Thinder

RIP.

Just ordered specifically for DCS: SAMSUNG 990 PRO PCIe 4.0 NVMe M.2 SSD Sequential Read Up to 7,450 MB/s Sequential Write Up to 6,900 MB/s ASUS TUF GAMING X570 has 2 x M.2 x4 PCIe Gen 4 & SATA mode, the new SSD will be running on Slot1. The last part of my system optimization, increasing the SSD bandwidth, this new SSD should contribute to obtain better performances, especially since the SSD I have now (970 EVO PLUS 500GB) while not being slow, is not on the same league: SAMSUNG 970 EVO PLUS 500GB Sequential Read Up to 3,500 MB/s 112.85% slower. Sequential Write Up to 3,200 MB/s 115.62% slower.

AMD released a video to explain the reasoning behind the X 3D series, more specifically the 7 5800X 3D and made it clear it was lower latency. +1, fast SSD read/write play a role, 3,500/3,300 MB/s for a 970 EVO, but you need the RAM to do its job as fast as possible This was one reason for my choice to stay with DDR4 as a proven and mature technical solution, I didn't think that 32GB of DDR4 Cl14 would be an issue, and after all the testing I've done, I still think it was the right choice for me, not saying other players are wrong to go for DDR5, but I am more comfortable with DDR4 for the time being. I had loads of issues with my new rig, a faulty PCI_E1 slot, Windows update messing up AMD drivers and preventing the GPU to boost, etc but now it's sorted and I'm getting there although I think there still is room for improvements, in this test I didn't set the RAM frequency properly it runs now at 3600MHz but it sure isn't as bottleneck. I'm waiting for further DCS MT updates to get back to it fully, for the moment I am looking at the quality of the graphic with the GPU setting I have and with Elite Dangerous, in some environments (inside a station) visuals are stunning, I run the game with everything maxed-up and I believe that maybe when MT will be really efficient it will be possible to do so in DCS too. For the moment out of the 8 cores, only one runs above 50% and this can be improved. PS I'm making inquieries to ASUS to know if it is possible to add a Samsung 990 Pro SSD, if it is, I will install DCS on it on M2_1, it should contribute to make this rig faster (7,450 MBps / 6,900 MBps vs 3,500 MB/s3,200 MB/s ). Originally I had installed it on the SSD I already have but during a fresh Windows 11 Install, didn't, the second drive (SATA) wasn't fitted.

To answer more precisely to your question, since you run an AMD and they have been designed for lower latency, I'd go for the RAM kit offering the combination of the lowest latency and highest bandwidth, this way you will take advantage of the 7800X3D cache. I don't know DDR5 RAM that well but if AMD increased their CPU bandwidth it is no reason to assume they increased the limits of the controller when it comes to the number of ranks it can handle, Corsair and GSkill pair them by two sticks, which seems to indicate that 4 ranks are the best combination for DDR5 systems. Better keep to their limitations and those of your CPU. If you're after performances, the G.Skill Trident Z5 Neo EXPO looks very good but like anything out of AMD recommended maximum frequency, you need to check that your CPU and motherboard supports it, just like a 3600MHz kit for DDR4 4 maxed-out at 3200NHz. QVL compatibility is for single memory kit only.

