Ramifications – (How to cope with building a reasonable budget gaming PC at this time)

[Silaz Carbryck]

So, if the manufacturers of gaming laptops are now switching to DDR4 DRAM modules to build more affordable computers then surely anyone who is trying to build a new desktop PC right now with the prices of DDR5 and the latest CPU’s being so high, should also be considering doing the same thing right?

The advice being offered here and which likely is still duplicated elsewhere on these forums in a few other places in slightly different ways ought to be that if one is stuck with a low budget to work with but you still need to build a decently reasonable-spec viable machine that can still handle most of the later games from recent years out there right now, then it ought not be too much of a problem using DDR4 DRAM to meet with your lower budget, as system RAM, the speed and capacity of which may be important for AI applications in those damned data centres, isn’t quite so important for gaming as your VRAM is going to be.

Certainly Elite: Dangerous, just like No Man's Sky initially started their development in 2012 and the custom engines for these games may have been in development prior to 2012, so for most Elite gameplay when in space flying spaceships my RTX 3050 6GB is fine. It does max out almost totally the available VRAM, but it’s just enough to do the job. However, if I want to play Operations or do other combat-based on-foot later incorporated into the game Odyssey-era activities, then I need a better GPU, and Morbads’ recommendation of the Intel Arc B580 GPU seems to fit the bill perfectly. It’s my assumption that actually VRAM capacity and quality of specification on board the GPU for gaming purposes is far more important than system DRAM clock speeds or capacity, meaning that one could, as I have done, get away with the minimum viable DRAM of 16GB DDR4 modules to make savings, which shaves off a packet as you then also don’t have to spend as much on your older generation CPU and motherboard, but it’s definitely best not to skimp on the GPU, even though the lower-end units are very highly-priced currently with the RAM crises being what it is.

So here is my personal recommendation for what would need to be your basic specs and costs for the key computing hardware necessary for a reasonable gaming PC build, on a reasonably low budget as currently priced on Amazon UK, which obviously you would still need to be competent enough with computer hardware to build for yourself to save even more money. (Building PC’s is like LEGO):

AMD Ryzen 7 5800X Processor (8 Cores/16 threads, 105W TDP, AM4 - £190
CORSAIR VENGEANCE LPX DDR4 RAM 16GB (2x8GB) - £127 (+£100 for 32GB)
ASUS TUF Gaming B550-PLUS Motherboard - £98
Sparkle Intel Arc B580 Titan OC 12GB GDDR6 - £339

The other gear that I went with for my build earlier this year after my old rig very inconveniently decided to blow up at the end of last year are more optional and down to your own personal choice and taste, but really one shouldn’t skimp too much on these more fundamental hardware components and you will need a large and deep computer case to fit the long GPU and tall CPU cooler:

Crucial BX500 SATA SSD 1TB, 2.5", Up to 540MB/s - £99
Seasonic Core GX ATX 3 2024 650W Fully Modular 80+ Gold PCIE 5.1 PSU - £89
Noctua NH-U14S, Quiet High-Performance Tower CPU Cooler 140mm Brown - £80
Fractal Design North XL PC Gaming Case - £145

Total cost of box only without peripherals and at current Amazon prices - £1167

I am no longer willing to recommend the Barclays Partner Finance company who operate on Amazon to use when buying on credit as from the experience I’ve had with them this year, they were terrible. They actually have got nothing to do with Barclays, only paying for a license to use the Barclays brand name for their dodgy financial product and they made it very difficult for me to make repayments each month, as well as them charging me more interest than they agreed to take over the six-month instalment plan.

Just use a basic credit card if you are lucky enough to have one of those that isn’t currently maxed-out, or dive into your savings, or try to save up quickly to take advantage of the current price reduction offers for some of these components if you are poor like me as money is tight.

Thanks for reading. And let me know if I have made a mistake recommending only 16GB DRAM if you think that 32GB ought to be the baseline standard. - Si

 

[DjVortex]

Sparkle Intel Arc B580 Titan OC 12GB GDDR6 - £339

According to videocardbenchmark.net, that card is approximately as powerful as an RTX 2070 and an RTX 3060.

I suppose that's perfectly good enough for ED at 1440p. (Perhaps even 4k, if you don't expect extremely high framerates.)

ED probably doesn't benefit from 12 GB VRAM compared to 8 GB, but the extra VRAM doesn't exactly hurt for other modern games.

 

[Silaz Carbryck]

According to videocardbenchmark.net, that card is approximately as powerful as an RTX 2070 and an RTX 3060.

I suppose that's perfectly good enough for ED at 1440p. (Perhaps even 4k, if you don't expect extremely high framerates.)

ED probably doesn't benefit from 12 GB VRAM compared to 8 GB, but the extra VRAM doesn't exactly hurt for other modern games.

