@@Kamil_OPhilip is a tech youtuber the same way Red Dead Redemption is a horse grooming simulator - as-in, yes, sure, but he's a lot of other things as well & that's brought to bear in his style of review & how he measures the product in a wider context than the strict price:performance equation.
This might be your unintentionally funniest video yet in my opinion. The switch between all the unexpected Fluffykins appearancs, existential questions about why we still strive for more, and the genuine insight into how this generation of cards and technologies will affect gaming just warms my heart for some reason. (Also when are you going to release the RTX Fluffykins remaster?)
thank you for the catmera angles, and moving the graphs and footage out of the way. excellent work as always dont remember anything you said or showed but i prefer this.
It's quite disappointing that no benchmark or review ever talks about VR. If you have a good quality headset you need to push rasterised games at crazy resolution and high framerate, and for that the 4090 was already limited and the 5090 might even be as well for upcoming headsets like the Pimax Crystal Super
@theowainwright7406 I seriously doubt it, relfex 2 looks close to what's done with motion reprojection, which gives the impression of higher framerate but with big visual artefacts, we'll have to see. Regarding frame gen, any upscaling or AI artefact is REALLY noticeable in VR, you basically have them right in front of your eyes. What does wonders is Dynamic Foveated Rendering but almost no game has native compatibility
@@theowainwright7406 i feel like mfg would give massive motion sickness in VR. Your movements/inputs are still being put out at whatever fps the game is actually running at. VR needs to be responsive else you'll feel like your arms are in jell-o
Shame that you get rude comments like that, I only found your channel in the last year or two but I greatly prefer your hardware reviews as they feel more... personal I guess. Any other channel is overly sanitized and... TH-cam-y
It makes him feel more attainable. On some level I know the likes of Gamer'sNexusSteve are too good for me even as a parasocial friend. Klikphilip is someone who went on caravan holidays in the early 00s, just like me.
Philip makes a good point in that many older cards cards (2 and 3 series) would benefit tremendously from MFG, yet they made it exclusive to the one series that DIDN'T need it !
As someone using a 2080, the lack of frame generation without using a hack to get it working is kinda stupid really, like 20 series absolutely could benefit from FG and of course, MFG, but yeah sure let's give it to the way more powerful cards.
DLSS 4 no longer relies on Optical Flow, and an NVIDIA Engineer has stated that it's now a question of optimization to see if they can port FG to older hardware
@@lukasg4807 sponsorblock is an extension that works by someone from the community marking some sections of a video, usually sponsored segments, so that you can skip it automatically. in this case, they used it to skip some "annoying" parts.
Good fellow, I must say that's a cracking review, old chap. It's an exquisite critique, if I do say so myself. The depth and breadth of analysis is simply spiffing, a truly thorough and ripping assessment, if you will. One couldn't ask for a more jolly good review, indeed.
Great review, Only thing I was kinda hoping you would cover is VR! VR is the only reason GPU's need to keep moving forward in raw raster power. The display port upgrade alone is going to drive the next 2-4 years of PCVR headsets. This GPU + A Headset that supports eye tracking will be insane for VR.
"you don't need it, you don't need that either it's ridiculous" is the best argument I've heard to make me want one. I am at least saved by the fact I can't afford one though!
One thing to keep in mind with the low-latency async reprojection feature -- which you may have mentioned in your original video on it -- is that it _only_ affects movements! Firing a weapon or doing any other sort of "gameplay action" will still show the "true" latency. I could see this feeling really weird and disjointed as the latency grows higher.
I cannot tell if the audio or visuals are out by 1/10 of a second, but CAN feel if there's even the slightest delay from my mouse inputs. I don't think the action on-screen being delayed will matter so long as my mouse adjustments feel instant.
I mean, a lot of games that really care about latency are already "reprojecting" a lot of gameplay actions - the CPU _can_ know you've done a thing before the GPU finishes rendering the next frame, after all.
I hope they'll keep pushing the envelope in the future, 6090 should be a 1200W GPU, absolute monster. Make my electricity bill hurt. I want to feel like I'm managing a data center when I get a new GPU.
i had this video open while benchmarking a bunch of tf2 maps to compare them against the map i'm currently building (you make for wonderful background audio when mapping btw) and i have to say. the few times i alt tabbed back into the video were all the times you remarked on fluffy doing something adorable. you really got me (and the retention figures) there!
