Używasz przestarzałej przeglądarki. Może ona nieprawidłowo wyświetlać tę lub inne strony. Powinieneś zaktualizować przeglądarkę lub użyć alternatywnej przeglądarki.
GPUs & CPUs & Enthusiast hardware: Questions, Discussion and fanboy slap-fights - Nvidia & AMD & Intel - Separe but Equal. Intel rides in the back of the bus.
Is it worth selling my Xbox Series S if I'm only going to have enough money (from the sale & savings) to deck out an OptiPlex and maybe get performance just barely higher than that of an Xbox One?
I don't even want to move to PC for mods, I'm just sick and tired of not being able to shitpost in the middle of playing games. I'd probably have about $400 USD to work with.
I honestly don't mind going far back as Haswell. 1080p/60 is all I need. Would like an OptiPlex in the small form factor, or basically the smallest one they sell that supports a low-profile card + has an optical drive if possible. One thing I really like about my Series S is that I can fit it in a backpack and take it with me to a hotel if I need to.
Keep in mind that the $400 I set as a limit isn't for eternity, I can upgrade stuff later if I need to. But right now base model XB1 tier performance is what I want.
Honestly just get an AMD business laptop with a zen4 CPU and game on the iGPU. The integrated Radeon 780M in my laptop’s 7840HS gives decent framerates at the monitor’s native 2k, it can definitely handle 1080p@60. Battery life isn’t great, but it sounds like you’re mostly going to use the thing plugged into the wall anyway, and you can’t get much more portable than a laptop.
Is it worth selling my Xbox Series S if I'm only going to have enough money (from the sale & savings) to deck out an OptiPlex and maybe get performance just barely higher than that of an Xbox One?
Honestly just get an AMD business laptop with a zen4 CPU and game on the iGPU. The integrated Radeon 780M in my laptop’s 7840HS gives decent framerates at the monitor’s native 2k, it can definitely handle 1080p@60. Battery life isn’t great, but it sounds like you’re mostly going to use the thing plugged into the wall anyway, and you can’t get much more portable than a laptop
I already have a laptop, a fairly weak one but it's good enough for going to college classes with. The thing I hate about laptops is that they have so many potential points of failure (battery, screen, RAM that's soldered, the potential to drop/otherwise damage it, yada yada). I do know about those Ryzen mini PCs though, maybe I'll look into those.
The most demanding game I play is Farming Simulator 25, which requires a 1050 TI and 8GB of RAM. Those are the minimum specs though, and unlike my Series S (GPU on par with a 1650 TI) I want to play the game at 60 FPS up from 30. Other than that title everything I play could easily run on an Xbox One (Cities: Skylines, KSP, Minecraft, Silent Hunter 3, Gmod).
Huh, didn't know that. I like that the one you linked doesn't have RGB, in my mind only the Founder's Edition cards were the ones without it. Thanks
I was keen on AMD since I'm not gonna use it for AI but tbf I haven't researched DLSS at all, my friends say it's pretty impressive.
Anyone who turns on DLSS is using the AI capabilities of their NVIDIA card. I have both AMD and NVIDIA GPUs around, and my experience with DLSS is that it effectively bumps the card up a tier. Like on paper, a Radeon 6600 has more fill rate than a 3060, but in practice, since the 3060 has tensor cores, it uses AI upscaling to get image quality on par with a 6700.
Example, in order. The first image is rendered in 1080p, and the next are rendered in 540p and upscaled using different methods: DLSS, XeSS, FSR2, and finally, linear interpolation.
As of the new transformer model, it is very difficult to tell the difference between DLSS and native res, other than the fact you'll run the game significantly faster (or run at the same fps with higher settings, take your pick).
EDIT: For completeness, I added in XeSS, also upscaled from 540p to 1080p. XeSS is an interesting animal because, like DLSS, it also AI-based, but it runs in a low-precision version of its algorithm on non-Intel cards. And yet, you can see the results are almost as good as DLSS. It turns out that one man's inferencing is about as good as another. Neural networks are actually a very old, solved technology. The big innovation lately has been how to use them to do useful things at useful speeds, and once one person figures out how to use them for one thing, everyone else copies their homework.
AI is really just a fancy curve fit. If you've ever done a regression fit with (x,y) data pairs and used that curve to estimate new values, you're not that conceptually far off from what AI does. Imagine your X data points are a bunch of images rendered at 540p, and your Y points are those same images rendered at 1080p, and we've got a very fancy multidimensional curve fitted to those pairs so that you can throw in any 540p image and get an estimated 1080p image back out. This isn't exactly what DLSS does, but it's conceptually close enough to get a basic idea. In other words, what DLSS does is more sophisticated than just that, but it's the same basic family of ideas. AI isn't "thinking like humans do," it's just a model fitted to past data that gets used to predict future data. When it comes to image enhancement, it's pretty damn good at it, even if it's dog shit at writing a legal brief or refactoring code.