I've always been quite happy with 60 FPS locked with my monitor set to 60Hz and just 1080p resolution. Really, our biological optical senses aren't capable of perceiving much more than about 30 FPS, but with video games you need more frames to get a smoother playing experience, especially when turning on a dime to spin around 180 degrees in order to face behind you to shoot at an approaching attacker. Above 60 FPS I don't think the quality of gameplay is truly improved all that much, and all you're doing is burning out your GPU more quickly. As for resolutions I think that 4K is probably really useful for large screens like projectors or 50"+ TV screens where you can perceive the improvements more than you can on a small 26-32" gaming screen that is only 50cm away from your eyes.

I was informed that Elite on-foot combat-based Odyssey gameplay would need about 8-10GB of VRAM to run smoothly, so that's why later on when I can afford it I will be upgrading to the Intel GPU because it's much better value than the Nvidia and Radeon offerings right now. As for other games, I don't really play a lot of other games, but it would be prudent to ensure that I'm covered with some decent hardware that can still handle other games anyway, and I was trying to indicate in the OP a potential build that would be suitable for gamers who want to play as many other more contemporary titles as possible on a smaller more reasonable budget.

 

[Morbad]

According to videocardbenchmark.net, that card is approximately as powerful as an RTX 2070 and an RTX 3060.

I suppose that's perfectly good enough for ED at 1440p. (Perhaps even 4k, if you don't expect extremely high framerates.)

Maybe in space, or in the middle of nowhere on planet surfaces, but go to surface settlements or hang around complex ruins and the B580 is going to be a slide show at 4k (without upscaling).

Still, it's generally closer to the 3060 Ti or 4060 non-Ti than to the 3060 or 2070 in performance, especially after the reduction in driver overhead that came last October.

ED probably doesn't benefit from 12 GB VRAM compared to 8 GB, but the extra VRAM doesn't exactly hurt for other modern games.

Odyssey is very borderline on 8GiB cards at settings that a B580 can handle well (1080p ultra, 1440p with some modest concessions). Again, I'm weighing worst-case areas much more heavily. Getting good performance in most places should be a given, it's the rough patches that stand out as problematic.

Really, our biological optical senses aren't capable of perceiving much more than about 30 FPS

That's misleading. You may not be able to count individual frames, but, despite diminishing returns past a fairly modest rate, there seem to be perceptible improvements to motion clarity and latency out to upwards of one-thousand frames per second and some strobic effects don't disappear until around 10kHz. The upper limits of human visual perception in this regard can really only be guessed at because most experiments haven't had access to sufficiently fast displays to find them (most of the stuff I see is a decade old and done with 480Hz projectors that were cutting edge at the time).

 

[Silaz Carbryck]

That's misleading. You may not be able to count individual frames, but, despite diminishing returns past a fairly modest rate, there seem to be perceptible improvements to motion clarity and latency out to upwards of one-thousand frames per second and some strobic effects don't disappear until around 10kHz. The upper limits of human visual perception in this regard can really only be guessed at because most experiments haven't had access to sufficiently fast displays to find them (most of the stuff I see is a decade old and done with 480Hz projectors that were cutting edge at the time).

I'm just going by what I learned back in the 1990's at college regarding the human biological visual cortex being able to perceive minimum frame rates long ago when film was first introduced at 24 FPS. With television, European PAL is 25 FPS and US NTSC TV standard was closer to 30 FPS, but these are just the minimum viable FPS rates for a smooth picture to be perceived and any lower then we will start to notice the lost frames, like in earlier last-century animations that might have only gone with 12 FPS. So I suppose if science has learned that we can perceive higher frame rates in more recent studies than we probably can, but shouldn't go below 24 frames as then we will start to see the lost frames.

I still would maintain the assertion that ludicrously high frame rates in video games are probably unnecessary and will just wear down our ludicrously expensive hardware investments more quickly than more moderate settings would, which is why I always stick to locking at 60FPS for my GPU and 60Hz for my screen as my practical standard.

 

[Morbad]

But I still would maintain the assertion that ludicrously high frame rates in video games are probably unnecessary and will just wear down our ludicrously expensive hardware investments more quickly than more moderate settings would.

'Unnecessary' is rather vague. I've played a lot of games at 15-30 fps and enjoyed them. I used to play dial-up Quake against my neighbor at 512x384 with about 25fps and had a blast. I beat MechWarrior 2 about a dozen times with an average frame rate of ~20 on the same system. Doom had a 35fps cap. It was tolerable, at the time.

Today, my fastest displays are 360Hz (though I spend more time on larger 240 and 144Hz displays) and 360 fps is not beyond the point where I can consciously, let alone subconsciously, perceive improvements.