I was like 'I've got a month before the 5000 series is released, plenty of time to finish it off!' then suddenly I didn't after my roof blew off, I attended an award show in Serbia, got hit HARD with covid or something and then suddenly found myself trying to test the 5090 in time before release lol. I'll get back to it, the 5090 actually makes the 8K topic more relevant
We really took the 4090 for granted. That was a 70% uplift over the 3090. Now we have to deal with harsh reality, 30% uplift, 30% power draw increase. Exactly the same as the 2080 Ti to the 3090 was. I guess the 40 series wasn't that bad after all. At least, the 4090 wasnt
I suspect this card will be a good target for undervolting, just like the 4090 is. I would not expect that the full 600W is the sweet spot for power efficiency on this card.
its pretty nice to see such a nice review put so succinct with direct bench's and comedy forefront without fluff. i say this with an LTT review the immediate recommended video. I think LTT are fantastic, but there are many reasons why ive been subbed for these channels for over a decade.
async reprojection shows us there is still improvements in graphics technologies. Imagine polishing and developing every published paper about better and faster render techniques. There is surely a better future.
One of the best and most insightful reviews I've ever watched, thank you. What about VR though? Wouldn't the RTX 5090 overkill performance be actually useful and able to be utilized there? Especially with high resolution (4k-ish) per eye headsets incoming
Back when I bought my 5700xt I was shocked that it was rated 225W. Cards today are absolutely insane, we really need to focus on reasonable power consumption.
I must say your review has made me appreciate the wide range of technologies the 50 series represents a lot more than I did based off of marketing and other reviews, in large part because it is very use-case focussed instead of numbers focussed. That in and of itself highlights the problem with these technologies though: there's no longer a good objective way to compare cards. FPS is easy, pacing and latency are doable, but the trade-off between smoothness, latency, and artefacts is nearly impossible to capture in a review and that ambiguity can be abused by marketing to push some very misleading figures. While I greatly respect the sheer scientific rigour of people like Gamers Nexus, this will be the review I'll share with others for this generation.
It feels like you're the only reviewer who actually tries looking at their camera instead of so many others who stare at their teleprompter ending up looking cross-eyed.
Can we also talk about idle power draw? It's not just when it's used, even idle it apparently already uses 40 watts doing barely anything! How high is the power draw, for example, when watching TH-cam? The majority of people probably don't game all the time.
Yeah there's clearly something wrong there. If you read all the new Max-Q functionality this card has it shouldn't do this, it should draw way less power than the 4090. I guess NVIDIA forced all the power saving tech off because it wasn't ready for primetime. Hoping for some postlaunch fine wine power draw optimizations.
The split between framerate and responsiveness happened even before frame generation thanks to Reflex. A little context, the render queue was a way to marginally improve performance by allowing the CPU to work ahead dumping instructions for the GPU to pick up so it never has to wait. The problem is that instructions sitting in the queue means the game state being rendered is stale. Reflex kills the render queue and enforces just in time draw call submission so draw calls are only generated right before the GPU is ready for them, this can slightly lower framerate especially in GPU bound scenarios but significantly improves responsiveness. Once upon a time Reflex was its own seperate SDK, a bit obscure it had a few e-sports titles using it. Being added to the requirements of DLSS3 was the best thing to ever happen to Reflex.
I realized you are on another level when you zoomed in on the reflection of your mouth in a mirror for a minute. Everyone else is struggling to benchmark this overkill card while you produce avant garde autistic art while benchmarking in a bathrobe.
Yes those 8k numbers are impressive indeed and I like that someone has tested that, its a far better data point than I had considered in the context of this card, or others for that matter.
Great video as always. It is cool to see the new dlss making super high resolutions playable and kind of gives me hope that they can plateau at this level and focus on power efficiency in the next generation, because while the idea of combining a space heater and PC is cool, maybe the carpet floors of my apartment might not take well to it.
