Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.debian.user > #221138 > unrolled thread

New RAM, does Debian has a tool to benchmark?

Back to article view | Back to linux.debian.user

Contents

  New RAM, does Debian has a tool to benchmark? dalios <dalios@eumx.net> - 2020-04-09 20:30 +0200
    Re: New RAM, does Debian has a tool to benchmark? Roberto C. Sánchez <roberto@debian.org> - 2020-04-09 20:40 +0200
    Re: New RAM, does Debian has a tool to benchmark? David Christensen <dpchrist@holgerdanske.com> - 2020-04-09 23:20 +0200
    Re: New RAM, does Debian has a tool to benchmark? dalios <dalios@eumx.net> - 2020-04-10 08:30 +0200
      Re: New RAM, does Debian has a tool to benchmark? Marco Möller <talby@debianlists.mobilxpress.net> - 2020-04-10 12:40 +0200
      Re: New RAM, does Debian has a tool to benchmark? Roberto C. Sánchez <roberto@debian.org> - 2020-04-10 14:40 +0200
      xpdf woes, was Re: New RAM, does Debian has a tool to benchmark? David Wright <deblis@lionunicorn.co.uk> - 2020-04-10 17:20 +0200
      Re: New RAM, does Debian has a tool to benchmark? David Christensen <dpchrist@holgerdanske.com> - 2020-04-11 05:10 +0200
        Re: New RAM, does Debian has a tool to benchmark? Tixy <tixy@yxit.co.uk> - 2020-04-11 11:50 +0200
          Improving performance: RAM or CPU speed (was: Re: New RAM, does Debian has a tool to benchmark?) rhkramer@gmail.com - 2020-04-11 15:10 +0200
            Re: Improving performance: RAM or CPU speed (was: Re: New RAM, does Debian has a tool to benchmark?) deloptes <deloptes@gmail.com> - 2020-04-11 15:40 +0200
              Re: Improving performance: RAM or CPU speed (was: Re: New RAM, does Debian has a tool to benchmark?) rhkramer@gmail.com - 2020-04-11 16:20 +0200
                Re: Improving performance: RAM or CPU speed John Hasler <jhasler@newsguy.com> - 2020-04-11 17:30 +0200
                  Re: Improving performance: RAM or CPU speed deloptes <deloptes@gmail.com> - 2020-04-11 21:10 +0200
                    Re: Improving performance: RAM or CPU speed John Hasler <jhasler@newsguy.com> - 2020-04-12 01:30 +0200
                  Re: Improving performance: RAM or CPU speed David Wright <deblis@lionunicorn.co.uk> - 2020-04-11 22:30 +0200
                    Re: Improving performance: RAM or CPU speed John Hasler <jhasler@newsguy.com> - 2020-04-12 01:50 +0200
                      Re: Improving performance: RAM or CPU speed Andrei POPESCU <andreimpopescu@gmail.com> - 2020-04-12 09:40 +0200
                        Re: Improving performance: RAM or CPU speed John Hasler <jhasler@newsguy.com> - 2020-04-12 20:40 +0200
                  Re: Improving performance: RAM or CPU speed "007@caiway.net" <007@caiway.net> - 2020-04-12 00:20 +0200
                  Re: Improving performance: RAM or CPU speed Celejar <celejar@gmail.com> - 2020-04-13 15:30 +0200
                    Re: Improving performance: RAM or CPU speed John Hasler <jhasler@newsguy.com> - 2020-04-13 17:50 +0200
                      Re: Improving performance: RAM or CPU speed Celejar <celejar@gmail.com> - 2020-04-13 18:30 +0200
                        Re: Improving performance: RAM or CPU speed John Hasler <jhasler@newsguy.com> - 2020-04-13 19:10 +0200
                          Re: Improving performance: RAM or CPU speed deloptes <deloptes@gmail.com> - 2020-04-13 20:20 +0200
                Re: Improving performance: RAM or CPU speed (was: Re: New RAM, does Debian has a tool to benchmark?) deloptes <deloptes@gmail.com> - 2020-04-11 21:20 +0200
          Re: New RAM, does Debian has a tool to benchmark? David Christensen <dpchrist@holgerdanske.com> - 2020-04-12 00:30 +0200
            Re: New RAM, does Debian has a tool to benchmark? deloptes <deloptes@gmail.com> - 2020-04-12 10:40 +0200

Page 1 of 2  [1] 2  Next page →

#221138 — New RAM, does Debian has a tool to benchmark?

Hi list!

I just purchased new RAM for my Debian Stable running PC to upgrade its
performance. I currently have 4GB and the new one is 8GB so I will have
12GB total.

My question is this: is there a tool in the repositories which I can use
to measure the system's performance with 4GB and then with 12GB of RAM
in order to see if it was a worthy choice?