It's not mere upscaling, i.e. they don't just take the base image and interpolate new pixels. The use of a model that is fitted to high-res data enables DLSS to reconstruct missing detail, which interpolation alone can't do. Older AMD cards run an interpolator called FSR2, which is more sophisticated than simple linear interpolation, as it uses data from the previous frame as well as the current one, but it is still just an interpolator and therefore not as capable as DLSS. You know how low-res looks a lot better in motion than in still images? FSR2 operates on the same principle; the movement between the two frames is what allows it to sort of "find" detail that upscaling a single frame alone can't do. In my experience, FSR 2 is quite blurry and prone to a lot of artifacts, so I almost never use it. DLSS is nearly identical to native res, so I always use it. FSR 4 is much closer to DLSS, but based on your budget, I doubt you'll be getting a 9060.
Even if you don’t like upscaling and call frame gen fake frames, DLSS is still a very nice feature to have, because it enables DLAA, which in my experience is by far the strongest antialias on the market. It runs much better than comparable supersampling and still looks great.
The most demanding game I play is Farming Simulator 25, which requires a 1050 TI and 8GB of RAM. Those are the minimum specs though, and unlike my Series S (GPU on par with a 1650 TI) I want to play the game at 60 FPS up from 30. Other than that title everything I play could easily run on an Xbox One (Cities: Skylines, KSP, Minecraft, Silent Hunter 3, Gmod).
there are tons of prebuilts anyway and most DDR5 stuff is cheap compared to DDR4 stuff.
check in sites for stuff you might like but be wary of shit like 8500G or 5600G/GT, they always cheap out on the memory and that's something not optimal for these apu's.
As of the new transformer model, it is very difficult to tell the difference between DLSS and native res, other than the fact you'll run the game significantly faster (or run at the same fps with higher settings, take your pick).
At what point would you say DLSS actually starts to degrade quality? I feel for 1080p, you need to use a higher setting, but for 4K, you can go lower. This is with a screen at optimal viewing distance, so 4K does provide additional information to the eye.
It seems as long as the ‘base’ resolution is sufficient enough, the AI is able to construct excellent scenes at either 1080p or 4K. This seems less intuitive from deterministic interpolation where we see break down at larger scale factors (the ‘old games look like shit on new TVs’ effect).
At what point would you say DLSS actually starts to degrade quality? I feel for 1080p, you need to use a higher setting, but for 4K, you can go lower. This is with a screen at optimal viewing distance, so 4K does provide additional information to the eye.
It seems as long as the ‘base’ resolution is sufficient enough, the AI is able to construct excellent scenes at either 1080p or 4K. This seems less intuitive from deterministic interpolation where we see break down at larger scale factors (the ‘old games look like shit on new TVs’ effect).
I think the answer is you don't want your base resolution to drop below 360p. 1080p Ultra Performance (360p -> 1080p) looks good, and 720p Performance (360p -> 720p) looks good, but 720p Ultra Performance (240p -> 720p) really does kinda look like shit.
As long as the base is over 360p, about the only time I notice is when there's some obvious render-to-texture effect whose resolution is calculated from the display res and is thus unaffected by the upscaling. Example from Darktide, 1080p DLAA left, 360p upscaled to the right. You can see that green hologram is a lower-res texture when the target res is 360p, and DLSS doesn't change that. However, on my laptop, this game runs at 20-25 fps at native 1080p, which is unplayable.
720p native, Performance (360p), Ultra Performance (240p). The 240p upscaling is pretty blurry, but then, it's starting the the same resolution the Nintendo 64 was working with, so you can only ask for so much. You can also see that the shadow maps' resolution appears to be derived from the screen resolution.
So I play this at 720p Performance with Fluid Motion Frames enabled (AMD's frame gen works with any supported game, regardless of what hardware you have). Put it all together, and it looks pretty good. Considering this is a 35W laptop GPU, I'm not complaining.
The most demanding game I play is Farming Simulator 25, which requires a 1050 TI and 8GB of RAM. Those are the minimum specs though, and unlike my Series S (GPU on par with a 1650 TI) I want to play the game at 60 FPS up from 30. Other than that title everything I play could easily run on an Xbox One (Cities: Skylines, KSP, Minecraft, Silent Hunter 3, Gmod).
At what point would you say DLSS actually starts to degrade quality? I feel for 1080p, you need to use a higher setting, but for 4K, you can go lower. This is with a screen at optimal viewing distance, so 4K does provide additional information to the eye.
