Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Starver]

Not that it adds much to the laptop v. desktop debate, but cat-owners-who-smoke clog up their cooling fans far more than most (for obvious reasons), whichever machine type. But you are certainly advised not to habitually use the laptop in bed[1] as being sat on the duvet/whatever tends to make a prime source for much more dust (microfibres, even if not bodily dust) just ripe to be sucked into the vents, which are more designed for operation on a smooth, flat, hard surface so make have to work harder.

Just some (additional) additional experience gained from sorting out diverse problems for others.


(The self-UPS capabity is a good additional point. My big excuse for missing it is that I use a laptop so old that the battery has died (and never got around to getting a replacement, knowing I'm using it as a handy-sized house-to-house portable legacy desktop anyway, and power supply being rarely off in this more fortunate place) so it's removed to prevent overheating and I always just plug it in.)


[1] Aside from the risk of snoozy crushing, or indeed any other reason you might imagine that involves the inbuilt cameras. Also, it's probably bad for sleep patterns. In both these issues. )

[delphonso]

My computer is by no means good, but something is up.

Playing DF on this laptop has required some sacrifices - small maps, pocket worlds, etc. It's manageable. It was, at least, on Windows. This summer I swapped to Ubuntu and didn't play DF until recently. Almost immediately I noticed a very sickening effect. The game chugs. I mean chugs. It runs at a normal, slowish speed, but then gets like...a pulse of speed. Everything ticks really quickly for half a second, then the game returns to normal. It's bad enough I had to lower the FPS limit - putting myself into self-induced FPS death.

Any clues what this could be and how I can...fix it?
I have 8 gigs of ram, of which about 1.5 is used.
My processor is a hearty 1.10GHz - 2 cores. One core is constantly at 95~% usage, even without DF running. The other fluctuates from 10% to 60% and I suspect is the source of this issue.

EDIT:
I installed the debian package of DF (out of date, 47.04...) and it does the same, so it's definitely the computer, not the files themselves.
The normal speed is satisfactory, but the speed-ups are actually nauseating. I limited FPS to 30 and still get the pulses occasionally, but it seems to be less frequent. A larger fortress doesn't get it, as it struggles to meet 30 FPS.

[dragdeler]

I know this opinion is entirely useless but whenever I hear biiig linux fanboys complain about bloat in windows, I think to myself: "yeah as if your distro is leveraging all system ressources with them generic drivers" ... You win some you loose some.


Also 8gig seems gigantic for that laptop you been upgrading havent you  ?

[wierd]

I would check these things:

1) Since this is a laptop, is CPU throttling happening?  Often, linux distros set rather.... ... ... aggressive... throttling settings for the CPU when it detects that it has been installed on a laptop.
  https://github.com/erpalma/throttled#static-fix

2) is ZRAM enabled? If it is, temporarily disable it, and see if your problem goes away.  (get output of [cat /proc/swaps], and see if /dev/zram0 is a swap device.)

[feelotraveller]

In terms of addressing the symptoms have you also tried reducing the G_FPS_CAP?  Usual advice is to set to a multiple of the FPS setting, so try 30 or even 15.

[delphonso]

In terms of addressing the symptoms have you also tried reducing the G_FPS_CAP?  Usual advice is to set to a multiple of the FPS setting, so try 30 or even 15.

Yeah, sorry for the lack of clarity, I lowered both to 30.

I would check these things:

1) Since this is a laptop, is CPU throttling happening?  Often, linux distros set rather.... ... ... aggressive... throttling settings for the CPU when it detects that it has been installed on a laptop.
  https://github.com/erpalma/throttled#static-fix

2) is ZRAM enabled? If it is, temporarily disable it, and see if your problem goes away.  (get output of [cat /proc/swaps], and see if /dev/zram0 is a swap device.)

I am running an update right now, but will check these two after the fact - I would guess it's throttling, as the RAM seems fine.

Also 8gig seems gigantic for that laptop you been upgrading havent you  ?

Such a thing would certainly void my warranty, so surely not, never.


EDIT:
Upgrading to Ubuntu 21.10 seems to have fixed whatever the issue is. CPU load is equal among the cores - so no more 100% on one and 10-60% on the other, with that the pulsing is gone. Wish I knew exactly what caused it, but happy to be rid of it. I'll keep an eye on it either way.

Now I can enjoy FPS death the old fashion way.

[Solifuge]

You will need to compile this driver, and install it.

https://www.realtek.com/en/directly-download?downloadid=c4676ea8d1adc3c6f28a4ae2c3e2787a

Forgot to follow up on this. Thanks so much! The link was broken, but I was able to nab the driver and get it working on Linux, just in time for my Windows 10 install to get critically unstable (TL;DR, aging GPU issues). Been happily working in Linux in a compatability mode that doesn't stress my GPU, without the system crashing all the time. Thanks a bunch!