9:54 Philip, can you please paint the corner of your wall? It kinda looks like the missing pixels from Mirage B apps and i cant unsee it. The seam has light spots...
Like I said, it comes with the added latency that standard FG already comes with. It is inherent for there to be some latency, given how it works- it would not be as low as if reflex were on and FG were off.
glad to know i'm not the weird one for letting my dear boy sylvesteri rest on top of my warm pc! also glad to know that the 1070ti still beats modern GPUs (at least in terms of the cat comfort benchmark)
This is gives me Deja Vu. Back in 2003 there was another GeForce 5900 card that fried PSUs, it was very power hungry for it's day. The GeForce FX 5900. It required a molex connector and it had 60 wats TDP. At the time almost all PSUs were 300 wats rated, but we didn't had the 80+ standard and in reality they were mote like 150 wats.
Thanks for the video, I enjoy your honest takes. There is simply too much fake feeling hype and rage out there. Your point about MFG and lowend/portable hardware is really good, I wish something like my steamdeck could have MFG, it would be more impactful there then on a high end gpu in a desktop.
But most games nowdays look blurry due to their shitty usage of noisy systems and then smearing TAA all over it to make it less flickering. We had great looking games 6 years ago.. look at Battlefield 1..
@tarakivu8861 Yes it's mostly the TAA causing the effects. It also helps that BF1 had a great engine specifically made for the game. MFG could work in all titles regardless of it's native output
Honestly I'm really glad you are doing these reviews, cause you seem to be the only one who's enthusiastic about new DLSS and Frame-Gen features. The general opinion I see on the internet is "Oh my god fake benchmarks/fake frames" etc. But I have played 500h of Cyberpunk with DLSS on and havent noticed any artifacts when I wasn't *actively* looking for them. So I think it's exciting that they are making DLSS even better and offer the opportunity use MFG to make even the most ridiculous settings work somehow. I dont know if this is the generation I'll be upgrading to from my 2070, but it sure does sound appealing.
If in the future, models to predict and reconstruct frames become good and fast enough, it would theoretically be possible to generate and display a frame before the traditionally rasterized frame has finished rendering. This could be used to double the framerate while keeping the original latency, instead of increasing it espite having more frames, like DLSS and co does now.
Regarding Reflex 2, I'm curious to see the specifics, and a bit skeptic about it. They haven't released the SDK or docs on it yet. From my understanding, Reflex 1 delays the start of the new frame (hence also gathering of input) until the current frame finishes rendering, solving the backpressure issue. But this creates a bubble in the rendering pipeline, because after dispatching the current frame to the GPU, the rendering thread and the GPU hardware scheduler idle out until the driver notifies the rendering thread for it to start producing the next frame. Seems Reflex 2 doesn't try to tackle this issue. Instead, it tries to bypass parts of the rendering pipeline by gathering camera data from the engine, feeding it to the GPU directly, which effectively splits the camera from the scene by updating the former while partially extrapolating the latter by reprojection. For frame generators, it looks like it'll be awesome, as access to more up-to-date camera position should significantly improve input latency. For us playing CS or Valorant, I'm not exactly sure about the effects. Since it's basically speculating the world around you, it'll rely on having high framerates made of actual frames in order to quickly correct the mistakes of the reprojections. The fact that most movement is horizontal and distances aren't really big in these 2 games also make them good candidates for this type of tech. But for games with lots of camera movement and/or are really fast paced, it could be a hinder. Even in CS and Valorant, if you already flick instinctively and precisely, you not only may not need this, but you also may not want to use it.
Great review as usual. The focus on the new tech is warranted and this card is clearly hamstrung by the immature software environment. Ray tracing performance uplift being lower than raster most of the time proves this. We need the testing for AW2 with the new RTX Mega Geometry update to know just how far NVIDIA have come. Looking forward to see what NVIDIA manages with Reflex 2 and there's nothing preventing NVIDIA from implementing a 3-4x mode here in the future. Imagine the latency savings and increases to responsiveness, but I guess NVIDIA properly needs to sort out the inpainting artifacts first.
cant wait to see the 8k results. if the 3090 was the beginning of 8k gaming, and the 4090 somewhere around playable, i hope the 5090 is truly the first 8k GPU
Your audience retention is very cute. Your arguments are very good. I'll buy a 5090 anyway because I explicitly want to have the absolute best in slot at least *once* but it'll go severely underutilized...
