IAF.Cheetah Posted December 25, 2018 Posted December 25, 2018 Hi guys, well.. as the title suggests, i'm buying new parts for a new build and I stumbled upon a video on youtube that was checking the READ/WRITE of these SSDs during gaming. The result was - NO CHANGE AT ALL. and all because today's games don't use more than 550 MB of read/write data. Can someone here concur this? did you upgrade from SATA to NVMe SSD and believe there is a noticeable change for the better? :helpsmilie: It can save me at least 220 $...
Demon_ Posted December 25, 2018 Posted December 25, 2018 https://forums.eagle.ru/showthread.php?p=3669407#post3669407 Attache ta tuque avec d'la broche.
Mars Exulte Posted December 25, 2018 Posted December 25, 2018 The only place something like this is bound to make much difference is during actual loading of a map or something. A 60gb map will load much faster on an NVMe than on a SATA SSD or conventional drive. For actual gameplay once everything is loaded? It really doesn't matter what you use 99.9% of the time. Де вороги, знайдуться козаки їх перемогти. 5800x3d * 3090 * 64gb * Reverb G2
hansangb Posted December 26, 2018 Posted December 26, 2018 Of course there *IS* a difference in speed for NVMe or SATA. Every test shows that. But probably not in DCS. I certainly didn't notice it when I changed to NVMe. But I use my PC for more than just DCS. Well, mostly for DCS, who am I kidding, but I do use my PC for other things. :) hsb HW Spec in Spoiler --- i7-10700K Direct-To-Die/OC'ed to 5.1GHz, MSI Z490 MB, 32GB DDR4 3200MHz, EVGA 2080 Ti FTW3, NVMe+SSD, Win 10 x64 Pro, MFG, Warthog, TM MFDs, Komodo Huey set, Rverbe G1
Gladman Posted December 26, 2018 Posted December 26, 2018 Faster load times with the NVME but that's where it ends. Still worth it in my opinion. i9 9900K @ 5.1Ghz - ASUS Maximus Hero XI - 32GB 4266 DDR4 RAM - ASUS RTX 2080Ti - 1 TB NVME - NZXT Kraken 62 Watercooling System - Thrustmaster Warthog Hotas (Virpil Base) - MFG Crosswind Pedals - Pimax 5K+ VFA-25 Fist Of The Fleet [sigpic]http://forums.eagle.ru/signaturepics/sigpic99190_2.gif[/sigpic] Virtual Carrier Strike Group 1 | Discord
hansangb Posted December 26, 2018 Posted December 26, 2018 Faster load times with the NVME but that's where it ends. Still worth it in my opinion. Hmm, now that you mention it, when you fire up the instant actions etc., it does seem faster to load. I was just thinking about the in-game performance. I guess I could temporarily move it to the SATA/SSD that I have to test. Maybe I'll do that if I get a chance. hsb HW Spec in Spoiler --- i7-10700K Direct-To-Die/OC'ed to 5.1GHz, MSI Z490 MB, 32GB DDR4 3200MHz, EVGA 2080 Ti FTW3, NVMe+SSD, Win 10 x64 Pro, MFG, Warthog, TM MFDs, Komodo Huey set, Rverbe G1
Demon_ Posted December 26, 2018 Posted December 26, 2018 He says, he can save $220, not save 4 sec in loading time. Attache ta tuque avec d'la broche.
FragBum Posted December 26, 2018 Posted December 26, 2018 He says, he can save $220, not save 4 sec in loading time. Agree save $220 on SSD. Yes nvme is faster but once DCS is running (it boots up quicker now on 2.5.4) and it doesn't seem to make any significant difference in game. You'd likely gain more in a CPU/GPU maybe even memory bump. :thumbup: Control is an illusion which usually shatters at the least expected moment. Gazelle Mini-gun version is endorphins with rotors. See above. Currently rolling with a Asus Z390 Prime, 9600K, 32GB RAM, SSD, 2080Ti and Windows 10Pro, Rift CV1. bu0836x and Scratch Built Pedals, Collective and Cyclic.
pagadi-zaiet Posted December 26, 2018 Posted December 26, 2018 I have purchased a Samsung Nv 2.0 970 Evo 2TB. Cost me £465. Had problems transferring my Windows 10 from SSD with Acronis. Managed in the end with Samsung file transfer. Cannot see much improvement, debatable stutters pos worse. I tryed a fresh Win10 install but the partition is not recognised, system reports I/O issues. Not sure yet to keep the drive or return. Doubt is faulty, presume some kind of incompatibility with mobo. Tryed installing driver and update Bios to lates version which states in prev version M2.0 Samsung issue sorted. So yes bit miffed and not quite sure what to do. System works ok atm.
