kksnowbear

I saw a listing for the same item on another site. "Open box" - but listing description says "used for less than 2 years, only selling because of upgrade."

Of course. That is, after all, the only sensible path

Right. Trust me, I'm very familiar with the 'scale of instances'... ...what I'm more referring to is why there would be such a delta - quite literally both ends of the spectrum, if you will. I just find it interesting that there would be such a marked difference in the experiences. And while I'm sure you are pleased with your own progress/result thus far, there are several - it would seem - that aren't as impressed. Although I can easily understand that there might be varying levels of positive (or negative) reports, I guess I am surprised to see that there *appear* as many of one as the other, on both sides of center.

You know, I don't do VR (likely never will)...and I haven't even as much as patched DCS to 2.9 yet, although I was as curious as everyone else what DLSS might bring...and I often read this (VR) subforum just out of technical curiosity. I have to say I am just amazed at the range of the reports here. Everything from totally amazing to completely unplayable, it would seem. Interesting. There must be some explanation for all this, though I'm damned if I know what that might be.

Ever since the power company increased their rates, I have come up with a different method...I may need more 'generators'.

I do genuinely appreciate your making the distinction I hope you can see that, in the context of the choice between new/used as stated, it seemed you favored the more expensive option as a means to avoid being dissatisfied. My perspective is that being satisfied doesn't necessarily require spending more, and spending more doesn't necessarily guarantee being satisfied

Fair enough - but let's not automatically equate buying used equipment with "always a loss". I sell a lot of used stuff, and everything is warranted. After 40+ years, the unsatisfied claims count is still 0. None. Nada. Zip. Zilch... If you buy from the right seller, it will fulfill the "desired cause" or you'll get 100% of your money back - you lose nothing. You do not automatically have to pay more to get value. In fact, I pride myself on exactly the opposite: Saving clients money while delivering comparable performance.

So ~1500 USD for the used machine vs ~2000 for the new one. (sorry, I convert so that I am more familiar with amounts/values) I'd say it's probably a better deal overall to go with the new unit - I am assuming it's within your budget. If you can afford the difference in cost, the new machine is (as LucShep said) a good deal more capable in terms of CPU, GPU, RAM and storage... ...And it's new, which I assume means full warranty and at least some support. BTW I am also assuming there's no room at all for negotiating either cost. Since it's sort of a 'close race', that could make a difference. If, for example, the used system's seller was willing to go down to 1300 (BRL 6500) then it would make the decision all the more difficult. But, at the prices you give, I'd go for the new one. HTH - best of luck to you.

I could easily be wrong, as this isn't exactly an area of expertise for me... ...but I was under the impression that the new DCS implementation was using DLSS 2.0, whereas frame generation is only supported in DLSS 3.0. Again, I could very well be mistaken.

Miners typically run undervoltages, yes...but I did want to specify a reason: I think the real/biggest reason they lower voltages, run custom VBIOS etc, isn't because of wear or heat (at least not directly). Mining can generate income, but how much depends on the electricity you use. Just like any business, your margin is affected by cost. So the idea for the miner is to do the greatest number of transactions while using the least power. And yes, indirectly this also means less heat (and less AC). Usually, this is why they lower voltages etc, for maximum efficiency as opposed to heat or wear. Just thought it was worth making the distinction. Of course, all things considered - and even though I deal with a *lot* of used hardware - I don't buy anything that I even remotely suspect has been mined on (with a very few and very specific exceptions which I won't go into here). This isn't as easy as it might sound, unless the source(s) are absolutely trustworthy. That's why it's worth it to deal with someone you can trust, even if it costs a little more. HTH

Broadly and generally, I'd agree with LucShep...however, that said, you've made things a little quirky by specifying the cost difference in percentage vs currency. For example: Used system costs 50000 BRL; new system therefore is 67,500 BRL: Both these are too expensive; look elsewhere (Yes, I realize this is absurd, I'm trying to make a point) Other extreme: Used system 1000 BRL, thus new = 1351. Buy the new one (this is also intentionally ridiculous to illustrate a point). Zeroing in a bit: Used system 6500BRL, new = 8785. Starts to get very tricky to see a "clear winner". That might be a good value (1750USD) for the new system, but what if it's simply beyond what you can spend? 1300USD is also a good price for the used machine (assuming good condition, whether there's any warranty, whether it's a seller you know and trust etc). TBH it would yield a more meaningful and accurate response if you specified cost (even if in BRL) than to just say one is x% more/less than the other. Once again, these are only examples I've concocted specifically to illustrate the point. It is often very difficult to judge things like this in other countries, because although the currency conversion is straightforward, the local economy is another matter. Often, specific things (like computers) can be grossly out of proportion in cost even compared to other things in the same country. A loaf of bread may be roughly comparable in cost to the USA (with currency conversion), but computers could be many times as much, and that difference must be considered. Is there some particular reason you use % difference as opposed to actual costs? It's your prerogative, of course - but that does make it harder to give a "good" answer.