So mostly getting 4000 series with a price cut is good i guess if i dont really plan t use MFG? I mean, 3000 adn 4000 series is getting better with each gpu releases I guess I agree with your mid point opinion,MFG will be great to make a GPU keep up with new and newer gen games, especially on mobile handheld
I appreciate the approach of subjective "feel" in terms of input latency with framegen and covering the mouse-frame disconnection technology thingy. I watched the reviews from GN and on technical standpoint theirs is of course objectively better - they have the tech to do the thermal imaging and all that, but there's tremendous value in a subjective look at the actual gameplay erxperience/
I'm only looking at the graph at 17:40 but maybe try and use the fat grid lines to distinguish different categories of what you're testing / showing. Might make it easier on the eye for a quick glance
ah here we see the fitting living room of someone who bought the 5090 at launch
philip is a strange guy
He just doesn't have a gf @@dantekiwi7926
@ why do you need a living room? The gaming room is my sleeping room too and there is no life in here lol
Can’t wait for the RandomgaminginHD review in 8 years
thats a very cultured joke
"OEM 5090 Blower style card is a BEAST...and very hot"
dont forget the iceberg tech video too
"How does 2025's 5090 perform in a CEX ex-business machine?"
@@Sinbad1999 8 slot blower card pulled right out of an alienware!
"I'm trying to be professional" he said in a bathrobe and pajama pants with a cat on his lap
Thank you mr klikman, you are like no other
So you're saying I shouldn't have sold my 4090 for $50 on marketplace?
Yeah, should've sold it for $5 and to me instead.
I'm not sending it back
do you have another one mate?
I would like one like this
If a cat trusts you, she gives you a headbutt. If you trust a cat, you let her walk around GPUs worth $4000 on the edge of a table.
Clicked your video instead of {insert tech youtuber here}.. love your down-to-earth takes on hardware
Yea it doesnt felt like that infomercial format
what do you mean? This is tech youtuber.
@@Kamil_OPhilip is a tech youtuber the same way Red Dead Redemption is a horse grooming simulator - as-in, yes, sure, but he's a lot of other things as well & that's brought to bear in his style of review & how he measures the product in a wider context than the strict price:performance equation.
Let's go with LTT since they used an old cpu - nullifying any data.
This might be your unintentionally funniest video yet in my opinion. The switch between all the unexpected Fluffykins appearancs, existential questions about why we still strive for more, and the genuine insight into how this generation of cards and technologies will affect gaming just warms my heart for some reason.
(Also when are you going to release the RTX Fluffykins remaster?)
i might be crazy but philip's humor is very "David lynch" like, a mambo between raw sincerity and surrealist sillyness
God tier.
"Too hot, even for a cat" Wow that's hot!
That's a real benchmark
If I heats I yeets - cat probably
Never stop making videos Phillip, you can make a video about any topic and it would be interesting. Big love from 🇦🇷
Warping interpolated frames to seem responsive would be an absolute game changer.
Game devs really need to start implementing asynchronous reprojection
thank you for the catmera angles, and moving the graphs and footage out of the way. excellent work as always dont remember anything you said or showed but i prefer this.
Cant wait to replace my PSU with a miniature nuclear reactor.
The RTX6090 is gonna come with its own external power supply like the Xbox360's power brick.
@@InnuendoXP * (sold separately batteries not included)
It's quite disappointing that no benchmark or review ever talks about VR. If you have a good quality headset you need to push rasterised games at crazy resolution and high framerate, and for that the 4090 was already limited and the 5090 might even be as well for upcoming headsets like the Pimax Crystal Super
Does reflex 2 and mfg help in that regard?
@theowainwright7406 I seriously doubt it, relfex 2 looks close to what's done with motion reprojection, which gives the impression of higher framerate but with big visual artefacts, we'll have to see.
Regarding frame gen, any upscaling or AI artefact is REALLY noticeable in VR, you basically have them right in front of your eyes.