Thanks in advance,
Dalios

[toc] | [next] | [standalone]

#221139

On Thu, Apr 09, 2020 at 09:26:06PM +0300, dalios wrote:
> Hi list!
> 
> I just purchased new RAM for my Debian Stable running PC to upgrade its
> performance. I currently have 4GB and the new one is 8GB so I will have
> 12GB total.
> 
> My question is this: is there a tool in the repositories which I can use
> to measure the system's performance with 4GB and then with 12GB of RAM
> in order to see if it was a worthy choice?
> 
> 
There is not enough information to adequately answer your question.

However, some things to consider are:

- Did you get on 8 GB module or two 4 GB modules, and is your existing 4
  GB in a single module or two 2 GB modules?  Many motherboard perform
  better with memory modules in matched pairs, so if you started with a
  single 4 GB module and added a single 8 GB module, you may not add as
  much performance as you might have epxected.
- Do you computing tasks include things that require lots of high speed
  access to main memory (e.g., manipulating large mathematical arrays)
  like in scientific computing?  If not, then you may find that a
  benchmark does not give you much useful information since you may not
  be using the RAM near its performance limit.
- Did you previously find that 4 GB was simply inadequate and that your
  machine spent lots of time swapping?  If so, then simply adding 8 GB
  more of memory may effectively eliminate the need for your system to
  swap.  A benchmark would be largely irrelevant there since your
  comparison would be between a spinning HDD or SSD and RAM; even the
  "slowest" RAM is literally orders of magnitude faster than any
  spinning HDD and still a great deal faster than SSD.

In any event, an idea of what sorts of information you are seeking from
the benchmark would make it possible to give more relevant advice.

Regards,

-Roberto
-- 
Roberto C. Sánchez

[toc] | [prev] | [next] | [standalone]

#221141

On 2020-04-09 11:26, dalios wrote:
> Hi list!
> 
> I just purchased new RAM for my Debian Stable running PC to upgrade its
> performance. I currently have 4GB and the new one is 8GB so I will have
> 12GB total.
> 
> My question is this: is there a tool in the repositories which I can use
> to measure the system's performance with 4GB and then with 12GB of RAM
> in order to see if it was a worthy choice?
> 
> 
> Thanks in advance,
> Dalios


As Roberto pointed out, hardware configuration is critical for memory 
performance.  If you would care to post your computer or motherboard 
make/ model, your memory module quantities/ makes/ part numbers, and 
your installation configuration (which modules are installed in which 
slots), people on this list can double-check.


I have been using Memtest86+, and it's predecessor Memtest86, for many 
years.  In addition to testing memory, Memtest86+ also displays 
fundamental memory information and metrics.  Once you know your modules 
are configured correctly, I suggest that run Memtest86+ for 24 hours to 
confirm everything is working correctly:

http://www.memtest.org/


I have also been using using dbench for many years, primarily to test 
disks and arrays.  It might measure some difference between 4 GB vs. 12 
GB of memory, but I would expect disk I/O to dominate:

https://dbench.samba.org/


There are other benchmarking tools available as Debian packages:

2020-04-09 13:02:34 dpchrist@tinkywinky ~
$ apt-cache search benchmark


Beyond that, and also as Roberto pointed out, your workload will 
determine what benchmarks have the most meaning.  But in general, adding 
memory increases overall performance.  Tripling your memory for $40 
should be a worthwhile upgrade.  You should notice it when you run a 
graphical desktop and open a lot of applications/ tabs.


David

[toc] | [prev] | [next] | [standalone]

#221143

Thank you to both of you for your answers. I am grateful for your time.

My PC is a desktop PC, see hardware details at the end of this email.


Here is more info, answering your questions:

* I purchased a single 8 GB module and currently the PC has a 4 GB
  module. My original plan was to buy two identical 8 GB modules but I
  decided to buy an 8 GB module now and the second later. I am aware
  that since the two modules (8+4) are not identical there is a
  possibility that they would not perform at their maximum speed.
  However I believe that 8+4 is better than 4 or 8 GB, am I right on
  this?
* The 8 GB module that I purchased is this:
  <https://www.crucial.com/memory/ddr3/ct102464bd160b/ct8815800>
* The computer is used for basic office tasks and the usual internet
  browsing (sometimes with many firefox tabs open). I am not using any
  particularly heavy programs like scientific computing, video rendering
  etc. However I am experimenting sometimes with virtual machines
  (usually with qemu-kvm). Not something professional, just as a hobby
  (I am trying to teach my self things by performing tasks that I
  wouldn't normally try on my main system).
* With the 4 GB RAM that I now have available, I have noticed my PC
  becoming slow and the memory usage going quite high. After booting,
  with just Gnome running, I see conky reporting that memory usage goes
  to something like 800 MB. Firefox adds ~800MB more and Thunderbird
  adds up to a total of more than 2 or maybe 2,5 GB! After launching
  Transmission (bit torrent client) I see that I have to restart Firefox
  quite often in order for the PC to be usable (especially after
  browsing a little and having many tabs open or even after some tabs
  have been closed). I have noticed that things got worst after
  upgrading to Stable (I was on oldstable until recently).
* I haven't been checking on swap usage a lot so I don't know how is
  that used.
* I have two HDD hard drives installed in the system.
* Up to now I have only been using Memtest86+ to check the memory's
  integrity but I didn't know that it can be used to see other
  performance metrics. I will check that now.
* I will also the dbench (which I haven't heard of before) as well as
  use the apt-cache command you suggested. If the information I give in
  this message helps you to suggest something specific please do, as
  hardware in general and especially benchmarks is not something I am
  particularly familiar with.

