What makes a computer "Snappy"?

[Jae Tee]

1 minute ago, AlexTheGreatish said:

There also seems to be a bit of an X-Factor.  Snip.

Memory controller is a good one, but in cpu speed do you mean the actual ghz?

[Jae Tee]

12 minutes ago, Mira Yurizaki said:

If we were to put an objective metric, " .snip.

So moving on to the 21st century, assuming I only need graphics for a general display would a discrete GPU be better or worse (or neither) then integrated? And on that note, Intel i-GPU vs. AMD APU? 

 

 

Ps. This is just a general example, NOT my whole question. 

[AlexTheGreatish]

10 minutes ago, Jae Tee said:

Memory controller is a good one, but in cpu speed do you mean the actual ghz?

Yeah GHz.  Like on some computers you can feel them wake up with an overclock (especially with like 6th or 7th gen if you got them to 5GHz-ish)

[GDRRiley]

1 minute ago, AlexTheGreatish said:

Yeah GHz.  Like on some computers you can feel them wake up with an overclock (especially with like 6th or 7th gen if you got them to 5GHz-ish)

good luck on 6th gen. a few 6700k will do that.

[Jae Tee]

11 minutes ago, AlexTheGreatish said:

Yeah GHz.  Like on some computers you can feel them wake up with an overclock (especially with like 6th or 7th gen if you got them to 5GHz-ish)

Cool.

 

BTW. Big fan of your work

[Kisai]

56 minutes ago, Jae Tee said:

So what actually makes a pc or laptop snappy? Obviously we'd be talking about high performance ssd's, and good ram. But i feel like there's more to it. Are "powerful" CPU'S part of it, or what?

It's dependent on a lot of things. What makes a PC feel sluggish are having all sorts of background software running.

 

Like a stock Win10 installation, or a Linux installation, or MacOS X installation feels snappy because nothing has been installed on it. 

 

The office machines I work on, don't feel snappy because they spend a good minute running scripts and loading all the business security software and such.

[MaZeeT]

Also be aware where your background software is storage... My system felt sluggish until I figured out launching steam + gog + battle.net from a HDD made the load into windows sluggish as F.

I moved the clients to the OS SSD but kept the game libraries on the HDD and it have been a lot better since.

 

[Mira Yurizaki]

8 hours ago, Jae Tee said:

So moving on to the 21st century, assuming I only need graphics for a general display would a discrete GPU be better or worse (or neither) then integrated? And on that note, Intel i-GPU vs. AMD APU?

An iGPU technically wins by locality, but I don't think it practically matters in the end.

[PianoPlayer88Key]

 

10 hours ago, Ryan_Vickers said:

…fast storage …not having enough RAM …

 

9 hours ago, Slottr said:

… how much ram(and it’s speed), what SSD, what type of SSD…

I've recently been thinking ... Which of these two scenarios would feel snappier?  First off, assume your current workload (which doesn't include gaming, but more like web browsing, office, maybe photo or light video editing or whatever) needs / uses about 12 GB of RAM.

• (Slow single-channel RAM, but having enough) - A single 16GB stick of DDR4-2133 RAM in your system, running at, say, CL22, OR,
• (fast dual-channel RAM, but NOT enough) - two 4GB sticks (totaling 8GB) of like DDR4-4500 CL18 or whatever, and your page file on a shingled 5400rpm <=2TB hard drive, or even on an IDE / PATA hard drive (via a PCIe to IDE controller)

 

Also speaking of not having enough RAM, where would the following types of devices on which to store a page file rank in relative "snappiness", compared to actually having enough RAM?  (For example you could set the baseline "having enough ram" to 100.00% performance / snappiness "feel", then rank the others below that accordingly.)

• IDE hard drive
• 5400rpm shingled SATA "budget/cheap" hard drive
• "performance" grade 7200rpm SATA / 10k RPM SAS hard drive
• SSHD
• budget QLC DRAMless SATA SSD (like Tier E)
• fairly premium SATA SSD (like a Crucial MX500 or Samsung 860 EVO)
• budget QLC DRAMless NVMe SSD (like Tier E)
• midrange TLC NVMe SSD (Tier C perhaps?)
• fairly decent MLC NVMe SSD (like the Samsung 970 Pro), or a PCIe 4.0 SSD or another one that claims half a million or more IOPS
• an Optane 905p or similar SSD

For example, would even a 1TB 970 Evo I already have (but am not booted off of) be fast enough so that if it was housing the pagefile, it'd "feel" close enough to the speed of RAM to mitigate the fact that the system I have it in is already maxed out on the amount of RAM that can be installed?  (I posted a screenshot in a spoiler a bit farther down, showing memory usage.)