If you play at lower resolutions with a DDR4 Cl16 RAM you might find a slight improvement but with Cl14, it's not gonna work at 4K, in fact you lose 24% in performance simply because your CPU controller is going to throttle down, they are not designed to support 4 ranks, only the 8GB sticks does have one rank per stick. So if you play at 4K, the gain over 64GB is obvious, capacity is one thing but a DDR4 32GB Cl14 kit is gonna recycle the same quantity of data (IF you have similar latency and frequency) PLUS the 32GB difference between the configurations nearly 3 times per mn on bandwidth and CPU throttling down alone, I'm curious to know how you'll manage to get 64 GB of DDR4 give you better performances in any game once you've set up a Cl14 4 x 1 stick kit. As I said I'm not into "several reports" especially when there is no precise specs and settings provided for the comparison, I tested those kits back to back in 3DMark Pro at 4K 2 X MSAA and play them in DCS too. Here is the complete list: When I test I play at tree-top level in environments that demands a lot of redraw work from the GPU, at high settings as well, if my RAM was an issue, my results would be a lot lower, it is not the case, the only limitation I get is GPU thermal limits, not RAM. Now I wont pretend that this will work with a DDR5 system but no one is gonna make me believe that 64GB of RAM works better than 32GB Cl14 DDR4, I know it doesn't and I know why. Here is what MSI Tech support replied to a request for information I made before I purchased my first GSkill kit, they replied a week after.

Endly someone bringing up the issue... People are desperate with RAM/VRAM usages solution but how many take system/OS optimization seriously? When you see the amount of processes your system is running which wouldn't be if it was actually set up for gaming, you understand, but I guess it's easier to buy more RAM or upgrade for a GPU with more VRAM, or at least that is what some are lead to believe. I've got rid of (nearly) everything I don't need in Windows and still have <> 55 apps running, some actually reinstall themselves, other are so hard to disable it's an issue on its own... Get MORE RAM? Nope, get the RAM your CPU will work better with and 32GB are enough, only it cost (new) the premium, at least for a DDR4 system. As for apps, how many of those do my system really need to run?

Believe it or not, I'm not a technerd as such, I learned the hard way, first having to build our PCs at the 3D Dpt where I was working as a CGI technician, as my boss was saying at the time: "Windows is designed to screw you" and "this job is about 90% problem solving and 10% creativity", which should be the opposite since we're not supposed to be those companies guinea pigs and were paying expensive software to do a creative job. Oh and I nearly forgot, I'm in the age group where you no longer learn easily but forget things faster than you learned them, which makes it a lot more challenging. The whole point is "they" don't work for us, they compete and even work against each other, best example being Microsoft and their sick level of compulsory intrusiveness: I wonder why nobody had a look at their procedure to launch a law suite against them, forcing updates is already pushing it, but sabotaging GPU drivers or literally replacing them without your consent is another level, Bill Gates sure left a strong legacy behind him, the triggers embedded in their recent updates literally hack people's PCs.

Just saying that nothing comes easy. Steam bugs sorted by reinstalling Radeon Sofware... Talk about standards, it's like every single programmer do their best to give you a hard time... Well, I reiterate, it's little use if you guys don't get your head into it, ask questions to support teams and read about it because the next time it happen (or something else), you could sort it out yourself then later on be in a position where you, in turn can assist by passing the little experience you gained... That's how it's supposed to be, not always the case since CPU Controllers RAM frequency limits are 3200MHz for both Intel and AMD DDR4 systems, a 3600 MHz might need some tweaking in BIOS. Btw, in the list of purgatorial karmatic experiences coming with my new gear, I forgot to mention MSI BIOS being unstable, making it nearly impossible to boot my system after swapping from one PCI_E slot to the other and back, "default" setting is default and didn't work once the GPU back to CPU_E1 slot, I had to swap again and set them both to Gen 3 or whichever worked. This one was the worst of all, I had complete boot failure down to the blue screen of hell forcing a clean install of Windows 10 Pro, which forced me to go to Windows 11Pro which itself came with more problems with Automatic Update minefield and AMD drivers settings take over, programmers would have wanted to screw my system, they wouldn't have done otherwise...

