And of course there wont be a huge boycott of nvidia for all the crap they pull because gamers, especially those with money, have zero backbone... and they dont care that nvidia screws everybody because they have the money to buy it so screw everyone else, they got theirs. The richer people are, the more sociopathic they are.
@@anonymous67856 256 bit bus compared to 512 bit bus... 84 SM compared to 160 SM. 10,000 cores vs 24,000... yeah, with these specs difference, I don't expect the 5080 to be less than 60% behind a 5090.
5080 will most likely gonna be 16GB of VRAM based on the buswidth unless nvidia uses 3GB modules instead of 2GB modules. (2x8=16) or (3x8=24). I doubt nvidia will sell a 5080 with 24GB of VRAM and it being as fast a 4090. it will cannibalize and tank the sales of 4090. I can only see nvidia being fine with this if they dont have much 4090 chips left.
@@VatiWah the flip side to that is they can take the 4090 chips and shunt them into prosumer/AI cards. So it might not be too bad from NVIDIAs point of view to have the 5080 trade blows with the 4090. Aibs might be pissed but not sure if NVIDIA will worry too much about that.
*Price to performance is no longer a metric to measure computer parts anynore. It worked in the 90s. It worked in the 2000s and even into the 2010s. But the pricing has chaotically shot all over the place.*
@@asmongoldsmouth9839 What do you mean? You can measure how much it costs to render a frame on average with any given GPU. There're plenty such benchmarks.
@@asmongoldsmouth9839 Actually the release prices are more in line with what your get now. When the none super cards launched the prices were practically in line USD per frame.
Those prices seem logical, so I would expect the split to be much smaller. You leave no incentive to upsell, the 5090 has to look expensive, but the 5080 has to look batshit insane. 60% for perf for ~20% more $, 80=$1600 + 90=$2000 if they have some software to chuck in to help justify the next price hike
just buy a 4090 when they go down in price, powerlimit it to 70% and you will not need anything else until PS6 era or you buy a 4k240 TV which still don't exist (only monitors, which suck)
I see the nvidia putting the 5090 at 2500-2600 MSRP and act like it's perfectly normal to have that ridiculous price, I've got a RX 6600 and I just want something competitive in the sub 400€ mark...
It's because that is what they are charging because big AI giants want those cards too, so they aren't about to drop price for consumers since they are a tiny fraction of market share for them. Personally I'm excited for amd STRIX apu line, because that has potential to really shake up the market.
@@ericcarlton87They would have more than likely did it regardless. People would be more inclined to buy Nvidia even if AMD was better in every category. That’s mindshare for you.
I just dont understand this crowd that wants 48GB of vram on like 5060’s lmao. BUY AMD FOR VRAM for fucks sake smh you are literally youre own worst enemy banging your head against the wall and expecting different results
@@NahBNah Nobody is expecting that much , that is a moronic take . But at this point for a midrange 5070 to have 12 GB of VRAM or higher end the 5080 to have 16 GB of VRAM is a joke , only a complete fanboy can defend that . The 5090 the highest end will likely have 28 GB of VRAM . Why is it so hard to understand that people want a longer lasting card than one generation unless you buy the 90 class card but that is exactly what Nvidia wants , that is why they are putting these cards with low bit buses . When the 80 class card has a 256 bit bus that tells you all you need to know , for the 4000 series it was fine , still a bit low but ok but for the 50 series it is a joke a complete joke .
I'm a bit disappointed if those 5070 specs are accurate, I know the GDDR7 memory will be faster, but it's still underwhelming to see it still with a 12GB 192-bit bus memory configuration, also seemingly less 10 less SM's? Though those have likely been improved too so maybe it balances out? it just feels like a GDDR7 sidegrade more than an upgrade if you already have a 4070S. Maybe it's my fault for setting my expectations too high, but channels like Redgamingtech regularly posted about how Blackwell was going to blow Lovelace out of the water. We'll really have to wait for the official announcements and real benchmarks to see if upgrading the memory to the newest generation will be all it needs to be substantially different from the previous generation
If youre upgrading every gen then it’s your own fault if youre expecting 500% uplift lol, also i think what is unique about 50 series isn’t so much the uplift but the huge increase in efficiency. I REALLY would love to see cards become less nuclear reactorish and if the rumors are true about 50 series being a two slot card? Hell yea that’s right up my alley plus it’s obviously better than 6x.
you can pretty much be certain that the 5080 will run at 4090D speeds. So that tells you how much the generational upgrade will be (look at tech powerup and take the ration between the CU's.)
@@NahBNah I'm not usually the kind of person who does upgrade every generation (before i got the 4070S i had been running a 1070 since 2016) but I felt like i might have made a mistake buying a 4070S if blackwell was going to be THAT MUCH better according to leaks, so i was considering selling the 4070S to jump onto the 5070 if the leap was big enough, however these new leaks make me think that might not be necessary. I was never expecting anything crazy like a 500% uplift, but early leaks were saying it could be more like a 50% uplift, which IS fairly considerable, but the new specs make be think it's maybe going to be closer to 20%
They couldn't sell the 4080 at it's MSRP. They dropped the 4080 Super @ $999 and the card started to move. Now they are selling below MSRP and still on shelves (except the founders edition cards). Nvidia can afford to sell the card at a high MSRP especially after gaining marketshare this quarter, but why not just get good press on their gpu's for once? Realistically it won't matter they'll charge us an arm an a leg, but if we can get a 5080 for $999 on launch I think Nvidia will be on everyone's good side.
The only good news Nvidia needs is high profit margins. The only way things get better is if people started buying more AMD cards and even that won't change anything now that Nvidia is focused more on the AI market.
Prices are softening in the used market now. My 3080 just died and I've been stalking 4080 supers and 4090s on eBay. They're not moving very fast and I keep getting offers from established refurb and open box sellers from my watch list.
Ive always said this: wouldn’t Nvidia rather sell 5,000,000 gpu’s at $100 less each or 2,000,000 gpu for $100 more each? Idk why they dont think that way because i can guarantee that with price reductions will beat “fat margins” any day of the week
@@NahBNah Depends on the profit margins. That said it seems like all companies and people have this mentality perceived value is better than actual value. You don't want to sell at at lower price because you want people to FEEL like they own something of value.
Why do you say that 3 extra months of waiting does not matter since those cards won't be cheap? it's not like if you wait longer they will lower the price... but hey at least if it means the 5090 has a 512bit bus then fine.. I'd rather wait 3 months for that than get it earlier but with 448 bit.
50 SMs for GB205 now? This is getting more pathetic day by day. They might as well cancel the 50 Series except for the flagship GPU, because I guarantee the rest of them will just sit on shelves and receive tonnes of criticism due to these absolutely laughable specs. Nvidia is just wasting their own money by producing this shitty, nerfed silicon that is going to receive nothing but criticism and just sit on shelves. If Nvidia had gone all out with the 50 Series lineup, I can almost guarantee they would sell like hot cakes. Nvidia could make serious amounts of money from the 50 Series if they had learned their lesson from what they did with the 40 Series. You could argue that Nvidia makes most of their profit from AI, but there is still tonnes of profit to be made in gaming if they do it right. It's still their second largest business, and it earns Nvidia billions. Nvidia taking gaming so lightly will backfire on them eventually.