Or another way to put it... would the performance of pagefile on the 970 Evo be fairly significantly closer to that of having enough RAM, than to that of having pagefile on a spinning hard drive?

BTW in the screenshot below with task manager open, drive C is a M.2 250GB Crucial MX200, drives D & E are 2.5" 1050GB Crucial MX300, and drive H is a M.2 1TB Samsung 970 Evo.

 

 

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

 

 

9 hours ago, Jae Tee said:

I like your version of normal, should run chrome given enough RAM.

Actually, I got a bit of a surprise recently when I was doing some browsing in FireFox.  (I normally use Chrome, but it gets sluggish when processes approach 2GB RAM and crash if a process ever hits 4GB, even though I have 64GB installed and might have had 20 or 30 GB still available.)  I was looking up houses in a few areas on a few search sites, then noticed my system was feeling a bit sluggish.  I clicked open Task Manager (was running in the background) and was greeted with:

Spoiler

 

 

Also, after I saved and closed a VM that was running (with 12GB RAM assigned to it), this is what happened to my RAM usage.

Spoiler

 

(While it was saving, drive E was pinned solid at 100%, basically storing the equivalent of the VM's hibernation file on it.)

 

 

 

====================

 

 

 

9 hours ago, Mira Yurizaki said:

…Ironically older systems, like 8-bit computers of the 80s, are actually snappier than modern systems. Why? It's a combination of …

 

This reminded me of another metric of performance I feel has gotten a LOT slower - that is, time to fill a relatively high capacity storage drive (relative to what's available when the drive is new) to full capacity, from completely empty.

It took me a full 24-hour day, maybe a bit longer, to do some task that wrote the entire capacity of one of my 10TB 7200rpm HGST Deskstar NAS hard drives.

An article on Tom's Hardware mentions that it took about 37 seconds to write 26 MB on a 40 MB hard drive.  That would work out to 57 seconds to fill the entire drive.

 

I wouldn't be surprised if some of the really old MFM hard drives from the early 1980s, like 5 and 10 MB capacities, would have taken even less time to fill to capacity.  For example, if a 5 megabyte HDD could saturate the interface (which was 5 megabits/second), it would fill it in 8 seconds.  I'd guess in the real world it was a bit slower than that, but still a LOT faster than a modern hard drive or even SSD in filling a drive to capacity from empty.

 

I wonder how long until we have fast enough SSDs so that even the highest capacity ones available at budget prices would be faster than the old MFM hard drives, in time to fill the entire drive?

 

 

 

===================

 

 

I also just remembered another situation in which performance was totally different than I had expected, and now I"m bringing games into the discussion.

Test system was my desktop, with an ASRock Z97 Extreme6 mobo, i7-4790K running at stock under a Hyper 212 Evo, 32GB (4x8GB) of DDR3-1600 (CL9 G.Skill Ares) RAM, booting from a 256GB Crucial M550 SSD, game might have been on a 4TB, 5TB, or 8TB 7200rpm hard drive, hooked up to a 24" 1080p 60Hz monitor (Dell U2414H). (I kinda doubt there was room on the SSD for it at the time.)

 

Game was GTA V, running with the following maxed out settings, which used about 9 GB VRAM.

Spoiler

First off, the highlighted setting in this screenshot had the biggest impact on VRAM usage.

 

 

The other 3 in the nested spoiler are there for completeness, to show ALL graphics settings used.

Spoiler

 

I ran the built-in benchmark on two different GPUs.

 

With the integrated HD 4600 graphics, I was getting about 1.8 fps at the beginning of the benchmark run (the part where it's swooping in on the houses).

With my EVGA Superclocked GTX 1060 3GB, I was getting ZERO POINT THREE (0.3) fps!!

 

Normally, if I run at more reasonable settings like 1080p high (in GTA V, or Witcher 3, Rise of the Tomb Raider, Fire Strike benchmark, etc), I get about 3-4 fps or so with the Intel HD 4600, and closer to 50-60 fps or so with the 1060.

 

 

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

 

 

Yes, I expected the 1060 to slow down a little when it ran out of VRAM, but I didn't expect it to tank that hard, AND I sure as heck didn't expect an INTEGRATED GRAPHICS to wipe the floor with it!  (True, single-digit frames per second, or even seconds per frame, isn't exactly what you'd consider playable. * )  I'm wondering why it was acting that way?