LMAO too true!!! (But, of course, the *sad* part give it three months tops and some freakin' idiot will be selling a cable that will allow that GPU to work with a PSU that was never intended for it)...and then there will be thread(s) here asking about buying one of these cables... ...because, you know, it will magically make a GPU with a power-limited VBIOS perform better... <smh>

Oh, one other thing (a free professional tip, if you will)... I wouldn't go with the Corsair RAM if you intend to actually use the RGB feature. To be brief, it won't cooperate with other lighting in that build (and I actually own/have used very similar or exact same parts in my own personal machine, so this is first-hand experience). For whatever idiotic reason, the manufacturers do not implement RGB functionality per any standard or spec. They all have their own, proprietary method. In rare cases, it might work together, but the odds of that happening diminish as you go to different manufacturers for various components. They all want you to install and run their stupid 'helper app' just to use RGB lighting. Bullsh!t, but there it is. I know of one exception to this rule that I've had decent luck with, but it's not perfect, either. If you want RGB functionality that works, use the Asus motherboard's control (as big of a bloated PITA as it is) and get memory from a more 'generic' source (i.e. NOT Corsair). I've had great experiences with XPG, Team and V-Color memory - and TBH, the internal components on RAM modules are often very close if not identical to bigger 'name brands' like Corsair. It's all made by one of a dozen or so manufacturers (and Corsair is not a manufacturer). I believe the three big companies actually making memory chips today are Hynix, Micron, and Samsung (there are a half-dozen other smaller outfits, I think, but those are the big three). Point is that brands like Corsair, AData/XPG, Team etc, etc...they all build modules containing chips from one of a few places, so you're getting the same guts even if the package is different. (Incidentally, this is the kind of thing that people pay me for, and get, when I'm building a system: Knowing what works best with what - and they don't have to know or worry about it, because if I can't make it work, they don't pay for it - as it should be).

Yessir...and also let me add: There is a huge difference in buying a complete used system, specifically built, tested, and warranteed/supported as a system, compared to buying parts separately (regardless of where). Imagine this scenario: You buy used parts (even if you're OK with building it yourself and have done it a few times)... After you get the whole thing running, something starts misbehaving. Now, even though you've built a PC maybe once or twice, you're not a professional tech. You have neither the tools, training, experience or time/patience to deal with a problem like this (we're talking the kind of problem that can give even experienced techs a very hard time). So you've bought a motherboard from one guy, a GPU from another, and a PSU from a third source.... The guy who sold you the motherboard, insists the problem is the GPU. The guy who sold you the GPU, says the problem is obviously the PSU. The guy who sold you the PSU says it's likely either the motherboard or GPU (or both...) (Of course, they all say it could also be your memory...or a driver...or, or, or...) (And BTW you also don't have two rooms full of parts to swap in and out to determine what the problem is.) What you have now for your money is a big finger-pointing contest. By comparison: When I've sold a system, in the rare event there's a problem, they bring the whole shebang back...if I can't fix it and show proof it's actually fixed, they can get 100% of their money back. They don't have to do anything else. And yes, I have also gone to customers' location or paid shipping BOTH WAYS when the rare problem did happen. I say 'rare' here because by the time these systems have left my shop they've been subject to solid days of testing, sometimes running 24 hours+ to make sure they're stable. And, oh yeah: I *do* have two rooms full of parts to test with Simply put, I stand behind what I sold, period. Not many used parts sellers are going to offer that level of support, at all. This is not intended to get off course too far, but I think these are significant points which change the equation considerably, and are often overlooked.