@@ZackSNetwork Fr... going from 192 to 160 is a 20% cut. If it was going from 192 to 172 that's fair since that's a 12% cut, same amount that the AD102 was cut in SM count (144 down to 128) and usually the full XX102 always gets cut down by 10-12% to save room for a TITAN card.
@@ZackSNetwork I hope that isn't true. It probably is, unfortunately. Even the 5090 isn't looking that exciting anymore. The 50 Series is going to be a complete flop, from the looks of it. 192 SMs is the full chip. 176 SMs would have been about right for the 5090. Like I said before, anything less than 168 SMs is a ripoff and nerfed. When the 50 Series were originally rumoured to be on a custom TSMC 3N node, the full GB202 was said to have 216 SMs, and the 5090 was said to have 192 SMs out of the full 216 SMs. It's a massive shame that this is no longer the case. There's no denying that if Nvidia had gone the bleeding edge route with the 50 Series, the profit margins would have been way higher. They are not even close to maximising the profit they can make from gaming. They are throwing so much money away in the gaming segment by cutting corners this much.
At this point, Nvidia should sell their gaming business to another company. It's a billion market, and it definitely has the potential to grow way bigger, and despite that, they don't give a shit about it. This whole AI crap is a bubble, and it eventually has to burst. Once or if the bubble bursts, the demand for AI accelerators will plummet. I am just hoping that day arrives soon.
@@laszlozsurka8991that’s not a fair cut historically. Kepler was 100% core. Maxwell was 92% core. Pascal was 93% core 92% bus. Turing was 94% core 92% bus. Ampere was 95-98% core (3080Ti-3090). Lovelace is 89% core 75% l2$. If we exclude the Lovelace and Kepler outliers we get 176-188 SMs and 480bit memory bus. Nvidia wants to drop down to 168 SMs or less because if there is a problem with a gpc they can cut it out and use the die. But at the size of the die there may be a couple other defects on a poorer yielding part so they would want to remove some more SMs and a bit of memory bus. I think 160 SMs and 448bit bus is likely as a result. 4090 loses 1 GPC, 4 more SMs, and 24MB l2$ which is pretty sizable in the die.
So 5080 is going to improve performance doing things that do not increase costs, while 5090 is going to increase costs a lot and then disable some of that to create consumer variant to lessen threat to professional cards priced over twice the 5090 price using same silicon.
5060/5060Ti with 128bit and 8Gb again. Another generation of Meh-midrange. The 5060 might be okay if the price is right (but it wont be). If they would just bump the bus for the Ti version to 160bit, you would get 10Gb vram which would be just enough to help with those otherwise terrible 1% lows frame rates. And then clamshell the memory for optional 20Gb versions for the production and creative users on budgets. Simple. If those specs hold, as a midrange user with a 4060Ti 16Gb, I see no reason to buy anything from the 50 series.
@@NahBNah a 160bit bus and 10Gb of vram would really help with the gaming performance issues, but you're right, AMD might still be a better choice. However, in non gaming applications that use features like ray tracing, cuda, and tensor cores, Nvidia is still the much better option, even in the midrange market. There's a lot more to consider in GPU abilities these days than just raster performance.
We are at the end of 12 GB of VRAM being sufficient for gaming. They do this to upsell, though. It is an old tactic. They want you to HAVE TO buy a 5080.
Gddr7 is way faster and more expensive than the gddr6 of last gen and rdna 4, why would they give you more especially when amd can't even keep up let alone compete
Its not like nvidia cares because the fascist tech bros at meta and x are buying billions dollars worth of their overpriced B200 AI GPUs. Nvidia doesnt give a damn about gamers, they threw gamers under the bus during cryptos, now they are doing it again with AI. Nvidia hates the people that made them who they are and gamers are too stupid to realize it.
@@Andrew-je8guMe too, but for 2500 its going to be worth it for sure. Im going with Zotac, they did the best job with 40 series, their cards consume least power for every given core clock.
This makes sense, Nvidia needs to give China the 5080. They already know that fan boys will pay 2K+ for a 90 class card so it only makes sense to price gauge before the scalpers do. Honestly it's a win-win because the people complaining can just buy lower tier cards.
I have a 3080 as well. But in 2025, we won't be able to play on better settings in games that well. 10-12 GB of VRAM is quickly falling short. I told everyone that 10-12 GB of VRAM is perfect for the next 3-5 years, and I was right. But we are at the end of that era.
I have a 3080 as well. It was great from 2020 to 2024. But 10-12 GB of VRAM is becoming a minimum requirement in high-end games now. 2025 and onward is going to thwart our game settings.
the price will depends on AMD if they can fill the gap in the RT side, the RT price tag on the nvidia side will be a non cence. there is still the DLSS price tag to cover. but because nvidia don't care about the desktop market so much. they are ready to put a big price on their GPU , they will make the same amount of money with lower GPU sold (and more AI product sold) we'll see but I don't have a good feeling for the pricing side.
I think pricing wise that they wont be much more expensive maybe like 10% more than the 40 series except the 5090 but that is why the performance uplift is sh!t . Not worth it at all .
So 5080 and below = almost no performance improvement worth the upgrade and same amount of V-Ram. That mean, 70 series still with 12GB and 60 still with 8GB. Noice. EDIT: 70/60 Class, my bad.
oc there will be speed improvements, just improvements in cache speed and latency will make massive performance improvements without any increases in cuda cores counts.
It does depend. If they shrink the L2 to a typical regular size and create a large L3 pool, there could be a "massive" perf uplift from that alone. If they do not increase L1 size and focus on speed and latency improvements, well that's where IPC generally is
$2250, sounds right, but that's with tax, $1830 tax excluded (american way) seems way too little, more like $2000-2100 without tax and $2460-2580 with tax
Im super happy with my Rtx 3060. Runs every games with decent setting and fps. Maybe if rtx 5060 comes with 16GB and 3x performance uplift compares to rtx 3060.
Not to want to disappoint you but the 4060 was only 15% better than 3060. The 5060 might be 30% better then that if nvidia is feeling generous. I will be thrilled if they give it more than 8gb vram
@@evers6214 usually Nvidia gives 2x perf uplift with every 2 generation but after they focusing more on Ray trace the raw performance getting less and less.
i really hope they give me the full GB202 512 bit die for my 5090. when i pay so much money, i hope for the best version of this die they can give us. i expect around 1999$ MSRP (possibly even lower, so they can claim they do something FOR the gamers), thats 400$ more than the 4090, which is already a massive 25% increase in price. all the 3000$ speculations are ridiculous, the 5090 will most likely be a massive 40-60%+ raw performance increase over 4090 (even if they cut down the GB202 die to the 448bit bus version), depending on workload. they cant increase price by 50% as well bc there would be zero price to performance generational gain, thats not going to happen for the flagship. remember the 4090 was basically the same MSRP as 3090, it was 'only' a 100$ increase over the 3090 to be exact .and all that while they increased 4090 efficiency A LOT, increased performance by - in some cases like 100% over the 3090 (path tracing comes to mind) - overall around 60%+ over the 3090 AND they even implemented a new feature which was frame generation. So if they sell the 5090 for 50% more money as some suggest, or even more, and make it around 50% faster than 4090, it wont get any of the praise as the 4090. It will be really bad press. It doesnt make sense for them to make such a high MSRP. It is WAY smarter to put it at like 1799, 1899 or even 1999$ absolute tops. This way Nvidia can say they only increased MSRP 'slightly' (if it will be priced at like 1799$ or so) bc of inflation and its not their fault that the street price is 2500$, which could very well be the actual street prices if Nvidia keeps the stock low enough, which they will likely do.
Here's a wild thought Nvidia's greed bites them in the ass finally. The 5080 going for more than $1200 is crazy there's no way you can explain that kind of expense to your wife.
I hear ya that things can change. Any chance in YOUR thoughts/guess that we'll get a version of the 5080 that has larger bus/more than 16GB VRAM? For 2024, at that level, 16GB seems weak for ai work and future-proofing. Total understanding they may be wanting to keep their gaming cards from cannibalizing their ai-specific cards?
*The 5080 will be similar performance to a 4090. THAT is what matters. I don't care about gaps. I don't care about any metrics to measure by. The 5080 will be MUCH cheaper than a 4090 and lower power consumption, but similar performance.*
@@blabla-fw9ix did you heard about the console cycle? The game industry move around that. Now we are in the ps5/seriesX cycle use 12gb of 16gbvram shared system for the ps5. Before when 8gb gpu are now 1080p cards they were 1440p like 2070super even the unlucky 3070. There is a video here of a 3080ti 12gb cant run well the latest rachet and clank game with ray tracing at 4k very bad 1%low fps the gpu is powerfull but 12gb is not enough for that game. 16gbs are ok for 4k until the cycle repeat again.
Are you really saying a $1,400 MSRP is optimistic for the 5080!? That's literally DOUBLE the the MSRP of the 3080, so realistically 5080s will be selling for around 2 grand, likely more? And that's Optimistically? Wow. PC gaming will be dead (or totally stagnated) in 3 generations tops, at that rate. RTX 7080s for the low price of $4K
5090, the only card we can look forward to, and it might be a paper launch. 5080 is going to be a nicer, less expensive 4090, but it's not a heavy hitter like we need to push 4K 240 OLEDs. AMD is looking to bring better value to gaming, but they are not pushing boundaries. I can see 5090 selling for $3000 plus from scalpers, it's the only card really worth paying more for. 5080 could be $1500 which makes some sense because it's only half the card 5090 is. Thumbs up.
Hardware rendering will be phased out eventually, dlss itself is AI render8ng and will literally be how all frames will be rendered, said nvidia anyways
Considering Nvidia is using 4nm I don't think there will be a price increase. 4000 series has sold not that well, and they really have reached the maximum of what can be charged. I'm itching to upgrade to a 5090 but anything more than 4090 MSRP and I'm out.
I suspect the 5090 will be around £1999-£2199 and I also suspect the 5080 will slot in and replace the 4090 at its £1599 msrp, and the 4090 will be dropped around £200-£300
I suspect they will price it in line with performance. They saw they couldn't ask too much for an 80 class card, while for the 90 they probably could have asked more. Now they make the performance difference between the 80 and 90 class even larger, so I think 1 K USD for the 80 class card and 2K USD for the 90 is possible.
I think they’ve dropped the 50 class cards for good because its easier to sell a $300 card as a 60 class. I don’t think Nvidia wants to compete under $300 anymore. We will eternally have 60 class cards with 50 class performance until eventually the 60 gets dropped and the 70 class turns into garbage.
That 5090 has to cost like $2500-3000. Unless Nvidia starts selling direct and ditches partners. They’ll sell out of all of them too because they’re essentially discount AI hardware.
Honestly if you want 16GB of VRAM and don't want to spend 1k for it , your best route would be to get an AMD card , the 60 class cards are pretty sad in terms of performance .
5060 and 5060 ti yet again going to be so bad 🤦♂️ 128 bit bus and 8gb vram is outrageous considering most if not all games need 10gb vram at 1080p and we've seen that since 2022 and later. It's almost like Nvidia is killing off the lower and mid tier GPUs because frame generation is basically useless on 8gb GPUs. In other words imma stick with my 3060 Ti And it's 256 bit bus if this is all true specifications ffs 🤔
The only way to fix this is speak with your wallet. If you buy it, they will keep doing it. If you refuse to buy it, they will be forced to change. Stop giving them money to shaft you with. I’ll keep my 3080Ti and nvidia can get bent.
Well 12GB VRAM GPU are going to have issues in late 2025. 16GB VRAM GPU’s will be good until around 2027 before having to lower settings. Honestly each tier needs at least 4 gb more VRAM. The 5060 should have at least 12GB VRaM on 192 bit memory bus The 5070 should have at least 16GB VRAM on 256 bit memory bus The 5080 should have at least 20GB VRAM on a 328 bit memory bus If I don’t see this then I’m either going to do three things. 1. Buy an AMD GPU 2. Buy the previous gen 4000 series for cheap (might go for 4090 if it hits the $1,000 mark) 3. Hit the used market and support the people and not corporations.
Good informative video for leaks that are rumored . The 5090 should have been a 4090 ti, which was never made. They had all of this planned out. Of course, it wouldn't have that many cuda cores or a 512 bit gddr7. Why did the other 4000 series get ti's but not the 4:014:01 4090 😮 As far as price goes, sure they will be more expensive than the previous generation. A 5090 could be around $2000-$2200, but that remains to be seen. I'll just stay with my 4090 for a few more years or even longer.
Heck yeah! I plan on keeping my 4090 until like 2027 These stupid 240Hz + crap is...well...crap. just to feel smoother in microseconds...I shall pass. Ones actual eyes can't tell the difference between 90 Hz and 360 Hz 😂 people actually fall for that snake oil You are smart!!! Don't waste money
@@ZackSNetwork not true at all it's lots of games 1080ti still can run games great ones that are optimized great or not demanding games Where did you get that lie from
Intel Battlemage will likely be out Q4 2024, people put too much emphasis on when products release, we are in a long economic recession, consumers are not going to buy new products. Consumers want deals when money is tight, Amd is on schedule with their releases but very few people have the money to buy them. The PC market is going to be in a rut for several years, and budget products will be key but I dont see big numbers for sales. Also all these pundits say Nvidia 4000 sells well, so why are 4060/ti and 4070/ti/super always in stock? 4000 series is selling well below what 3000 ssries did. So its very disingenuous to present it like that.
We miss the Christmas shopping season, this means a loss in sales. Mind you I would be getting a 60 series card (really a 50 series with more cores) if it has at least 12gb ram.
skalpers gonna have a field day for us gamers that want the best, you can speculate as much as you want but i am going to say MSRP will be around 2000USD for the 5090 skalpers get there hand on this and they will then i can see 5000USD its a sad truth personally i am happy with the 4090 took a while but i finally have one and not going to stress over the 5090 unless its 2 times the performance of the 4090.
I know Nvidia is greedy but there is no way the 5080 can cost that much for the VRAM and for the performance it has , I mean if it costs as much as a 4090 the people would much rather buy a 4090 . It will be like 1200 bucks likely but still terrible for the performance it has .
Spot on pricing. What Blabla doesn't understand is the 5080 has to be batshit insanely overpriced to create a reason to upsell to the next tier. Logic has no place here. 1.6k and 2k is my bet too.
At this point it is likely all 5080 models will be gimped again with a 256 bit memory bus and all models below 5080 will go down hill. I have a wee bit of a byte of hope that Nvidia will release a Workstation 5090 with a 512 bus, but there will be plenty of refurbished stock also where you can start tinkering with. Get out the soddering iron and thermal pads.
The 4090 is so powerful, even having a 4k 120 TV, that I would just justify a 5090 for VR and flight sim. I was excited for it but now I am considering investing in a 7.2 audio system instead... Games needing more power 4090 are not coming out and no I don't care about raytracing at all.
Is GB207 the 5050? That looks like it could be beat by an amd igpu before the 60 series comes out. The lineup looks fairly disappointing based on the specs but it is a new arch so perhaps it will be better then this slide shows but so far the only good thing NVIDIA seems to be doing is GDDR7.
Why bother with so much rumors and what we do not know ? It is NVIDIA. It will be super expensive, hot as the sun, heavy as led 😊 Why even anticipate all this? It will not change the specs, delivery dates or anything else ... It can only arise disappointment 😊
5090 (skimmed GB202 - 448bit - 28GB) - 5090 Ti (full die - if they ever feel the need - 32GB) - 5080 Ti (further skimmed GB202 - 320bit - again, if they ever feel the need - 20GB) 5080 (full die GB203 - 256bit - 16GB) - 5070 Ti (skimmed GB203 - 256bit - wishful thinking? - 16GB) 5070 (full die GB205 - 192bit - 12GB) - 5060 Ti (skimmed GB205 - 12GB) 5060 (full die GB206 - 128bit - hopefully with 3GB memory modules for still 12GB VRAM? - otherwise 8GB 🤮) And 5050 Laptop (GB207 - 8GB entry - yay). I expect this, just because it pleases my expectation OCD lol 🤣
Cant wait to upgrade my 3060 12GB to a 5060 8GB ddr6 🤡 🤣 The huge generational uplift is reserved for the 3000$ card that will be a paper launch and distributed to tech youtubers and other shills and maybe to less than 1% of the gaming playerbase world wide.
Like yeah why not the 5080 have the up in BW like the 5090! 5080 Should be 384-bit and NOT 256-bit again! My 3070 was nerfed on the amount of RAM should be 3070 16GB like they did over seas. NIVIDA Stop Nerfing the GPUS . lol
You really think the 5080 will cost even more than last gen? Why would nvidia make the same mistake twice in a row ? The 4080 sold poorly at its ridiculous price and the 4080 super has done much better while being $200usd cheaper ( the 4080 super can found $300-400aud below msrp here in Australia) no one wants an 80 class card above $1,000....especially if the specs are accurate and it's going to be way beyond the 5090. It needs to be $899 imo. But I think they'll price it at $999. I can't imagine nvidia being dumb enough to complete ruin the 80 series class. Twice in a row would just about do that, it's in no many land above $1,000. Not powerful enough and not cheap enough and only 16gb, I'll be shocked If they even try to charge $1,200. It would be hilarious if they do 😂 annoying because I was somewhat looking forward to it before the leaked specs but it will be funny to see the reactions from media and consumers lol
If Navi 48 best chip have 256 bit bus and the same VRAM than 7900Xt and Leaks say performance from 7900XT to 7900XTX. The price is importent but can the RNDA 4 Raytracing power be the same like a 4070 Ti super or even 4080. For 599$ i am total in and pay my 700-750€ for Custom. Nvida way to greed for 2000 RTX to 4000 RTX and this strategies to give low amount of vram to force customers to buy every 2 years a new gpu no thx not with me.
Glad I downgraded several years ago from 4k back to 1440p, Don't care what they release with insane prices, I'll still get well over 80fps in every game I play. 5080 looks like another skip generation and they are not going to entice me to move up to 90 class with stupid pricing.
I don't think nvidia can charge more than $1k for the 5080 unless it has more than 16gb of vram somehow. I would expect the gpu to last until at least 2029 or 2030. With next gen consoles maybe around 2 years away I see vram usage jumping in the near future. I'm not spending more than $1k on a gpu then being forced to turn down textures in just 2 years cause next gen games and consoles will want more than 16gb.
they don't have to change the lineup if they charge $1200. 5090>4090=5080 but less ram>4080 super(no card competing with another). the 5080 is probably a 4090 in performance but less ram and $400 cheaper. personally the 5080 looks like a bad card no matter what because of its ram, If it beats the 4090 in path tracing I will revise my opinion as that speaks to a lot of improvement and optimization so that it can run on less ram.
@@TV-xv1le If NVIDIA puts out a $2,200 5090... there will be people who buy it. Remember the 3090 Ti? That GPU costed $2,000 yet people still bought it. Plus AMD is not even competing in the high end so what else you're gonna buy anyway if you want a solid 4K setup? The 5080 will be your only choice so expect NVIDIA to take full advantage of that. Sure you can buy a 4090 used but by the time the 4090 comes down below $1,600... the 5080 will be long launched.
Considering they're giving you gddr7 when no one else is and nothing from amd or intel will even come close to competing they can more or less charge what they want. Esp if it has a big RT uplift
Can Amd Stop launching the XTX first? or atleaste the name, like were going to have the same problems as RX7000s where there is a 7900"gre" a 7900xt and a 7900xtx Like wtf AMD? Just call it a 7800 7800xt and 7900 and then release your top end XT later if you really want to compete with the top end of the stack with nvidia. They keep doing this gpu number class mindshare marketing stupidity that isnt working
Given the size gap between the GB202 and GB203, there is so much room for cutdown GB202 to deliver a 448-bit 5090, a 480-bit 5090ti, a 384-bit config, and a 320-bit config before dropping into the GB203. Given the risks of defects that can lead to such drastic cut downs, it is safe for Nvidia to have both 5090 and 5080 line ups to be handled by GB202. This opens GB203 for 5070, GB205 for 5060ti, GB206 for 5060, and GB207 for 5050. My thoughts: GB202 - 448 bit -> 5090: $2,200 GB202 - 320 bit -> 5080: $1,500 (RTX 4090 replacement) GB203 - 224-bit -> 5070: $1,000 (RTX 4080 replacement) GB205 - 192-bit -> 5060ti: $700 (RTX 4070 replacement) GB206 - 128-bit -> 5060: $500 (RTX 4060 replacement) GB207 - 128-bit -> 5050: $300 (Budget tier, budget gaming laptop) Honestly, Nvidia would be better setting the GB206 as the 5050 and turn GB2070 into the 5040 because there are plenty of low numbers. I am looking at speculated VRAM bandwidth, and if the RTX 5080 will be an RTX 4090 replacement, it needs to be like a GB202-250 cutdown just for bandwidth. And as far as prices, I feel optimistic.
With those performances ? Not a chance in hell regardless of Nvidia's greed . I wouldn't pay over 1k for that card and even that seems too much for it , its a joke . It will likely be 1200 bucks .
So instead of an 30% gap between the 80 and 90 class we're gonna see a 60-80% gap between the 80 and 90 class... fantastic...
And of course there wont be a huge boycott of nvidia for all the crap they pull because gamers, especially those with money, have zero backbone... and they dont care that nvidia screws everybody because they have the money to buy it so screw everyone else, they got theirs. The richer people are, the more sociopathic they are.
Cores don't scale at this point. The 4090 should be 60% better than the 4080 by TFLOPs, but it's not
@@anonymous67856 256 bit bus compared to 512 bit bus... 84 SM compared to 160 SM. 10,000 cores vs 24,000... yeah, with these specs difference, I don't expect the 5080 to be less than 60% behind a 5090.
5080 will most likely gonna be 16GB of VRAM based on the buswidth unless nvidia uses 3GB modules instead of 2GB modules. (2x8=16) or (3x8=24). I doubt nvidia will sell a 5080 with 24GB of VRAM and it being as fast a 4090. it will cannibalize and tank the sales of 4090. I can only see nvidia being fine with this if they dont have much 4090 chips left.
@@VatiWah the flip side to that is they can take the 4090 chips and shunt them into prosumer/AI cards. So it might not be too bad from NVIDIAs point of view to have the 5080 trade blows with the 4090. Aibs might be pissed but not sure if NVIDIA will worry too much about that.
So the 5080 will be the worst price to performance again...
@@SeanM487 If the leaked specs are true that seems highly unlikely. I lost any optimism regarding GPU launches given the recent years.
*Price to performance is no longer a metric to measure computer parts anynore. It worked in the 90s. It worked in the 2000s and even into the 2010s. But the pricing has chaotically shot all over the place.*
@@asmongoldsmouth9839 What do you mean? You can measure how much it costs to render a frame on average with any given GPU. There're plenty such benchmarks.
@@asmongoldsmouth9839 Actually the release prices are more in line with what your get now. When the none super cards launched the prices were practically in line USD per frame.
Classic nGreeDia. Why change a policy if it's not broken? (For the simps).
5080 - 1199$
5090 - 1799$
Agreed i think there is too much fear driving people to say like 5k for the 5090 lol. I really think its going to be around that ballpark too
Those prices seem logical, so I would expect the split to be much smaller. You leave no incentive to upsell, the 5090 has to look expensive, but the 5080 has to look batshit insane. 60% for perf for ~20% more $, 80=$1600 + 90=$2000 if they have some software to chuck in to help justify the next price hike
I will wait till 5090 would drop it's cost to 1000$ and then would buy.
@@stanpikaliri1621 bre prepared to wait for a decade lol, even 4090 price hasn't given up yet
@@stanpikaliri1621 you will have to wait 2.5-3 years for that 😂
Hard to be excited for Nvidia cards unless you’re buying the highest end 90 card.
just buy a 4090 when they go down in price, powerlimit it to 70% and you will not need anything else until PS6 era or you buy a 4k240 TV which still don't exist (only monitors, which suck)
@@dieglhix What a terrible take
NVidia has you exactly where they want you 😞😞😞😞😞😞😞
@@xblur17 optimal advise
Why power limit it if it's the most efficient card ,just undervolt it and declock 5 percent my 4070 super undervolted runs on 150 w@@dieglhix
Value died at the GTX 1080 series.
I see the nvidia putting the 5090 at 2500-2600 MSRP and act like it's perfectly normal to have that ridiculous price, I've got a RX 6600 and I just want something competitive in the sub 400€ mark...
No you dont 🤪
@@attepatte8485 yes he does :)
@@attepatte8485 I don't what?
Can't you get a 6800xt for that price? Should give quite a performance boost over the 6600xt
It's because that is what they are charging because big AI giants want those cards too, so they aren't about to drop price for consumers since they are a tiny fraction of market share for them. Personally I'm excited for amd STRIX apu line, because that has potential to really shake up the market.
I really hope Intel sticks with their GPU line.
Think of what they can do in five years.
Be good for competition
They cant do anything. They cant catch ada lovelace in 5 years when is already almost 2 years old now.
Go bankrupt
Nvidia needs competition. AMD isn’t releasing a high end GPU. Nvidia seems to be able to do and charge whatever they want
@@ericcarlton87They would have more than likely did it regardless. People would be more inclined to buy Nvidia even if AMD was better in every category. That’s mindshare for you.
i swear if the 5070 ti/super/base doesnt have 16gb...
Probably wont
Silent!
You will buy anyway.
Then cry online in your corner how we took your money, see you within 2 years to repeat
Signed Nvidia
I just dont understand this crowd that wants 48GB of vram on like 5060’s lmao. BUY AMD FOR VRAM for fucks sake smh you are literally youre own worst enemy banging your head against the wall and expecting different results
@@NahBNah it's not buy amd for vram, it's that nvidia are deliberate reducing vram so people buy the high end 5080
@@NahBNah Nobody is expecting that much , that is a moronic take . But at this point for a midrange 5070 to have 12 GB of VRAM or higher end the 5080 to have 16 GB of VRAM is a joke , only a complete fanboy can defend that . The 5090 the highest end will likely have 28 GB of VRAM . Why is it so hard to understand that people want a longer lasting card than one generation unless you buy the 90 class card but that is exactly what Nvidia wants , that is why they are putting these cards with low bit buses .
When the 80 class card has a 256 bit bus that tells you all you need to know , for the 4000 series it was fine , still a bit low but ok but for the 50 series it is a joke a complete joke .
I'm a bit disappointed if those 5070 specs are accurate, I know the GDDR7 memory will be faster, but it's still underwhelming to see it still with a 12GB 192-bit bus memory configuration, also seemingly less 10 less SM's? Though those have likely been improved too so maybe it balances out?
it just feels like a GDDR7 sidegrade more than an upgrade if you already have a 4070S. Maybe it's my fault for setting my expectations too high, but channels like Redgamingtech regularly posted about how Blackwell was going to blow Lovelace out of the water.
We'll really have to wait for the official announcements and real benchmarks to see if upgrading the memory to the newest generation will be all it needs to be substantially different from the previous generation
theoretically they could give it 18gb, bc with gddr7 there is supposed to be 3gb modules too now, but we all know its gonna have 12gb again...
If youre upgrading every gen then it’s your own fault if youre expecting 500% uplift lol, also i think what is unique about 50 series isn’t so much the uplift but the huge increase in efficiency. I REALLY would love to see cards become less nuclear reactorish and if the rumors are true about 50 series being a two slot card? Hell yea that’s right up my alley plus it’s obviously better than 6x.
you can pretty much be certain that the 5080 will run at 4090D speeds. So that tells you how much the generational upgrade will be (look at tech powerup and take the ration between the CU's.)
@@NahBNah I'm not usually the kind of person who does upgrade every generation (before i got the 4070S i had been running a 1070 since 2016) but I felt like i might have made a mistake buying a 4070S if blackwell was going to be THAT MUCH better according to leaks, so i was considering selling the 4070S to jump onto the 5070 if the leap was big enough, however these new leaks make me think that might not be necessary.
I was never expecting anything crazy like a 500% uplift, but early leaks were saying it could be more like a 50% uplift, which IS fairly considerable, but the new specs make be think it's maybe going to be closer to 20%
They couldn't sell the 4080 at it's MSRP. They dropped the 4080 Super @ $999 and the card started to move. Now they are selling below MSRP and still on shelves (except the founders edition cards).
Nvidia can afford to sell the card at a high MSRP especially after gaining marketshare this quarter, but why not just get good press on their gpu's for once? Realistically it won't matter they'll charge us an arm an a leg, but if we can get a 5080 for $999 on launch I think Nvidia will be on everyone's good side.
I'm thinking that as well, the 5090 will cost an arm and a leg but the 4080 sales showed Nvidia that there is a leash they have
The only good news Nvidia needs is high profit margins. The only way things get better is if people started buying more AMD cards and even that won't change anything now that Nvidia is focused more on the AI market.
Prices are softening in the used market now. My 3080 just died and I've been stalking 4080 supers and 4090s on eBay. They're not moving very fast and I keep getting offers from established refurb and open box sellers from my watch list.