FragBum Posted December 26, 2018 Posted December 26, 2018 I have purchased a Samsung Nv 2.0 970 Evo 2TB. Cost me £465. Had problems transferring my Windows 10 from SSD with Acronis. Managed in the end with Samsung file transfer. Cannot see much improvement, debatable stutters pos worse. I tryed a fresh Win10 install but the partition is not recognised, system reports I/O issues. Not sure yet to keep the drive or return. Doubt is faulty, presume some kind of incompatibility with mobo. Tryed installing driver and update Bios to lates version which states in prev version M2.0 Samsung issue sorted. So yes bit miffed and not quite sure what to do. System works ok atm. Unless it's changed you need to install the drivers when you install windows at least that is the case with Intel 750 series. Stutters or drop frames are very much linked to CPU or GPU bottlenecks as far as I can tell. There isn't much drive or virtual memory activity going on in usual game play. :) Control is an illusion which usually shatters at the least expected moment. Gazelle Mini-gun version is endorphins with rotors. See above. Currently rolling with a Asus Z390 Prime, 9600K, 32GB RAM, SSD, 2080Ti and Windows 10Pro, Rift CV1. bu0836x and Scratch Built Pedals, Collective and Cyclic.
etherbattx Posted December 27, 2018 Posted December 27, 2018 NVMe SSD over SATA SSD ??? Faster load times with the NVME but that's where it ends. Still worth it in my opinion. i have not seen any improvement in games or sims with nvme. can you post your load time improvements ? it will be interesting to see.
Gladman Posted December 27, 2018 Posted December 27, 2018 Sorry, I've used that ssd for another system.. i9 9900K @ 5.1Ghz - ASUS Maximus Hero XI - 32GB 4266 DDR4 RAM - ASUS RTX 2080Ti - 1 TB NVME - NZXT Kraken 62 Watercooling System - Thrustmaster Warthog Hotas (Virpil Base) - MFG Crosswind Pedals - Pimax 5K+ VFA-25 Fist Of The Fleet [sigpic]http://forums.eagle.ru/signaturepics/sigpic99190_2.gif[/sigpic] Virtual Carrier Strike Group 1 | Discord
etherbattx Posted December 28, 2018 Posted December 28, 2018 (edited) NVMe SSD over SATA SSD ??? nvme is a faster technology than sata but most users, machines and most workloads can’t take advantage of it. unless you have specialized workloads, the speed difference between nvme and sata is 10x less than the time it takes to blink your eyes. or put another way, unless you are loading multi 1000 of files or a few 500Mb ones, you cant even measure the difference. Edited December 28, 2018 by etherbattx
hansangb Posted December 29, 2018 Posted December 29, 2018 (edited) UPDATE: See post 17 OK, ran some quick tests. On the same system. I opened an closed DCS 3 times. Ran the two built in tests with Caucus and one with Persian Gulf. On the opening speed, there is network traffic to Eagle.ru and DigitalCombatSeries that lasts about 24 seconds or so. Confirmed twice. So the start up time may be a function of the network traffic. Meaning no matter how fast your disks are, it may not matter. SSD is Samsung 850 Evo 1TB NVMe is Samsung 960 Pro 512GB /* results deleted for now. See EDIT below */ EDIT: Hmm, I think this is a flawed test. The shader files are still on my C:\users... directory if I recall. I'm away from my DCS computer so I'm not positive. And I'm sure accessing the shaders\meta files are a big part of it. For this test, I just made a copy of my entire DCS folder to the SSD and ran the EXEs from there. I *think* there is a way to move the c:\users\...directory for DCS. If I figure it out, I'll update the test. EDIT AGAIN. I just realized that I can symbolic link to the C:\users... OK, that's easier, let me give that a shot in a few days. Stay tuned. Edited December 30, 2018 by hansangb hsb HW Spec in Spoiler --- i7-10700K Direct-To-Die/OC'ed to 5.1GHz, MSI Z490 MB, 32GB DDR4 3200MHz, EVGA 2080 Ti FTW3, NVMe+SSD, Win 10 x64 Pro, MFG, Warthog, TM MFDs, Komodo Huey set, Rverbe G1