You know, as someone who actually buys/sells used and new components and systems, I have to jump in here... First, let me say I admire you (LucShep) and have found your advice to be generally helpful and accurate... ...but I also have to ask, if we're talking used parts here, is there any guarantee of support, warranty, and history on these parts? Is it a trusted seller who can offer references from known sources who can be verified? Do they offer trade-in for used parts on the initial sale, as well as future upgrades? Do they provide free labor for life on any upgrades or repairs? An absolute iron-clad, 100% satisfaction or money back guarantee? Not to complicate matters, but these things all have definite value and cannot be ignored. If you shop/buy based on cost alone, you may well find yourself regretting that purchase sooner or later. The second-hand market can be a useful resource for saving money when buying a PC. Unfortunately, the reality is that there are also opportunities for problems and dishonest people in the world. I tell people who are considering stuff I offer that they might get a better price - but I honestly feel that, if we're looking at actual value, I offer a better deal. Just a different perspective that is worth considering. To the OP: Best of luck As I've said above, LucShep knows what he's talking about (his advice has helped me in the past, and I've been doing this professionally for many decades).

Beat me to it But yes, this is pretty much the way it is. It's the cache that gives these X3D chips their advantage, and only half of the 'cores' in a 7950/50X3D have access to the cache. And for whatever it's worth, although the MT patch has helped, DCS is still a long way from what I'd call "good" core utilization...so having a lot of extra cores doesn't really make a lot of difference. The majority of a CPU's ability is sitting there bored, based on what I've seen first hand.

FWIW I think you have the right idea about AMD at the current time. However, I am somewhat surprised to hear you're concerned about the 5800X3D beating a 7800X3D. I'd be interested to hear more about that. Also, just to be clear: The 5800X3D isn't AM5 - so that would address any concerns about future 'upgradability'. You are correct in saying the AM5 has a future in front of it - but AM4 is already "maxed out" with a 5800X3D. (I think you probably know this, but just to be sure). So, I wouldn't be considering an AM4 platform if the intent is to do a total replacement. Even if we assume 5800X3D is comparable to a 7800X3D, the latter has an upgrade path where the former does not. HTH

Below is a video showing 3840x1600 at High settings (not 'max') and he gets 50's-70 with a 2080Ti. There are times it is down below 50, depending on the map and down around the deck (which is typical). 3840x1600 is only about 75% the pixel count of 4k, so... seems fairly likely you'd get less performance with a resolution that's 35% higher... As I said it depends on a lot of things. I didn't specifically mention altitude, but that's a big factor as well. Frame rate can drop 30% just by going down at 'tree top' especially when the area is populated with lots of buildings. Another/different example: I have a friend, we both got 3090s at about the same time (I since went to a 4090). He has run DCS tests at very high settings at 1440p and frame rate dropping to ~70 at times on both Normandy and Syria maps. Not 4k obviously, but if his settings drop FPS to 70 at only 1440p with a 3090, it seems really unlikely he'd get even 60FPS at 4k. Somehow I doubt a 2080Ti is going to do better than a 3090.

Without going into a ton of detail: Bottom line is 'the more the better' and that's just the way it is. That being said, anything over ~11G (as with 1080Ti, 2080Ti...) is *adequate* in terms of VRAM. However, it depends. The settings you use will have an impact on VRAM used, and on performance as well. The question is not strictly "How much VRAM do I need at 4k" As a comparative example: A 3090 is *far* better than a 2080Ti in terms of performance. The 2080Ti has 11G VRAM and will run 4k - but we must ask "At what frame rate, and with what settings?" I think it should be obvious the 3090 will run 4k at far higher frame rates and higher settings. So, while it is definitely a good idea to consider minimum VRAM, you should also seriously consider what kind of performance you're expecting/will get. Following are some estimates: There is a 4060Ti 16G model, but I believe it's performance at 4k would be fairly poor (<20FPS??) A 4070 has 12G, and better performance than the 4060Ti 12G - but still, frame rates at 4k won't be too impressive (~25 FPS??) (The reason I have ??s after the estimated FPS above is because - again - it depends on settings, the map, MP or SP, etc etc etc....please keep in mind these are broad estimates. Also, read below about DLSS.) Realistically, to even approach 60FPS at 4k, you're looking at cards that inherently have more than 12G: 3090 (24G) 3090Ti (24G) 4080 (16G) 4090 (24G) ...and *maybe* a 4070Ti at 12G (though the frame rate would probably be ~30) (There is also a 12G 3080Ti, though it's 4k frame rate is likely to be in the 20's.) Kindly note the estimates above are for higher settings. If you are OK with lower settings, you could probably improve on these estimates a bit. Something else important to consider: If you intend to use an Nvidia card, then I believe it is accurate to say that any Nvidia card with 12G VRAM or more will also support DLSS. Wihtout going into a ton of detail here, a forthcoming DCS patch is supposed to add DLSS support. So, *if* you can live with the trade-offs in terms of quality, DLSS can increase frame rates substantially. It depends, of course, on how well it's implemented, as well as a lot of other factors...but it is remarkable technology that can improve frame rates quite a lot (at the expense of some quality). The potential downside is that you might not care for the 'quality cost' of how DLSS performs in DCS, and that's a factor that, unfortunately, cannot be predicted. But it is something to be aware of, and it's significant. I believe people start off thinking they're going to game at 4k, up until they realize what kind of hardware is needed to deliver *decent* performance (and the cost thereof). In fact, this is the second thread I recall with this very subject, fairly recently. Nothing wrong with it, mind you, just I think people truly don't realize why 4k gaming is still really *very* ambitious (and very expensive).