Ive always said this: wouldn’t Nvidia rather sell 5,000,000 gpu’s at $100 less each or 2,000,000 gpu for $100 more each? Idk why they dont think that way because i can guarantee that with price reductions will beat “fat margins” any day of the week
@@NahBNah Depends on the profit margins. That said it seems like all companies and people have this mentality perceived value is better than actual value.
You don't want to sell at at lower price because you want people to FEEL like they own something of value.
Why do you say that 3 extra months of waiting does not matter since those cards won't be cheap? it's not like if you wait longer they will lower the price... but hey at least if it means the 5090 has a 512bit bus then fine.. I'd rather wait 3 months for that than get it earlier but with 448 bit.
50 SMs for GB205 now? This is getting more pathetic day by day. They might as well cancel the 50 Series except for the flagship GPU, because I guarantee the rest of them will just sit on shelves and receive tonnes of criticism due to these absolutely laughable specs. Nvidia is just wasting their own money by producing this shitty, nerfed silicon that is going to receive nothing but criticism and just sit on shelves. If Nvidia had gone all out with the 50 Series lineup, I can almost guarantee they would sell like hot cakes. Nvidia could make serious amounts of money from the 50 Series if they had learned their lesson from what they did with the 40 Series. You could argue that Nvidia makes most of their profit from AI, but there is still tonnes of profit to be made in gaming if they do it right. It's still their second largest business, and it earns Nvidia billions. Nvidia taking gaming so lightly will backfire on them eventually.
Look how cut down the 5090 will be. Paul said 150 to 160 SM at most. The full die is 192 SM.
@@ZackSNetwork Fr... going from 192 to 160 is a 20% cut. If it was going from 192 to 172 that's fair since that's a 12% cut, same amount that the AD102 was cut in SM count (144 down to 128) and usually the full XX102 always gets cut down by 10-12% to save room for a TITAN card.
@@ZackSNetwork I hope that isn't true. It probably is, unfortunately. Even the 5090 isn't looking that exciting anymore. The 50 Series is going to be a complete flop, from the looks of it. 192 SMs is the full chip. 176 SMs would have been about right for the 5090. Like I said before, anything less than 168 SMs is a ripoff and nerfed. When the 50 Series were originally rumoured to be on a custom TSMC 3N node, the full GB202 was said to have 216 SMs, and the 5090 was said to have 192 SMs out of the full 216 SMs. It's a massive shame that this is no longer the case. There's no denying that if Nvidia had gone the bleeding edge route with the 50 Series, the profit margins would have been way higher. They are not even close to maximising the profit they can make from gaming. They are throwing so much money away in the gaming segment by cutting corners this much.
At this point, Nvidia should sell their gaming business to another company. It's a billion market, and it definitely has the potential to grow way bigger, and despite that, they don't give a shit about it. This whole AI crap is a bubble, and it eventually has to burst. Once or if the bubble bursts, the demand for AI accelerators will plummet. I am just hoping that day arrives soon.
@@laszlozsurka8991that’s not a fair cut historically. Kepler was 100% core. Maxwell was 92% core. Pascal was 93% core 92% bus. Turing was 94% core 92% bus. Ampere was 95-98% core (3080Ti-3090). Lovelace is 89% core 75% l2$. If we exclude the Lovelace and Kepler outliers we get 176-188 SMs and 480bit memory bus.
Nvidia wants to drop down to 168 SMs or less because if there is a problem with a gpc they can cut it out and use the die. But at the size of the die there may be a couple other defects on a poorer yielding part so they would want to remove some more SMs and a bit of memory bus. I think 160 SMs and 448bit bus is likely as a result. 4090 loses 1 GPC, 4 more SMs, and 24MB l2$ which is pretty sizable in the die.
You won't be able to get a 5080 for under 1400, especially the first few months
So 5080 is going to improve performance doing things that do not increase costs, while 5090 is going to increase costs a lot and then disable some of that to create consumer variant to lessen threat to professional cards priced over twice the 5090 price using same silicon.
5060/5060Ti with 128bit and 8Gb again. Another generation of Meh-midrange. The 5060 might be okay if the price is right (but it wont be). If they would just bump the bus for the Ti version to 160bit, you would get 10Gb vram which would be just enough to help with those otherwise terrible 1% lows frame rates. And then clamshell the memory for optional 20Gb versions for the production and creative users on budgets. Simple. If those specs hold, as a midrange user with a 4060Ti 16Gb, I see no reason to buy anything from the 50 series.
And if youre buying a midrange card from nvidia youre dumb because AMD is the better choice every single time but yet people want to screech about it
@@NahBNah a 160bit bus and 10Gb of vram would really help with the gaming performance issues, but you're right, AMD might still be a better choice. However, in non gaming applications that use features like ray tracing, cuda, and tensor cores, Nvidia is still the much better option, even in the midrange market. There's a lot more to consider in GPU abilities these days than just raster performance.
5070 12gb vram is baaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaad
lol translate to english just adds aaaaa
We are at the end of 12 GB of VRAM being sufficient for gaming.
They do this to upsell, though. It is an old tactic. They want you to HAVE TO buy a 5080.
Anyone who thinks that the 4070 or 5070 needs more than 12 GB VRAM for gaming is smoking crack
Gddr7 is way faster and more expensive than the gddr6 of last gen and rdna 4, why would they give you more especially when amd can't even keep up let alone compete
@rodiculous9464 gddr7 yes its faster but we need more capacity. If one let me choose between 8 ddr5 or 16 ddr4 I will choose ddr4
It is going to be a paper launch anyway
@@SeanM487 The initial MSRP is gonna be so high even Scalpers are gonna see it as risky.
Its not like nvidia cares because the fascist tech bros at meta and x are buying billions dollars worth of their overpriced B200 AI GPUs. Nvidia doesnt give a damn about gamers, they threw gamers under the bus during cryptos, now they are doing it again with AI. Nvidia hates the people that made them who they are and gamers are too stupid to realize it.
@@sir1junior same with 4090, I paid 1700 Q4 2023. Now it's at 2300 USD.
My max is 2.5k for 5090 if they hit 3.5k i will go with 5080...
@@Andrew-je8guMe too, but for 2500 its going to be worth it for sure. Im going with Zotac, they did the best job with 40 series, their cards consume least power for every given core clock.
This makes sense, Nvidia needs to give China the 5080. They already know that fan boys will pay 2K+ for a 90 class card so it only makes sense to price gauge before the scalpers do. Honestly it's a win-win because the people complaining can just buy lower tier cards.
I think you mean people can just buy AMD
Got me 3080 and unless i can get 4090 level performance for under 1000usd in my country imma hold out another generation
I have a 3080 as well. But in 2025, we won't be able to play on better settings in games that well. 10-12 GB of VRAM is quickly falling short.
I told everyone that 10-12 GB of VRAM is perfect for the next 3-5 years, and I was right. But we are at the end of that era.
I have a 3080 as well. It was great from 2020 to 2024. But 10-12 GB of VRAM is becoming a minimum requirement in high-end games now.
2025 and onward is going to thwart our game settings.
the price will depends on AMD
if they can fill the gap in the RT side, the RT price tag on the nvidia side will be a non cence.
there is still the DLSS price tag to cover.
but because nvidia don't care about the desktop market so much. they are ready to put a big price on their GPU , they will make the same amount of money with lower GPU sold (and more AI product sold)
we'll see but I don't have a good feeling for the pricing side.