BitMaster Posted December 29, 2018 Posted December 29, 2018 You SEE !!!! IT'S 0.7 secs faster !!!! Kiddin' aside. It is the better drive when it comes to fast I/O...if DCS will benefit of such ultra low latency..I doubt it, DCS is no db. Gigabyte Aorus X570S Master - Ryzen 5900X - Gskill 64GB 3200/CL14@3600/CL14 - Sapphire Nitro+ 7800XT - 4x Samsung 980Pro 1TB - 1x Samsung 870 Evo 1TB - 1x SanDisc 120GB SSD - Heatkiller IV - MoRa3-360LT@9x120mm Noctua F12 - Corsair AXi-1200 - TiR5-Pro - Warthog Hotas - Saitek Combat Pedals - Asus XG27ACG QHD 180Hz - Corsair K70 RGB Pro - Win11 Pro/Linux - Phanteks Evolv-X
etherbattx Posted December 30, 2018 Posted December 30, 2018 NVMe SSD over SATA SSD ??? nvme is great tech. it’s newer tech. it’s faster tech. but its not “make DCS faster/better” tech. if that’s your goal, your time and money is better spent elsewhere
hansangb Posted December 30, 2018 Posted December 30, 2018 (edited) OK, I reran the test. So, DCS Beta installed on C: and D: drive. C: is NVMe and D: is SSD. First post has the HW spec. I deleted the "meta shaders", "meta shaders 2" and "fxo" sub directory in both SSD and NVMe tests. When running the D:/SSD test, I made sure the C:\users\..\SAVED GAMES folder had a symbolic link to the D: In other words, the SAVED GAMES folder only existed on the D:/SSD drive. This was to make sure that NVMe's (C drives) C:\users...folder was not used. Remember, I created as symbolic link to the D:/SSD drive. For lines 1, 4, 6, the times show how long it takes firing up the game/missions the first time. DCS took the time to build out the shaders. Lines 2 & 3 are just me quitting and starting the game. Lines 5 and 7 are me quitting the game, restarting the game, and firing up the instant mission again. So NVMe for DCS makes no difference whatsoever. ............................................SSD......................NVMe 1- Open DCS (in seconds) .........36.23..............37.67 2- Open DCS (in seconds) .........22.09..............24.33 3- Open DCS (in seconds) .........22.53...............23.77 4- SU25T-Caucus-Cold&Dark .....55.95 ..............54.85 5- SU25T-Caucus-Cold&Dark .....19.41 ..............23.15 6- T51D-Persian-Cold Start ........44.11 ..............41.35 7- T51D-Persian-Cold Start ........16.66 ..............18.23 Edited December 30, 2018 by hansangb Grrrrrr. I didn't use tabs, and it still screwed up the columns hsb HW Spec in Spoiler --- i7-10700K Direct-To-Die/OC'ed to 5.1GHz, MSI Z490 MB, 32GB DDR4 3200MHz, EVGA 2080 Ti FTW3, NVMe+SSD, Win 10 x64 Pro, MFG, Warthog, TM MFDs, Komodo Huey set, Rverbe G1
FragBum Posted December 30, 2018 Posted December 30, 2018 Question? Is the slightly longer times for NVMe because of PCIe bus overhead or lanes assigned? I realise even SATA uses a subset of the PCIe system but NVMe slots in at PCIe level as best I understand it. Control is an illusion which usually shatters at the least expected moment. Gazelle Mini-gun version is endorphins with rotors. See above. Currently rolling with a Asus Z390 Prime, 9600K, 32GB RAM, SSD, 2080Ti and Windows 10Pro, Rift CV1. bu0836x and Scratch Built Pedals, Collective and Cyclic.
