I impulse bought this right before the manufacturer's rebate ended and promptly found out my wife is pregnant, thankfully Costco has a great return policy

EDIT: The baby is planned, just thought it would take longer to get pregnant. It's also just about priorities. We could still afford the baby and keep the computer but the money will be better spent on baby necessities and I can just use my 3080 rig a bit longer

Doesn't feel 8 years old at all

NVIDIA

• RTX 5050 - MSRP $249 → US price $300 - +20%
• RTX 5060 - MSRP $299 → US price $450 - +51%
• RTX 5060 Ti 8GB - MSRP $379 → US price $500 - +32% US / +14% global
• RTX 5060 Ti 16GB - MSRP $429 → US price $730 - +70% US / +53% global
• RTX 5070 - MSRP $549 → US price $800 - +46% US / +31% global
• RTX 5070 Ti - MSRP $749 → US price $1,050 - +40% US / +44% global
• RTX 5080 - MSRP $999 → US price $1,550 - +55% US / +38% global
• RTX 5090 - MSRP $1,999 → US price $4,900 - +145% US / +136% global

AMD

• RX 9060 XT 8GB - MSRP $299 → US price $360 - +20% US / +26% global
• RX 9060 XT 16GB - MSRP $349 → US price $460 - +32% US / +28% global
• RX 9070 - MSRP $549 → US price $630 - +15% US / +13% global
• RX 9070 XT - MSRP $599 → US price $750 - +25% US / +21% global

Intel

• Arc B570 - MSRP $219 → US price $260 - +19%
• Arc B580 - MSRP $249 → US price $310 - +25% US / +25% global

>Buckley examines the "huge success" of Palworld 1.0. "Thank you, gamers," he says. "As cliche as it might sound, we didn't expect it to go that well. Genuinely. 900,000 CCU, insane amount of sales yet again. We're really, really happy. Again, very life changing for the studio."

Second and third pick for the result.
I tried to bend it back but it’s very hard and it could snap.
This is unacceptable for a 1200€ card

Credits: ODONATA CINEMA, KLCBoss

ngl this is like download more ram at physical hardware level instead of software. Anyone willing to try? it cost like 250 dollar… wtf

So basically they are comparing native fps vs dlss 5 + frame gen set to 6x, and showing like dlss 5 increases your fps, not lowers it. Very very lame way to lie

source of the info > HU

I did some testing in Goat Simulator using a mouse because people in this group have told me that GFN is pure magic. I'm using Wifi 7 in both ends and a $300 router. I get that some single player games might work, but come on. This is unplayable for anything that uses a mouse or fast reactions. I'm 800 miles from the server. It passes all the check points according to Nvidia.

I'm ready to get downvoted in a PERSONAL computer subreddit.

Edit: this is 25 ms. Apparently recommended Ms is 40 according to Nvidia.

For like two weeks my PC would randomly black screen and reboot maybe once every day or two. Temps were fine, stress tests were fine, RAM passed everything and event viewer wasnt telling me much.

Was looking to upgrade my PSU because that seemed like the next obvious suspect.

Then yesterday it happened literally as I leaned back in my chair. Turned the PC back on, leaned back pretty hard again and it instantly rebooted.

Turns out one of the wheels of my chair sits right beside the leg of my desk. When I leaned back it would slightly push the desk sideways. My front USB C header wasnt seated properly and that tiny movement was enough to make the cable move and occasionally cause the whole PC to freak out.

Unplugged the header completely last night and I havent been able to make it crash since.

I spent two weeks testing RAM and GPU drivers because my chair was basically rebooting my computer.

The load seems pretty heavy for older gen hardware, but still impressive due to some compatibility.

Even for RTX 3000 series gpus, the performance overhead is high, but what's impressive is that Nvidia did not go out of their way to block it officially because there are mods to unlock it (optiscaler I reckon).

Still, Nvidia could block it IF they want. But given the performance overhead, I think this new tech is not worth it for 3000 and older series users even though there is a good bit of photo realistic look (which Nvidia is claiming). I understand photorealism less. Here owning an Rx 590, honestly, I'd STOP at dlss 4.5 if I can ever own an Nvidia GPU in the future. Why I said, "ever own"? Then oh boy, you have gotta check out my previous two posts in this sub.