What does wonders is Dynamic Foveated Rendering but almost no game has native compatibility
@@theowainwright7406 i feel like mfg would give massive motion sickness in VR. Your movements/inputs are still being put out at whatever fps the game is actually running at. VR needs to be responsive else you'll feel like your arms are in jell-o
yeah we need more of these VR channels get involved with Benchmarking VR.
Shame that you get rude comments like that, I only found your channel in the last year or two but I greatly prefer your hardware reviews as they feel more... personal I guess. Any other channel is overly sanitized and... TH-cam-y
Thank you :) Yeah just part of the job of being shown online these days unfortunately. I get lots of nice comments too!
It makes him feel more attainable. On some level I know the likes of Gamer'sNexusSteve are too good for me even as a parasocial friend. Klikphilip is someone who went on caravan holidays in the early 00s, just like me.
Philip makes a good point in that many older cards cards (2 and 3 series) would benefit tremendously from MFG, yet they made it exclusive to the one series that DIDN'T need it !
As someone using a 2080, the lack of frame generation without using a hack to get it working is kinda stupid really, like 20 series absolutely could benefit from FG and of course, MFG, but yeah sure let's give it to the way more powerful cards.
Does not have optical flow.
@@starbez optical flow is just some matrix pipeline that accelerates by 0.17x, not needed
DLSS 4 no longer relies on Optical Flow, and an NVIDIA Engineer has stated that it's now a question of optimization to see if they can port FG to older hardware
But only with Reflex 2, otherwise the latency would still feel poor
the sponsorblock highlight just goes to caboosing that's golden
Sponsorblock the goat
Wdym?
It always does :)
@@lukasg4807 sponsorblock is an extension that works by someone from the community marking some sections of a video, usually sponsored segments, so that you can skip it automatically. in this case, they used it to skip some "annoying" parts.
Your current presentation style is certainly Inspiring.
Love the lighting in this one, Phil. It's got a real early 2000s CIA black site vibe to it.
Took me a few seconds to realize that FE stands for "founder's edition" and not "fan edition"
My favourite cat channel on TH-cam, you just have to put up with the guy talking about graphics cards
i love your perspective on tech so much more than dedicated hardware folks' often hivemind-ish views, please continue doing these!!!
10:41 phillips has turned into an old cat lady and i'm for it
Good fellow, I must say that's a cracking review, old chap. It's an exquisite critique, if I do say so myself. The depth and breadth of analysis is simply spiffing, a truly thorough and ripping assessment, if you will. One couldn't ask for a more jolly good review, indeed.
Great review, Only thing I was kinda hoping you would cover is VR! VR is the only reason GPU's need to keep moving forward in raw raster power. The display port upgrade alone is going to drive the next 2-4 years of PCVR headsets. This GPU + A Headset that supports eye tracking will be insane for VR.
Thought I wasn't too interested in those "pure tech" videos after building a new PC; Congratz on proving me wrong. Truly captivating review.
"you don't need it, you don't need that either it's ridiculous" is the best argument I've heard to make me want one. I am at least saved by the fact I can't afford one though!
None of Linus' employees lick his legs! That's why I come to 2kliksphilip for all my needs.
This should have more likes
One thing to keep in mind with the low-latency async reprojection feature -- which you may have mentioned in your original video on it -- is that it _only_ affects movements! Firing a weapon or doing any other sort of "gameplay action" will still show the "true" latency. I could see this feeling really weird and disjointed as the latency grows higher.
I cannot tell if the audio or visuals are out by 1/10 of a second, but CAN feel if there's even the slightest delay from my mouse inputs. I don't think the action on-screen being delayed will matter so long as my mouse adjustments feel instant.
I mean, a lot of games that really care about latency are already "reprojecting" a lot of gameplay actions - the CPU _can_ know you've done a thing before the GPU finishes rendering the next frame, after all.
Your camera angles were 11/10 in this review haha :D
Glad the 4090 is still looking like a good investment long term.
