ACC graphics settings help

[Steve Gordon]

Can anyone help with the graoic settings for ACC. At 1080p the game looks average but any higher settings I cannot get good framerate. I have a Gtx 1060 and intel i7 2600 CPU @ 3.40ghz. I know its not the fastest rig. I know its ACC is beta but just looking of some advise. Thanks

 

[jven]

Surprised on the above screen shots that ACC is only consuming 2GB of RAM considering it seems to be calculating its rear off.

 

[Case_]

I up front totally disagree with you that the 1060 is a "low-end" GPU. I, in fact, would call it mid-to-high end

By all means disagree if you want to. But honestly, I don't really care how many people own a 1060, it is still a fairly low end card (I'd argue that's precisely part of the reason why so many people use it). Call it mid range if it makes you feel better, but calling it high end is really pushing it. But the fact alone that it is a 1060 tells you it is a low to mid range card at best - this is how nVidia has been numbering their cards for years now. A x50 series is the absolute low end, x60 is low to mid range, x70 is the mid range enthusiast card and then there's the high end x80 series.

As for the CPU, I posted the note about CPU load without digging further because I was rather shocked. No other game, rFactor2, PC2, or anything else I have comes close to working the CPUs like that.

Then I guess the games behave differently on your PC than what is my experience with two (well, three if I include my old i5 2500k) different platforms, because I certainly wouldn't say that nothing comes close to ACC. Be it rF2, AC or Raceroom, they can get pretty CPU intense at times, and pretty much all of the current sims (including ACC) mostly depend on 2 threads maximum, the rest is just tagging along if it's there at all. What makes a bit of a difference in case of ACC is the fact that unlike the rest of the sims, there are 3-4 moderately utilized threads on top of those primary two, while the other games mentioned barely use more than those 1-2 main threads. The most threads I've seen used was with PCars 2, which seems to be using 3 threads fairly decently plus around 4-6 additional fairly lightly utilized ones. So I'm really not sure why ACC behaves so differently for you - it's not that much more thread intensive than the rest in my experience.

But to get to the facts, the CPU has a Passmark benchmark of 12,460. ACC's recommended CPU is benchmarking ~12,700. So it's inaccurate at best to portray my CPU as under-powered or to say, "oh, it's a mobile"....

I honestly don't care much for benchmarks, it is just a synthetic number that often has very little in common with the actual real world performance. I don't even know much about Passmark. I would guess the score you quote is a multithreaded performance number, which, given that your CPU is 6c/12t, will quite likely cause the CPU score higher simply for that reason alone - it has a lot of cores and a lot of threads, so the number will be higher. And indeed, looking at some benchmarks, it seems to perform about 25 or so percent faster in multithreaded benchmarks than a non-overclocked i7 2600k, which is not exactly great to say the least, given that it has 2 more cores and 4 more threads to work with, while its single core performance is only marginally better than the non-overclocked 2600k. That alone is telling. But not that surprising, really - it has a base frequency of only 2.2 GHz, which is pretty low, and while its turbo frequency is 4.1 GHz, given that it's a mobile CPU, it really is hard to tell how often (and on how many cores) does it actually reach turbo. So, yeah, from what I see, the assessment "oh it's a mobile CPU, that might explain a few things" seems fair. It's not a completely bad CPU it seems, but it's fairly underwhelming for an 8th gen i7 in 2019 in my opinion and its real-world performance will not likely be all that great, especially since we are talking about load that requires both single and multicore performance to be fairly high.

I would guess that when they turn on CPU optimizations that should make the code ~30% more efficient hopefully

Yeah, I'm not holding my breath. The Unreal engine is apparently quite bad with multithreading as it relies heavily on a single main thread, and there's only so much they can do with the rest. (And, let's be honest, the old AC could be pretty CPU intensive as well.)

 

[Joakim Stensnæs]

That's definitely not the case for me. What CPU and GPU do you have?

i7 6700k + 1070ti

 

[RasmusP]

i7 6700k + 1070ti

Did you ever touch the ini and changed the taa config with custom parameters? If not it might be worth a try.
Do you have the taa setting in acc set to epic?
I do see what you mean but from within the cockpit it's not visible for me. Chase + hairpin on the other hand give me some ghosting too.
Just guessing and trying to help

BTW, qhd=1440p. Although 1440p can also be the wider screen resolutions like 3440x1440, while qhd is always 2560x1440.