Planning on going Radeon next (given the current prices), but not sure when I can afford to. I REALLY HOPE Radeon's FSR 4.5 (no talk about it on the internet; I'm conjecturing) is at least indiscernibly near DLSS 4.5 so that there is 1. Great competition between the two and 2. that there is no regret buying Radeon for me in the future XD. Personally, I'm not sold on the photorealism of this bleeding edge tech. Maybe it's great! And you can opine that way, and I'd gladly read about it, but ME AM personally not liking it (so far) based on what I have seen.

But hey, we all want tech to progress to make our experience better. So, hoping for the best.

UPDATE: I've had to rethink this substantially - the final product will be different, but viable. I'm attempting upscaling now directly on the V100 in WDDM mode, and that'll pipe through the igpu. Should mostly kill latency but will be complicated. This post still serves as a proof of concept, and I'll continue development and updating. The goalpost has shifted towards a true temporal upscaler directly on the WDDM card that takes advantage of its capacity. The goal is to get a high performance upscaler taking advantage of tensor cores to run natively on either a v100 16gb or a titan v. In theory, a V100 card absolutely blows the rtx 20xx line out of the water in both memory bandwidth (huge for ML speed) and raw compute, but the difficulty is going to be getting it to work, and actually compete.

I am going to continue to add development updates to this post until it's time for a new thread

Disclaimer: I'm still working on holding framerate always above 60 and still improving the upscaling quality (playing around with a bunch of different neural net implementations) - so take this as a working demo to back up the rest of this post and that I'm not just talking out my ass. I've only been working on this for a few days and have a long way to go. I think reddit is cranking down the res, so the youtube video is at https://www.youtube.com/watch?v=6-y3H98Y7hg&feature=youtu.be

TL-DR: I've developed a way to use an old, cheap nvidia datacenter card to run machine learning accelerated upscaling on any game you already own, without replacing your GPU or your setup.

Hi everyone,

I'm a software developer who's currently focused on AI/ML workloads. I tend to be thrifty and use hardware that's a little less mainstream to get extra performance.

I own a Tesla V100 - a nvidia chip from 2017/2018 that was the first chip to ever include tensor cores. The 16gb PCIe edition is currently $230 on ebay. I only own it in the first place for running and training LLMs.

I'm developing an AMD-FSR esque temporal upscaling solution that uses a small neural network on the V100's. I actually already have a working (alpha, buggy of course) implementation running. This began as a personal hobbyist project, as I actually own TWO v100s - I wanted to see if I could put the extra one to work doing upscaling while the main one does the gaming workflow. The way I have the display setup is funky and hacky - one v100 has had its driver modified to enable wddm mode, and I stream the output from that through an iGPU.

The realization I've come to is that what I'm developing should already work as a drop in with any video card, provided you have an extra PCIe x16 lane on your machine that's not in use. Essentially, the V100 intercepts the frames being rendered in real time, upscales them, and hands them back off to your main GPU.

I feel fairly confident there'll be interest in a fully developed version of this, considering any card that actually has AI upscaling is still crazy expensive. As I mention, a 16gb V100 can be bought for about $230 and dropped into any existing rig. This should also work with any existing graphics card you already have installed.

I'm not ready to release this yet, just gauging interest in a fully developed version. I'd be remiss if I only told you about it though, so here's an early demo of upscaling tomb raider 2013 720p to 1080p.

GPU temp decreased by 5~7° (used to be 74°) and the CPU temp decreased by 15° (used to be 75°)

I let half of the air conditioning like this to cool the room.

I’ve been playing A Plague Tale Legacy: Resonance and honestly, it looks absolutely amazing.

The atmosphere, environments and overall visual presentation are beautiful.

Jokes aside, the game currently seems to struggle with freezes in certain areas. I’ve encountered a few noticeable freezes while playing, so there are definitely still some technical issues that need to be ironed out.

Asobo Studio seems to already be working on fixes, so hopefully we’ll see a patch addressing these problems soon.
Apart from that, I’m really enjoying it so far. ❤️

Tested on RTX 5090