[nickbii]

I'm interested in people who have done speed tests. I know that RAM Speed is very important for DF, but I don't know whether it's RAM Timings or the actual MhZ rating of the RAM stick that counts. I'm assuming that dual-channel is also a no-brainer.

Also: how do the Apple M1 processors stack up?

[wierd]

The frontside bus speed has not really changed all that horrible much for a decade or more.

What has changed, is the width of the memory read window, and the number of concurrent operations that occur on each memory read operation.


This is is what things like "quad channel memory" are all about.

[nickbii]

The frontside bus speed has not really changed all that horrible much for a decade or more.

What has changed, is the width of the memory read window, and the number of concurrent operations that occur on each memory read operation.


This is is what things like "quad channel memory" are all about.

So getting a motherboard that supports quad-channel, and then filling the four slots with the standard 1600, gonna max out FPS? Or does the new DDR5 with 5200 do better? Or should I focus on timings?

Or do we just not have data that detailed yet?

[Schmaven]

The frontside bus speed has not really changed all that horrible much for a decade or more.

What has changed, is the width of the memory read window, and the number of concurrent operations that occur on each memory read operation.


This is is what things like "quad channel memory" are all about.

So getting a motherboard that supports quad-channel, and then filling the four slots with the standard 1600, gonna max out FPS? Or does the new DDR5 with 5200 do better? Or should I focus on timings?

Or do we just not have data that detailed yet?

I may be wrong, but it was my understanding that because the data DF processes is so small, that it's more a timing + speed issue than a size issue when it comes to RAM.  And clock speed + timing are definitely related.  If your ram operates at 2GHz, with an 8-8-8-20 timing, that would be roughly equivalent to RAM operatimg at just 1GHz but with 4-4-4-10 timing.  Similarly, 2GHz RAM with 7-7-7-15 timing would be quicker, as would 3GHz RAM with 11-10-10-28 timing.  You typically have to work out the algebra yourself as the manufacturer's specs typically just give raw measurements.

I don't know how multiple sticks of RAM factor in, this is just a stick to stick comparison.

[wierd]

modern CPUs use prefetching to get data into the cpu before the program actually wants it, so that it is in cache, and ready to roll.

This is where multi-channel memory comes in, as the processor can fetch more than just one data element during the mem-read operation, and have it waiting in the wings and ready to go.  The size of the channel can easily offset a slower rate.

Think, "timing rate == how fast a bucket can be scooped in and out" and "channel width == how many buckets can be moved in and out at the same time".

There are functional limits to how fast you can drive silicon, so intel and pals have been improving the width, and the prefetching computation logistics.  As long as DF has the data it wants in the processor cache, the issue will not be memory bound IO contention, but processor bound computation contention on a single thread.

[LordBaal]

Lets say you are moving places. Which will you chose, a single moving truck up which can fit many things but you are stuck to a single trip each time or two or four pick up trucks?

Also be mindful that not all CPUs and motherboards support dual and quad. Some support dual only. Or at least back in my time. I'm ridiculously out of date when it comes to hardware.

[Starver]

It also depends on what  memory you want to read/write, how much and from where.

If you have an Panamamax containership then it can have containers fairly quickly unloaded/loaded from it (ignore stacking order, quayside or onboard, for this analogy) with maybe 20t or so of contents at a time, perhaps several such containers at a time by several of the parallel servicing cranes, which is a lot of shifting around well within usual capabilities, and well worth the time to pack the container elsewhere and use road/rail links to get it to the port, and at the receiving port get it onto another flatbed truck/carriage to get to the distribution depot.

But if you only want to work with several disparate items that might have been scattered across many containers because it seemed logical at the time (e.g. you need to check/replace a single batch-line of goods that were constructed to the wrong spec on one of your manufacturing conveyors, but mixed in with the rest) then it might be easier (but not usually authorised!) to crack the container on the ship, or in the port grounds, and use wheelbarrows/pallette-trucks to move the requisite items out/any replacements in that you sent along in a transit van.

The highly efficient 20t-lifting cargo-handles may be overkill for the job, or even take more time.


Of course, if you're normally happy to get the sheer throughput of mass-tramsfer of data without many niggly exceptions such as this to spoil the perfection of it, then it's good. And the predictive(/look-ahead) mechanisms can probably help even when oddities are called for (needs 'understanding' by everything from the compiler to the on-chip microcode to work best, somehow anticipating things that the programmer may even expect to happen, just by examining what he ended up asking for (which might imperfectly express his expectations, or he could be wrong... How advanced is the meta-programmers' art such that they allow for this, I don't know!).


Not that it's a bad thing, in general, I reemphisise, I'm just positing a feasible outlier situation.

[King Zultan]

I got a ZTE Zmax Z970 at a car boot sale that I was intending to use it as something I can put videos on it to watch when I don't have my laptop and was wondering if there is something I can do to get around the 32Gig Micro SD card limit?