I thought someone had asked to leave the PSU subject alone. Don't preach at me, you have no idea what you're talking about. I build computers professionally, and have for over 40 years. The people who pay me to do this vote with their dollars about whether I know what I'm doing. And plenty of people besides me care about things like efficiency - although it's nice of you to decide what everyone cares about. Just because you don't get it, doesn't mean no one else does. I already (accurately) described the one-time cost difference vs. continuing savings. And my original point about cost of the PSU had nothing to do with efficiency of itself, regardless. Apparently you're incapable of understanding that. Sorry for your lack of understanding, but it doesn't change the facts. Now, since someone was nice enough to ask, here's an idea: Leave it be.

Not sure where you saw me say anything about saving the difference in cost of a more expensive PSU. (Oh, wait...that would be because I didn't say that.) However since you brought it up, you will absolutely save money by running a PSU more efficiently. This applies regardless of how much or how little your power costs (and regardless of how much or little the actual amount you save is, either percentage or cost). Using less costs less. And, unlike saving money *once* by buying a cheaper power supply, the efficiency savings will continue indefinitely (for the life of the power supply, and these days, even the warranty is 10 years for high-quality units). The more you run the PC, the more you save. One-time cost difference in the purchase of a cheaper power supply cannot do that. And that's a fact.

This is true, provided of course the 4-cable power "tail" is configured correctly. (There have been cables provided - with 1000W PSUs - that still limit the GPU to 450W). There are also three-cable variants that signal the GPU that the PSU can provide 600W. I know this because I have one As explained above, it is wise to leave room as 'overhead' in sizing a power supply - how much will depend on a lot of factors including desired efficiency and overall system configuration. Certainly nothing excessive about a 1000W unit (or even 1200). As noted, the manufacturer's recommendation for some 4090s is 1000W, and for good reason. Actual load on a PC switched-mode PSU cannot be measured effectively by a UPS or a plug-in 'wall wart' type watt meter. They are useful, I'm not saying they're not - I have two of the watt meters and at least a dozen UPS units. But they have limitations as applies to measuring power, and those limitations are often overlooked or misunderstood (as above in this thread).