0:20 and increased price lol
I hope they'll keep pushing the envelope in the future, 6090 should be a 1200W GPU, absolute monster. Make my electricity bill hurt. I want to feel like I'm managing a data center when I get a new GPU.
i had this video open while benchmarking a bunch of tf2 maps to compare them against the map i'm currently building (you make for wonderful background audio when mapping btw) and i have to say. the few times i alt tabbed back into the video were all the times you remarked on fluffy doing something adorable. you really got me (and the retention figures) there!
Continuation of your 8k series when? :O
I was like 'I've got a month before the 5000 series is released, plenty of time to finish it off!' then suddenly I didn't after my roof blew off, I attended an award show in Serbia, got hit HARD with covid or something and then suddenly found myself trying to test the 5090 in time before release lol. I'll get back to it, the 5090 actually makes the 8K topic more relevant
@2kliksphilip yes especially with the Samsung Odyssey G9 57 inch being 8K too
I love the casual presentation style you’ve gone with, it’s so entertaining
I like your cozy way of presenting stuff
thank you for the foot cam phil
I love the editing! You are way more casual and less strict that in the older videos! i hope it simplifies production!
Damn, I was really looking forward to you talking about reflex 2. I really think that's gonna change gaming forever
I listem to your videos before bed, your narration is great. V good analysis!
We really took the 4090 for granted. That was a 70% uplift over the 3090. Now we have to deal with harsh reality, 30% uplift, 30% power draw increase. Exactly the same as the 2080 Ti to the 3090 was. I guess the 40 series wasn't that bad after all. At least, the 4090 wasnt
4090 jumped several nodes. The 5090 not so much. Looks like they've focused on it's physical design and software features this generation
40 series was Pascal level of evolution, it was just behind big price increase, unlike Pascal
I'm enjoying the new formats of these videos!! Keep it up Phil:)
VR TESTING. DAYUM this thing will drive some headsets with high resolutions!
I suspect this card will be a good target for undervolting, just like the 4090 is. I would not expect that the full 600W is the sweet spot for power efficiency on this card.
its pretty nice to see such a nice review put so succinct with direct bench's and comedy forefront without fluff. i say this with an LTT review the immediate recommended video. I think LTT are fantastic, but there are many reasons why ive been subbed for these channels for over a decade.
async reprojection shows us there is still improvements in graphics technologies. Imagine polishing and developing every published paper about better and faster render techniques. There is surely a better future.
Thanks for this chill style of video lol.. never change
such a smart way to feature the mirror, graphs and STILL show your mouth moving!
Glad you tried to give a nuanced approach. Great stuff cold man.
Never change Mr Phillips
I'm so excited about this, and I don't mind if some may qualify it as "too hot."
Oh, and there's a graphics card too.
your reviews are the best, im hoping to see the next benchmark if the reflex thingy :333
this switch to a more comfortable setting makes these reviews far superior to any other review on youtube. OnlyKliks when?
You're kinda making me actually start to like and become interested in fake frames... 😳
Fluffykins seems to be stuck at a lower LOD than usual.
thanks for a thoughtful and nuanced exploration of the new hardware/software.
11:34 Is this the new standard for commentary videos? Cuz Im here for it
One of the best and most insightful reviews I've ever watched, thank you. What about VR though? Wouldn't the RTX 5090 overkill performance be actually useful and able to be utilized there? Especially with high resolution (4k-ish) per eye headsets incoming
Back when I bought my 5700xt I was shocked that it was rated 225W.
Cards today are absolutely insane, we really need to focus on reasonable power consumption.
I must say your review has made me appreciate the wide range of technologies the 50 series represents a lot more than I did based off of marketing and other reviews, in large part because it is very use-case focussed instead of numbers focussed. That in and of itself highlights the problem with these technologies though: there's no longer a good objective way to compare cards. FPS is easy, pacing and latency are doable, but the trade-off between smoothness, latency, and artefacts is nearly impossible to capture in a review and that ambiguity can be abused by marketing to push some very misleading figures. While I greatly respect the sheer scientific rigour of people like Gamers Nexus, this will be the review I'll share with others for this generation.
It feels like you're the only reviewer who actually tries looking at their camera instead of so many others who stare at their teleprompter ending up looking cross-eyed.
