Messages in this thread | | | Subject | Re: utility for testing ram? | Date | Fri, 2 Oct 1998 11:04:34 +0200 (MEST) | From | (Rogier Wolff) |
| |
Doug Ledford wrote: > Stephen C. Tweedie wrote: > > > > Hi, > > > > On Mon, 28 Sep 1998 23:30:18 -0500, Doug Ledford <dledford@dialnet.net> > > said: > > > > >> Start it running overnight, if it's still running when you wake up, your > > >> memory's likely to be ok. > > > > > No, no, and NO! If you want to test your RAM, you can't run some test that > > > is CPU power limited. You'll never access your RAM here faster than the CPU > > > can compile the kernel > > > > Actually, gcc has such a large working set that it turns out to be a > > very good test indeed at least for cache memory. On a 166MHz alpha UDB, > > anything which fits into L1 cache absolutely flies, but kernel compiles > > take about 4 times as long as on my 486-DX100 simply due the the UDB's > > crippled L2 cache size and off-CPU memory bandwidth. Unless you have a > > massive L1 cache like the new 21264s have, gcc speed is quite likely to > > be influenced as much by L2 memory performance as by raw CPU speed. > > I'll stand by my claim that my test will trounce a gcc compile test any day > of the week for finding bad RAM %^) The compile test may touch a lot of > RAM, may have a large working set, and may actually compute things fairly > quickly, but gcc can't compile, k for k, as fast as gzip can decompress the > kernel. Then add in the significantly larger amount of data you are running > through RAM in a short period of time via DMA from the disk sub-system and > the compile test won't be able to match the stress the test I listed gives > your machine. I can say this much from experience. Find a machine that > fails my test on one out of every four passes, and I'll show you a machine > that will compile kernels all day long without a hiccup (x86 arch anyway).
What is "my test" again? A memory test? Could I possibly have your test?
I have all memory tests from the memory testing book from A.J. van de Goor in "chipmunk" and they find very little problems with PCs that gcc crashes on...
Roger.
-- | Most people would die sooner than think.... | R.E.Wolff@BitWizard.nl | in fact, most do. -- Bertrand Russsell | phone: +31-15-2137555 We write Linux device drivers for any device you may have! fax: ..-2138217
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |