Has NVIDIA Lied about the RTX 5000 Performance?

[@HardwareBusters]

► If you want to support us, please consider becoming our patrons. www.patreon.com/hardwarebusters ► Best PSU Picks 2024: hwbusters.com/best_picks/best-atxv3-pcie5-ready-psus-picks-hardware-busters/ ► Best PC Fan Picks 2024: hwbusters.com/best_picks/best-case-heatsink-radiator-fan-picks-hardware-busters/ ► Best CPU Cooler Picks 2024: hwbusters.com/best_picks/best-cpu-cooler-picks-hardware-busters/ ► For shopping at Amazon please use the following affiliate link to help us do more reviews: amzn.to/3v4D5IG ► Join our discord server: discord.gg/TqKDeXGt ► Our site: www.hwbusters.com

[@ZeebraZee]

7:50 PC World asked Frank Azor about why AMD hasn't switched to the 12V-2x6 and he said not at this point, "8 Pin works great, it's reliable, there has been no issues, no melting components, none of these things" "There is no major benefit, when you look at the risk vs reward benefit, there is risk there which has been proven and the reward is not really a big reward, so just go with what works, it's simple, it's reliable and customers don't have to deal with adaptors or new cables" Basically boils down to risk vs reward.

[@bgtubber]

I've had melted PCI-E connectors on two different PSUs using the same video card - MSI Suprim X RTX 3090. Both were quality PSUs - Corsair HXi 1200W. The video card was never overclocked. I'm not sure why that happened as I always make sure the connectors are plugged in all the way until they click. So yes, PCI-E connectors can and do melt.

[@arenzricodexd4409]

More like they are not adopting it because of the internet sentiment around the connector. Else AMD always proud themselves to adopt new standard ahead of competition. Like they did with PCIE 2.1 before. And that move end up causing them losing sales.

[@kevinerbs2778]

@@bgtubber The issues is that it's 2 years 4% malfunction rate on the 12V-2x6 vs 15 years 5% malfunction rate for the 8 pin pci-express plug. That number volume is massively different & proves that the 12V-2x6 has extremely poor survival rate, by comparision.

[@fracturedlife1393]

PSU manufacturers are slowly forcing AMD into it anyway. Everything has a 12Vhpwr and lost 1 or 2 PCIE. have to go beyond 1250w for 2 x eps and 4 x pcie

[@GameslordXY]

​@@bgtubber Did you use cable extensions?

[@CasualGamers]

1) framegen won't fix input lag if you have a baseline of low fps before MFG, like turning path tracing in a game such as Alan wake 2 in a 5070. The 4090 is a much faster gfx card that will handle this without input lag issues. 2) not everyone has psu with the new power connector. I much rather not have to use an adapter if I can. Plus, only high end oc models will need 3x8pin, for those I agree it makes sense to move on.

[@MikahRiveria]

Nvidia framegen thing is supposed to be used in conjunction with DLSS upscaling, first you upscale to get essentially double or more the FPS, 30FPS upscaling to 60FPS, this is real actual FPS since the game is rendering at lower than native resolution, then after you get 60fps(reasonable latency 16ms) you multi frame gen that, there will definitely be added latency, so it won't feel like garbage low frame rate latency but at least you can stomach it.....that's the end rosy goal....

[@GameslordXY]

​@@MikahRiveria Or 60 to 100. Or 90 to 150 and then slap on Multgen

[@nossy232323]

600 watt for a GPU is already crazy. Going over that? No thanks, even my RTX 4090 is already heating up my room.

[@justhomas83]

Yep under-volting my RTX4090 was the only way to keep temps in normal range.

[@barnun2525]

Not sure what 4090 you have. Mine never goes over 65c. My 3080ti got hotter than my 4090 ever did!

[@ZondaRbg]

@@barnun2525 Even if u have water cooled 4090 all that heat from radiator will go in your room and again will heating up your room. I have a heater 400W that im using right now. :D So imagine RTX 4090 can heat more than this heater...

[@barnun2525]

@@ZondaRbg I get what you saying. But my point is it runs cooler. Now what did run hotter was upgrading my cpu to intel 13900. That’s where I saw the largest change in my room and case temperatures.

[@nossy232323]

@ Sure, the cooler moves the heath from the GPU into the air of your room. There is no question that the cooler of the RTX 4090 is more powerful than the one of the RTX 3080.

[@datriaxsondor590]

The performance uplift is garbage in all those games. The only difference is, the frame generation garbage hides it in the DLSS games. The raw performance on those, is still trash.

[@sturmmagnunstein1008]

Fg doesnt even hide it that well because the frame time is absolutely awful when the base line frame rate is bad.

[@Chaos_Rifle]

i dont personally mind if AI is invoved, like anything its a tool. I do despise how everything is AI this and AI that, for no reason. its just dumb. Even more so, I despise frame generation because it looks terrible, isn't representative of the games ACTUAL state, produces input lag in certain regimes, and doesnt do what its marketed for: fix low framerates. (it works, but is a terrible experience until running at high frame rates, where its no longer necessary and just making stuff blurry or overtly wrong) Frame generation is a cancer on games, we already see some games using it by default to justify running at 30fps without it, like stalker2.

[@GameslordXY]

If anything is cancer , it's you and players like you are a cancer NOT FrameGeneration. Unwiling to accept new OPTIONAL things. I used FrameGen to enjoy Raytracing on my RX6600 in CP2077. Now if CPR introduced FSR3.1 ... . Before using it. I was almost trembling due to horror stories about FrameGen. Know what I did? More like what I didn't do. I didn't turn on onscreen stats and ENJOYED the game . Sure visuals do to FSR3 weren't super amazing picture quality, but I was able to play Very High settings with RT Reflections at 1200p(because there exist resolutions between 1080p&1440p)+ Super Resolution + AFMF2 (AMD Fluid Motion Frames 2) which AMD version of Reflex. It's not perfect, but I am very much enjoying it. For the most part. Yes, picture quality Could be much better, but after many hours of experimentation with settings and resolutions I concluded FrameGeneration is like a gift from heavens or something.

[@dahahaka]

Asking for double the performance after 2 years is kinda out of touch lol there's the limits of physics still

[@hinchlikescake7592]

Yeah thats asking a lot. Blackwell is still on the same 4nm node as Ada LL. Bumping up specs, TDP and adding GDDR7 to get that uplift. 30-40% is pretty normal generational leap unless you're jumping to a smaller process.

[@GameslordXY]

Perhaps they should wait 6-8 years. Like console manufacturers. And then release 300% better rasterisation card with DLSS Omeha + Multi Mega Ultra Tera Zeta Frame Generation where on every one frame , there is 7000 SI(Supreme Inteligence)fps

[@charleyweinhardt]

yes and he has 17,000 subscribers. This world is braindead.

[@arenzricodexd4409]

​@@GameslordXYconsole maker got money from game sales. Gpu maker does not. During the early days of 8th gen console sony said they hope the original PS4 hardware can last them even longer than PS3 did. AMD on the other hand expect console life cycle will be much shorter than it was in the past. They expect major console generation will be shorten to 5 years instead the usual 7 to 8 years. So there is definitely a conflict there. In the end we somehow get something in the middle where mid gen refresh happen.

[@kevinerbs2778]

Not really Multi-GPU is still possible with mGPU so gaining double is possible. Just that no body wants to implement it any more

[@JimCKD]

Thanx for your insight. I think most of us are "disappointed" because of the marketing "5070 is 4090" more than anything else. But if I had the money a 5090 would have been my choice. Thumbs up for the vid!

[@ram64man]

Absolutely, the use of ai features is something I am not interested in , I care about raw performance I don’t like hdr due to the huge performance hit , all this talk about requiring raytrace and procedural frames on new titles just do they can save development time and more importantly budget cuts , there problems are not our concern, people power matters and this more than ever is where consumers refusal to buy is an absolute must , or titles like Indiana Jones which demand more performance for vanity reasons by default should be hit where it matters in there wallet by refusing to buy titles that require global illumination features etc over raster performance only or all that will happen is the budget gamers will be forced into paying more and more for less

[@MikeBob2023]

Thank you, Aris. 🙏🏼 *Substantial* performance gains (gen over gen) eventually come at the cost of using *too much power,* and I think we've been at this 'diminishing returns' threshold since _Ampere._ I've been watching this trend develop for years (in both CPU and GPU design), and I'm happy to see someone exposing NVIDIA's 'bait and switch' way of trying to get around it. I have no doubt that their 50 series will sell like crazy, but I don't care one wit about frame generation so I'm officially bailing on team green.

[@tonnylins]

now we wait for the reviews

[@gglovato]

Nvidia has turned into an AI company, it was to no one surprise they would heavily push their AI-schtik to the graphic stack

[@asitrai9965]

That's not how any of this works. They are pushing AI because that is the only way to get meaningful improvements gen on gen. It would be hard to sell a card if all it was doing was 30% more than last gen.

[@MikahRiveria]

Well luckily AMD has finally added ML/AI hardware into RDNA4 for ML based upscaling, FSR4 looks good, there goes Nvidia fanboys excuse, framegen and asynchronous time warp thing can be added in the future much easily compared to RDNA3 and earlier which lacks any ML/AI based hardware.....let's multi frame 200x so that 60fps MFG 200x 12000FPS, since Sky is the limit and Nvidia wants to have a "my fake frames is higher" co ck measuring contest......Nvidia themselves seems to think Framegen is the future why not 60FPS frame gen 200000000x since fanboys everywhere is slurping everything Nvidia has, like a religious cult fanatic....

[@gglovato]

@@asitrai9965 no its not, its their solution looking for a problem, the real solution is improving raster

[@The_Noticer.]

Frame generation is not free, it is more efficient, but it does cost. Which is why the native rendering fps drops as you increase to x3 and x4

[@VGShrine]

I hope Aris can help me with this question. I already have a Seasonic Prime TX-1600 (ATX 2.5 version). Do you think that PSU could handle the RTX 5090 or do I need to move to the ATX 3.x PSU version with the native 12VHPWR/12V-2x6 connectors on the PSU side? Thanks so much for all the informative content that you release.

[@HardwareBusters]

It can surely handle it, just use or buy the proper cable for it. Ask Seasonic about it.

[@VGShrine]

@HardwareBusters thanks for the quick response. I already contacted Seasonic and they answered that I can purchase their 2 PCI connector to 12V-2x6 cable (90° version) but they also mentioned that I can still use the regular 12VHPWR cable as long as it is connected properly. Both cables are rated for 600W max.

[@richardbenke9687]

Thanks. I’ve watched a bunch of videos about the 5000 series but you’re the first to explain how power consumption plays into it.

[@bakakafka4428]

Right on the money. Thanks for speaking out when other influencers refuse to!

[@Bulu2898]

Guys if u fortunate to have a 4090 KEEP IT for the next year. This 50series is just to make money 💰.

[@notwhatitwasbefore]

For gaming on Nvidia cards the SM (Streaming Multiprocessor) are what determines how many pixels can be produced at one time and thus how many true/real/actual/normal frames it can generate. The Tensor cores do vector space math which has no traditional place in the render pipeline but because it was part of the chip design (nvidia are mostly not a gaming company) they found something for that silicon to do in some games namely dlss which requires vector space calculations to function and then they sold that to gamers as a good thing instead to just wasted die space (which it was and frankly is) This trend of developing a hardware path for other math types than those required to do gaming has continued with the RT and AI sections of silicon which have purpose in the profession market segments but then has to be software ported to the gaming market and hyped to get people to just buy it instead of think for 2 seconds and realise the die space used to do this task is at the cost of more traditional SMs that drive real performance in a game rendering workflow. So on one hand no nvidia didn't lie it does do (I suspect) what they have shown but it is a bait and switch becuase most people who have never heard of a tensor (despite that one being 7 years ago now) because they didn't major in math don't have the technical understanding of what performance is on a gpu. Without all the 'ai' fudge tech only measuring the SM perf it looks like this gen is around 5-10% faster in a traditional render pipeline which is (as a long term gamer) what 99% of gamers want/need (they just often don't have the knowledge to know that yet). I personally like native rendering with low latency which is why I have switched off of Nvidia as they are not heading in that path (doubt AMD are either but they are not so far down the wrong path yet), low internal render resolution then artifacted (no other word for it) extrapolation to fill the monitor with essentially optical flow frames to make number bigger is not usefull to me and under those terms yes Nvidia is lieing because the 5070 is margin of error a 4070s not a 4090 in genuine game performance. I hope when reviews come out and the stagnant progression is shown that the people in the market for a new GPU reject these cards and they rot on shelves because otherwise (and I sadly have been saying this for a while) then next gen of cards will be even worse (and they are) and more AI slop will be the only option.

[@GSP-76]

According to Nvidia, Ai is absolutely needed for PC gaming going forward. So blurry video, terrible input lag are necessary and you have to pay upwards of $2000 for a flagship to do it. I bought the 4090 because of the massive raster performance uplift. The 50 series is simply a joke and chock full of gimmicks.

[@sturmmagnunstein1008]

Yeah let me just get lossless scaling instead od upgrading. Next theyll just sue the dev of lossless scaling for stealing their frames!

[@themachine4008]

I have a question.... I have a pcie 4.0 mobo and the 5000 series is pcie 5.0. What performance difference will I see if I use pcie 4.0 for gaming on a 5070 vs the 4070. Ty

[@arenzricodexd4409]

None. 5070 will work the same regardless the PCIE version is 4 or 5.

[@Dex-of7ri]

50 series seems to be so disappointing...

[@Entheo_Wolf]

I can’t make use of any of that AI Fram Gen (DLSS) because I only play on VR. Fram Gen doesn’t work in VR. I also heard from another KZbinr that the 5090 is gimped by its proportionately low wattage. So overclockinng will be necessary to get an adaquet Performance uplift..

[@Entheo_Wolf]

kzbin.info/www/bejne/boa2lY2piN-hZqcsi=c8eMLu8K29V-uNm8

[@nossy232323]

Overclocking is limited. They are already close to the 600 watt.

[@Entheo_Wolf]

@@nossy232323 My mb can suply an extra 200 watt over the PCI slot to the GPU.

[@paulc5389]

Maybe it's time to switch to measuring performance in latency instead of fps. Frame interpolation is image smoothing, and there's nothing wrong with that if you like it and don't mind the artifacts it produces. But it does not increase the actual performance of a game and Nvidia are straight up lying saying otherwise. The issue isn't frame gen itself it's Nvidia lying about the performance of their cards. So ask yourself WHY they feel the need to do this. I'm sure the reviews will tell the real story. I'm guessing 10% increased performance from the 4070 to the 5070.

[@aggelos468]

So what will happen with the next series with over 600w?

[@fracturedlife1393]

I'm still cool with the 3x8. Especially when hwinfo reckons xtx spikes 700w+. The idea of a smaller connector is great, so start putting it at the back of the cards or on the end and hiding it away! At least they've made an interesting cooler though.

[@weirdodude1173]

I think it is almost a lie to say that 5070 has 4090 performance. In that one way of using it, with their multi frame generation, it might be true that it gets the same FPS number or higher than 4090 does, but that isn't total performance. I think it is pretty sleazy and it is misleading in a number of ways, but not a total lie. Those marketing people are snakes. Thumbs up for the video.

[@dikkie1120]

I wanna build a new pc do you think the rtx 5080 is worth to buy?

[@OGSuperDirty]

I share the same thoughts as you. The 5090 will be power limited and OC potential limited. With an estimated 25-30% raster difference, an overclocked 4090 can close that gap quite a bit. It brings into question the value of the 5090, especially at $2000 + tax for just the FE model. Partner cards like Asus Astral will be $2500+ tax which is simply absurd.

[@lon3wolf642]

Same as 40 series over 30 series, gains were not that big until you turned on FG.

[@yajnalgibno6536]

You think melting connectors would still be a thing in 5000 series?

[@HardwareBusters]

I don't believe it will be a major issue but I cannot know for sure.

[@user-wh8jm6sb1o]

I think if nvidia says i Want to use the technology of DLSS, ai and frame trace generation to the older card series with a software update and make a good optimization at the new cards, i think the results will be a win win for the community of gamers and the company.

[@scottxiong5844]

I too want to see how efficient the new GPU will be.

[@Aktuvor]

Thank you for your insights, im in the market for a new gpu (not 5090 mind you), and have always been happy with Nvidia. But all the news around these cards is giving some doubt about the strength of the new stuff. Do you or others think the new 5080 is even worth it or is it more the 5070-ti or 9070xt that I should look forward to?

[@HardwareBusters]

we all have to wait and see how this will go. Still expecting review samples.

[@portwest400]

Yes AMD should move along with the times, the x3 ugly power connectors should be forgotten by now

[@monsterrun]

Nvidia changed how they calculate performance...because they have only 5 fps performance gain from thr 4090...

[@AcePunk9]

dunno why but i am not comfortable when i see my 4080 uses more than 300W. and i definitely prefer it to be around 200W using dlls quality + undervolt. seeing that 40c temps makes me relax.

[@maxyakovi1423]

The rtx 5080 super will have 24 vram and it will be the smartest card to buy. Because you can overclock it and pull it above 400+ and it will approach the performance of the 5090 for much less money

[@ozzie2545]

Nvidia home to all of your fake frames....

[@XMatt5X]

I think that 30% is a good perfomance jump and that frame generation is a cancer, the industry is sick with lazy people. The major problem here is ray tracing if you turn off ray tracing everything will run very well, Nvidia choose to use software to deal with ray tracing since in the last years they were focused on IA market so just took advantage from other market segment technology instead of put some investment to find a way deal with ray tracing using pure power. I would like to see gpus with more RT cores instead of Tensor Cores

[@arenzricodexd4409]

Nvidia can put more power to RT core but it is useless since CPU will still going to be the bottleneck. Frame gen exist to bypass cpu.

[@XMatt5X]

@@arenzricodexd4409You're wrong. Even a 5600 can make a 4080S use 100% of the power, search benchmarks on google.

[@GameslordXY]

Well Digital Foundry just uplouded Deep Dive + Reaction since Richard received 5080 soon after(if not before, but bound with NDA)the pressentation. I assume the rest of them also received something . That being said , I do not know if that means the technical first hand analalysis . Benchmarks and such until I start watching. Wish me luck🤣🤣🤣

[@TerraWare]

They didnt lie but some of the marketing could be percieved as deceptive or missleading to your average non tech savy consumer. For example when Jensen said 5070 is as fast as a 4090 there was a pause and a cheer. Quite a few people that live streamed the event where like wow no way. Than after the pause Jensen said "only through the power of AI" or something like that. Did he lie? No and if the generared frames look really good that is an impressive thing of course but yeah. When it comes to marketing Nvidia are really agressive.

[@HardwareBusters]

I was there, these cheering boys and girls are usually brought by the brands. And they have instructions to scream every now and then :)

[@ChrisKadaver]

If you'll rely on frame generation to a point where game devs develop the games with frame generation in mind... In other words, program games where input lag won't matter much for the experience I can't see why you can't play at 4K with B580 if you don't mind tweaking settings. If frame generation with lossless scaling is just slightly worse than with nvidias multi frame generation I wouldn't mind using it if I'm already forced go go down that route. By that time I would've already given up on using mouse and keyboard anyway. With a stable base framerate at 30fps without spikes in frame times and lets say 3x or 4x frame generation... Well I see very little benefits of using nvidia to get like 240fps of more with an expensive graphics card. I only need high frame rates for the low respons times in shooters where graphics doesn't matter. And in high fidelity games 30fps base frame rate is fine. Atleast if I can get away not spending more than 250-350 USD.

[@ToneRetroGaming]

Anyone who has driven a turbocharged vehicle, especially an older one and compared it to a naturally aspirated vehicle with the same horsepower completely picked up on his analogy. It was spot on.

[@mitchmitchelsheimer36]

the 5090 is powerlimited, it would need aprox 730w to bring the horsepower of this card on the road, thats maybe true for the othe rmodels as well

[@Intrested_in_Intresting]

Great, thank you !!!

[@paulthebeardedonedowning6820]

I noticed some board partners showing off how they have mitigated the connector issue by having an led light up to show you you're not connected properly...... My 4090 already had this I know because I didn't connect it properly on the first install think maybe AMD is playing to the concerns of the gamers in this respect if something has a little bit of negative press steer clear of it a mistake imo because if you go try and buy a PSU today a lot have the 12VHPWR connector

[@whitecobranldaniel958]

Its a 2 in 1 for nvidia. They make a big ai cart (5090) for datacenters or mining etc. Use the same technoligy for gamers and the only they do is make marketing that ai is so good and the gamers believe. Ai use a lot of memory so fill the memory with ai frame generator for a fake result. I buy 20 roses. 2 of them are real, the others made from plastic. The salesman say: they look all the same so i change nothing about the price. That is A.I.

[@residentCJ]

Would be easy to get more Power with less Energy, just use 3NM Process 🍀🧙🏿‍♀️ like BlackW. Fiddy series is only 4NM which is old Ada Build. Raw Perf upgrade NEVER was 100% gen to gen. 3090-4090 was 70% at the best. Typically gen to gen its 40% .

[@janneliiti8898]

Nvidia could have just list the 5080 and 5090 with small price decrease. For example for VR only the pure performance counts. Personally i dont undetstand people with under 1600w powers. eight years ago you could buy a power rated like that. ask your self how many powers have you bought since then? .. still runnin my 1600w super flower

[@SlinkySlyde]

3090ti to 4090 was only like 25%. Now they can't rely on process node shrinks it's much harder to get those gains from architecture alone. Nothing to do with being lazy. I like the channel in general but I think this is a bad take overall.

[@charleyweinhardt]

1:25 EVERYONE is dumb

[@Bobovoicu]

Software Don t beat power horse ever

[@arenzricodexd4409]

But node shrink no longer allow you add more raw performance without raising power. Keep pushing for raw performance then at one point the gpu alone will be using 1000w. I doubt going with 3nm can significantly improve efficiency that much over 5nm. Most people already forgotten what happen with 28nm to 20nm transition before.

[@FellTheSky]

Answering your question: They did, but since they have no competition, it won't matter.

[@HardwareBusters]

unfortunately so true..

[@delfinigor]

The 8-pin connector is good and should not be changed. So what if there are two or three 8-pin connectors on the graphics card?

[@arenzricodexd4409]

The new power connector increase the safety over the old 8 pin. Just that they go crazy with it to push 600w on a single cable.

[@sudd3660]

of course nvidia lied, only raw performance should be in charts. those extra features is separate, most times those are not in any game we play. take my 4090 as an example, my most played game over its lifetime has no dlss or frame generation, tensor core acceleration. and in new games that i need the fps and all the features i can get to be playable, the raw performance is not enough to get me to playable anyway.....