Guys, if you think I'm not stressed at time having to cope with stuff that don't work as intended because the standards are different, the new gear is defectuous or Microsoft decide to play havoc with AMD Drivers and the GPU settings.... With this new rig I had to deal with: A PCI_E1 Slot that didn't work out of the box (MSI Mobo), DCS first VR update that completely destroyed the best settings I had in VR with the Pico (See topics, it was epic), AMD sending me messages asking to check on drivers because Microsoft update was replacing them with their own, the very same Microsoft Update causing the GPU to lose <> 10% in performances because they included MPO settings that prevents the GPU to boost, following DCS instructions for new settings to get my VR to work again after yet another update, test 4 different AMD drivers, setting up DCS Beta MT, etc etc. All of that while being told that what I experienced before it all turned sour never happened despite uploading 4 different videos while testing my new GPU, being trolled by Intel users telling me I was "all wind" because I pointed out that their Youtube gurus were making loads of noise about a few mishandled AMD GPU, being trolled by some members of the same "community" about how bad and prone to explode our 7900 were, being depressed by everyone crying at AMD in their forum etc. So believe me when I tell you that I needed more faith than the Pope to sort it all out and figure out what was wrong because in some case (Windows update) no one knew what was going on, today I recovered my GPU performances, figured out my new motherboard settings, got my system to run fast and cool but I'm still struggling to get Steam Virtual Desktop to work stable because their update is also screwed up. That's how it goes, with new technologies comes new standards, drivers, apps and so on, so if you stop learning, your PC dies or crawl at best, anyway I'm always trying to help people with stuff I know but I did ask GSkill, AMD and MSI support about the RAM kits, what worked and what doesn't, even before swapping from Crucial Cl16. I'm always happy to help but my tip; ask technical supports and do your own home work, the more you understand how your rig works, the better.

I don't think it will suck but why don't you try to conduct a back to back test with a free version of 3DMark? Test the old kit before swapping and do the same test with the new one, compare the results, if possible at 4K 2 X MSAA because ultimately, that's what DCS demands if you want to play at higher res, there I'm talking VR or more demanding scenarios. One benchmark for all your hardware Information about 3DMark Download 3DMark Basic edition for Windows. Link to 3DMark Time Spy download. From the moment I decided to upgrade my PC I started to use 3DMark Pro at those resolutions because I wanted to play DCS in VR and I have tested 2 CPUs, 4 RAM kits, 2 GPUs, BIOS settings and even cooling upgrades (Thermal limits), so I know exactly what does what and which effect each upgrade has on my rig's performances, it is worth the time and effort. And btw, if you're serious about your gaming you'll optimize your PC the way I described first, memory usage is not only a question of RAM, the OS apps also play a large role in it especially if not taken care of.

Just an example of how cheap a better RAM kit can be. GSkill Trident Z RGB - 32gb (4 x 8 gb) DDR4 3200mhz CL14 £75.00 And you also can sell yours too.

OK, that's better. Both are non B.Die kits, Cl16. I'm not familiar with the Corsair Kits but chances are the 8GB sticks are one rank vs two for the 16GB, they both have quite a high latency of 77/77.5ns, the 32GB kit is newer so it is possible that they improved them to increase bandwidth compared to previous kits, its scores 50.5 GB/s vs 43.8 GB/s in MC Read Peak Score which is not bad for a Cl16 kit. You might experience a slight improvement but not at 4K or/and under heavy load because of the limits of your CPU controller, when I was talking about sink and hole it's easy to see, a Trident Z (Cl14) will run at 53 GB/s, meaning every 21 second it will recycle the equivalent of your 64GB RAM kit more than your Cl16 kit, and it will do it at 4K, when your combo will throttle down, that's what lower latency do for you. Cl14 kit eliminates the bottleneck you'll experience at 4K with a Cl16 and overall greatly limits the need for higher capacity, it also work very well with AMD processors because they are designed for lower latency, especially the X 3D. So all in all it is an upgrade, OK, but not what I would have gone for even so the cost is much lower than paying for the premium of a new Cl14 kit, from my humble PoV simply increasing the capacity of a Cl16 kit is not doing enough for a system intended to use playing DCS, even less in VR. Consider if you have the funds, looking for a second hand Cl14 kit, 4 x 1 sticks are best, then you will notice a difference under load as I did after upgrading from a Crucial Cl16 kit.