I think pricing wise that they wont be much more expensive maybe like 10% more than the 40 series except the 5090 but that is why the performance uplift is sh!t . Not worth it at all .
Could you imagine coughing 1k usd for a 70 class card? 😂😂
So 5080 and below = almost no performance improvement worth the upgrade and same amount of V-Ram. That mean, 70 series still with 12GB and 60 still with 8GB. Noice. EDIT: 70/60 Class, my bad.
Yep , unless you are buying a 5090 you might as well not bother .
oc there will be speed improvements, just improvements in cache speed and latency will make massive performance improvements without any increases in cuda cores counts.
@@minus3dbintheteens60 Better performances ? Of course but massive is a big Word.
It does depend. If they shrink the L2 to a typical regular size and create a large L3 pool, there could be a "massive" perf uplift from that alone. If they do not increase L1 size and focus on speed and latency improvements, well that's where IPC generally is
My Guess Ref Model 5090 will be £1799
$2250, sounds right, but that's with tax, $1830 tax excluded (american way) seems way too little, more like $2000-2100 without tax and $2460-2580 with tax
No 2K+ for sure It has to be more than what the 4090 is now.
Im super happy with my Rtx 3060. Runs every games with decent setting and fps. Maybe if rtx 5060 comes with 16GB and 3x performance uplift compares to rtx 3060.
Not to want to disappoint you but the 4060 was only 15% better than 3060. The 5060 might be 30% better then that if nvidia is feeling generous. I will be thrilled if they give it more than 8gb vram
50% uplift and 12GB is more realistic id say
@@JanM2 if Paul's sources are anything to go by, 5060 has a 128-bit bus again, so either 8 gb or 16gb config.
Are you open to AMD gpus ? They might give you better performance for the price.
@@evers6214 usually Nvidia gives 2x perf uplift with every 2 generation but after they focusing more on Ray trace the raw performance getting less and less.
do to economic problems, PRICE really matters. most people have RTX 4090, 3090 don't have to buy 5090
i really hope they give me the full GB202 512 bit die for my 5090. when i pay so much money, i hope for the best version of this die they can give us. i expect around 1999$ MSRP (possibly even lower, so they can claim they do something FOR the gamers), thats 400$ more than the 4090, which is already a massive 25% increase in price. all the 3000$ speculations are ridiculous, the 5090 will most likely be a massive 40-60%+ raw performance increase over 4090 (even if they cut down the GB202 die to the 448bit bus version), depending on workload. they cant increase price by 50% as well bc there would be zero price to performance generational gain, thats not going to happen for the flagship.
remember the 4090 was basically the same MSRP as 3090, it was 'only' a 100$ increase over the 3090 to be exact .and all that while they increased 4090 efficiency A LOT, increased performance by - in some cases like 100% over the 3090 (path tracing comes to mind) - overall around 60%+ over the 3090 AND they even implemented a new feature which was frame generation.
So if they sell the 5090 for 50% more money as some suggest, or even more, and make it around 50% faster than 4090, it wont get any of the praise as the 4090. It will be really bad press. It doesnt make sense for them to make such a high MSRP. It is WAY smarter to put it at like 1799, 1899 or even 1999$ absolute tops. This way Nvidia can say they only increased MSRP 'slightly' (if it will be priced at like 1799$ or so) bc of inflation and its not their fault that the street price is 2500$, which could very well be the actual street prices if Nvidia keeps the stock low enough, which they will likely do.
Here's a wild thought Nvidia's greed bites them in the ass finally. The 5080 going for more than $1200 is crazy there's no way you can explain that kind of expense to your wife.
I hear ya that things can change. Any chance in YOUR thoughts/guess that we'll get a version of the 5080 that has larger bus/more than 16GB VRAM? For 2024, at that level, 16GB seems weak for ai work and future-proofing. Total understanding they may be wanting to keep their gaming cards from cannibalizing their ai-specific cards?
5080 - $1300-1500
5090 - $2000 since there's not competition
If the leaks are true, with the 4090 slotting in between the two.
*The 5080 will be similar performance to a 4090. THAT is what matters. I don't care about gaps. I don't care about any metrics to measure by. The 5080 will be MUCH cheaper than a 4090 and lower power consumption, but similar performance.*
Yeah and same VRAM , oh wait only 16 GB , yeah lets see how long that will last you in 4k .........
@@blabla-fw9ix 16gb is ok at 4k until ps6 came out in 2027
@@robertopazsoldan9641 The PS6 will be much weaker than the 5080 , why is that relevant ?
@@blabla-fw9ix did you heard about the console cycle? The game industry move around that. Now we are in the ps5/seriesX cycle use 12gb of 16gbvram shared system for the ps5. Before when 8gb gpu are now 1080p cards they were 1440p like 2070super even the unlucky 3070. There is a video here of a 3080ti 12gb cant run well the latest rachet and clank game with ray tracing at 4k very bad 1%low fps the gpu is powerfull but 12gb is not enough for that game. 16gbs are ok for 4k until the cycle repeat again.
Honestly the 5080 should get a VRAM bump to 20GB.
Are you really saying a $1,400 MSRP is optimistic for the 5080!? That's literally DOUBLE the the MSRP of the 3080, so realistically 5080s will be selling for around 2 grand, likely more? And that's Optimistically? Wow.
PC gaming will be dead (or totally stagnated) in 3 generations tops, at that rate. RTX 7080s for the low price of $4K
People will expect 24,000 cuda cores to cost 2k+, but they will be shook when 10,000 is inevitably priced at 1.5k.
5090, the only card we can look forward to, and it might be a paper launch. 5080 is going to be a nicer, less expensive 4090, but it's not a heavy hitter like we need to push 4K 240 OLEDs. AMD is looking to bring better value to gaming, but they are not pushing boundaries. I can see 5090 selling for $3000 plus from scalpers, it's the only card really worth paying more for. 5080 could be $1500 which makes some sense because it's only half the card 5090 is. Thumbs up.
Idc about dlss performance uplift, I want pure GPU raw power rasterization increase over last generation. I do not use dlss...
Hardware rendering will be phased out eventually, dlss itself is AI render8ng and will literally be how all frames will be rendered, said nvidia anyways
@@Boots3962 as of today it's still a blurry Mess, once they fix that im OK.
Considering Nvidia is using 4nm I don't think there will be a price increase. 4000 series has sold not that well, and they really have reached the maximum of what can be charged. I'm itching to upgrade to a 5090 but anything more than 4090 MSRP and I'm out.
That bigger memory controller + new VRAM will up the price a bit.
I doubt they care, AI will carry them
I am running a RTX4090 but I am praying RDNA5, RDNA6, and Battlemage, Celestial are really competitive.
I suspect the 5090 will be around £1999-£2199 and I also suspect the 5080 will slot in and replace the 4090 at its £1599 msrp, and the 4090 will be dropped around £200-£300
I suspect they will price it in line with performance. They saw they couldn't ask too much for an 80 class card, while for the 90 they probably could have asked more. Now they make the performance difference between the 80 and 90 class even larger, so I think 1 K USD for the 80 class card and 2K USD for the 90 is possible.
I think they maintain prices.
