Hi £1042, I take you have cl36 6000 2x16gb ram. If so i have the same ram. samsung bdie. Can you tell me what dram voltages you set for this as the Zentimmings screenshoot omitts it. also was there another other settings in bios that you used . Gigabyte b650m gaming x ax , 7800x3d, corsair 2x16gb 6000mhz
@@kaisersolo76 ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-utVaTD2S6mA.html that's the original video- I don't think I messed around with the dram voltages too much! The video should run you through most of the changes I made.
Hey I'm super CPU bound, do you know if changing the options "Scattered Object Distance" or "Terrain Tessellation Distance" will have any effect ? I feel like SC needs a little tag on these settings that says Affect CPU or GPU...
@@Kalyptic Hey! That's a very good question! I've never been able to work out if they actually do anything- or like you say, what they effect. A tag/description is a very good idea!
Properties and disable full-screen optimisation and set to run as admin And run the game with high CPU priority and you can set this using a registry key so you do it once and it stays Ask and i will add it to this comment [in will add it when i can my pc isnt setup right now]
It's important to emphasize that you can only inject FSR 1 in SC, which is greatly inferior to FSR 2 and DLSS 2, so people aren't flabbergasted about the lower quality. Not all FSR is created equal.
Fortunately, if you use Lossless Scaling, you can use its LS1 upscaler which does a better job on antialiazed games like SC than FSR 1 does. Still nowhere near as good as FSR 2.1 or DLSS, but it can still look surprisingly good and provide great performance gains if you're GPU limited.
Very helpful! One little addition/correction: SSD speed DOES matter when you're on 16GB of RAM, because page-filing. I clearly observe this on my second pc.
Don't turn off vsync. Switch to Fastsync (Nvidia) or Enhanced sync (AMD) and set double buffer instead of triple buffer. All in your driver's control panel. No tearing, no lag.
@@WillieBeamin270 You set up everything in control panel and then you can turn vsync off in Star Citizen just to be sure it won't interfere with driver settings.
Another thing to keep in mind is cooling, ESPECIALLY if you disabled e-cores. Your cooling software only reads total die temp, so while you might be seeing 80c and think “that’s well within my hardware’s spec,” your individual p-cores could be throttling at 100c+. Switching to a more aggressive fan curve that keeps the overall die temp at something like 60c under load helps drag the core temps down to values within their operating rating. When I did this I saw a DRAMATIC improvement in performance. Also keep in mind that, in most scenarios, your frame rate is tied to the netcode & server performance. So if the server is strained you’ll see performance degradation locally regardless of what you do client-side.
Absolute legend m8. This is the new definitive ( and tested) guide. Thanks for all those hours and hours of rerunning tests. Must have been brutal. o7 sir
For the high settings bit, I noticed a difference on my old computer between low and high. It was small, but it was there. When I switched to my newer and more powerful computer, I noticed no difference whatsoever. Also, this was all before Gen 12 was added, so it's quite possible that if I went back to my old PC that there would no longer be a difference.
Saw this video yesterday and I have a 12th generation Intel i5 processor. Turned off my e-cores this morning to check if there was a difference because I was suffering from severe jittery behaviour when going to mobiglass or even inventory. Turning off e-cores made such a big difference! All jitteriness is gone. (with 3.23.a for reference)
@@tenpoundfortytwo maybe it is a system by system issue or even for certain cpu's, I have no idea. I could try to switch a few times back and forth to gather some more information but not sure if it would actually be worth the effort atm. I'm just glad I can play without having to wait several seconds each time I want to open my mobiglass or check my inventory.
Excellent summary, big thanks for all your research. Few things I've found work for me: I've been using RSR in the AMD software to upscale to 1440p and it looks pretty decent. Switching the game from a Sata SSD to a 970 NVMe significantly reduced stuttering, especially for the first few mins of gameplay. Made a big positive difference in smoothness for me.
A great video, comprehensive and honest. Some of those will definetly help with 1% lows and make the game feel better even if the general fps don't increase by much. Ultimately we'll all need better CPUs if we ever want to play this game at 4k on high without dipping below 60. Or CIG could optimize it ... but then again I think that is a long ways away.