David Christensen wrote:
> [...]
> Tripling your memory for $40
> should be a worthwhile upgrade.  You should notice it when you run a
> graphical desktop and open a lot of applications/ tabs.

That sounds encouraging and that is exactly what I am after but I
thought I could ask here for benchmark software suggestion to see if I
can get something measurable. (and hopefully learns something on the way).


Thanks again,
Dalios


-------------------------------------

Motherboard
    Manufacturer: Pegatron
    Form factor: uATX - 19.3x18.0 cm (7.6x7.1 in)
    Chipset: Intel H81
    Memory sockets: 2 x DDR3
    Processor socket: LGA 1150

Processor
Intel Core i3-4170
    CPU speed: 3.7 GHz
    CPU cores: 2
    CPU Cache: 3 MB
    TDP: 54 W
    Socket: LGA 1150

Memory
    Amount: 4 GB
    Speed: PC3-12800 MB/s
    Type: DDR3-1600

Video Graphics
Intel HD Integrated Graphics

Hard Drive
    Size: 1 TB
    Interface: SATA
    Rotational Speed: 7200 rpm

[toc] | [prev] | [next] | [standalone]

#221146

On 10.04.20 08:26, dalios wrote:
> (...)
>      CPU Cache: 3 MB
> (...)
Working with much older hardware than you and using Debian as a Desktop 
with GUI (KDE) and not as a Server, I found the CPU Cache to be the most 
important piece of hardware letting you "feel" your desktop "responding" 
slow or fast. Nowadays I claim that 1 MB of cache is insufficient for 
experiencing a satisfying "responsiveness" when interactively working 
with the desktop. I do have a system with 2 MB cache and it is "usable", 
although not "comfortable". My 3 MB cache system is fine and I am 
working with it as my main system for my daily work, using heavily 
Firefox with usually some 5 to 15 tabs open AND additionally having a 
second Firefox with a different Firefox profile open for there running 
web apps of Microsoft Office 365 in usually 3 tabs. In parallel I always 
have open Thunderbird, RIOT (the matrix.org messenger client), 
Cherrytree, Kate, and some times also LibreOffice Writer or even ImageJ 
for some scientific image processing. This system has (only) the 
following hardware specifications and runs all this:

CPU: Intel Core2Duo P7450 2.13 GHz
Cache: 3 MB
RAM: 4 GB

Note that my 1 MB, 2 MB and 3 MB CPU Cache systems all have an almost 
identical CPU of type Core2Duo and CPU speed, and I also have another 2 
MB CPU cache system with a slower CPU speed. I clearly "feel" strongly 
the different amount of cache at my hands, but I do not feel the CPU 
frequency differences!

The last hardware upgrade was to recently extend it to now have 2x 4 GB 
RAM, but this was more because of having received the RAM free of cost 
than having felt a need to search for it. Instead, what really becomes 
an every day annoyance is a missing USB3 port, with the desire to have 
USB3 Gen2 available for the necessary (highly recommended) everyday data 
backup to some external storage media.

What I am about to say: with your hardware I would not worry too much 
about benchmarking the RAM. Simply add as much RAM as you can, following 
other people recommendations on how to best combine them (which RAM 
mounted to which slot of your motherboard), but then instead of loosing 
time with benchmarking better spend time on fine tuning some parameters 
which are needed to adjust in order to let Debian run with nice 
"responsiveness" as a desktop on not so fast hardware. Keep in mind, 
that not used RAM is a waste of resources. If your system is constantly 
using as much RAM as possible, but of course without ever to start to 
swap data to slow bulk storage media, then it best uses your hardware 
for you. Linux is intelligent towards this goal. It uses free RAM and 
there caches data which you might need again, in order to then be able 
to deliver you this data quickest. If your activity, some application, 
needs the RAM, then immediately some cached data is thrown away (cached 
data is considered to be able to become loaded in newly, if needed) and 
the freed RAM is immediately available for your activity. So, the cache 
will not impact your application's needs. Cache is used when possible 
and not used when there is no vacancy for it. Now, this explains why you 
will always see high RAM usage on your system. Whatever is unused by the 
user activity will be used for a performance increase and for this will 
be filled by the cache system. If you have 4 GB mounted to your 
hardware, then it will be used almost completely, and if you now mount 8 
GB (or 12 GB) to it, then you will (and want to) see that the 8 GB (or 
12 GB) are used almost completely. Nice!