A good demonstration of how higher frame rate equals more visual information that can directly improve perception is Blur Buster's persistence of vision test: https://testufo.com/persistence

Higher the frame rate, the more clear the image is. At a sufficiently high frame rate, the detail is virtually the same as the original, static, image.

In games higher frame rates translate directly into more motion information for a given time interval. Stuff that is literally invisible, because it's not on screen long enough for sufficient information to be delivered at lower framerates, can become perfectly clear at higher frame rates. Frame rate alone can be the difference between not seeing a grenade at all; seeing a single frame of a grenade, but not being able to infer it's trajectory; or seeing several frames of that grenade's path, allowing one to avoid it while simultaneously directing return fire at it's source. Indeed, one of the reasons so much stuff in Odyssey needs to be telegraphed with high contrast (or outright glowing) trails/tracers and artificially low projectile velocities, is surely that the game cannot maintain the frame rates that would be expected from most tactical shooters.

As for hardware, running a lower frame rate, all other things being equal, will absolutely reduce power consumption, temperatures, and the stress placed on components. That said, I expect non-defective hardware to be able to at it's specified power and thermal limits essentially indefinitely, without issue. On my own hardware (which I've usually binned to for well above average samples) I expect even more.

 

[Silaz Carbryck]

In games higher frame rates translate directly into more motion information for a given time interval. Stuff that is literally invisible, because it's not on screen long enough for sufficient information to be delivered at lower framerates, can become perfectly clear at higher frame rates. Frame rate alone can be the difference between not seeing a grenade at all; seeing a single frame of a grenade, but not being able to infer it's trajectory; or seeing several frames of that grenade's path, allowing one to avoid it while simultaneously directing return fire at it's source. Indeed, one of the reasons so much stuff in Odyssey needs to be telegraphed with high contrast (or outright glowing) trails/tracers and artificially low projectile velocities, is surely that the game cannot maintain the frame rates that would be expected from most tactical shooters.

I just Googled this to check:

"The average human reaction time is about 150 to 200 milliseconds for touch and sound, and about 200 to 250 milliseconds for sight.

Average Times by Sense:

Auditory (Sound): 140 to 170 milliseconds (0.14 to 0.17 seconds).
Tactile (Touch): 150 to 170 milliseconds (0.15 to 0.17 seconds).
Visual (Sight): 180 to 250 milliseconds (0.18 to 0.25 seconds).

According to Orange Neurosciences, “Auditory stimuli reach the cortex faster than visual stimuli.”"

So if our visual reaction times from eye to hand are limited to as much as a quarter of a second (as they get worse after we leave our teenage years) it doesn't really make much difference if we still see that grenade being thrown at us by the attacker a quarter of a second earlier, just because we had a few more hundred FPS to see it coming because we likely wouldn't be able to react in time to move away from it's effect when it's detonated anyway, even if we do have better motion information being delivered by that much higher frame rate.

I'm still content with locking at 60 FPS because I think it's fine and still fit for purpose, even for a fast-paced shooter.

And I used to play the first Unreal game online in multiplayer mode with just a 28kbit dial-up modem in software rendering mode at about 20FPS with a 350ms ping, so I know what lower frame rates and high latency lag in older games used to be like in the earlier days as well. We just made do with what we had back then, but the games were still very enjoyable as they were still quite cutting-edge for the day.

 

[Morbad]

I just Googled this to check:

"The average human reaction time is about 150 to 200 milliseconds for touch and sound, and about 200 to 250 milliseconds for sight.

Average Times by Sense:

Auditory (Sound): 140 to 170 milliseconds (0.14 to 0.17 seconds).
Tactile (Touch): 150 to 170 milliseconds (0.15 to 0.17 seconds).
Visual (Sight): 180 to 250 milliseconds (0.18 to 0.25 seconds).

Another case of true, but beside the point and largely irrelevant.

So if our visual reaction times from eye to hand are limited to as much as a quarter of a second (as they get worse after we leave our teenage years) it doesn't really make much difference if we still see that grenade being thrown at us by the attacker a quarter of a second earlier

That statement is both false in and of itself--any handicap to reaction speed is a handicap to performance and even small differentials can frequently be significant enough to be deciding factors when the median reaction times all cluster around the same figures (a single ~17ms frame is almost ten percent of reaction time, and a large fraction of total latency)---and also not the scenario I was describing.

Even if end-to-end latency is somehow identical, the player with more frames within that time window has more information. Noting a trajectory of an attack and making a counter attack upon it's source requires enough points to judge the arc of a throw, a constantly shifting movement vector.

we likely wouldn't be able to react in time to move away from it's effect when it's detonated anyway, even if we do have better motion information being delivered by that much higher frame rate.

Yet, I can still do so regularly. Having the information to know whether one is moving toward it or away from a relatively slow moving ballistic threat, and it's point of origin, should almost be a given in tactical shooters...but will be needlessly difficult at ~60 fps, absent cartoony visual aids.