Can we also talk about idle power draw? It's not just when it's used, even idle it apparently already uses 40 watts doing barely anything! How high is the power draw, for example, when watching TH-cam? The majority of people probably don't game all the time.
Yeah there's clearly something wrong there. If you read all the new Max-Q functionality this card has it shouldn't do this, it should draw way less power than the 4090. I guess NVIDIA forced all the power saving tech off because it wasn't ready for primetime. Hoping for some postlaunch fine wine power draw optimizations.
The split between framerate and responsiveness happened even before frame generation thanks to Reflex. A little context, the render queue was a way to marginally improve performance by allowing the CPU to work ahead dumping instructions for the GPU to pick up so it never has to wait. The problem is that instructions sitting in the queue means the game state being rendered is stale. Reflex kills the render queue and enforces just in time draw call submission so draw calls are only generated right before the GPU is ready for them, this can slightly lower framerate especially in GPU bound scenarios but significantly improves responsiveness.
Once upon a time Reflex was its own seperate SDK, a bit obscure it had a few e-sports titles using it. Being added to the requirements of DLSS3 was the best thing to ever happen to Reflex.
I realized you are on another level when you zoomed in on the reflection of your mouth in a mirror for a minute. Everyone else is struggling to benchmark this overkill card while you produce avant garde autistic art while benchmarking in a bathrobe.
Yes those 8k numbers are impressive indeed and I like that someone has tested that, its a far better data point than I had considered in the context of this card, or others for that matter.
Clicking here first next time since linus's 130-person merch company couldn't find a fucking 9800x3d to test with.
Great video as always. It is cool to see the new dlss making super high resolutions playable and kind of gives me hope that they can plateau at this level and focus on power efficiency in the next generation, because while the idea of combining a space heater and PC is cool, maybe the carpet floors of my apartment might not take well to it.
you look different , then I have expected 👍👍
The 50 series of NVIDIA GPUs feel like they're going all out on working smarter instead of harder.
Question is, is smarter better?
Is smarter actually dumber?
9:54 Philip, can you please paint the corner of your wall? It kinda looks like the missing pixels from Mirage B apps and i cant unsee it. The seam has light spots...
11:30 apparently the multi frame gen doesnt add much latency, maybe a few extra miliseconds
i think its on gamernexus or optimums review
Like I said, it comes with the added latency that standard FG already comes with. It is inherent for there to be some latency, given how it works- it would not be as low as if reflex were on and FG were off.
I like the perspective of the reviews you make. Most reviewers are more technical than realistic. This seems to ground this halo product
please never stop using your music in vids, its so good
glad to know i'm not the weird one for letting my dear boy sylvesteri rest on top of my warm pc! also glad to know that the 1070ti still beats modern GPUs (at least in terms of the cat comfort benchmark)
I like the presentation style couch video style
Very good video bro
This is gives me Deja Vu. Back in 2003 there was another GeForce 5900 card that fried PSUs, it was very power hungry for it's day. The GeForce FX 5900. It required a molex connector and it had 60 wats TDP. At the time almost all PSUs were 300 wats rated, but we didn't had the 80+ standard and in reality they were mote like 150 wats.
8:43 i miss cache so much😭
Thanks for the video, I enjoy your honest takes. There is simply too much fake feeling hype and rage out there. Your point about MFG and lowend/portable hardware is really good, I wish something like my steamdeck could have MFG, it would be more impactful there then on a high end gpu in a desktop.
Came for the breakdown
Stayed for the introductory sniff
That hogwarts comparison with MFG really sold it for me. Insane how choppy and motion-blurred most games look in comparison.
But most games nowdays look blurry due to their shitty usage of noisy systems and then smearing TAA all over it to make it less flickering.
We had great looking games 6 years ago.. look at Battlefield 1..
@tarakivu8861 Yes it's mostly the TAA causing the effects. It also helps that BF1 had a great engine specifically made for the game. MFG could work in all titles regardless of it's native output
i almost clicked off the video until fluffykins came to rescue
thank you fluffykins!
Honestly I'm really glad you are doing these reviews, cause you seem to be the only one who's enthusiastic about new DLSS and Frame-Gen features.