Thanks for the video, overclock RAM apparently improve a lot. I try to use default XMP(3600) to use memory try it to 4266. Thought I don't record the FPS, it really feel much better . My set is 13400F, RTX3060 12G, XPG D50 3600 DDR4
Awesome video, a full comprehensive list! I did find turning fTPM off with core isolation and memory integrity both off to improve 1% lows quite a bit, and averages by about 3%. This is on Zen 4 only. ReBAR might help a lot if you are running out of VRAM, but that's unlikely to happen on cards that support ReBAR, as I don't think there are a lot of 6 GB cards, and most people wouldn't be playing at 4K with those :D
Very interesting- I'll have to have another look at TPM! I seem to remember something about some motherboards not having a hardware module and using software, but I can't remember properly!
Hmm, just to be clear TPM is not used at all by any system (hardware or software based TPM) in Windows by default. It is a in module that are only being called like when you encrypt your drive or using like biometrics with your login. Really struggle to see how a TPM can hinder anything in a game... unless there is some kind of bug in Windows of cause.
@@muf6861 fTPM is a "simulated" TPM inside the CPU as opposed to a physical chip on the motherboard. On intel platforms, there is no reason to disable it, it doesn't affect performance. With Zen 4, there still seems to be a bug with it, even though AMD claimed it was fixed last year.
I noticed a smoother experience in SC when I switched from an i7-5820k to an i7-6950 Extreme. Despite the slightly higher Ghz value of the 5820 and very similar fps values reported from the rtx3070. Bumping the Extreme up from 3Ghz to 4.2Ghz made a difference also. So many comments in threads telling me I'm wasting the rtx4090 on such an old CPU (7 yrs) Yet this old CPU can run 38 - 72 frames in SC 3.19 across three 4k screens 🤔 Rtx4090 set to 95% voltage. Seeing Gpu Temps hitting as high as 58C. My buddy is running an i7-9900k, 32GB ddr4 3200 with an rtx4090 on an Alienware 34" wide-screen oled (3440x1440p) averages in the 70 Fps. His GPU Temps are in the low 50's. (Both are Zotak Trinity OC)
Thanks for the info as always a great video and a lot of help the only thing I am not happy with is .. when Gen 12 is done blar blar but its no difference lol
@tenpoundfortytwo Have you ever explored 2 vs. 4 sticks of RAM on AM4? Gamers Nexus explored this through Shadow of the Tomb Raider benchmark and found significant improvement. Given your RAM timing profile, RAM speed and capacity all make a large improvement in Star Citizen, this could be worth a geeze for us running on AM4? Love your work!
I love when big companies, advertisements, and helpful streaming/RU-vidrs tip providers do this because they set themselves up for the criticism I am about to give. This can be taken as constructive criticism because many big-name companies, streamers, and RU-vidrs all like to cut corners or neglect doing what should obviously be common sense. When you show a split screen regarding performance comparisons, all screens should be in motion, displaying the same environments. What you did, and so many have done, is only indicate motion in the HDD performance screen. High-End and Low-End SSD display screens are focused on one spot. As a result, many assumptions can be made about why you neglected to correctly display all comparison windows with the same or similar motion rather than the facts you wanted to convey. Little mistakes like this can also affect credibility or validity in some of your viewer's minds. This is constructive criticism, so I hope you are open to it.
What people often dont know in regards to SSDs is that the transfer rates given on the box are the maximum the drive can achieve in direct data transfers. In normal operation like in a game youre limited to how fast the ssd controller can actually find the data it wants. In those cases even though they usually got better controllers the fastest pcie nvmes arent much faster than a good sata.
@@tenpoundfortytwo When I was reaching 100% memory the performance went down whole PC started to stutter. Since increasing to 64 gig no noticeable difference in performance from before except no longer hitting 100% mem and stuttering.
Don't know how much you paid for the RAM upgrade but they've confirmed this is due to memory leaks, the RAM usage is never cleared if unused assets and it is a bug, in the future the game will never reach beyond 24-26gb of ram in average.
I DO recommend trying the quality settings, very high in fact absolute boosted my fps by at least 30 fps in area 18, i have found that way long ago where this works on multiple games, even on planetside 2 setting textures to ultra makes it so to cpu doesn't have to downsize textures, only do this if you can see your gpu utilization being low while cpu is at 100%
About the "delete your shader cache" bit: The shaders are already split into subfolders based on game version. So when you update from say 3.20 to 3.21, you already get a fresh shader cache. So how would deleting the old shader cache help anything? Hint: It won't. The game client wouldn't use it anyway. This might have helped some time ago, when perhaps the shaders were not separated by game version. But nowadays cleaning this folder won't help one bit, except free up a few hundered megabytes of disk space.