 

One possibility I"m thinking is .... consider the 1060 has 3GB VRAM, and the settings I was using in the game were wanting 9GB.  When the GPU runs out of VRAM, it has to swap out with system RAM.  Therefore it has to go out the PCIe slot, across the motherboard traces, into the system RAM, do it's swap, go back to the GPU, and considering my workload was wanting TRIPLE the VRAM as my card had, it had to basically make TWO (or 3) swaps PER FRAME!

By contrast, the integrated graphics uses system memory, so if it runs out of its assigned VRAM (2GB - apparently my screenshots above were done on the iGPU), that extra RAM it needs to grab is right there, so no waiting around to swap it out.

 

Like someone said as I was typing this up....

25 minutes ago, Mira Yurizaki said:

An iGPU technically wins by locality, but I don't think it practically matters in the end.

 

 

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

 

I also tested on my laptop (Clevo P750DM-G, 15.6" 60Hz 1080p display, i7-6700K, 40GB DDR4-2133 (1x8+2x16GB) RAM installed at the time, and a GTX 970M with 6GB VRAM.  With that setup and the same settings that used about 9GB VRAM, I was getting about 6 fps at the start of the GTA V benchmark.  (At more "normal" high / ultra 1080p settings, that 6GB 970M gets about 30-35 fps or so in the same games and benchmarks I mentioned above.)

 

I wonder if there's some actual way to scientifically test that, and figure out why the dedicated GPU's performance tanks so hard, like the 3GB 1060 did when asked to play a game with settings needing 9GB VRAM.  (Maybe prove, or disprove my theory on the time / latency for doing the VRAM swapping.)  I don't think I have the proper equipment to do the test myself though.

 

Another idea I had to test (and also don't have the necessary equipment) ... Pick a game that with normal settings at a given resolution would run about 60 or so fps on a 2080 Ti, BUT, run the game at insanely high settings to go way beyond maxing out that card's VRAM, and observe the performance tankage.  Then, swap the 2080 Ti for a Quadro RTX 6000 or 8000 that's got like 48GB (or is it 64GB or 96GB) VRAM, and re-run the game.  I wonder if that might be one of the few situations when a Quadro could wipe the floor with an x80 Ti or a Titan in gaming?  

One example would be to do that experiment using GTA V, and the same settings I used, but at 4K or 8K, and maybe dual or triple monitors.

 

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

 

 

* yes, I put an asterisk when I hinted earlier that ~3 fps may not be considered playable, because ...  I ran Witcher 3 a couple times on the Intel HD 4600 graphics, at 1080p ultra settings.  I was getting about 3 fps, but the actual pace of the game was significantly slowed down by at least a few times, giving me much more time to react to things that were happening, like being attacked by various monsters, etc.

 

Late last night (or early this morning) I was watching @LinusTech streaming, I think it was Beat Saber or whatever, on Twitch.  He was doing this one song that had a lot of difficult moves that were very fast, and wasn't getting very far into the song.  I wonder if a technique like the one I mentioned above (running the game with such a combination of high settings and low-end hardware that the fps drops to single digits, but the game's pace itself also slows way down) might have made that sequence playable?  

 

 

 

========

 

P.S. I hope my separating the various things I was replying to / mentioning (with spaced out paragraphs and dashes between them) makes my not-super-short post a bit more readable, and not just a solid unformatted wall of text.  I wish I knew which post it was, but somewhere else on the forum (think it was within the last week or so) I saw someone write something like "you should take as much effort into making your post as you would hope people would take into replying", or something like that.  

 

[Mira Yurizaki]

@PianoPlayer88Key

 

That reminds me, I recall when I was building and setting up my Windows 98 machine, it has a Pentium III running at 800 MHz, PC-133 RAM (256MB I think), and a 7,200 RPM drive. So probably upper high-end for that time period. But holy moly, it's freaking fast. Thing boots up almost as fast, if not faster, than any of my modern machines.

 

Regarding the VRAM thing you were testing, I found an app that allocates VRAM and forever-waits to see what happens when I run some games. It does try its best to remain resident in memory, but occasionally Windows does shove some of it out of VRAM. I'd have to go back and play around with it though.

[vanished]

1 hour ago, PianoPlayer88Key said:

Also speaking of not having enough RAM, [snip]

I believe having enough, even if it's slow RAM, will be massively better than not having enough RAM, even if the RAM is fast and the pagefile is on a quick device like an NVMe SSD.  The slower the device the page file is on though (particularly in random IO but also sequential), the worse that gap will be.