It won't explode but it might not boot up, that's a possibility, some combination of O.S release (early Windows 10 Pro can't handle 64GB despite Microsoft claims of the contrary) doesn't work that well if you fit the RAM just after a fresh install. We still have no idea of the latency of your previous and future RAM kit, it is important to have a better picture, when you have time, post your all specs in your signature, it makes things easier for people to assist you. But just try to figure this: Filling up a sink with a small hole, you may double the size of the sink, but the hole size remains the same, that's what doubling RAM capacity can lead to. On the other hand, increasing the hole size by 30%+ will help emptying it, which is the same principle than RAM processing data to the Processor, for the same frequency and capacity, 30% increase is a huge improvement. Read Intel's article on How to Overclock RAM

+100. I'm very happy with my ASUS mobo, the MSI X570 ACE MAX it replaces was a complete catastrophe, not only its PCIE_E 1 was faulty, it is also a mediocre design with bad accessibility (impossible to have access to the battery with an SSD fitted for example). On the other hand I had similar very good experience with thew MSI B450 GAMING PLUS, so by experience I can say it is not a question of brand but which model within a brand, some are good, some are bad, my ASUS TUF Gaming X570-PLUS runs perfect, I only have to get use to the different BIOS interface to make sure I get the settings I want. I couldn't agree more, it is particularly true in the case of the X3D CPUs, O.C them it not how you optimize their performances, in the case of the 5800X 3D it is by proper RAM-CPU bounding, a lot more efficient than throwing more capacity at them, the Controllers set the tempo not the forum talks.

It all depends on the latency. There is no 32GB kits with 1 single rank, only the 8GB has them and this solution is only efficient at Cl14, so at least with 2 X sticks you're on the safe, but if you think about swapping a 4 X 1 Cl14 stick for those, you won't see any improvement at 4K, more likely the opposite. Gskill 64GB 3200/CL14 is slower than 32GB at 4K, the CPU Controllers won't run the channels at full speed, as simple as that, as tested, 24% difference between the 2 X 1 and 4 X 1, where is the gain? The whole point here is that some people seems to think that larger capacity make your PC faster, in reality, you can have 64GB slower than 32 simply because your CPU won't run at full throttle under load or because the latency is too low compared to a Cl14 kit. Instead of insisting on the opposite, they should splash the money and put it where their writing is, I've done it, tested them and I know what the score really is, you don't need 64GB with a bottleneck, you need RAM that process data faster. In the case of a 4 X 1 stick Cl14 kit, latency recycles your RAM capacity way faster than Cl16, that's 30.88% minimum and I guaranty you that a 63GB kit won't be 30.88% faster than a 32GB on capacity only, not even going from 3200 to 3600MHz. I noticed you don't specify the latency of your RAM, and I doubt that you will spend over 500 quids on a Cl14 kit of this capacity, from where I'm standing it might just turn out to be a complete waste.

Just a remark in passing. Unless you fly at tree top level, you are not pushing your system, the best test in my case is flying the Mirage 2000 in the Caucasus map in the valley surrounded by trees, that's when your system has to pull the gloves off to get you the performances at 4K, not above the sea or at 500Ft, it doesn't have to redraw all the environment details the same way.

