Best bang for the buck.. mem or cpu?

  • Thread starter Thread starter Clive
  • Start date Start date
C

Clive

Brian Link said:
Just bought an Abit KD7A, and a new Seagate 7200rpm 60g HD. I used the
1gig Athlon from my old mainboard, and put a 512M PC2700 memory module
in.

I'd like to boost the performance of the machine a bit, mostly for
multitasking, redraws, and gaming wouldn't hurt. I've only got a 100
bucks left to spend.. should I spring for another 512 module, or get a
faster Athlon?

BLink
Brian Link in St. Paul

Depends a little on your OS, but I would go for a faster CPU.

Clive
 
Brian said:
Just bought an Abit KD7A, and a new Seagate 7200rpm 60g HD. I used the
1gig Athlon from my old mainboard, and put a 512M PC2700 memory module
in.

I'd like to boost the performance of the machine a bit, mostly for
multitasking, redraws, and gaming wouldn't hurt. I've only got a 100
bucks left to spend.. should I spring for another 512 module, or get a
faster Athlon?

I was in the same situation and I just bought a faster cpu, for what
that's worth.
 
Brian said:
Just bought an Abit KD7A, and a new Seagate 7200rpm 60g HD. I used the
1gig Athlon from my old mainboard, and put a 512M PC2700 memory module
in.

I'd like to boost the performance of the machine a bit, mostly for
multitasking, redraws, and gaming wouldn't hurt. I've only got a 100
bucks left to spend.. should I spring for another 512 module, or get a
faster Athlon?

BLink
Brian Link in St. Paul

Tough call. A gig of memory is sweet, but an Athlon XP 2600+ will give
you a noticeable boost...I'd go the CPU (but I'm not familiar with that
motherboard, but I assume you are and know what CPUs it can take).
 
Brian said:
I'd like to boost the performance of the machine a bit, mostly for
multitasking, redraws, and gaming wouldn't hurt. I've only got a 100
bucks left to spend.. should I spring for another 512 module, or get a
faster Athlon?

the serious thing would be try to measure things and find your bottlenecks.

which os? i mostly run unixes, so i'd use "top" to see if i was hitting
swap, and to see what sorts of jobs pegged by cpu for long. if neither of
those things happen, maybe the money should be spent on a raid? there used
to be an old "rule" that most people underspend on disks in workstations.
it seems to me that should get more true as we concentrate more storage on a
single spindle. gotta move that one arm to seek for all files in
succession.

oh wait, you said games, can't be a unix ;-). maybe the task manager can
give you enough of an idea.

as a pure guess, i'd think cpu would be the way to go ... unless you think
your slow tasks could be disk bound.
 
Lefty said:
Brian Link wrote:




the serious thing would be try to measure things and find your bottlenecks.

which os? i mostly run unixes, so i'd use "top" to see if i was hitting
swap, and to see what sorts of jobs pegged by cpu for long.

I just tried that command, and it says 10% of mym memory is being used
for running X...but it also says that 497mb out of 512mb of memory is
used...what the?? heh Am I interpretting this wrong? :-)
 
I think that most windows OS's use as much memory you have until something else wants it...then passes the ram to the requesting
program and uses the swap file for system use.
 
Sooky said:
Lefty wrote:

I just tried that command, and it says 10% of mym memory is being used
for running X...but it also says that 497mb out of 512mb of memory is
used...what the?? heh Am I interpretting this wrong? :-)

can you post the first few lines? mine looks like this right now ... LOL i
forgot i was running windows! (now, that is funny)

here's an old one snagged off another post:

8:00pm up 1:22, 0 users, load average: 2.84, 2.07, 1.54
71 processes: 65 sleeping, 5 running, 1 zombie, 0 stopped
CPU0 states: 43.0% user, 56.0% system, 41.0% nice, 0.0% idle
CPU1 states: 40.0% user, 59.0% system, 40.0% nice, 0.0% idle
Mem: 257480K av, 64600K used, 192880K free, 70828K shrd, 12404K buff
Swap: 265032K av, 0K used, 265032K free 24432K cached

the two most important things imo are the "swap" used and the "idle"
percentage. this guy has his system pegged (no idle time, high load
average), but he hasn't hit swap (0K used). if this was a normal situation,
he could benefit from a faster cpu. more memory wouldn't make him go faster
because he hasn't run out.

i don't think anyone normally stays at 0% idle though. generally it will go
up when you do something intensive and then fall way off ... modern systems
in a desktop role are typically 95% idle when doing stuff like typing usenet
posts.
 