I always consider Debian to be the Linux distribution which 
out-of-the-box comes well tuned for server applications, while not being 
preconfigured nicely for desktop usage. The configurations which in this 
regard I am mainly having in mind are the ones which adjust how the 
Linux cache system should work, how the RAM which is not used by your 
applications is meanwhile used. Afterwards it might also be worth to 
consider the many logs and statistics which by the Debian default 
configuration become generated, and how I/O to the bulk storage media is 
used, but for the moment I won't have a look at that.
For now, maybe simply consider to adjust the following parameters for 
the cache system and see if your system comforts you better:


create a file  /etc/sysctl.d/99-SettingsByMyself.conf  with the 
following content, which will adjust the 2 main parameters 
"vm.swappiness" and "vm.vfs_cache_pressure" influencing the cache system 
towards the needs of running a Desktop with GUI on not so fast hardware:


# responsiveness of system preferred over data treatment speed:
# decrease from default=60 towards 0
# or
# data treatment speed preferred over responsiveness of system:
# increase from default=60 towards 100
#
# 0 = disable any swap activity
# 1 = minimum swap activity, helps to avoid running out of RAM
vm.swappiness = 1

# lower value in order to prefer inode/dentry cache to stay in cache
# for finding files faster; (default=100) (could be > 100)
vm.vfs_cache_pressure=25



Good Luck! Marco.

[toc] | [prev] | [next] | [standalone]

#221148

On Fri, Apr 10, 2020 at 09:26:44AM +0300, dalios wrote:
> * The computer is used for basic office tasks and the usual internet
>   browsing (sometimes with many firefox tabs open). I am not using any
>   particularly heavy programs like scientific computing, video rendering
>   etc. However I am experimenting sometimes with virtual machines
>   (usually with qemu-kvm). Not something professional, just as a hobby
>   (I am trying to teach my self things by performing tasks that I
>   wouldn't normally try on my main system).
> * With the 4 GB RAM that I now have available, I have noticed my PC
>   becoming slow and the memory usage going quite high. After booting,
>   with just Gnome running, I see conky reporting that memory usage goes
>   to something like 800 MB. Firefox adds ~800MB more and Thunderbird
>   adds up to a total of more than 2 or maybe 2,5 GB! After launching
>   Transmission (bit torrent client) I see that I have to restart Firefox
>   quite often in order for the PC to be usable (especially after
>   browsing a little and having many tabs open or even after some tabs
>   have been closed). I have noticed that things got worst after
>   upgrading to Stable (I was on oldstable until recently).
> * I haven't been checking on swap usage a lot so I don't know how is
>   that used.

Given the above, the odds are that a benchmark will not provide you any
meaningful absolute metrics.  That said, tripling your memory will
almost certainly result in your computer rarely touching swap.  That
alone, will make a huge difference in your day-to-day computing
experience.  Additionally, the benchmark may still be a fun "gee whiz"
sort of thing to do.  It can be nice to get some idea of how your
machine performs in a quantitative sense.  It will certainly let you
establish a baseline against which you can compare future upgrades or
future replacement systems.

Regards,

-Roberto
-- 
Roberto C. Sánchez

[toc] | [prev] | [next] | [standalone]

#221156 — xpdf woes, was Re: New RAM, does Debian has a tool to benchmark?

On Fri 10 Apr 2020 at 09:26:44 (+0300), dalios wrote:

[…]
> * With the 4 GB RAM that I now have available, I have noticed my PC
>   becoming slow and the memory usage going quite high. After booting,
>   with just Gnome running, I see conky reporting that memory usage goes
>   to something like 800 MB. Firefox adds ~800MB more and Thunderbird
>   adds up to a total of more than 2 or maybe 2,5 GB! After launching
>   Transmission (bit torrent client) I see that I have to restart Firefox
>   quite often in order for the PC to be usable (especially after
>   browsing a little and having many tabs open or even after some tabs
>   have been closed). I have noticed that things got worst after
>   upgrading to Stable (I was on oldstable until recently).

If things have got worse with stable/buster, is it possible that you
use xpdf to browse sizeable documents?

I find 4GB very adequate for browsing with lots of tabs etc, but
reading PDFs (like many people, I've been reading some large reports
and court filings over the last year or two) on buster can bring the
computer to its knees quite rapidly. When I switch back to Firefox,
fvwm displays its window, but Firefox can take half a minute or more
just to rerender itself.

The bugs concerned are
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=926501
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=942086
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=945188

As my workaround, I've modified mc's mc.ext file to open PDFs
with xpdf when I press Return, but with zathura when I press F3.
(But I miss my customisations of xpdf with the latter.)