Microsoft Automatic Updates have a lot to do with the XTX loss of performances, AMD also messaged me regarding Microsoft replacing their native drivers with their own. After an update, I lost <> 10% score in 3DMark Pro scores; Firestrike 4K 2 X MSAA, the date corresponded with the DCS "VR" update so I thought all the troubles I had running my gear in VR after the first tests came from it, but it was only partly true, same for AMD drivers, I tested 4 of them including the last one (Release Date 4/27/2023), they are not the source of the problem. My system runs DCS MT in VR perfectly today but not before I bumped into some article explaining how Intel CPU users were falling victim of settings embedded in one of Microsoft Updates, the culprit is MPO, ansd there is a fix. mpo_disable.reg When I tried to OC my GPU it crashed with mention "Driver Time Out", so I investigated and find this solution used by an AMD user, it seems to work at many levels but I am sure it is not the only thing in Windows 11 holding GPU performances. Here are my actual settings: The card runs smooth, no crashes, no flickering, image quality and clarity are good, performance is <> a few % of what I had before losing to whatever it was, but this app seems to be working, note that I have a very good case cooling (5 case fans + 1 CPU fan) and the temperature of the GPU stays within reasonable limits (max 81°C Junction under heavy load). I didn't conduct further testings after taking those screenshots, with high settings in DCS MT, it managed to run smoothly and cool: 99% GPU usage at 2823MHz vs "Up to 2450 MHz" stock, so definitively, there is a lot more potential in this GPU that what one can get with those hurdles still holding it, as I figured during my first tests.

I tested Cl14 GSkill kits at 3200MHZ and 3600MHz, all but one 32GB capacity. Configurations were 4 X 1 or 2 X 1 sticks, even if you use BDie, 64GB might actually lead to a loss of performances: In my case <> 24% at 4K MSAA X 2 in Fire Strike 3DMark Pro test, in a 4 X 1 16GB configuration, the reason is that only the 8GB sticks are one rank and the maximum your average high street CPU Controller can handle is 4, we're talking Zen 3 and the Intel equivalent and DDR4 RAM. So basically, the 2 X 16GB kit with the same sticks from the 64GB kit is way faster than 4 X 16GB, the fastest configuration being the 4 X 1 8 GB thanks to interleaving (4 sticks) all at the same stock frequency of 3600MHz. From my Pov, you'd be much better off optimizing your PC RAM usage, disabling and/or uninstalling every app you don't need because they are chewing up your RAM capacity only running in the background, I've done that but I still have about 52 apps doing just that, you also can set your Start Up Items options to minimize the number of apps kicking in when you start. Setting up your paging files to a different partition and giving them the priority for Windows apps is also helping the process, in short, there is no miracle when it comes to RAM, the Controllers are setting the pace (Bus bandwidth) taking into account the load they come under and in gaming, better look at 4K to make sure you know when your CPU is struggling. The RAM will let your CPU run in an optimum way or not, Cl16 RAM or higher for a DDR4 kit will already insure the presence of a CPU-RAM bottleneck, the B.Die Cl14 kits will avoid that and allow you to run at optimum frequency thanks to a much wider range of timing, be it 3200 or 3600MHz, what matters here is latency. For more on the subject have a look at what Intel Techies were doing with basically the same RAM kits than mine (now I use the NEO instead of the RGB but they are very similar): How to Overclock RAM You don't need to O.C a Cl14 kit, when I used the3200MHz I never O.Ced it, but they are designed for this so it is possible to run a 3200MHZ Cl14 kit at 3600MHz without problems, I chose the stock 3600MHz and that's as fast as my CPU Controller would actually run RAM. Now, under load in DCS high settings during tests my RAM usage was 17729MB vs a total capacity of 37798MB split between 32GB of RAM and 5.120MB of Virtual Memory (paging files), (corrected from in idle to under heavy load) my system uses about the same percentage of RAM than yours, 23.833MB but if you don't use Cl14 it will do it faster. In short, I don't have a RAM issue right now, better, I don't have a RAM-CPU bottleneck and the low latency make sure the data is recycled fast enough to avoid a bandwidth issue, because at the end of the day, it is the RAM speed and latency which matters when it comes to processing data. In Intel test, they managed to increase their bandwidth So with the same capacity, the bandwidth raised from 32 to 53 GB/s (21 GB/s difference), I'm sure there are Cl 16 RAM kits that would run your system at 31 GB/s when the CPU controller throttle down under load, even more so with 8 ranks, this shows the reason why lower latency is so important for DDR4 systems. According to Crucial, their DDR4 3200MHz have a Clock Cycle Time(ns) of 0.62.