[Curious Pineapple]

I've recently switched from having dual CPU's and a single 8GB stick of RAM on each to a single CPU with 48GB of triple channel memory. Old setup felt snappier to use, game performance is exactly the same (GPU bound) and the room is noticeably cooler over in this corner. Same machine, just removed a CPU and swapped out memory.

[Bramimond]

Good software will go a long way. I don't think we have enough fancy hardware to make Windows snappy. Doherty treshold says 'productivity soars when a computer and its users interact at a pace (<400ms) that ensures that neither has to wait on the other.' and anything above that just won't be snappy.

 

Imagine you could actually boot your PC and start whatever software you'd wanted in less than 400ms. And the only reason that's not a reality is because developers are lazy bums prioritizing developing time over end user experience.

 

For example, try opening a say, 30GB logfile with your Windows text editor or word processor of choice and try searching for stuff inside. The experience will be painful. Open the same file in for example vim and everything will stay snappy.

[Kisai]

11 hours ago, PianoPlayer88Key said:

 

 

I've recently been thinking ... Which of these two scenarios would feel snappier?  First off, assume your current workload (which doesn't include gaming, but more like web browsing, office, maybe photo or light video editing or whatever) needs / uses about 12 GB of RAM.

• (Slow single-channel RAM, but having enough) - A single 16GB stick of DDR4-2133 RAM in your system, running at, say, CL22, OR,
• (fast dual-channel RAM, but NOT enough) - two 4GB sticks (totaling 8GB) of like DDR4-4500 CL18 or whatever, and your page file on a shingled 5400rpm <=2TB hard drive, or even on an IDE / PATA hard drive (via a PCIe to IDE controller)

 

Also speaking of not having enough RAM, where would the following types of devices on which to store a page file rank in relative "snappiness", compared to actually having enough RAM?  (For example you could set the baseline "having enough ram" to 100.00% performance / snappiness "feel", then rank the others below that accordingly.)

• IDE hard drive
• 5400rpm shingled SATA "budget/cheap" hard drive
• "performance" grade 7200rpm SATA / 10k RPM SAS hard drive
• SSHD
• budget QLC DRAMless SATA SSD (like Tier E)
• fairly premium SATA SSD (like a Crucial MX500 or Samsung 860 EVO)
• budget QLC DRAMless NVMe SSD (like Tier E)
• midrange TLC NVMe SSD (Tier C perhaps?)
• fairly decent MLC NVMe SSD (like the Samsung 970 Pro), or a PCIe 4.0 SSD or another one that claims half a million or more IOPS
• an Optane 905p or similar SSD

 

 

Dual channel memory will always outperform single channel memory unless you allow page swapping. The second you hit the page file, the machine with less memory will tank first. If you have 32GB or 64GB of ram, just turn the page file off and see what circumstances actually exaust the OS. I've only ever been able to get FFXV to do it on cue. Chrome running Twitter/Slack/Discord will also do it because those frameworks are sloppy. For a while there, I'd go to bed and come back to every "infinite scroll" tab would be crashed. It was exceptionally bad when Chrome refused to have a 64-bit version.

 

7200RPM > 5400 RPM. A 2.5" drive will be faster than a 3.5" drive in seek time, not data transfer.

QLC > TLC > MLC > SLC in terms of "speed" since it has more bits to read/write at once, however the tradeoff is never worth it. You'll also notice that the bigger the SSD, the faster the SSD is.

 

So SSD's are a compromise between speed, price, and durability. Cheaper drives are less durable, but faster. More expensive drives are more durable, but slower*. SSD's of higher capacity are faster than lower capacity drives (also applies to USB flash drives). 

 

Basically because the SSD capacity can be stacked you also gain performance in the same way you gain performance with dual channel RAM. So 2016-2018 SSD's the most common SSD's were 256GB or 512GB. 2019-2020 will be 1TB or 2TB, especially with PCIe4 drives, 1TB drives will hit the pcie 3 x4 performance limit pretty quickly.

 

Also be careful of "common knowledge" statements in computer builds. The ideal build is always to max out the specs on a certain part UNLESS that comes with a performance/quality compromise. So when you have larger amounts of RAM, you typically have to pick slower memory because you will have less headroom to overclock it. When you pick CPU's with more cores, you are picking CPU's with slower cores for much the same reason, the reliability goes down at higher speeds. Larger, faster drives are sometimes not even worth considering if the drive's purpose is bulk storage (eg external drives), because the difference between transferring 4TB of data to a SATA SSD in an external enclosure and transferring 4TB to a 5400RPM drive is time and reliability. Unless you're going to need to read from that drive as frequently as a local drive, that $80 drive is a better value than spending $700 on the SSD.