Anyway, this topic has cropped up several times before and I'm already repeating my last posts on the topic:

Post in thread 'When can we expect performance optimizations?'

May 27, 2021

All I'm pointing out is that the scientific evidence seems to suggest that human visual performance runs out at 75hz.

Not in a whole slew of relevant metrics it doesn't.

When it comes to flicker, the recent consensus is somewhere between 500 and 10000Hz, depending on individual and type of of lighting, before strobic effects are wholly imperceptible:

https://www.lrc.rpi.edu/programs/solidstate/assist/pdf/AR-Flicker.pdf#page=6

Client Challenge

To eliminate the perception of motion blur on sample and hold displays, ~1000Hz will probably do the...

• Morbad

Post in thread 'What fps is acceptable?'

May 23, 2021

There are always going to be trade offs between frame rate and other things, and serious ones for the foreseeable future, especially in Odyssey.

Personally, I'm not a framerate junky, not in these kinds of games. For the last several years, I've tend toward relatively slow VA panels because bad contrast bothers me a lot more in space games than a bit of blur. The fastest display I currently own barely does justice to 120 fps/Hz (it's advertised as 165Hz, and I run it at 144Hz to take advantage of VRR, but the true pixel response means a lot of transistions can't keep up). Still, I...

• Morbad

• 

Post in thread 'Interdiction Disruptor - new defensive module idea'

Nov 17, 2025

Bit of a necro, but I feel this this more recent video demonstrates latency compensation better than the videos I linked previously. Once one has a grasp of latency compensation, it should be easy to infer how we get the asynchronous behavior, generally favoring the attacker (most blatant in the case of shadow rams and the interdiction tunnel game), we often see between CMDRs, where the network model is tuned to tolerate fairly extreme latency between peers.

Source: https://www.youtube.com/watch?v=WFw4F2AyaP4


Peer-to-peer models aren't that different from...

• Morbad

If your subjective perception is that 60Hz is enough for you, I have no reason to doubt you, especially where Elite: Dangerous Odyssey is concerned. However, the idea that human perception, and thus gaming performance, is capped anywhere near these levels is not something that can be supported by either formal studies (some of which I have been linked to in the above threads), nor by less formal performance testing (a good one).

 

[Silaz Carbryck]

I couldn't find any actual science studies quickly, but here is a journalistic article concerning perception of frames per second:

How Many Frames Per Second Can the Human Eye See?

So yes, you are right, in more recent years it's become known that some people can perceive much higher frame rates than was previously considered possible. But there's going to be a lot of variation between individuals. Most people are probably not going to be able to process 200 or more FPS, and we have to take into account the human reaction time for our visual cortex to register an image and then for our body's to physically react to it, you shouldn't say that's not important because it's going to be the governing factor for fast-paced gameplay reasons, where not everyone is going to be able to compete on a level playing field based on their individual personal biological capabilities, not the capabilities of their computer. Maybe the fighter pilot gets the job because they were more gifted than the average individual and can react faster than normal as well as have a higher perception of faster-moving visual stimuli. But most of us are not going to be so well-gifted, and we have the law of averages to consider.

Still, the vast majority of computer monitors run at 120Hz or lower, not all are high refresh-rate gaming ones, and also we have to consider not just our slow biological reaction times but also the normal 'clock' speeds at which the human brain actually processes information, which presumably is what contributes to our slow reaction times, as this is only in the range of between 1 and possibly slightly over 100Hz, again depending on the individual as well as their state of mind at the time. (Stimulant drugs can augment and improve gamma frequencies of our brains temporarily) We certainly don't process information very quickly when compared to computers which is one of the things that machines excel at compared to our biological minds. So even if your GPU were to be outputting more than 200 FPS, your biological senses and brain processing speeds are likely to limit your capabilities in a fast-paced game the parameters of which are defined at birth and have nothing to do with how expensive your gaming PC and GPU was. Not everyone is a fighter pilot, as most of us that play video games are going to have average biological limitations that we can't improve drastically just by gaining more skills from greater experience of playing games more often.

Well, that's my take on it. As I said I'm happy to play games at a lower FPS that seems reasonable to myself, because perhaps I just don't personally perceive as high a number of frames than others might, and we should accept that some people are going to have better reaction times and perception of frames per second than others and that most people are going to be average with there only being edge-cases at the extreme end of that bell-curve distribution graph when it comes to biological capabilities regarding registering and matching what our computers can potentially manage to display.

