I saved the midi file and crashed my machine. All this while P95 was still testing. It was the first time I've seen a count-off after "beginning dump of physical memory..."
1) Does this matter? Is my system ok, since I ran the test for 12 hours w/o a crash, as sam said? Or is it intended that you test while actually using large amounts of data, which would mean that I failed?
2) [sort of just a rewording] Should I run it again, and not mess around?
I assume you ran the prime95 'blend' torture test; that's the default one and says something like "uses lots of RAM". If this ran for 12 hours, there's a high probability that your RAM is perfect, so I bet your GS problems lie elsewhere. Corrupt Gigs, some buggy driver, software conflict, OS corruption, hard to say, but yuck!
It's odd that your RAM tests perfect but your problems only appear when you add memory. I'm about out of suggestions and sorry to hear you've got trouble, very frustrating I'm sure! In my experience GS is stable and reliable on a solid machine, but maybe there are some setups it just doesn't like?
A couple more things. You didn't say you were testing with prime95 with the full 2 gigs in, but I assume you went back to that setup? I wonder if prime95 consumes more physical (and then virtual) memory as it runs longer, maybe 12 hours isn't enough? Maybe it was running in the gooder lower RAM but forced GS to allocate from the badder upper RAM? Try a longer test like 24 hours maybe.
Anyway IMO there's no way a RAM change will help you if there's no problem with your RAM, here's hoping you get to the root of the problem.
Sorry to drag this out, but I appreciate your interest.
I tested with the same one gig that's been quite stable. The crash happened while this test was about 12 hours in, but the cause was my USING GS while the test was running. I'm pretty sure my current ram steup is fine.
My last question was whether or not the test should be run concurrently with real work. Assuming this is not the case, then I'm fine so far.
The next step is to pop in the othere gig and test that.
I'm not really having trouble, just trying to get things really tip top while I'm not working on a particular project.