> * I haven't been checking on swap usage a lot so I don't know how is
>   that used.

Swap growth is a clear symptom of the above if it's happening.

> * I have two HDD hard drives installed in the system.
> * Up to now I have only been using Memtest86+ to check the memory's
>   integrity but I didn't know that it can be used to see other
>   performance metrics. I will check that now.
> * I will also the dbench (which I haven't heard of before) as well as
>   use the apt-cache command you suggested. If the information I give in
>   this message helps you to suggest something specific please do, as
>   hardware in general and especially benchmarks is not something I am
>   particularly familiar with.

Personally, I've never bothered; I rely on the performance
I experience while doing my habitual tasks. But that's partly
because I don't have an easy way to remedy matters.

Cheers,
David.

[toc] | [prev] | [next] | [standalone]

#221180

On 2020-04-09 23:26, dalios wrote:
> Thank you to both of you for your answers. I am grateful for your time.
> 
> My PC is a desktop PC, see hardware details at the end of this email.
> 
> 
> Here is more info, answering your questions:
> 
> * I purchased a single 8 GB module and currently the PC has a 4 GB
>    module. 

Get an anti-static wrist strap and practice static-safe techniques 
whenever handling memory modules, motherboards, etc..


Also, never touch electrical contacts -- the acid in your skin oil will 
corrode them.


Manufacturer and part number of the 4 GB module?


> My original plan was to buy two identical 8 GB modules but I
>    decided to buy an 8 GB module now and the second later. I am aware
>    that since the two modules (8+4) are not identical there is a
>    possibility that they would not perform at their maximum speed.
>    However I believe that 8+4 is better than 4 or 8 GB, am I right on
>    this?
> * The 8 GB module that I purchased is this:
>    <https://www.crucial.com/memory/ddr3/ct102464bd160b/ct8815800>

Crucial 8GB DDR3L-1600 UDIMM
CT102464BD160B


How did you select this particular part?


> * The computer is used for basic office tasks and the usual internet
>    browsing (sometimes with many firefox tabs open). I am not using any
>    particularly heavy programs like scientific computing, video rendering
>    etc. However I am experimenting sometimes with virtual machines
>    (usually with qemu-kvm). Not something professional, just as a hobby
>    (I am trying to teach my self things by performing tasks that I
>    wouldn't normally try on my main system).
> * With the 4 GB RAM that I now have available, I have noticed my PC
>    becoming slow and the memory usage going quite high. After booting,
>    with just Gnome running, I see conky reporting that memory usage goes
>    to something like 800 MB. Firefox adds ~800MB more and Thunderbird
>    adds up to a total of more than 2 or maybe 2,5 GB! After launching
>    Transmission (bit torrent client) I see that I have to restart Firefox
>    quite often in order for the PC to be usable (especially after
>    browsing a little and having many tabs open or even after some tabs
>    have been closed). I have noticed that things got worst after
>    upgrading to Stable (I was on oldstable until recently).

https://en.wikipedia.org/wiki/Wirth%27s_law


My laptop is maxed out at 2 GB.  If I open more than a few browser 
windows with heavy JavaScript, the computer slows to a snail's pace.


> * I haven't been checking on swap usage a lot so I don't know how is
>    that used.

top(1) displays swap statistics.


I use Xfce.  The System Load Monitor panel plug-in provides a real-time 
bargraph display.


> * I have two HDD hard drives installed in the system.

Manufacturers and part numbers?


> * Up to now I have only been using Memtest86+ to check the memory's
>    integrity but I didn't know that it can be used to see other
>    performance metrics. I will check that now.
> * I will also the dbench (which I haven't heard of before) as well as
>    use the apt-cache command you suggested. If the information I give in
>    this message helps you to suggest something specific please do, as
>    hardware in general and especially benchmarks is not something I am
>    particularly familiar with.

Save dbench and the others for later.


> 
> David Christensen wrote:
>> [...]
>> Tripling your memory for $40
>> should be a worthwhile upgrade.  You should notice it when you run a
>> graphical desktop and open a lot of applications/ tabs.
> 
> That sounds encouraging and that is exactly what I am after but I
> thought I could ask here for benchmark software suggestion to see if I
> can get something measurable. (and hopefully learns something on the way).
> 
> 
> Thanks again,
> Dalios
> 
> 
> -------------------------------------
> 
> Motherboard
>      Manufacturer: Pegatron
>      Form factor: uATX - 19.3x18.0 cm (7.6x7.1 in)
>      Chipset: Intel H81
>      Memory sockets: 2 x DDR3
>      Processor socket: LGA 1150

Pegatron appears to be an OEM manufacturer.  I cannot find 
specifications given the above information.  What is the part number on 
the motherboard?  Was the computer assembled from parts, or bought as a 
system?  If the latter, what is the make and model number on the system 
case?