EDIT: Going by the first Tactuallabs paper latency study experiment that I read from the link you provided above I have copy/pasted the final paragraph referring to the results of the first latency study and comparing their results to previous studies:

“As a whole, these results bring some clarity to a
disagreement in the literature over latency JNDs since they
suggest that form-factor and input task must be considered.
Ng et al. [19], who report a JND of 6.04 ms, argue that the
textbook threshold of 100 ms (a value based on Miller’s
work [16] and guidelines for audio in the
telecommunications industry [4], among other sources) is
incorrect and an artifact of the limited performance of input
devices and measurement techniques available at the time
of its origin. Our results suggest that both are correct, as Ng
et al. report on a direct input dragging task and most others
report on a variety of indirect input tapping tasks (menu
selection, key typing, path drawing, etc.). Indeed our
reported mean JNDs of 11 ms, 69 ms, and 96 ms for direct
dragging, direct tapping, and indirect tapping correspond
very closely to the reported JNDs of 6.04 ms (Ng et al.,
direct dragging), 64 ms (Jota et al. [8], direct tapping), and
100 ms (textbook guideline) – with all of these previously
reported values falling well within the 95% confidence
intervals of our estimated means.”

So obviously with the upper end being about a tenth of a second and the lower end being more like a hundredth, there is great variability between peoples' reaction times that would fit with my assertion that not everyone is going to be able to benefit from, perceive and react to 100Hz FPS/refresh rates and the spectrum between low and high is considerable if the above mean values are to be understood. With computer games we are mostly tapping keys, squeezing triggers or pressing mouse buttons, so the direct tapping test part of the study is more relevant to our video games playing.

 

[Morbad]

Most people are probably not going to be able to process 200 or more FPS

That depends on what we're testing. Presence of something? Trajectories? Silhouette identification? Fully processing details?

I might be able to reliably identify the basic contents of a scene flashed for that 13ms figure from the MIT study cited in that article (which is the fastest they tested, not the limit of all the test subjects), but I can tell the difference between an FDL and Clipper that are on screen for significantly less than that, and almost everyone will be able to point out the location of a bright flash of light that's only 0.1ms in duration.

and we have to take into account the human reaction time for our visual cortex to register an image and then for our body's to physically react to it, you shouldn't say that's not important because it's going to be the governing factor for fast-paced gameplay reasons, where not everyone is going to be able to compete on a level playing field based on their individual personal biological capabilities, not the capabilities of their computer

It's not important in that it doesn't detract from the impact of other delays, nor does personal reaction time set a floor for the minimum duration a stimulus has to be to deliver visual information. One can easily have time to react, without having the information to react appropriately.

Still, the vast majority of computer monitors run at 120Hz or lower, not all are high refresh-rate gaming ones

Sure, but it's not unusual for people playing video games to possess high refresh rate displays. Nowadays, high refresh rates aren't a premium or niche feature, until you get past 280Hz or so.There are at least ten displays in my house that can do 144Hz or higher and the least expensive one of them was about 80 USD.

EDIT: Going by the first Tactuallabs paper latency study experiment that I read from the link you provided above I have copy/pasted the final paragraph referring to the results of the first latency study and comparing their results to previous studies:

“As a whole, these results bring some clarity to a
disagreement in the literature over latency JNDs since they
suggest that form-factor and input task must be considered.
Ng et al. [19], who report a JND of 6.04 ms, argue that the
textbook threshold of 100 ms (a value based on Miller’s
work [16] and guidelines for audio in the
telecommunications industry [4], among other sources) is
incorrect and an artifact of the limited performance of input
devices and measurement techniques available at the time
of its origin. Our results suggest that both are correct, as Ng
et al. report on a direct input dragging task and most others
report on a variety of indirect input tapping tasks (menu
selection, key typing, path drawing, etc.). Indeed our
reported mean JNDs of 11 ms, 69 ms, and 96 ms for direct
dragging, direct tapping, and indirect tapping correspond
very closely to the reported JNDs of 6.04 ms (Ng et al.,
direct dragging), 64 ms (Jota et al. [8], direct tapping), and
100 ms (textbook guideline) – with all of these previously
reported values falling well within the 95% confidence
intervals of our estimated means.”

So obviously with the upper end being about a tenth of a second and the lower end being more like a hundredth, there is great variability between peoples' reaction times that would fit with my assertion that not everyone is going to be able to benefit from, perceive and react to 100Hz FPS/refresh rates and the spectrum between low and high is considerable if the above mean values are to be understood. With computer games we are mostly tapping keys, squeezing triggers or pressing mouse buttons, so the direct tapping test part of the study is more relevant to our video games playing.

One need not be aware of reduced latency to benefit from it, and latency isn't the only variable impacted by frame rate.

I would be astonished if the overwhelming majority of people (lets say 95%+) could not benefit from significantly more than 100Hz, in at least some gameplay scenarios. Some people will see diminishing returns sooner and we'd no doubt see quite a spread, but 100Hz is pretty damn low on the ranges of some of these metrics.