It seems as long as the ‘base’ resolution is sufficient enough, the AI is able to construct excellent scenes at either 1080p or 4K. This seems less intuitive from deterministic interpolation where we see break down at larger scale factors (the ‘old games look like shit on new TVs’ effect).
I have one, its not terrible but I'm using it in a HP SFF with an i5-2400 for retro collections on big TV. The overall compatibility is pretty good for oldfagging, don't play anything new on it the cpu can be choked by just attempting ps3 emulation so I've never really put the actual GPU through its paces enough to be disappointed.
Meant to find a tiny lease dumped MFF replacement with M2 and ddr4 but forgot midsearch. Got the RX 6400 during the great gpu shortage to replace a 730... a 3050 low profile was more like a fever dream or mirage then.
There is one catch: Zeus can only beat the RTX 5090 GPU in path tracing and FP64 compute workloads. It's not clear how well it will handle traditional rendering techniques, as that was less of a focus. In speaking with Bolt Graphics, the card does support rasterization, but there was less emphasis on that aspect of the GPU, and it may struggle to compete with the best graphics cards when it comes to gaming. And when it comes to data center options like Nvidia's Blackwell B200, it's an entirely different matter.
I would also venture that unless something absolutely revolutionary is attempted and succeeds, like a 100x jump in perf/watt, the big names will catch up. RDNA5 could easily double raytracing performance from RDNA4, 10x is not long off and will happen naturally within the next 5-10 years.
Latest Windows 11 security patch might be breaking SSDs under heavy workloads — Phison investigating reports of disappearing drives following file transfers, including some that cannot be recovered after a reboot
Here are a list of tested and affected drives that did and did not get fucked in the Windows 11 security update
Notably WD Blue M.2 SSD's got turbofucked and couldn't be retrieved.
The error occurs when downloading 50gb or more at once. Seems to affect drives without NAND more.
Now my rig is all SSD's, 1 SATA, 3 M.2. I haven't noticed any issues. I also haven't been making heavy downloads to my one MLC Kingston NV2 drive I use for file storage so... idk , everything else has 3D NAND. But this is something to be aware of if you have Windows 11. Hopefully they fix this fast.
A few quick questions. My situation is as follows; I'm thinking about getting a Minisforum with the HX7945. My use case is 75% just general use, non AAA gaming, browsing, watching movies, etc. The other 25% is messing around in VMs, video editing and encoding with DaVinci Resolve, and compressing/decompressing files that I upload to/download for work. I just like the idea of having most of the 7950x performance, but only 65-100w power draw.
1. Is it worth getting an X3D CPU if I never really plan on getting a high end GPU? The highest end GPU I'll end up with is likely the 7900 GRE.
2. Am I missing out on significant real world performance (for my use case) with DDR5-5600 CL40 or DDR5-5200 CL36 SODIMMs over a traditional desktop with DDR5-6000 CL30?
A few quick questions. My situation is as follows; I'm thinking about getting a Minisforum with the HX7945. My use case is 75% just general use, non AAA gaming, browsing, watching movies, etc. The other 25% is messing around in VMs, video editing and encoding with DaVinci Resolve, and compressing/decompressing files that I upload to/download for work. I just like the idea of having most of the 7950x performance, but only 65-100w power draw.
1. Is it worth getting an X3D CPU if I never really plan on getting a high end GPU? The highest end GPU I'll end up with is likely the 7900 GRE.
2. Am I missing out on significant real world performance (for my use case) with DDR5-5600 CL40 or DDR5-5200 CL36 SODIMMs over a traditional desktop with DDR5-6000 CL30?
yesn't, go for the basic bitch DDR5 with pcie-4.0/5.0 stuff, also on the CL/DDR frequency the results aren't really different unless you do some calculations and factor in the procssor ghz that you will use, overlcock and shieet. also also X3D is literally made for gaming according to AMD themselves but they show some benchmarks for stuff regarding their flagship which is the 9950 series, do you want to know the other ones? though luck, alls you get is comparisons to gaming, i'd rather tell you to look at the processor you want in the amd comparison site and work from there.
The non-X3D 16-core boards are much cheaper and it doesn't sound like you'll be pushing it. Some non-gaming workloads benefit from 3D cache but you have to look around for them. You can check the Phoronix review of the 7950X3D for some examples. Compression may benefit, while decompression probably doesn't.
i hope new ryzens don't go the route of just filling the L3 like crazy and forget about the core clock, leaving for you to rely on OC shit, intel already has the computer heater section covered up pretty well, they don't need competition in that part.
i hope new ryzens don't go the route of just filling the L3 like crazy and forget about the core clock, leaving for you to rely on OC shit, intel already has the computer heater section covered up pretty well, they don't need competition in that part.
, everything else has 3D NAND. But this is something to be aware of if you have Windows 11. Hopefully they fix this fast.