> 
> Processor
> Intel Core i3-4170
>      CPU speed: 3.7 GHz
>      CPU cores: 2
>      CPU Cache: 3 MB
>      TDP: 54 W
>      Socket: LGA 1150

https://ark.intel.com/content/www/us/en/ark/products/77490/intel-core-i3-4170-processor-3m-cache-3-70-ghz.html


That processor should be more than adequate for Debian graphical desktop 
use.


Comparing the processor Memory Specifications against the Crucial 8 GB 
Product Specifications, that memory module is probably okay.  But, the 
motherboard specifications would be more authoritative.


That processor supports VT-x and VT-x with EPT, but not VT-d.  It should 
be adequate for virtual machines that are not I/O heavy.


> 
> Memory
>      Amount: 4 GB
>      Speed: PC3-12800 MB/s
>      Type: DDR3-1600
> 
> Video Graphics
> Intel HD Integrated Graphics

Intel® HD Graphics 4400


That should be adequate for desktop use.


Because memory is shared between the processor and graphics, higher 
graphics resolutions and/or refresh rates will show down the processor.


> Hard Drive
>      Size: 1 TB
>      Interface: SATA
>      Rotational Speed: 7200 rpm
> 


David

[toc] | [prev] | [next] | [standalone]

#221188

On Fri, 2020-04-10 at 20:01 -0700, David Christensen wrote:
> My laptop is maxed out at 2 GB.  If I open more than a few browser 
> windows with heavy JavaScript, the computer slows to a snail's pace.

Would the amount of RAM affect browser JavaScript performance? Is that
due to JIT compiler results not being able to be cached or something
like that? I've just assumed slow performance it due to CPU throughput.

-- 
Tixy

[toc] | [prev] | [next] | [standalone]

#221191 — Improving performance: RAM or CPU speed (was: Re: New RAM, does Debian has a tool to benchmark?)

On Saturday, April 11, 2020 05:45:56 AM Tixy wrote:
> On Fri, 2020-04-10 at 20:01 -0700, David Christensen wrote:
> > My laptop is maxed out at 2 GB.  If I open more than a few browser
> > windows with heavy JavaScript, the computer slows to a snail's pace.
> 
> Would the amount of RAM affect browser JavaScript performance? Is that
> due to JIT compiler results not being able to be cached or something
> like that? I've just assumed slow performance it due to CPU throughput.

Back in the old days (when I found my computers slower than I liked), I found 
that adding more RAM was usually a better approach to improving performance 
than getting a faster CPU.  Not sure how to quantify that, but I like to have 
at least 12 GB in any computer I use.

That was especially true when CPUs started to reach the point (around 3 GHz., 
iirc) that their improvement in speed slowed down.  (But it was also true 
before that.)

If I had a computer that I was dissatisfied with the performance, I would 
definitely add RAM to try to get up to 16 GB or so.  Only then would I think 
about getting a faster CPU.

Aside: The idea of extra cores sounds good to me, but I don't know how much 
software actually takes advantage of this (or if the OSs / kernels) are 
typically smart enough to distribute tasks across multiple cores -- I hope 
they are, but really haven't been paying attention.

I think my bottleneck these days is again RAM, on my daily driver I have 16 
GB, but sometimes have 1000 or more tabs "open" in Firefox (on Wheezy).  
("Open" is a little misleading -- occasionally Firefox crashes.  When it does, 
I restart it and choose the option to restore all tabs (or whatever it says), 
the old tabs show up with their names, but the content from those URLS is not 
actually loaded (or fetched) until I click on one of those tabs to view it 
again.

(Actually sort of a nice feature -- it lets me keep easy access to lots of 
pages without filling up RAM.)

[toc] | [prev] | [next] | [standalone]

#221192 — Re: Improving performance: RAM or CPU speed (was: Re: New RAM, does Debian has a tool to benchmark?)

rhkramer@gmail.com wrote:

> I think my bottleneck these days is again RAM, on my daily driver I have
> 16 GB, but sometimes have 1000 or more tabs "open" in Firefox (on Wheezy).
> ("Open" is a little misleading -- occasionally Firefox crashes.  When it
> does, I restart it and choose the option to restore all tabs (or whatever
> it says), the old tabs show up with their names, but the content from
> those URLS is not actually loaded (or fetched) until I click on one of
> those tabs to view it again.

Look at the L1 L2 L3 cache. Many people underestimate this

here are two examples from different pcs

# lscpu | grep cache
L1d cache:           16K
L1i cache:           64K
L2 cache:            2048K
L3 cache:            8192K


# lscpu | grep cache
L1d cache:           32K
L1i cache:           32K
L2 cache:            256K
L3 cache:            6144K

[toc] | [prev] | [next] | [standalone]

#221193 — Re: Improving performance: RAM or CPU speed (was: Re: New RAM, does Debian has a tool to benchmark?)