[Bombastinator]

I’m going to go a touch milder I think.  The SSD, yes.  But I think to some degree the CPU is a bit like the ram.  It’s not having as much as possible so much as having enough for what you are doing.  If you’re just watching YouTube. You’re not going to be able to tell the difference between a 3000g and a 3950x.  It’s when you start a dozen YouTube instances where the 3000g chokes and dies and the 3950x keeps going that you see the difference.  I would say the word is “ample”. Ample IPs, ample ram, ample gb/s storage.  Computers can work with less than ample, but they have to shuffle stuff around and wait for things, and that’s when snappiness dies.

[ARikozuM]

"Jazz hands" gets my votes. 

[imreloadin]

MacOS

[ARikozuM]

48 minutes ago, CircleTech said:

Mostly just a new SSD and any CPU made after 2007 with a clock speed of over 3GHz.

 

A core 2 duo PC from 12 years ago with an SSD can feel snappy while an i9 PC with a 1TB WD green can feel like a slug.

I don't know. Those slugs can be pretty harsh at full speed. 

[rabbitanarchy14]

piono88. first off omg there is way to many tabs open in firefox. second you have firefox and chrome open i would assume with all the same million stupid tabs. second you have vm open. third there is a wackton of programs you have running in the background no shit your ram is gone.

[Jae Tee]

15 hours ago, PianoPlayer88Key said:

 

 

I've recently been thinking ... Which of these two scenarios would feel snappier?  First off, assume your current workload (which doesn't include

 

.snip.

 

wow, thats alot of writing. Now i have to read all that.

 

That being said, most of the important info has alredy been mentioned, but it's always good to hear from someone with some hands on experience. 

All in all: Good CPU, Good RAM, Good SSD (good mobo) as well as minimal bloatware. This IS all kinda obvious, so if there is anything ells specific that someone know please speak up.

[Bombastinator]

4 hours ago, Jae Tee said:

wow, thats alot of writing. Now i have to read all that.

 

That being said, most of the important info has alredy been mentioned, but it's always good to hear from someone with some hands on experience. 

All in all: Good CPU, Good RAM, Good SSD (good mobo) as well as minimal bloatware. This IS all kinda obvious, so if there is anything ells specific that someone know please speak up.

Sounds like you’re trying to find some “secret”. Basically there isn’t one.  Avoid hitting the wall with the resources you’ve got.  Someone mentioned running lots of tables of both firefox and chrome.  You’re probably hitting walls with that.  I suspect the wall is likely memory.  Easy enough to check though.

[NineEyeRon]

Right click on desktop>View>Align icons to grid

[alexdoe]

Upgrading to SSD can boost PC performance drastically. You will be able to play games very smoothly. Gamers and game developers are quite rich nowadays. More and more people are starting to play games as smartphones and console's prices not really increasing with inflation. Every streamer should know how to do their taxes. They should know their smartphone or computer come under business expenses and don't have to pay tax for it. They can save a lot of their money by hiring a money advice service like ours. Every rich person learns about taxes to use it in their favor.

On 2/5/2020 at 10:47 AM, Jae Tee said:

So what actually makes a pc or laptop snappy? Obviously we'd be talking about high performance ssd's, and good ram. But i feel like there's more to it. Are "powerful" CPU'S part of it, or what?

[Brennan Price]

I feel like there is something that a lot of people are wildly missing off for some reason... and that would be the specs to run something smoothly... which of course is down to the OS desired specifications. Sure, a quad core Intel Pentium and 2GB of RAM may struggle to run Windows 10 effectively but it would still work... whilst if Lubuntu or Slax Linux (for example) was installed on that same machine then it would run much better, not to mention that if the Windows machine had more RAM or a better CPU then it would also run better as well but... I'm coming onto my point now... 

 

It is not just down to the hardware, it is down to how much better the hardware spec is compared to the operating system required/recommended specs. 

 

(Of course an SSD is just simply helping this to be achieved due to faster load and read/write times but I'm assuming for each of these cases, an SSD would be in use anyway to make the best of the situation, I haven't used a computer without an SSD in years). 

 

PS: Not a dig at Windows users, neither am I trying to promote Linux, just trying to make a point  

[Cinderz]

Good modern(ish) processor that can handle either a good sata based ssd or nvme ssd.