 

[Silaz Carbryck]

Ok. Here's my hypothesis regarding frames per second perception, processing and human reaction times.

If we are to go by the data gathered from that study I quoted that you had posted previously then the mean average latency for button-pressing reactions is around 64-69ms.

That's about one 15th of a second. For me, at 60 FPS it would take, if this was my personal latency, 4 frames for me to recognise my target in sight and press my mouse button in time to win the frag with my sniper rifle.

For the individual who is running at 120 FPS, and if they also have an average latency, then they get 8 frames per second of visual data to process before they press the button in time.

It may well be Morbad that you are just more able than I am to perceive the extra frames, process that visual information and then react to it more effectively because your brains' 'clock' frequency is running at the upper end of the spectrum. So, closer to 100Hz with your reaction latency being lower than average as well from the reports you have given me before about your gaming performance and the advantage of having more frames per second being rendered. Therefore you would be more likely considered by the Air Force to take on the role of a fighter pilot for them.

Whereas with me it's more likely that the reason I'm not able to report better gaming results from all those extra frames per second is simply because my brains' 'clock' frequency might be as little as half what yours runs at, so perhaps more like 50Hz, which may be the average, IDK for certain if that's the actual average, but I guess it's probably close enough to. So I'm not so easily able to perceive, process and react to all those extra frames of visual information so effectively for this reason.

Perhaps it is that both perception and processing of more frames per second, as well as an individuals' reaction times are all closely linked to that 'clock' frequency of our brain that is going to drastically vary between individuals.

This would then infer that someone with a lower than average 'clock' frequency that their brain is running at, isn't going to benefit much from a more expensive gaming rig that can handle many more frames per second, as it's not going to improve their game the way it would if they were positioned at the opposite higher end of the 'clock' frequency spectrum.

I take about twice as long to read and write than the average person, so when I read a passage I read at the same speed that I would speak out loud. It's very likely that this is one of those annoying aspects of my ADHD Autism Spectrum Disorder and learning disability. But just because I may only have an average 'clock' frequency when I process information doesn't mean that I can't think. And just because I read and write more slowly than most people it doesn't mean I can't read and write. It just means that I take longer than most people to think, or read and write. And it's a sort of slight handicap I suppose, which I have gotten used to.

There maybe a study that is relevant to my argument out there somewhere that a team worked on once already, but I don't know what to search for on Google Scholar. And if it exists it might well be locked-off behind a subscription paywall with just the abstract available to read.

So what are you thoughts about this hypothesis Morbad?

 

[FrankXjr[G911]]

2080 ti graphic card is stiill strong for ED

 

[Morbad]

So what are you thoughts about this hypothesis Morbad?

It's entirely likely, almost a given really, that there is a fairly wide spread of individual response to any given visual stimulus. However, there are all sorts of thresholds of perception of different visual stimuli of different duration. Even if your do happen to be well below the mean, there are probably relevant stimuli that would scale way past 60 or even 100Hz.

The example I've been focusing on is motion clarity, because it's both easy to demonstrate and has clear gameplay implications. It would take some pretty extraordinary evidence to convince me that a more or less health human stopped seeing motion clarity improvement at 100 fps. Look at TestUFO's eye tracking test; if you can track the alien across the screen, but it doesn't look as clear as the static alien image, then you'd benefit from a faster display. Same thing is applicable any time you move your view in a game, or an in game object transits your field of view sufficiently fast...you lose visual details that may matter.

Here are a couple of more recent videos that explain some of the different variables that frame rate affects:

Source: https://www.youtube.com/watch?v=7zky-smR_ZY

Source: https://www.youtube.com/watch?v=Sb_7uN7sfTw


Even if your sensitivity to every variable, every threshold, is somehow half mine, there is going to be some kind of benefit from much higher frame rates than ~60. It may very well be that ~60 is still more than good enough, once all things are considered, but one should not mistake 'good enough' for 'no improvement to be had'.

2080 ti graphic card is stiill strong for ED

And still moderately faster than a B580.

I don't think the OP is looking for a used card, however.

 

[Tyres O'Flaherty]

'Unnecessary' is rather vague. I've played a lot of games at 15-30 fps and enjoyed them. I used to play dial-up Quake against my neighbor at 512x384 with about 25fps and had a blast. I beat MechWarrior 2 about a dozen times with an average frame rate of ~20 on the same system. Doom had a 35fps cap. It was tolerable, at the time.

Today, my fastest displays are 360Hz (though I spend more time on larger 240 and 144Hz displays) and 360 fps is not beyond the point where I can consciously, let alone subconsciously, perceive improvements.