[@R4ZORLIGHT]

Blackwell over Ada has no improvement on the "traditional" GPU side (rasterization) they only improved the AI stuff (Tensor cores/Software). Guess Nvidia is now completely on the AI train. I hope AMD can catch up sooner than later. I doubt we will see big uplifts in rasterization from Nvidia ever again in the foreseeable future. I am absolutely fine with stuff like DLSS (upscaling) but frame generation is something totally different, the latency alone is very problematic.

[@gamingfromjohnwayne]

Nvidia is like our gov.there tricksters.

[@AngryChineseWoman]

They lied by omission

[@stavroskaz]

Why i believe that there will be a hacking way to unlock dlss 4.0 on a 40x0 card and make them like 5000 series. Time show

[@GameslordXY]

No need since it's availible on ALL RTX generations. Striped down off course, but availible. Or will be. I imagine driver updates shall be required.

[@vincentvega3093]

You are confusing DLSS4 with MFG. That's not your fault, it's Nvidia's mess

[@VladThyKing]

I’m about to get the 5080 ohh yeah

[@JDD_Tech_MODS]

Stick with ur 4090.

[@XX-121]

buy amd and teach nvidia a lesson. unless you're buying xx90 card you can find an amd card that performs at that level and prob at a better price. i switched a while back and haven't had any problems and feel proud to give nvidia the finger. they can take their fake frames and shove them where the sun don't shine.

[@shmingussss]

Yes

[@ArturoTabera]

0:50 "He didn't exactly lie." Question answered. Video end.

[@Roman00744]

Sorry buddy but I will never trust the garbage 12 whatever connector that has thin wires with 600W limit being pushed to the limit over the 8pin thick wire connector that can do 300W but limited to 150W for safety reasons. One 12v connector can do 600W. Three 8pin connectors can do 900W but limited to 450W. Compromising safety to have less connectors and less wires is IDIOTIC and I will never support it, using three 8pins on my GPU now and had no problems so why would I change it.

[@HardwareBusters]

16AWG is thin?

[@Roman00744]

@@HardwareBusters Thinner than the 10 or 12AWG in the 8pin

[@Stanislav_EU-s5t]

I’m wondering why AMD didn’t show its GPU’s ! I’m surely will got for AMD,nvidia disappointed me with their AI generated crap.

[@hardware_freak_on_a_leash]

I have to agree Aris. AMD needs to start implementing 12v2x6. The new standard is here to stay. Additionally 3x8pin seems total overpowered for the RX9070. As if AMD/AIBs want to insinuate that the 9070 is a super powerful, high end GPU...

[@MrMeanh]

As someone playing on a 4k120Hz OLED TV both MFGx3 and MFGx4 are pretty much pointless, even regular FG is of limited usefulness as I prefer getting 90fps without FG to 120fps with FG. I can see MFG being useful for some players with 180Hz panels, but most likely you would need a 240Hz+ monitor to use MFG fully.

[@GameslordXY]

Have YOU driven 500HP cars? Turbo and otherwise. Nvidia should have simply made 1000W 5090(Recommended 1300W minimum,but with everything else that goes to PC 1600W - 2000W PSU),750W 5080(Recomended 1200W minimum, but 1300-1500 ideal recommended),620W 5070Ti(Recommended minimum 1050W(Thermaltake would thank them😂) , with 1200 - 1300W ideal Recommended) and 450W 5070(Recommended minimum 850W , ideal recommended 1050W-1200W) . 5090 would be priced 3500$ MSRP for FE and FE would as big as KFA2 HOF 4090, while partner cards would be up to 600mm at least and 5- 7 slots FE 5070 would be about the same size as 5090 is now, but partner cards would be around 500-550mm . Production of 1600W and 2000W PSUs would ramp up . Test benches would become EXTREMELY popular until case manufacturers regroup and adjust lol . And to be perfectly frank, we don't truly need PC cases or test benches. Components can just lie around once connected, can they not?😁