cichlidfan Posted December 31, 2018 Posted December 31, 2018 The test is only valid if the drive is the only one in the system. When mixing SATA and NVMe drives the motherboard used provides a lot of variables. ASUS ROG Maximus VIII Hero, i7-6700K, Noctua NH-D14 Cooler, Crucial 32GB DDR4 2133, Samsung 950 Pro NVMe 256GB, Samsung EVO 250GB & 500GB SSD, 2TB Caviar Black, Zotac GTX 1080 AMP! Extreme 8GB, Corsair HX1000i, Phillips BDM4065UC 40" 4k monitor, VX2258 TouchScreen, TIR 5 w/ProClip, TM Warthog, VKB Gladiator Pro, Saitek X56, et. al., MFG Crosswind Pedals #1199, VolairSim Pit, Rift CV1 :thumbup:
hansangb Posted December 31, 2018 Posted December 31, 2018 Yup, it's too much of a pain to use one drive at a time. So I would leave it up to user error, other systems, etc. I ran additional tests and NVMe came in faster by a few seconds etc. I didn't total it all up though. hsb HW Spec in Spoiler --- i7-10700K Direct-To-Die/OC'ed to 5.1GHz, MSI Z490 MB, 32GB DDR4 3200MHz, EVGA 2080 Ti FTW3, NVMe+SSD, Win 10 x64 Pro, MFG, Warthog, TM MFDs, Komodo Huey set, Rverbe G1
Lao Fei Mao Posted December 31, 2018 Posted December 31, 2018 (edited) Since NVME SSD makes no obvious difference with SATA3 SSD on dcs. For those who want to kill the stutters, if your CPU or GPU is not outdated, then I suggest you to add more physical RAM, at least 32GB. The windows virtual RAM = physical RAM+pagefile on disk.And the pagefile MIS Afterburner monitored is actually the submitted parts of virtual RAM in windows task manager. So the bigger your physcial RAM is the less the pagefile used on the disk, on the condition you let the OS automatically manage the pagefile size. For example, I have 32GB RAM(31.9G in task manager), no more than 12GB used in game, it seems 19.9GB idled. No, it's not the truth, the 19.9GB are actually acting as the so called pagefile(MIS after burner monitored), sometime 19.9GB is not enough for DCS, then the pagefile on disk(7GB automatically allocated) would take part in. So if your RAM is less than 16GB, then the pagefile on disk would enlarge to 20+GB for some complicated scene.Great burden to disk. And the speed of disk couldn't match with the RAM, even the NVME disk. Edited December 31, 2018 by Lao Fei Mao
etherbattx Posted December 31, 2018 Posted December 31, 2018 NVMe SSD over SATA SSD ??? virtual memory and page file usage is not that simple. first thing to remember, monitoring tools only show what windows tells them. and windows does not always tell the truth. for example, can your tool tell the difference between a memory page only in ram, and one that is in ram and used from ram, but backed up in the page file and causing it to grow in size? second thing to remember, working set is different than allocated, different than committed, different than private. third thing to remember, data can only get into your ram via a hard page fault. so monitoring the number of hard page faults can easily be misleading. forth thing to remember. DCS is not the only process running on your machine using CPU, using ram, hitting your storage device or downloading megabytes through your network card. so when you are looking for the causes of stutter, be sure to look at the other 179 programs you have running at the same time. oh, and don’t forget the extra processes used by vr. they take cpu, ram and hard page faults too.
BitMaster Posted December 31, 2018 Posted December 31, 2018 very complicated, we could discuss for years... The answer is still very simple: SSD + 32GB Gigabyte Aorus X570S Master - Ryzen 5900X - Gskill 64GB 3200/CL14@3600/CL14 - Sapphire Nitro+ 7800XT - 4x Samsung 980Pro 1TB - 1x Samsung 870 Evo 1TB - 1x SanDisc 120GB SSD - Heatkiller IV - MoRa3-360LT@9x120mm Noctua F12 - Corsair AXi-1200 - TiR5-Pro - Warthog Hotas - Saitek Combat Pedals - Asus XG27ACG QHD 180Hz - Corsair K70 RGB Pro - Win11 Pro/Linux - Phanteks Evolv-X
FragBum Posted December 31, 2018 Posted December 31, 2018 virtual memory and page file usage is not that simple. first thing to remember, monitoring tools only show what windows tells them. and windows does not always tell the truth. for example, can your tool tell the difference between a memory page only in ram, and one that is in ram and used from ram, but backed up in the page file and causing it to grow in size? second thing to remember, working set is different than allocated, different than committed, different than private. third thing to remember, data can only get into your ram via a hard page fault. so monitoring the number of hard page faults can easily be misleading. forth thing to remember. DCS is not the only process running on your machine using CPU, using ram, hitting your storage device or downloading megabytes through your network card. so when you are looking for the causes of stutter, be sure to look at the other 179 programs you have running at the same time. oh, and don’t forget the extra processes used by vr. they take cpu, ram and hard page faults too. Agree however from my observations there isn't that much VM/SSD/network stuff happening during play when frame rates drop, with Rift I mainly see GPU hit the wall at 100% and or CPU hit the wall @ around 30% for me now running 2.5.4 OB. TBH I think the Win10/Rift/maybe even DCS* etc is now better at shearing out the multicore usage. :thumbup: *Perhaps by default from say Oculus app updates? Control is an illusion which usually shatters at the least expected moment. Gazelle Mini-gun version is endorphins with rotors. See above. Currently rolling with a Asus Z390 Prime, 9600K, 32GB RAM, SSD, 2080Ti and Windows 10Pro, Rift CV1. bu0836x and Scratch Built Pedals, Collective and Cyclic.
Recommended Posts