I've been playing SC for 7 or 8 years on a home built I5 Dell Optiplex with a Zotac GT1030 and maxed out ram slots. I did get a solid state drive dedicated to SC though, eventually. I am currently using it to play 3.23, I shit you not. I just ordered a proper rig a few days ago. Used, but proper.
I'm running an undervolted 5800X and an overclocked 3070 and they seem to be a good match for each other in SC. They both hit 100% utilization right around the same time. During the worst that SC can throw at me I'm running 82-83c on the CPU and 80-81c on the GPU. That's just below the thermal throttling limits so I'm maxing out each of them, CPU at 4.8-4.9 GHz, GPU at 2070 MHz. Getting 110-120 frames in quantum travel, getting 40-70 most other places with dips into the 20's-30's in cities (1440p).
Just gotta say that optane drives are on a 100%different level than nand flash nvmes. It's latency is much closer to that of ram and it's ability to handle multiple writes and reads at once is unaffected by how much capacity remains on the drive. I use an optane as my c drive and have 16 gb of it set as as virtual memory on top of 32 gb or ram. In my experience, optane helps reduce the duration of long stutters. I've also seen framerate increase on optane, including a 100% increase in framerate in mwo, another cry engine game. When I watched the drive access of various games in the windows hardware monitor, I noticed that some games heavily rely on the speed of swap memory storage and others are constantly reading driver files from the boot drive. This is why I think optane has helped me so much. Other than my optane my pc is kind of a mess using an old e5-2690 v4 with 32 gb of single rank ddr4 2133 run in quad channel mode. GPU is an RTX A5000 and I keep my games on a separate nvme that uses a 2nd stick of optane as a primocache. I'm constantly cpu bottlenecked but the use of optane seems to lessen that a little. I also plan to upgrade my ram to 128 gb of dual rank ram running in quad channel mode. Needless to say, Star citizen has been a fun game to benchmark my pc with. It's the first game I've ever played that uses all 32 gb of my ram. =(
Well now, I feel like an idiot. Checked my RAM and it was running 2133 and I built my PC! I forgot I did a bios flash a few months back and never OC'd my RAM back to 3600.
Yeah, I just watched your video comparing both cards again. 4080 is slightly ahead on almost every test you’ve made. I thought the 7900 xtx was going to be the poor man’s 4090… But it turns out the 4080 is. That’s what I have atm. Also 4090 is very power hungry…
Awesome video! What would a 5800x/6800xt w/48g3000mhz(2x16+2x8 matched spec) be bound by? Would a X3D be a better upgrade over 5900/5950? Would 32gb @3600 be better? Target is 100ish avg fps max settings @1440p.
100 fps is almost impossible for even the latest and greatest CPU's unfortunately! For that combo I think the 5800X would be the bottleneck in most areas, but at 1440p there might be parts of the game where the 6800XT is struggling. 3600 ram would definitely help, but you're talking a few fps. The 5800X3D would be the best upgrade option for am4.
Good video. I just wish I could play the game in 1440p without the game crashing on me lol. Everything is just fine in 1080p but it soft resets my PC shortly after changing it to 1440p (commonly when exiting the hanger in my ship or arriving to a planet for a landing). This is the only game that soft resets my entire PC and on occasion will completely boot my AMD gpu driver in the process.
Can you do a vid about memory and Pagefile.sys? with 16GB, what settings should you pagefile.sys be? Should you let Windows set it automatically or do it manually? second, if you have 32GB RAM, should you change it too? 64GB? Maybe a video not about performance, but rather about Stability for SC.
10:28 good to know for 4K, but I am running a laptop with a laptop version of the RTX3080 in it and for 1440p it runs well enough 40+ FPS with very high clouds on. Granted they may have improved things with the build since this video, 3.22.0
Hi you mentioned turning off SMT in BIOS; I used the batch file method instead to isolate Star Citizen to the first six Ryzen cores. Would this work also?
Good guide, but 50% of hints are for desktop. Maybe was better to aknowledge before each point. Unfortunately, RAM timing and E-cores off aren't a thing on notebooks, for example.