When I asked where and by what means the load was being measured, I meant at what physical point, and using what type device. You are not measuring load physically at the GPU itself, and you're using a UPS display (or software) to read the load at the line side of your PSU. This is outside the chassis, and in fact isn't even measuring the actual load on the PSU itself. Again, it is not appropriate to size a PSU on anything other than full system load, definitely not while running a single game (even if that's the only game you ever play on the PC). You cannot (to repeat, cannot) determine actual PSU load by reading UPS software or meters/panel displays - such displays/software are not designed to sample or 'catch' max transients, as I already explained. It's just not possible using that measurement. Among other things, your UPS is not on the load side of a switch-mode power supply that is always running at <100% efficiency, so it's not even capable of telling you the exact load on your PSU - and certainly not in real-time. (BTW, if the "high accuracy" of measurement you're referring to is by using the free PowerPanel Personal software that your UPS supports, then I'd say it's not likely anywhere nearly as accurate nor capable as you think it is). If you have a 4090 that's power limited to 450W, great. But that doesn't mean that every 4090 will be limited the same way. As a specific example, my Asus TUF unit is set up at 600W (and that's both in VBIOS and a proper 12VHPWR cable). Again, there have been documented tests showing 4090 power excursions to 700W - which your UPS is unable to catch and report. The power reported by these types of measurement devices is absolutely limited by how fast they can sample, and the transient excursions can happen and be gone faster than the sample rate of those devices. Not only that, but transients like this are absorbed/obscured by the electronics inside the PSU; anyone who's ever waited on DC LEDs inside a PC to extinguish after switching off a PSU has witnessed this. Such GPU power excursions will never be felt at the line side of a PSU, much less the measuring circuit in a UPS that powers the PC - but they will be felt at the 12v (low voltage/load side) of the PSU itself, which is why the PSU must be able to handle them. The reason your system might not experience this type of GPU load is precisely because your GPU is power limited to 450W. Typically, that's done because the manufacturer knows they used weaker VRMs in the hardware, and they limit the power to what those weaker components can handle. Not all 4090s are that way. This is also the exact same reason you cannot base any/every 4090 scenario on your experience: You don't know what kind of PSU, cables, or the exact model of 4090 someone else might have, and you cannot say that your setup is representative of all others. As I describe above, my own 4090 doesn't have the same power limit yours does - and neither do many others. That is as factual and "real world" as it gets. Also, 1200W is by no means "overkill" for a 4090-based system. It allows for the maximum total system load x2, which means the PSU itself will run at max efficiency, per the 80Plus spec. Factually, if you run a PSU at a higher load than 50%, it's efficiency decreases. If my device has actually measured 658W max, and if I were using an 850W PSU then I'm certainly not operating at 50% load and thus definitely not max efficiency. That means more heat, higher cost, etc. These may be small factors, but they are absolutely meaningful nonetheless. This isn't to say a 1200W PSU is required, but it's certainly not overkill if you're trying to achieve max efficiency - and that's a fact. You're also dismissing that all three of the biggest GPU manufacturers in the US have recommended 1000W PSUs for some of their 4090 models. So if someone happens to get one of those models of 4090, and follows your intimation that 850W is enough, they're actually disregarding the manufacturer's recommendations - and thus could be denied warranty service if there's ever a problem. In fact, if the manufacturer recommends a 1000W PSU and someone has an 850W, then the manufacturer's support team would absolutely be within their rights to refuse any support. As much as you dismiss the specs, they exist for a reason. It's foolish to ignore that reason when designing power systems, and professionals know better than doing that. Never mind that it makes zero sense to spend thousands on a PC with a 4090, but cheap out on the PSU to save 2% of the cost.

Yup lol couldn't agree more - what a PITA. Especially if you've gone to great lengths to clean it all up and tie everything down. I just went through that not too long ago, having changed cables from the hideous 4090 'tail' to a custom 600W 12VHPWR cable. To be clear, though, there are models of the 4090 for which an 850W PSU is adequate - they will likely be power limited in VBIOS, and this is a 'hard' limitation in that it is usually based on the GPU's voltage regulators (VRMs), which of course cannot be readily changed. Some people have realized the limits in VBIOS can be 'cheated' by loading different VBIOS with higher limits, but it should be obvious why that can be dangerous. So if you picked your 4090 carefully...no need to change PSUs for adequate capacity (though both efficiency and overhead are different matters which should be considered). Best of luck!

You don't specify where 340w is being measured, or by what means...and "full load in VR" sounds like you mean in-game (i.e., while running DCS) which is not the same thing as "full load" for an entire system, nor necessarily even full load for only the GPU. Regardless, it's not appropriate to use a figure like average power, or one game, to properly size a PSU. Among other things, regardless of efficiency, the TDP of a 4090 is fully 100w higher than a 3090, and that's per Nvidia specs. I've measured 658W total system load, albeit with a comparatively unsophisticated measuring device. Again, these devices cannot accurately sample the type of excursions that are documented to occur with a 4090. But if my "basic" device shows 658W, it is entirely likely the max transients are higher. And if we're talking about efficiency, you need a PSU with roughly twice the capacity of the typical max load to run at its highest efficiency. In my case, that works out to over 1200w. EDIT: Sorry, don't mean to hijack the thread and this PSU discussion is somewhat off the thread topic. That said, the question of power supply was raised by someone else, and it does seem prudent to correct inaccurate information.