The general opinion I see on the internet is "Oh my god fake benchmarks/fake frames" etc. But I have played 500h of Cyberpunk with DLSS on and havent noticed any artifacts when I wasn't *actively* looking for them. So I think it's exciting that they are making DLSS even better and offer the opportunity use MFG to make even the most ridiculous settings work somehow.
I dont know if this is the generation I'll be upgrading to from my 2070, but it sure does sound appealing.
Thank you for the vids ❤
Bruh the power consumption graph perfectly mirroring the fps graph
what is the game at 3:40?
Alan Wake 2
If in the future, models to predict and reconstruct frames become good and fast enough, it would theoretically be possible to generate and display a frame before the traditionally rasterized frame has finished rendering. This could be used to double the framerate while keeping the original latency, instead of increasing it espite having more frames, like DLSS and co does now.
Regarding Reflex 2, I'm curious to see the specifics, and a bit skeptic about it. They haven't released the SDK or docs on it yet.
From my understanding, Reflex 1 delays the start of the new frame (hence also gathering of input) until the current frame finishes rendering, solving the backpressure issue. But this creates a bubble in the rendering pipeline, because after dispatching the current frame to the GPU, the rendering thread and the GPU hardware scheduler idle out until the driver notifies the rendering thread for it to start producing the next frame.
Seems Reflex 2 doesn't try to tackle this issue. Instead, it tries to bypass parts of the rendering pipeline by gathering camera data from the engine, feeding it to the GPU directly, which effectively splits the camera from the scene by updating the former while partially extrapolating the latter by reprojection. For frame generators, it looks like it'll be awesome, as access to more up-to-date camera position should significantly improve input latency. For us playing CS or Valorant, I'm not exactly sure about the effects. Since it's basically speculating the world around you, it'll rely on having high framerates made of actual frames in order to quickly correct the mistakes of the reprojections. The fact that most movement is horizontal and distances aren't really big in these 2 games also make them good candidates for this type of tech. But for games with lots of camera movement and/or are really fast paced, it could be a hinder. Even in CS and Valorant, if you already flick instinctively and precisely, you not only may not need this, but you also may not want to use it.
Title updated before my very own two eyes. Thanks fluffykins 🐈
Great review as usual. The focus on the new tech is warranted and this card is clearly hamstrung by the immature software environment. Ray tracing performance uplift being lower than raster most of the time proves this. We need the testing for AW2 with the new RTX Mega Geometry update to know just how far NVIDIA have come.
Looking forward to see what NVIDIA manages with Reflex 2 and there's nothing preventing NVIDIA from implementing a 3-4x mode here in the future. Imagine the latency savings and increases to responsiveness, but I guess NVIDIA properly needs to sort out the inpainting artifacts first.
cant wait to see the 8k results. if the 3090 was the beginning of 8k gaming, and the 4090 somewhere around playable, i hope the 5090 is truly the first 8k GPU
Your audience retention is very cute.
Your arguments are very good. I'll buy a 5090 anyway because I explicitly want to have the absolute best in slot at least *once* but it'll go severely underutilized...
2:16 most important info right here. Every cat owner must be informed.
The new pattern for moore's law is Performance one year and Efficiency the next, and both are going to be relying on NPU's more each time.
So mostly getting 4000 series with a price cut is good i guess if i dont really plan t use MFG?
I mean, 3000 adn 4000 series is getting better with each gpu releases
I guess I agree with your mid point opinion,MFG will be great to make a GPU keep up with new and newer gen games, especially on mobile handheld
Personally playing on a PSP feels more comfortable
I appreciate the approach of subjective "feel" in terms of input latency with framegen and covering the mouse-frame disconnection technology thingy. I watched the reviews from GN and on technical standpoint theirs is of course objectively better - they have the tech to do the thermal imaging and all that, but there's tremendous value in a subjective look at the actual gameplay erxperience/
Love the target paper on the left where it's only the civilian shot
Can't wait for 2kliksphilip to reveal Fluffy was always an AI generated cat
everh sngle comment about ai is ai generated
I'm only looking at the graph at 17:40 but maybe try and use the fat grid lines to distinguish different categories of what you're testing / showing. Might make it easier on the eye for a quick glance