On Saturday, April 11, 2020 09:39:06 AM deloptes wrote:
> rhkramer@gmail.com wrote:
> > I think my bottleneck these days is again RAM

> Look at the L1 L2 L3 cache. Many people underestimate this
> 
> here are two examples from different pcs
> 
> # lscpu | grep cache
> L1d cache:           16K
> L1i cache:           64K
> L2 cache:            2048K
> L3 cache:            8192K
> 
> 
> # lscpu | grep cache
> L1d cache:           32K
> L1i cache:           32K
> L2 cache:            256K
> L3 cache:            6144K

Thanks for the reply -- interesting.  

Two of my machines (Wheezy and Buster) don't have L3 cache, my Jessie machine 
does.

I see varying definitions / explanations of L3 cache -- some say it is on the 
CPU die, and some say it is on the motherboard (but, in either case, one level 
further away from the CPU(s) (mcow).)

And, also in either case, not easy to upgrade short of replacing the CPU or 
motherboard.

RAM is still (to me) the most cost effective upgrade to an existing system.

(I guess I could mention that I don't play games, so video performance is not 
a key for me (and, when I do / did play. games, video performance was not a 
key -- games like FreeCiv and such).

I am considering a new machine at this time, a laptop with a good docking 
system (I'm looking for one with a USB-C port) so that I can easily undock it 
and take it to meetings, but use it at home for development with GCC 8 (or 
high 7) -- I want to do some work on a program (scintilla) that now uses 
features of C++ versions 11, 14, and 17.  (And I was thinking of taking the 
laptop to meetings where I might get help, but who knows when any such 
meetings will take place again.)

(I actually bought a machine (mail order that advertised a USB-C port but have 
to return it as it doesn't have a USB-C port.)

I guess I should say that I don't have objectional performance with any of the 
machines I have -- well, maybe I should take that back -- I keep the Wheezy 
system as my daily driver for several reasons -- including laziness, but also 
it seems better performance with kmail and firefox.

When I have as few as 10 to 15 tabs open on the Firefox on my Jessie system, 
Firefox crashes (I mentioned in a previous post in this thread having a 
thousand or more tabs "open" in Firefox on Wheezy with minimal problems -- yes 
occasional crashes (maybe once every 2 to 4 weeks?), but I couldn't come near 
that on Jessie, and I expect the same problem on Buster.

I guess that is all related to Wirth's law (and variations, Gates', ...).

[toc] | [prev] | [next] | [standalone]

#221206 — Re: Improving performance: RAM or CPU speed

rhkramer writes:
> When I have as few as 10 to 15 tabs open on the Firefox on my Jessie
> system, Firefox crashes (I mentioned in a previous post in this thread
> having a thousand or more tabs "open" in Firefox on Wheezy with
> minimal problems -- yes occasional crashes (maybe once every 2 to 4
> weeks?), but I couldn't come near that on Jessie, and I expect the
> same problem on Buster.

What Firefox version?  I used to see frequent Firefox crashes
(apparently memory leaks: my system is always up) but that stopped a few
years ago (though the irritation of that has been replaced by the
irritation of being required to create a new profile every time I
upgrade Firefox).
-- 
John Hasler 
jhasler@newsguy.com
Elmwood, WI USA

[toc] | [prev] | [next] | [standalone]

#221227 — Re: Improving performance: RAM or CPU speed

John Hasler wrote:

> What Firefox version?  I used to see frequent Firefox crashes
> (apparently memory leaks: my system is always up) but that stopped a few
> years ago (though the irritation of that has been replaced by the
> irritation of being required to create a new profile every time I
> upgrade Firefox).

but you can also import and activate your old profile

https://support.mozilla.org/en-US/kb/profile-manager-create-remove-switch-firefox-profiles

[toc] | [prev] | [next] | [standalone]

#221240 — Re: Improving performance: RAM or CPU speed

I wrote:
> What Firefox version?  I used to see frequent Firefox crashes
> (apparently memory leaks: my system is always up) but that stopped a few
> years ago (though the irritation of that has been replaced by the
> irritation of being required to create a new profile every time I
> upgrade Firefox).

deloptes writes:
> but you can also import and activate your old profile

That's what I'm talking about.  The process consists essentially of
creating a new profile and importing the old one into it.  A pointless
song and dance that Firefox could do automatically in the background.
At the very least they could offer a simple dialog along the lines of
"Which of these old profiles do you wish to import?"