One note on memory over clocking, if using an AMD 7000 series CPU, you won’t be able to have a stable system with frequencies much higher than 6000 MHz even if your memory kit says otherwise. 6000 is the sweet spot. Timings and Buildzoid tuning are even better. I can tell SC is smoother without even looking at FPS.
Not sure why that would happen- did you definitely save the changes in the BIOS? I suppose it's possible that the XMP profile isn't stable and so it's reverting back to something that will run.
Great video 10-42. I can only speak for the 13th gen Intel (i9-13900k) but I've tried turning e-cores off in the bios, using Process Lasso, and just leaving them on, and there is a significance difference from 3.18 to 3.19 when it comes to FPS. I leave them on, and I get the best FPS that way. I don't think the 12th gen CPUs got as much of a boost as the 13th gens, but in max 2k settings, I have hit 175+ FPS in quantum, and 115 FPS in Area18 with the e-cores on. Every other configuration resulted in lower 1% lows (stuttering, etc.) Now, I'm also running an MSI RTX 4080 Gaming X Trio OC GPU, 64 GB Corsair Dominator DDR5 6200mhz, 2x SN850X SSDs, all on a Asus RoG STRIX Z790 mobo. So I genuinely think something was changed from 3.18 to 3.19 when it comes to the hybrid instruction set code (or there were several small Microsoft Windows 11 updates as well during that time, that could have had an effect on it). But that's just me. I've noticed 3.19 running better on my i7-11700k w/ 4070ti and 64gb DDR4 memory, on Samsumg 980 Pro SSD as well). Just my experience.
It seems like the problem is the combination of e-cores and Windows 10. I just played SC for the first time tonight (because free!) on a i9-13000K with a 12GB RTX 3080 on Windows 10 and it was unplayable. It took me about an hour of trying different in-game settings (none helped) and googling to figure out the problem. It still doesn't run great (compared to the buttery smoothness of No Man's Sky, for instance) but it's playable at least. But great video! I love the fact that you debunked several of the myths. Fyi, here is the command to run SC without e-cores on the i9-13000K WITHOUT disabling them in BIOS: %ComSpec% /C Start "RSI Launcher" /High /Affinity 0000FFFF "C:\Program Files\Roberts Space Industries\RSI Launcher\RSI Launcher.exe"
What I would REALLY want to see is how does this game run on a server with 2x AMD EPYC 7371 CPU paired with a high-end CPU and tonns of RAM. Something tells me, that would be the best performance, hands down. Or maybe a Threadripper 1920x?
Unfortunately it probably wouldn't run as well as you'd think! The advantage those parts have is lots of cores- but SC really only cares about fast single core performance. Seeing as those chips tend to clock lower, they don't perform that well.
@@tenpoundfortytwo Thank you for your kind reply. Since you are reading this I can thank you for all your work with these videos 🙂They are a handful to watch. Especially the latest since I'm on ryzen 3600 planing to upgrade to 5800X3D so it is hyper relevant for me. Thanks again! However! I think Star Citizen would run suprisingly good because looking at not only the FPS of the game, but also the task manager and the /core load of the CPU, I can see, that Star Citizen is using the cores irregularly. That shows me, that CPU cores are being "turned on and off" using between 80-90% when "used" and 60-70% when "not used". This tells me, that the game is pooling a significant amount of data (from I assume the memory) and when it is done, then start ramping up the core again. Until that, it is used for "general gameplay usage". I have no clue what does the game do to mimic this behaviour but it is clearly not optimized, as we all know. Another upside for e Threadripper or Epyc is, the fact, you can use like 2TB of RAM if you wish. Now that would be for Star Citizen, like a birthday cake for a 3 years old 😀 I know, this is not relevant (again, feedbacking to your video), but CIG I believe are lying also about the telemetry. I can't really imagine someone hitting a stable 60 FPS with something like a GTX 1080 (non Ti). That bugs me a lot. I hope, in the future, you'll have the chance to test the game in a system like what I've been talking about and I'm really really sorry for the long post. It is just rare to find interactive RU-vid creators. 🙂
@ImreBertalan86 no problem, thanks for the kind words! Yeah I would like to test sc running completely on! One day! The telemetry page is a strange one, I'm not sure they're lying- i think the page is just completely broken and the way they calculate the average is very strange!