JAD said:
I think that most windows OS's use as much memory you have until
something else wants it...then passes the ram to the requesting
program and uses the swap file for system use.

yeah, we could probably talk details, but if you are swapping a lot that is
a reason to get more memory.

i thought of something after the last post. in a unix system, if you have a
"lot" of free memory (before you hit swap) the system will slup it up for
use as disk buffers. that will speed you up as well.

so maybe looking at free mem, and the amount dedicated to "buff" will hlep
too.
 
Lefty said:
can you post the first few lines? mine looks like this right now ... LOL i
forgot i was running windows! (now, that is funny)

here's an old one snagged off another post:

8:00pm up 1:22, 0 users, load average: 2.84, 2.07, 1.54
71 processes: 65 sleeping, 5 running, 1 zombie, 0 stopped
CPU0 states: 43.0% user, 56.0% system, 41.0% nice, 0.0% idle
CPU1 states: 40.0% user, 59.0% system, 40.0% nice, 0.0% idle
Mem: 257480K av, 64600K used, 192880K free, 70828K shrd, 12404K buff
Swap: 265032K av, 0K used, 265032K free 24432K cached

the two most important things imo are the "swap" used and the "idle"
percentage. this guy has his system pegged (no idle time, high load
average), but he hasn't hit swap (0K used). if this was a normal situation,
he could benefit from a faster cpu. more memory wouldn't make him go faster
because he hasn't run out.

i don't think anyone normally stays at 0% idle though. generally it will go
up when you do something intensive and then fall way off ... modern systems
in a desktop role are typically 95% idle when doing stuff like typing usenet
posts.

12:45am up 1 day, 8:25, 1 user, load average: 0.12, 0.09, 0.04
89 processes: 87 sleeping, 2 running, 0 zombie, 0 stopped
CPU states: 1.2% user, 0.0% system, 0.0% nice, 98.8% idle
Mem: 513984K av, 500620K used, 13364K free, 0K shrd, 110720K buff
Swap: 1020088K av, 3052K used, 1017036K free 225432K cached
 
Sooky said:
12:45am up 1 day, 8:25, 1 user, load average: 0.12, 0.09, 0.04
89 processes: 87 sleeping, 2 running, 0 zombie, 0 stopped
CPU states: 1.2% user, 0.0% system, 0.0% nice, 98.8% idle
Mem: 513984K av, 500620K used, 13364K free, 0K shrd, 110720K buff
Swap: 1020088K av, 3052K used, 1017036K free 225432K cached

PS - I'm running rehat 8, as a regular user, with X and mozilla as the
major processes running...
 
Sooky said:
Lefty wrote:

12:45am up 1 day, 8:25, 1 user, load average: 0.12, 0.09, 0.04
89 processes: 87 sleeping, 2 running, 0 zombie, 0 stopped
CPU states: 1.2% user, 0.0% system, 0.0% nice, 98.8% idle
Mem: 513984K av, 500620K used, 13364K free, 0K shrd, 110720K
buff Swap: 1020088K av, 3052K used, 1017036K free 225432K
cached

well, you touched swap (3052K used) in this session. that makes it more
complicated than if it just said 0K used. on the other hand, you've got
more free currently (13364K free) than you do in swap. and a lot in buffers
(110720K buff). i'd say load averages are pretty low.

i wouldn't say this system was in trouble, but i'm surprised it hit swap
with 512M.

we might have to treat this as a group project ;-), i think it's interesting
but i don't feel like i see a clear answer.

did you actually get that high in memory usage just running mozilla, or did
you do something more intensive earlier? you might want to compare to the
system after a reboot, and then go looking for the things you do that are
the most memory and cpu intensive respectively.
 
Lefty said:
well, you touched swap (3052K used) in this session. that makes it more
complicated than if it just said 0K used. on the other hand, you've got
more free currently (13364K free) than you do in swap. and a lot in buffers
(110720K buff). i'd say load averages are pretty low.

i wouldn't say this system was in trouble, but i'm surprised it hit swap
with 512M.

we might have to treat this as a group project ;-), i think it's interesting
but i don't feel like i see a clear answer.

did you actually get that high in memory usage just running mozilla, or did
you do something more intensive earlier? you might want to compare to the
system after a reboot, and then go looking for the things you do that are
the most memory and cpu intensive respectively.