There was no warning when they started doing this.  The first time it
hit I was told that I should import my old profile via my nonexistent
"Firefox account".  Not going to happen.
-- 
John Hasler 
jhasler@newsguy.com
Elmwood, WI USA

[toc] | [prev] | [next] | [standalone]

#221233 — Re: Improving performance: RAM or CPU speed

On Sat 11 Apr 2020 at 10:21:42 (-0500), John Hasler wrote:
> rhkramer writes:
> > When I have as few as 10 to 15 tabs open on the Firefox on my Jessie
> > system, Firefox crashes (I mentioned in a previous post in this thread
> > having a thousand or more tabs "open" in Firefox on Wheezy with
> > minimal problems -- yes occasional crashes (maybe once every 2 to 4
> > weeks?), but I couldn't come near that on Jessie, and I expect the
> > same problem on Buster.
> 
> What Firefox version?  I used to see frequent Firefox crashes
> (apparently memory leaks: my system is always up) but that stopped a few
> years ago (though the irritation of that has been replaced by the
> irritation of being required to create a new profile every time I
> upgrade Firefox).

What sort of upgrade? Just the regular security fixes (which seem to
be quite frequent recently), or point-releases, or what?
And what parts of your profile do you lose? Bookmarks, cookies (like
the ones you might want to keep for logins) or what?

It's just that I've been using the same two profiles (two users) for
years. I've copied them from machine to machine as my principal
workstation has changed, and I copy them to other machines so that
I don't have to find lots of websites I use fairly regularly. (They
gradually diverge, of course, from the canonical copy on my principal
workstation.)

Some of the files in the profiles are pretty ancient and obviously
were converted into other formats but not cleaned up. I mean, what
do these do?

  1592146 Jul  8  2013 urlclassifier.pset
 54059008 Jul  8  2013 urlclassifier3.sqlite
     7099 Feb 25  2013 sessionstore-2.js
   101823 Jan 14  2013 xpti.dat
   151170 Jan 14  2013 compreg.dat
     1024 May  7  2012 old-pluginreg.dat
    29034 Apr 18  2012 sessionstore-1.js
    16384 Aug 22  2011 secmod.db

Cheers,
David.

[toc] | [prev] | [next] | [standalone]

#221241 — Re: Improving performance: RAM or CPU speed

David Wright writes:
> What sort of upgrade? Just the regular security fixes (which seem to
> be quite frequent recently), or point-releases, or what?  And what
> parts of your profile do you lose? Bookmarks, cookies (like the ones
> you might want to keep for logins) or what?

Upstream releases (these evidently happen about every fifteen minutes
but I only upgrade every few months or so).

I don't lose anything.  I just have to go through a silly song and dance
to import my old profile (which took a bit of research to discover as I
have never made use of the multiple profiles feature).  It's not clear
to me why importing profiles when upgrading should require any user
input, especially when there is only one profile.
-- 
John Hasler 
jhasler@newsguy.com
Elmwood, WI USA

[toc] | [prev] | [next] | [standalone]

#221255 — Re: Improving performance: RAM or CPU speed

[Multipart message — attachments visible in raw view] — view raw

On Sb, 11 apr 20, 18:43:38, John Hasler wrote:
> David Wright writes:
> > What sort of upgrade? Just the regular security fixes (which seem to
> > be quite frequent recently), or point-releases, or what?  And what
> > parts of your profile do you lose? Bookmarks, cookies (like the ones
> > you might want to keep for logins) or what?
> 
> Upstream releases (these evidently happen about every fifteen minutes
> but I only upgrade every few months or so).

How do you upgrade / from where do you get the packages? (apt, manually 
installed deb, etc.)

Kind regards,
Andrei
-- 
http://wiki.debian.org/FAQsFromDebianUser

[toc] | [prev] | [next] | [standalone]

#221344 — Re: Improving performance: RAM or CPU speed

Andrei writes:
> How do you upgrade / from where do you get the packages? (apt,
> manually installed deb, etc.)

Apt from Debian/Sid.  Just did it a couple of days ago.
-- 
John Hasler 
jhasler@newsguy.com
Elmwood, WI USA

[toc] | [prev] | [next] | [standalone]

#221236 — Re: Improving performance: RAM or CPU speed

On Sat, 11 Apr 2020 10:21:42 -0500
John Hasler <jhasler@newsguy.com> wrote:


> What Firefox version?  I used to see frequent Firefox crashes
> (apparently memory leaks: my system is always up) but that stopped a
> few years ago (though the irritation of that has been replaced by the
> irritation of being required to create a new profile every time I
> upgrade Firefox)

My firefox profile is never lost.
 
I copy my .mozilla directory and take that with me to another town.
On the PC there I replace my .mozilla directory and start firefox:

Result:
Everything is restored: history, bookmarks, settings, logins/passwords, 
even the open tabs.

When I return to my town I take with me my new up to date .mozilla
directory.

Firefox is a very convenient browser nowadays.

I do not take my entire home directory with me, for some reasons.

PS. I believe you have to use the same version of firefox on both
machines for this to work.

[toc] | [prev] | [next] | [standalone]

Page 1 of 2  [1] 2  Next page →

Back to top | Article view | linux.debian.user

csiph-web