A question about clouds in 3.19 and VRAM, and does having more VRAM help? I have 1 PC with a 3070 Ti with 8GB and a RX 6750 XT with 12GB. Seeing that the performance levels are somewhat similar, do you think I have something to gain from having more VRAM? They are both running PCI-E 4.0
Interesting question! What cpu do you have? That could actually be the bottleneck and the reason why both cards are performing the same. In the testing that I've done, VRAM usage isn't actually out of control in SC. But I need to do some more up to date testing!
"Very high" settings are not a myth. I checked it just yesterday. I have an old PC and 80% CPU usage. If I turn off the "very high" settings - I immediately get 100% bottleneck. I return "very high" and the CPU usage is again around 80%. Tested several times in different (urban) areas.
running rx5500xt and r5 3600 its mostly gpu bound, playing at 720p low graphics only getting 30fps to 40. 16gb ram but ram util is not as bad as gpu usage
I just started two days ago and would love this game as is if it would run for more than 15 minutes without bugging out or "unexpected error"... will it be better later this week after "the event" is over?
I9 13900KS has no issues with E-Cores, no stutter at all. If I turn them off with process lasso I have worse performance almost 5% loss. Same with HT, HT On gives me more performance.
Just reinstalled SC and what I noticed was sort of ghosting/screen tearing. Changed Nvidia setting to 10gb cache and vsync fast and it looks much smoother. Not sure if it’s visible in fps, but for the eyes it’s a million times better
Something else that can help mitigate ghosting is black frame insertion, which is called different things by different manufacturers, samsung calls it MBR or motion blur reduction. It's gonna dim the screen a bit, but you can turn up brightness and contrast to mitigate that
Last year I built a new PC but didn't have enough money for a new GPU and had to sue the 10 year old one! 5800X Kingston Fury Beast 2x 16 ddr4 3600Mhz I can install the game on one of these the SSDs, but as you said there is no differenc -Samsung 870 EVO 500Gb -Samsung 970 EVO Plus 1Tb NVMe GIGABYTE GTX 760 2Gb WindForce 3X OC Asus ROG Strix B550-E Gaming I'm worried about my GPU if the game is going to run ok.
It's in the PC right now as I type- but very tricky to do any testing while free flys and events are on. Planning on getting testing once 3.19 is back to normal!
I disagree on setting graphics quality to "very high" being a myth. Both with my previous 3900x/2070 super build and the current 7950x3d/6950xt build setting quality to very high resulted in the highest average framerate, with very high giving almost double the average framerates as the medium setting does. The only time this won't be the case is if you're running a relatively low end GPU, in which case you're not really offloading enough work off the CPU to have an effect.
I have a 5700xt with r9 5900x 32g ram at 3200 and m2 ssd. when I start playing I have 70 to 90fps which is great. When I land on outpost for a mission I have huge drops. About 15 to 20fps maybe less. I m playing with friends with lower specs and at these situations they have more fps than me. Any suggestions?
About turning VBS off, this thing makes computer vulnerable and anti-cheat systems are kicking players who disabled this. So how to deal with this then ?
do not recommend turning off the default memory security, not having this on BRICKED armored core 6, it refused to run because of EAC and issues to do with the memory security.
"high setting myth" its true. mayby depending on the system,but for me ,it instantly giving me 10 fps. from around 25-30 fps in area 18 to 36-40 fps. i haVE I7 7700K
If you only have the game open then no it wont But if you have some-stuff in the background it can help i see upto 60% ram usage with the game and a couple other bits open like discord and playing music on youtube But what i will say is if you did upgrade to 64GB it would need to be as a 2x32GB kit as reaching 3600MHz with 4 sticks of 16GB is typically not possible So that being said if you opt for an upgrade to 64GB you would be silly not to buy a 4000MHz kit seeing as the price of DDR4 has plummeted [again 2x32GB]
Not on your platform, for longevity and if the price isn't absurd, upgrade to 64Gb when you'll build on a new DDR5 platform, even then 64Gb won't help themselves for star citizen, there's a few memory leaks in 3.19, it will not stay like that long term, and will probably hover under 30gbs at all times.
@mav_4404 I was wondering if you'd manually overclocked your gpu or cpu- but sounds like you haven't. Tricky to offer any advice other than make sure your drivers and bios etc are up to date. Have any other games started crashing?