 

[RasmusP]

Well, OK, Martin, perhaps there is more than 1 person here qualified to talk about this.
I up front totally disagree with you that the 1060 is a "low-end" GPU. I, in fact, would call it mid-to-high end, discounting the new and not-debugged-enough RTX series, it is only out-performed by the 1070 and 1080 series. Here's the GPU Benchmarks to prove my point:
GTX1080 .... 12,385
GTX1070 .... 11,291 .....< ACC's recommended GPU
GTX1060 .... 9,061
GTX1060 .... 7,734 .....<(MaxQ 6GB) my GPU and 3x faster than the minimum
GTX 1050Ti...5,981
GTX460 ..... 2,654 ..... < This is ACC's minimum GPU

Wow!! A lot of people out there with a 1050Ti or less. Most people, that is. They won't be racing on this game in multiplayer with more than 10 cars displaying even with turning off all the good graphics.

As for the CPU, I posted the note about CPU load without digging further because I was rather shocked. No other game, rFactor2, PC2, or anything else I have comes close to working the CPUs like that. That's really all you need to know to know it's surprising.

But to get to the facts, the CPU has a Passmark benchmark of 12,460. ACC's recommended CPU is benchmarking ~12,700. So it's inaccurate at best to portray my CPU as under-powered or to say, "oh, it's a mobile"....

As someone whose career includes over 20 years coding, though these days mostly web sites but including 2 years of PC Assembler programming, yes, I am shocked at the CPU load here (compare it to rFactor2 which is DX11, too).
Check for yourself the 2 screen prints I'm posting to give the background and I think my initial shock is more than well supported. I would guess that when they turn on CPU optimizations that should make the code ~30% more efficient hopefully, and not sure if some of that can be offloaded to the GPU, but my GPU is already relatively maxed out, too. This is amazing given that, in my opinion, Paul Ricard on rFactor2 looks better than Paul Ricard on ACC (which is a shame).
I'm pretty sure CPU optimizations alone won't solve the load problem, meaning they might have to drop some calculations on lower cpus. I'd say they've under-specced this on the Steam Store Page.

Before posting the screen captures, there's still the 2 horrendous graphics problems I described. Anyone else seeing the full spread of grain/marbles on the grass near the track, where the marbles don't move with the grass? I've only had this since 0.5.1 or so and it's very distracting.
--------------------------------------------------------------------------
System Load graph from Task Manager:

You can see on the next graph where I stopped ACC in the middle (about 2 minutes prior to taking the pic). The CPU cores were obviously heavily loaded; though 69% in the above graph at that moment.

View attachment 288363 View attachment 288364

First: 1060 is mid range at the moment. 70/80/80ti are above while 50/50ti are below. Problem is that its performance is similar to the 970 and that was the entry high end card in late 2014.
Anyway a 1060 should be fine with fullHD in all modern games if you have a mix of low/mid and a few high settings for sure! Acc is a bit too heavy at the moment for what I looks like!

Regarding cpu: the problem here isn't that your mobile cpu is bad, the problem is that by saying it's a 2018 i7 and the number is something with 87x0 everyone is thinking it's a modification of the 8700k and not that it's a mobile.
Just misleading. Imo that's a fecking beast of a cpu for laptop!

Anyway about performance numbers though:

Problem is that you probably can't overclock the laptop i7 due to heat.
Friend of mine got an i7 8700k and only raised the multiplier (5 GHz) while checking if the voltage is staying below what's recommended. Changes the numbers to this:

My I7 2600k is running at 4.4 GHz easily since a few years. Changes the numbers to this:

So your 8750 beats mine in the overall score but at single thread performance they are the same while the 8700k is a massive jump ahead.
And since Single thread performance is still the limiting factor for simracing, it was a bit misleading without the "h" at the end or mentioning that it's a mobile cpu.
So you've got a cpu that has the per core performance of my i7 from 2011 but with 2 more cores, no overclocking and it might be throttling due to heat.
Its still an absolute beast for a mobile cpu, don't get me wrong. But it's nowhere near the standard 8700.

 

[Case_]

I still don't think a 1060 should be considered a mid range GPU in 2019. It mostly performs perfectly fine if all you need is 1080p/60 medium-ish detail and couple it with adequate CPU, sure. But 1080p/60 on medium detail really isn't mid range anymore IMO - that's more like 1440p/60 or possibly even 1440p/120.

