On Tue, 2017-04-25 at 18:47 +0100, Ken Moffat wrote: > On Tue, Apr 25, 2017 at 11:10:26AM -0600, Alan Feuerbacher wrote: > > > > Interesting points. Until recently I didn't concern myself with > > such speeds > > much, but used all default settings. I'm now retired and have time > > on my > > hands, and am getting a new CPU, MB and memory. I've been a bit > > confused by > > all the advertising numbers and hype, but think I understand what's > > going > > on. > > > > My new Intel CPU (i7 7700K) is spec'd at 2400MHz memory speed, but > > can be > > overclocked. Apparently it makes little sense to buy memory that is > > spec'd > > at a higher speed than the CPU is spec'd at, but advertising hype > > touts > > stuff like "3600MHz memory!" I stumbled on a bargain at Amazon for > > 3333MHz > > memory (less money for Prime members than lower spec'd memory for > > non-Prime > > members), so now I have an ASUS MB spec'd at 3866MHz, memory at > > 3333MHz and > > CPU at 2400MHz+. I don't have a clear idea what to do with all this > > new > > technology. > > > > It would be interesting to play with various memory speed settings > > as I > > compile LFS programs over the next few weeks. > > > > Comments? > > > > Alan > > > > First, make the most of having time on your hands - the rest of us > will be doing all we can to encourage you to contribute, and then > suddenly you will have less time on your hands ;-) > > But seriously, playing with new/current hardware is always > interesting. I've had two iterations of A10 Kaveri hardware (ok, > you intel guys can snigger at that) - on the first one it was fun > and games getting it to run the memory at its rated speed. > Eventually that machine died (motherboard or capacitor fault, I > think). The replacement worked briefly, then became unreliable - in > the end I swapped in the memoryfrom the broken one and it now runs > ok. An A10 is an APU (integrated graphics), for which faster memory > is supposed to improve the graphics (no idea if that is general, or > only for gamers). > > BUT: memory is not just the clock, it's the memory timings > https://en.wikipedia.org/wiki/Memory_timings and various > DIMMs from different manufacturers might have different timings. > > And at the end of the day it is questionable how much real > difference the variations will make. For semi-recent but not i7 > hardware, I would be extremely happy to get a 5% speed improvement > :) But like Bruce, I find that each new toolchain makes the build > slower. We are close to the point where a desktop i7 is the minimum > for comfortable compile times. I hope to try a Ryzen soon-ish, but > I need to change monitors before I do that. > > For me, my basic test - on a booted LFS system - is to rerun pass 1 > binutils. I haven't played with memory settings beyond persuading > htat original A10 to use what it had, but I also use this to compare > the times with various cpufreq governors. If the memory speed does > actually make a difference, I'm sure it might show upo there - but > for that I am dubious about how repeatable the results are (unless > you use an rt kernel). > > Whatever, find something to explore, and have fun. > > ĸen > -- > Error: ( : 1) not enough arguments > > Don't you just love Tiny scheme ?
"We are close to the point where a desktop i7 is the minimum for comfortable compile times." Not necessarily - it's possible to adjust the optimization settings to increase compilation speed at the expense of program speed. For example, you can compile binutils with minimal optimization, and then, once the system is running, have it compile with maximized optimizations in the background. Where you have a dual-core for instance, setting it to max out one core and leave the other alone would mean no noticeable slowdown for typical programs (text editor,terminal, etc). Doing this for all programs would boost compilation speed, and if you rebuild each part with the same version and the only settings touched are the optimizations, there shouldn't be any problems (at least, not that I'm aware of). It might be a bit inconvenient, but it could be worse. I'm getting a new $500 laptop soon. Depending on how long it takes to build binutils the first time, I may end up doing this on some programs. It shouldn't be a problem (7th Generation i5, 8 GB DDR4 RAM), but I'd rather have the system built within a day or two. It shouldn't be too unreasonable, I managed to build an LFS system on a dual-core Intel i-5 (or as it's officially called, Celeron) within a week, and it had DDR3 RAM, and the max speed was 2.16 GHz. I'll test if dropping in optimized builds causes problems when it arrives. -- http://lists.linuxfromscratch.org/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/blfs/faq.html Unsubscribe: See the above information page Do not top post on this list. A: Because it messes up the order in which people normally read text. Q: Why is top-posting such a bad thing? A: Top-posting. Q: What is the most annoying thing in e-mail? http://en.wikipedia.org/wiki/Posting_style