A good demonstration of how higher frame rate equals more visual information that can directly improve perception is Blur Buster's persistence of vision test: https://testufo.com/persistence

Higher the frame rate, the more clear the image is. At a sufficiently high frame rate, the detail is virtually the same as the original, static, image.

In games higher frame rates translate directly into more motion information for a given time interval. Stuff that is literally invisible, because it's not on screen long enough for sufficient information to be delivered at lower framerates, can become perfectly clear at higher frame rates. Frame rate alone can be the difference between not seeing a grenade at all; seeing a single frame of a grenade, but not being able to infer it's trajectory; or seeing several frames of that grenade's path, allowing one to avoid it while simultaneously directing return fire at it's source. Indeed, one of the reasons so much stuff in Odyssey needs to be telegraphed with high contrast (or outright glowing) trails/tracers and artificially low projectile velocities, is surely that the game cannot maintain the frame rates that would be expected from most tactical shooters.

As for hardware, running a lower frame rate, all other things being equal, will absolutely reduce power consumption, temperatures, and the stress placed on components. That said, I expect non-defective hardware to be able to at it's specified power and thermal limits essentially indefinitely, without issue. On my own hardware (which I've usually binned to for well above average samples) I expect even more.

As a matter of interest, what are the specs of your rig(s)?

 

[DjVortex]

Really, our biological optical senses aren't capable of perceiving much more than about 30 FPS, but with video games you need more frames to get a smoother playing experience

I'm not very sensitive to framerates, but having a g-sync display, I can tell when the framerate drops below about 45 FPS, give or take. It becomes visibly choppy.

Above that, it's not so much a question of choppiness but about blurriness. The famous "UFO test" demonstrates it perfectly: Eg. the 120 Hz version is much sharper than the 60 Hz version, even though the sprite itself is exactly the same for both. I can't see any choppiness in the 60 Hz version, but the difference in blurriness is extremely obvious.

 

[Markov Chaney Again]

@Morbad (or anyone else who wants to contribute evidence-based knowledge)

If one's rig IS bound on main memory bandwidth, what is the single biggest lever that you can pull in the settings to get more framerate and or less laggy jitter or jittery lag?

I'm not getting into the depths (haha) of vision and perception. In ED that is the wrong debate because the problem with low FPS on ED is that it also affects how the engine behaves so you get input lag and whatnot. And below 20FPS or so it just stops doing some secondary things at all so you will have geology missing and such. And as Morbad has said above, really the debate is latency and nothing to do with whether you can "see" 1000FPS.

 

[DjVortex]

If one's rig IS bound on main memory bandwidth, what is the single biggest lever that you can pull in the settings to get more framerate and or less laggy jitter or jittery lag?

It is my understanding that PCI-Express 4 at x16 actually about matches typical DDR4 RAM in speed (which I actually find it surprising), but only if the RAM is single-channel. In dual-channel mode the read speed of the RAM almost doubles, in which case for a typical 3200 MT/s DDR4 dual-channel RAM kit it's the PCI-Express bus that's the limiting bottleneck.

That being said, if you have one single stick of slow DDR4, it can apparently actually be slower than the PCIe bus. Who would have guessed. (I suppose it's the 16 parallel channels that makes it so fast. Every individual channel is vastly slower than RAM, but doing transfers in parallel allows it to catch up nicely.)

Would highly recommend having two sticks (and putting them in the slots recommended by the motherboard for dual-channel operation). Even slower RAM ought to stop being the bottleneck at that point.

 

[Morbad]

As a matter of interest, what are the specs of your rig(s)?

I have a lot of rigs, whose specific components are usually in flux.

My usual main rig is a Ryzen 9800X3D paired with an RTX 4090. My HTPC setup is a 9700X paired with an RX 9070 XT, but I've been moving around the 9070 XT to my other systems to get better platform comparisons. However, I have many other systems that range from stuff that was high-end just a few years ago all the way down to my super socket 7 retro/nostalgia build with a K6-III+ 450 (overclocked to 600MHz, of course).

I've posted some performance comparisons in Odyssey of a few of my setups in one of my recent threads.

@Morbad (or anyone else who wants to contribute evidence-based knowledge)

If one's rig IS bound on main memory bandwidth, what is the single biggest lever that you can pull in the settings to get more framerate and or less laggy jitter or jittery lag?

Honestly, there isn't much that can be done if the performance floor isn't a GPU bound setting. You can reduce shadow quality or the LOD slider, which are about the only two settings that can meaningfully help in CPU or system memory bound scenarios, or you can make the CPU and/or memory subsystems faster.

Among the settings that are in the config files, but not the game menus, adjusting thread counts, stack sizes, and queue sizes can provide a very modest improvement, sometimes. You can also essentially disable shadows, if you really need to...has to be a real potato of a system for this to be worth doing, however.