The only things I've done, since rebooting yesterday with the new kernel
supplied by redhat, is run X, run mozilla and mozilla mail, view a 2mb
real video file, and update the net-snmp packages with rpm like 12 hours
ago...I'll take it down for a reboot now and post the results.
 
The only things I've done, since rebooting yesterday with the new kernel
supplied by redhat, is run X, run mozilla and mozilla mail, view a 2mb
real video file, and update the net-snmp packages with rpm like 12 hours
ago...I'll take it down for a reboot now and post the results.

i got out my old o'reilly "system performance tuning" book, and it reminds
me that there are two kinds of swapping: housekeeping and emergency. in
housekeeping, it may be that some process was idle for so long that the os
figured "why waste memory" and put it on disk. that's not bad. on the
other hand, emergency swapping is when you run out of real mem and the
system has to swap (flog) to disk to get out of a hole. that can hurt
you, but it's better than nothing ;-).

fwiw, i'm on my fedora system (900mhz duron, 256m) now. here is a shot
after i boot (switching to a text window before i log into the gui), after
the gui login, and after a little surfing:

09:12:22 up 1 min, 1 user, load average: 0.25, 0.15, 0.05
43 processes: 42 sleeping, 1 running, 0 zombie, 0 stopped
CPU states: cpu user nice system irq softirq iowait idle
total 0.0% 0.0% 0.0% 33.1% 33.3% 33.5% 0.0%
Mem: 255528k av, 82724k used, 172804k free, 0k shrd, 7652k buff
26296k active, 45552k inactive
Swap: 522104k av, 0k used, 522104k free 46368k cached

after gnome startup:

09:13:19 up 2 min, 2 users, load average: 0.98, 0.33, 0.12
61 processes: 60 sleeping, 1 running, 0 zombie, 0 stopped
CPU states: cpu user nice system irq softirq iowait idle
total 0.0% 0.0% 1.9% 0.0% 0.0% 0.0% 98.0%
Mem: 255528k av, 154292k used, 101236k free, 0k shrd, 10976k buff
40488k active, 98484k inactive
Swap: 522104k av, 0k used, 522104k free 70864k cached

after surfing:

09:19:08 up 8 min, 2 users, load average: 1.14, 0.46, 0.20
64 processes: 61 sleeping, 2 running, 1 zombie, 0 stopped
CPU states: cpu user nice system irq softirq iowait idle
total 0.9% 0.0% 0.9% 0.0% 0.0% 0.0% 98.0%
Mem: 255528k av, 233972k used, 21556k free, 0k shrd, 12252k buff
47672k active, 168524k inactive
Swap: 522104k av, 0k used, 522104k free 120368k cached

as you can see, i'm getting a little tight with 256m, but i haven't hit
swap. (i really should investigate why fedora reports all my time in
softirq and iowait sometimes ... but i don't think it's real. the system
seems responsive)

on the other hand my load averages go up when i get cpu-bound on the
activity.

btw, another reminder from the oreilly book ... if you really do get into
true emergency swapping situations you can use vmstat to look at it. the
"si" and "so" are the critical bits.
 
lefty said:
i got out my old o'reilly "system performance tuning" book, and it reminds
me that there are two kinds of swapping: housekeeping and emergency. in
housekeeping, it may be that some process was idle for so long that the os
figured "why waste memory" and put it on disk. that's not bad. on the
other hand, emergency swapping is when you run out of real mem and the
system has to swap (flog) to disk to get out of a hole. that can hurt
you, but it's better than nothing ;-).

fwiw, i'm on my fedora system (900mhz duron, 256m) now. here is a shot
after i boot (switching to a text window before i log into the gui), after
the gui login, and after a little surfing:

09:12:22 up 1 min, 1 user, load average: 0.25, 0.15, 0.05
43 processes: 42 sleeping, 1 running, 0 zombie, 0 stopped
CPU states: cpu user nice system irq softirq iowait idle
total 0.0% 0.0% 0.0% 33.1% 33.3% 33.5% 0.0%
Mem: 255528k av, 82724k used, 172804k free, 0k shrd, 7652k buff
26296k active, 45552k inactive
Swap: 522104k av, 0k used, 522104k free 46368k cached

after gnome startup:

09:13:19 up 2 min, 2 users, load average: 0.98, 0.33, 0.12
61 processes: 60 sleeping, 1 running, 0 zombie, 0 stopped
CPU states: cpu user nice system irq softirq iowait idle
total 0.0% 0.0% 1.9% 0.0% 0.0% 0.0% 98.0%
Mem: 255528k av, 154292k used, 101236k free, 0k shrd, 10976k buff
40488k active, 98484k inactive
Swap: 522104k av, 0k used, 522104k free 70864k cached

after surfing:

09:19:08 up 8 min, 2 users, load average: 1.14, 0.46, 0.20
64 processes: 61 sleeping, 2 running, 1 zombie, 0 stopped
CPU states: cpu user nice system irq softirq iowait idle
total 0.9% 0.0% 0.9% 0.0% 0.0% 0.0% 98.0%
Mem: 255528k av, 233972k used, 21556k free, 0k shrd, 12252k buff
47672k active, 168524k inactive
Swap: 522104k av, 0k used, 522104k free 120368k cached

as you can see, i'm getting a little tight with 256m, but i haven't hit
swap. (i really should investigate why fedora reports all my time in
softirq and iowait sometimes ... but i don't think it's real. the system
seems responsive)

on the other hand my load averages go up when i get cpu-bound on the
activity.

btw, another reminder from the oreilly book ... if you really do get into
true emergency swapping situations you can use vmstat to look at it. the
"si" and "so" are the critical bits.

results from after the reboot:

1:51am up 42 min, 1 user, load average: 0.13, 0.05, 0.01
74 processes: 72 sleeping, 2 running, 0 zombie, 0 stopped
CPU states: 10.0% user, 0.4% system, 0.0% nice, 89.6% idle
Mem: 513984K av, 205428K used, 308556K free, 0K shrd, 15656K buff
Swap: 1020088K av, 0K used, 1020088K free 106608K cached


same programs running now: mozilla and x (and bash)...I wonder if it was
rpm that wasn't releasing memory?
 
lefty wrote:

results from after the reboot:

1:51am up 42 min, 1 user, load average: 0.13, 0.05, 0.01
74 processes: 72 sleeping, 2 running, 0 zombie, 0 stopped
CPU states: 10.0% user, 0.4% system, 0.0% nice, 89.6% idle
Mem: 513984K av, 205428K used, 308556K free, 0K shrd, 15656K buff
Swap: 1020088K av, 0K used, 1020088K free 106608K cached


same programs running now: mozilla and x (and bash)...I wonder if it was
rpm that wasn't releasing memory?

i don't know. it might do something funny when you update packages that
are running. anyway, with 308556K free you're in fat city.
 
Brian said:
Just bought an Abit KD7A, and a new Seagate 7200rpm 60g HD. I used the
1gig Athlon from my old mainboard, and put a 512M PC2700 memory module
in.

I'd like to boost the performance of the machine a bit, mostly for
multitasking, redraws, and gaming wouldn't hurt. I've only got a 100
bucks left to spend.. should I spring for another 512 module, or get a
faster Athlon?

BLink
Brian Link in St. Paul

CPU or graphics card if you game
 
Just bought an Abit KD7A, and a new Seagate 7200rpm 60g HD. I used the
1gig Athlon from my old mainboard, and put a 512M PC2700 memory module
in.

I'd like to boost the performance of the machine a bit, mostly for
multitasking, redraws, and gaming wouldn't hurt. I've only got a 100
bucks left to spend.. should I spring for another 512 module, or get a
faster Athlon?

BLink
Brian Link in St. Paul
 
Just bought an Abit KD7A, and a new Seagate 7200rpm 60g HD. I used the
1gig Athlon from my old mainboard, and put a 512M PC2700 memory module
in.

I'd like to boost the performance of the machine a bit, mostly for
multitasking, redraws, and gaming wouldn't hurt. I've only got a 100
bucks left to spend.. should I spring for another 512 module, or get a
faster Athlon?

BLink
Brian Link in St. Paul

Since you don't have a ram starved system, I would stick the $
in a cpu.

:-) Greg B.
 
Back
Top