And with some recent games, you will be struggling even with the 1080p/60 on medium detail to not drop below 60. Just look at AC: Origins/Odyssey (which are BTW the most multithreaded games I've seen so far, with moderate to heavy usage of up to 9 threads). Or ACC, since we're already talking about it - even if your CPU can manage to keep you at 60+ fps in clear weather and you stay CPU bound for the most part, throw in some night racing and/or rain and a 1060 range card drops down to its knees with framerates possibly dropping to low 40's.

 

[Neilski]

Regarding the 8750H: the TDP is very low relative to a desktop CPU, which means limitations even when everything is going perfectly. However, many laptops aren't capable of shifting all of the heat from the CPU on its own, never mind the combined load of CPU+GPU when both are running hard, and then you get throttling.

Also, I'm with Martin on the loading thing: the fact that the i7 seems to be saying that all 12 execution units are pegged is a clear sign that something is wrong - ACC isn't capable of doing that by itself, so I recommend checking what else is sucking cycles. I don't even see a way for throttling (alone) could cause that behaviour, but I would still strongly recommend checking for throttling if you haven't already, @jven - see @RasmusP's suggestion to check the clock (I normally use Win10's Resource Monitor, accessible with a single click from the Win10 Task Manager's 'Performance' tab).

 

[RasmusP]

I still don't think a 1060 should be considered a mid range GPU in 2019. It mostly performs perfectly fine if all you need is 1080p/60 medium-ish detail and couple it with adequate CPU, sure. But 1080p/60 on medium detail really isn't mid range anymore IMO - that's more like 1440p/60 or possibly even 1440p/120.

And with some recent games, you will be struggling even with the 1080p/60 on medium detail to not drop below 60. Just look at AC: Origins/Odyssey (which are BTW the most multithreaded games I've seen so far, with moderate to heavy usage of up to 9 threads). Or ACC, since we're already talking about it - even if your CPU can manage to keep you at 60+ fps in clear weather and you stay CPU bound for the most part, throw in some night racing and/or rain and a 1060 range card drops down to its knees with framerates possibly dropping to low 40's.

I think we can agree to "disagree". I was half asleep when writing my post above. I meant: "from my personal perspective I also think the 1060 can be called mid-range, still".
Totally fine that you have the opinion of it being low-end.
We both know the performance of basically all current cards so we really don't need to discuss anything. Our perspectives on that are influenced by our surroundings/virtual surroundings and they are just that

For me the term "mid-range" means the card I recommend to people who want to be able to play everything just fine but don't want to spend big money. 1060 still fits into this I'd think. 1070 or 2060 and above are all 350€ and more here in Germany and that's too expensive for mid-range from my perspective.
Also I really don't know anyone in real life with a 1440 or 4k monitor. Sure, they got 4k TVs as you basically can't buy FullHD TVs anymore over here but monitors... A few even bought new monitors lately, also just FullHD. Maybe that's different for your environment?`

One could argue that just because the average person buys low-end stuff, we shouldn't call that low-end "mid-range" as the term "mid-range" should be purely based on raw performance. And I totally respect that view on it. For me the term is just something different.

Long post, but I felt like writing it. It really doesn't matter for both of us how we call what. We know the facts and that's the important thing. Our decisions aren't based on what others might call how

 

[hozes]

for me; the catagorization of gpus depands on the resolution, for 4k rtx2080 or gtx1080ti, for 1440 gtx1080 vega64, for 1080p gtx1070 vega54 shouldbe high end gpus for maxing out settings ( 60fps of couse)
i am on my 1360x768 screen, i can almost max every game i play; pubg, rf2 ( i always turn down shadows one click from ultra, shadows are fps hungry nomatter what game is)
so i think for 1080p/60fps, with a gtx1060 or as in my case on1360x768 with a rx570 (shadows on high) we can play acc easily, but cpu optimization is somehow not figured out yet, so we hit big fps racing while ai.

 

[maelstrom]

If you don't already have it, get MSI Afterburner. You can set the options to cover almost all aspects of system monitoring and the subsequent graphs give you about 10 minutes or so of performance data. It also allows you to create a custom fan curve for your GPU which is always a good thing. I agree with much of what's been said about laptops and heat. Because of the restricted space, it's difficult to keep the CPU at a sensible temperature and, no matter what the max turbo speed is, it's unlikely that it will be running at it more than just occasionally. Most laptops will clock down fairly quickly when they start to get warm and will rarely produce their theoretical maximum performance - it's just the nature of the beast.