I'm not getting into the depths (haha) of vision and perception. In ED that is the wrong debate because the problem with low FPS on ED is that it also affects how the engine behaves so you get input lag and whatnot. And below 20FPS or so it just stops doing some secondary things at all so you will have geology missing and such. And as Morbad has said above, really the debate is latency and nothing to do with whether you can "see" 1000FPS.

Elite: Dangerous is a weird one and does have a ton of stuff that is inexplicably still tied to render frame rate. One that I only noticed recently is that using the Maverick Suit's cutter tool is much easier at higher frame rates because it's basically a gun that shoots damage decals one per frame, so the higher the frame rate the less likely there are to be gaps and the more reliably the panel falls away after one quick pass. It's really dumb.

Latency is one facet of the debate, but latency gains have rapidly diminishing returns, especially if one has any control over the depth of the swap chain/flip queue (which is a large part of what Anti-Lag, ULL, Reflex and similar features exist to control). The most obvious facet that scales best with frame rate is motion clarity. The more frames you have, assuming the display can keep up (both in refresh rates and pixel transistion time) the sharper moving stuff looks...and that scales, for very normal people, well into the middle to upper three-figures of frame rates.

 

[Markov Chaney Again]

I meant software settings, as in "what one ED tunable can I turn down one notch for small cosmetic impairment but great justice and moar performance"

But I didn't write that, which is on me... and you made some very good points there. I do have dual-channel 3200 DDR4 so like you, I am now surprised to hear the bus is probably the bound. I did already know dual-channel was worth it so I chose the motherboard carefully (and then got very lucky with a one-day offer that happened to be the day I was looking!)

I have been digging into a different weird problem which is that when I leave calc.exe running by mistake, that causes jitter. After finding some threads on the nVidia forum that I am willing to lend credence, it seems some apps do ask for GPU they have no intention of using, and calc.exe is cited. The asking is enough to push memory-pressured GPU over the edge and cause jitter, because various parts of the DX11 and older nVidia drivers are absolutely doo-doo at memory management.

So I think part of "build a budget PC" should be "find yourself a decent general launcher utility that will kill everything you don't need to be running when playing ED"

i5 12400F [Alder]
1060 6GB [I might actually be up against the 192-bit bus here?]
Corsair/Samsung 2 x 8GB sticks, XMP 2.0 @ 3200

Probably right on the edge of core-bound anyway so possibly I just have a very balanced system and will have to upgrade either the lot or nothing.

 

[Morbad]

Would highly recommend having two sticks (and putting them in the slots recommended by the motherboard for dual-channel operation). Even slower RAM ought to stop being the bottleneck at that point.

If one isn't GPU limited and doesn't have a CPU that is quite slow by modern standards, chances are that RAM is the overriding bottleneck, no matter how many channels you have or how fast the memory is. Though the extra bandwidth and parallelism of extra channels definitely helps, the overriding problem with system memory latency, which mostly scales with memory clock speed and timings and is ultimately limited by physical proximity of the memory.

In comparing my 9800X3D and 9700X in a controlled Odyssey surface test, the former proved to be a full third faster (across average, 1% low, and 0.1% low frame times), despite running ~175MHz slower on the cores and having only a minor main memory performance disadvantage. The reason for this is that with a 96MiB L3 cache, L3 hit rates were in the mid-80% range, while the CPU with only a 32MiB L3 cache had hit rates in the mid 50% range. Every cache miss has to wait on main memory and the 9700X has three times as many cache misses (15% vs. 45% miss rate) while playing EDO, pushing the bottleneck squarely on the reasonably fast DDR5 I'm using.

I meant software settings, as in "what one ED tunable can I turn down one notch for small cosmetic impairment but great justice and moar performance"

Nothing, with caveats. Shadows and LOD are the only graphics options in the game that alter the CPU/memory bound frame rate floor to any meaningful degree and turning them down too far has major cosmetic implications.

i5 12400F [Alder]
1060 6GB [I might actually be up against the 192-bit bus here?]
Corsair/Samsung 2 x 8GB sticks, XMP 2.0 @ 3200

Probably right on the edge of core-bound anyway so possibly I just have a very balanced system and will have to upgrade either the lot or nothing.

That 12400F isn't core bound. My E-2678v3 often is, because of it's Haswell cores running at only 3.3GHz. A hex-core 4Ghz+ Alder Lake with DDR4-3200 is limited by cache capacity, cache/ring clock, and system memory performance, in this game, in non-GPU limited scenarios.

The GPU's issue is less the bandwidth than the memory capacity. EDO on ultra quality wants 7-9GiB of VRAM at 1080p. However, even without a VRAM constraint, the GPU is probably the bottleneck in Odyssey, unless you run modest settings.

If your GPU utilization is falling along with performance, I would check VRAM usage before suspecting CPU or memory performance issues.