 

[jven]

All, very good discussion in this thread.
Thanks for all the comments above - useful and I usually have a few other things open using a little bit of resources, but as I showed, ACC was consuming 69% of the CPU itself at the one point in time.

I'm not so concerned about this for myself as I am for others - even if my hardware is lowest of the low end it is still nearly at the ACC Recommended Hardware level, which means they are understating what is needed to potential race opponents.

When buying this laptop last year, I decided to make the tradeoff for slightly lesser hardware but instead having a laptop with all my racing games installed for when that day ever comes I have to spend the night in a hospital, as long as my laptop is there maybe I can still race! Um...well, drive is a more accurate term for me, race implies I'm competitive or something.

 

[Case_]

Thanks for all the comments above - useful and I usually have a few other things open using a little bit of resources, but as I showed, ACC was consuming 69% of the CPU itself at the one point in time.

You know what, I had another look at how my CPU usage is while running ACC (Ryzen 2600 at 3.9 GHz all cores, meaning it's also a 6c/12t CPU just like yours, and still below recommended CPU for the game, which is 2600x, though the overclock likely compensates for that), and it is generally around 50 %, spread over all cores (because, as we discussed, that's what usually happens). Now granted, that's with various stuff running in the background like a browser with quite a few tabs and also with OBS recording (but using NVENC, so CPU usage around 5-8 percent), but if we consider that especially with the overclock, it's a somewhat more powerful CPU than yours, and that yours might really be throttling quite a lot (you really should check the frequencies and temps you run ACC at), I guess 69 % usage isn't that out of reach.

I'm not so concerned about this for myself as I am for others - even if my hardware is lowest of the low end it is still nearly at the ACC Recommended Hardware level, which means they are understating what is needed to potential race opponents.

See above. I really think you should look into what's actually going on on your own hardware to get more data to work with. You are judging the performance of your CPU based on some Passmark score we know very little about, and - unless I missed something, you don't really know if your own hardware even reaches those benchmark scores and/or how far off it is. Mobile CPUs really can be quite underwhelming under stress fairly often, because, as also mentioned before, their TDP is quite low so they throttle quite early, and they often run fairly hot due to limited cooling, which further complicates things.

Also, it only just occured to me that your GPU also is the mobile version of 1060, which means it's clocked fairly lower than a desktop 1060, and that in turn means its performance in a demanding game like ACC can easily also not be on par with a regular desktop 1060. Looking at some benchmarks, it seems it can range from being almost on par with a regular 1060 to being almost 30 % slower in some situations. So this is likely another piece into the puzzle of the low-ish performance you are seeing, and one more reason to be somewhat conservative when comparing your setup with a regular desktop gaming PC.

 

[jven]

You know, I have a totally different viewpoint on this. Personally, I think that while what you suggest is theoretically possible, the odds of it being the actual problem are miniscule. The reason is because the "preponderance of the evidence" is already against it.

While ACC comes to complete freezes of probably 1 frame per 3 seconds at times during a 20-car single player race (it seems worst during turns) and the graphics look like crap with 6 million marbles in the grass but "stuck to the screen??", none of the other apps/games I use exhibit even 5% of the extremity of ACCs problems. Therefore, this is ACC's problems and I don't feel the urge to even begin to investigate what I can just assume is inadequate and oversold development. After all, my hardware measures up well to their "Recommended", not minimum but Recommended specifications.

And before you make further broad generalizations about mobile hardware remember - that hardware is successfully running many other fairly heavy duty 3d apps/games. rF2 in DX11 single-player with 40 cars, no problem! DCS, Project Cars 2, F1 2018 - all good. Clearly ACC is failing. And so I'll leave it to them to fix it. And I expect they will optimize it out if means dropping some detail. But I don't care - if they don't manage it then I won't recommend the game and will just use others; plenty on the market.

We can jaw about the efficiency of mobile architectures all year, but the mass of evidence described already simply says you're wrong. And based on that evidence, I don't feel it's my problem to pursue. I guess we're done until ACC rolls out v0.6 in 2 weeks.

 

[Case_]

Ah. OK. So you don't really want to solve anything or know what's actually happening, you basically just want to complain about it.

Fair enough, I guess. Sorry, I just didn't realize we've just been wasting a fair amount of each other's time here.

 

[John-Eric Saxén]

After all, my hardware measures up well to their "Recommended", not minimum but Recommended specifications.

It does not, because as Martin explained to you, mobile hardware and desktop hardware are two completely different things. Mobile hardware has to adhere to TDP, heat and battery limitations, which are much stricter than for desktop. Unless you actually bother to run benchmarks to compare your hardware against similar desktop components, you can't compare it against the official recommendations.

 

[jven]

It does not, because as Martin explained to you, mobile hardware and desktop hardware are two completely different things. Mobile hardware has to adhere to TDP, heat and battery limitations, which are much stricter than for desktop. Unless you actually bother to run benchmarks to compare your hardware against similar desktop components, you can't compare it against the official recommendations.

Boy, it's a reading comprehension thing, what can I say? I appreciate any efforts at help but all along I've been doing exactly what you said I should do - comparing using the most standardized, best available benchmarks. Are they perfect? No, of course, no benchmark ever will be. But they are the best available. And yes, I do still say, even after considering these comments, that they measure up well.

As for "similar desktop hardware", that makes no sense - we have the benchmark of my specific hardware and we have the benchmark of the specific hardware that Kunos recommends. In fact, according to the benchmarks I'd say they are similar hardware. And again, although my hardware is close to Kunos' recommendations, I'm not running maxed out which their "recommended hardware" should readily support.

And overall remember, I am running 1080p, not 4k. And rememeber, too, ; ACC is supporting 4k, which I can run it at; but switching down to 1080p should give some extra room for a lesser system. Yet, still the issues...

Given the above, I feel like the "your mobile hardware is insufficient" is well settled in it's [my hardware] favor - especially when the issue with stopping frame rates on a 20-car playing field happen on lap 1. That's no time for thermal issues to kick in too hard. And yes, I have a video capture of a such a race and the frame lockups occurring only 1 minute in - as I had mentioned they seem to be worse in the big turns, but that's probably just because they'll be most visual there as distant objects will appear to move more from one frame to the next.

 

[jven]

Ah. OK. So you don't really want to solve anything or know what's actually happening, you basically just want to complain about it.

Fair enough, I guess. Sorry, I just didn't realize we've just been wasting a fair amount of each other's time here.

OK, Martin, let's tackle this. We're into a bit of rut here and it's not productive at all. As much as I want ACC to run beautifully on my pc, I also hope others can gain something from this thread which has probably gotten a bit too focused.

We can discuss the finer points of the thermodynamics of processors all day, and we can lose sight of why I posted here in the 1st place. In fact, I did 1st post about a couple issues and was asking whether other people were seeing the same issues (especially with the graining/marbling which I suspect is tightly related to video settings in ACC) and if there were known "accommodations" or workarounds. In fact, I asked about this several times. I guess you got so focused on the thermal/hardware issues that you discounted that part. As for threading, this is almost 2020 and I can't help if idiot project managers are buying into monolithic architectures, that will always be a problem for their product down the road and only get worse. AC 2.0 will need to be more open-threaded, threading libraries are old now (and open source).

Even if my hardware is fully insufficient, I am running the game, and the answer I am sure is not "you're insufficient" but "well, quick fixes might be to turn this, this, and that down...".

So getting past all that, and noting that I am sure I am going to try to beat you in an online race one day , it seems no one else commented in this thread about the issues I mentioned so let's just leave it at that and wait for v0.6.

TBH, at this point in my life I just want to race on my happy beautiful race-view screen seeing glorious race cars all over; and I want to dig into hardware specifics about as much as I want to learn about caster, camber, toe-in (-out?), and bumpstops.

So I'll remain optimistic I'll get this all worked out by the time ACC is public (it seems they still have a lot to deliver after build v0.6?) and see you on a server sometime. And I promise I'll try to quit typing so much!

 

[Albert McSaltens]

Does anyone know how to get rid of the vinyete effect on replays? other than turning post proccessing to low?

 

[Durge Driven]

It is all relative, RTX2060 beats a 1070Ti and a 1060 beat a GTX970 so midrange beats hi-end in both cases

What game in same era as any 60 series does not run at least medium/high settings

 

[Zed2018]

ACC has an optimization problem and it is a fact.
I've reading other forums aswell, its not just a one guy thing.
For example I thought I could play it at minimum details at 1080p, with an i5 6500 4x3.2 GHz, 16 GB and an RX 470 4 GB. It turns out it eats my cpu like prime95 would do, cpu is in the menu at 100%, not variating but completely pushed to the limit. Changing Resolution doesnt change anything. Playing ACC is done in 40 FPS hotlap mode, lowest possible details selected. My PC runs all other games pretty solid.
Cheers