I hope so
We will see a comeback of XX50
GB202 5090 32GB GDDR7
GB203 5080 16GB GDDR7
GB205 5070 12GB GDDR7
GB206 5060 8GB GDDR7
GB207 5050 8GB GDDR6
oh god, not again, well let's hope this time it doesn't cost a kidney for the shitty performance it'll deliver
I think they’ve dropped the 50 class cards for good because its easier to sell a $300 card as a 60 class. I don’t think Nvidia wants to compete under $300 anymore. We will eternally have 60 class cards with 50 class performance until eventually the 60 gets dropped and the 70 class turns into garbage.
5080 $1200 5090 $1800
That 5090 has to cost like $2500-3000. Unless Nvidia starts selling direct and ditches partners. They’ll sell out of all of them too because they’re essentially discount AI hardware.
Should I wait for 5060ti 16gb or just buy 4060Ti 16gb?
Honestly if you want 16GB of VRAM and don't want to spend 1k for it , your best route would be to get an AMD card , the 60 class cards are pretty sad in terms of performance .
5060 and 5060 ti yet again going to be so bad 🤦♂️ 128 bit bus and 8gb vram is outrageous considering most if not all games need 10gb vram at 1080p and we've seen that since 2022 and later. It's almost like Nvidia is killing off the lower and mid tier GPUs because frame generation is basically useless on 8gb GPUs. In other words imma stick with my 3060 Ti And it's 256 bit bus if this is all true specifications ffs 🤔
So it looks like raytacing gets the most benefit from gddr7 which is interesting because it should have been used for rdna4
5080 - Kidney
5090 - Soul
The only way to fix this is speak with your wallet. If you buy it, they will keep doing it. If you refuse to buy it, they will be forced to change. Stop giving them money to shaft you with.
I’ll keep my 3080Ti and nvidia can get bent.
Well 12GB VRAM GPU are going to have issues in late 2025. 16GB VRAM GPU’s will be good until around 2027 before having to lower settings. Honestly each tier needs at least 4 gb more VRAM.
The 5060 should have at least 12GB VRaM on 192 bit memory bus
The 5070 should have at least 16GB VRAM on 256 bit memory bus
The 5080 should have at least 20GB VRAM on a 328 bit memory bus
If I don’t see this then I’m either going to do three things. 1. Buy an AMD GPU 2. Buy the previous gen 4000 series for cheap (might go for 4090 if it hits the $1,000 mark) 3. Hit the used market and support the people and not corporations.
Given all the configuration, i gonna skip this gen, I need X070 or GX205 to be at least 32 TMU or 8192 cores
The more you buy, the more you save.
AI is driving the future. Glad to see the faster memory in this gen. Doubt anyone but a serious (and rich) enthusiast will even buy a 50 though...
Good informative video for leaks that are rumored . The 5090 should have been a 4090 ti, which was never made. They had all of this planned out. Of course, it wouldn't have that many cuda cores or a 512 bit gddr7.
Why did the other 4000 series get ti's but not the 4:01 4:01 4090 😮
As far as price goes, sure they will be more expensive than the previous generation. A 5090 could be around $2000-$2200, but that remains to be seen. I'll just stay with my 4090 for a few more years or even longer.
Meanwhile my 1080ti is still chugging along.
It's a monster it was made to last all these cards 3060 a upgraded 1080,3060-3080-3090-2080 were all made strong to dominate 1440/1080 for years
Heck yeah! I plan on keeping my 4090 until like 2027
These stupid 240Hz + crap is...well...crap. just to feel smoother in microseconds...I shall pass. Ones actual eyes can't tell the difference between 90 Hz and 360 Hz 😂 people actually fall for that snake oil
You are smart!!! Don't waste money
If all you play is very old games and retro games sure. The 1080ti is a dinosaur in 2024.
i'm still playing with 980ti, 1080p is fine for most games, 4k is ok for older stuff
@@ZackSNetwork not true at all it's lots of games 1080ti still can run games great ones that are optimized great or not demanding games
Where did you get that lie from
RTX 5080 MSRP “should” be $900, but of course it won’t be, but it’d be nice if it was.
Intel Battlemage will likely be out Q4 2024, people put too much emphasis on when products release, we are in a long economic recession, consumers are not going to buy new products. Consumers want deals when money is tight, Amd is on schedule with their releases but very few people have the money to buy them.
The PC market is going to be in a rut for several years, and budget products will be key but I dont see big numbers for sales.
Also all these pundits say Nvidia 4000 sells well, so why are 4060/ti and 4070/ti/super always in stock? 4000 series is selling well below what 3000 ssries did. So its very disingenuous to present it like that.
We miss the Christmas shopping season, this means a loss in sales. Mind you I would be getting a 60 series card (really a 50 series with more cores) if it has at least 12gb ram.
skalpers gonna have a field day for us gamers that want the best, you can speculate as much as you want but i am going to say MSRP will be around 2000USD for the 5090 skalpers get there hand on this and they will then i can see 5000USD its a sad truth personally i am happy with the 4090 took a while but i finally have one and not going to stress over the 5090 unless its 2 times the performance of the 4090.
5060 will be $5.6B USD, 5070 and up will be substantially more.
2k plus - 5090
1600 - 5080
I know Nvidia is greedy but there is no way the 5080 can cost that much for the VRAM and for the performance it has , I mean if it costs as much as a 4090 the people would much rather buy a 4090 . It will be like 1200 bucks likely but still terrible for the performance it has .
Spot on pricing. What Blabla doesn't understand is the 5080 has to be batshit insanely overpriced to create a reason to upsell to the next tier. Logic has no place here. 1.6k and 2k is my bet too.
At this point, nvidia is a monopoly, and can charge whatever they want.
3080 ti was the last good nvidia gpu. 40 series is trash unless you can waste 2k for a 4090. Blackwell is about to repeat that scheme
At this point it is likely all 5080 models will be gimped again with a 256 bit memory bus and all models below 5080 will go down hill. I have a wee bit of a byte of hope that Nvidia will release a Workstation 5090 with a 512 bus, but there will be plenty of refurbished stock also where you can start tinkering with. Get out the soddering iron and thermal pads.
The 4090 is so powerful, even having a 4k 120 TV, that I would just justify a 5090 for VR and flight sim. I was excited for it but now I am considering investing in a 7.2 audio system instead... Games needing more power 4090 are not coming out and no I don't care about raytracing at all.
Seems crappy.
Is GB207 the 5050? That looks like it could be beat by an amd igpu before the 60 series comes out. The lineup looks fairly disappointing based on the specs but it is a new arch so perhaps it will be better then this slide shows but so far the only good thing NVIDIA seems to be doing is GDDR7.
from my understanding, yes
My biggest question…
How much can you bench?
60 cards are mainstream not mid range (excluding price) that the 70 cards.
5090 $1699 i guesses, just random though
Why bother with so much rumors and what we do not know ? It is NVIDIA. It will be super expensive, hot as the sun, heavy as led 😊 Why even anticipate all this? It will not change the specs, delivery dates or anything else ... It can only arise disappointment 😊
Rtx fell to as low as $1519 fyi.
5090 (skimmed GB202 - 448bit - 28GB) - 5090 Ti (full die - if they ever feel the need - 32GB) - 5080 Ti (further skimmed GB202 - 320bit - again, if they ever feel the need - 20GB)
5080 (full die GB203 - 256bit - 16GB) - 5070 Ti (skimmed GB203 - 256bit - wishful thinking? - 16GB)
5070 (full die GB205 - 192bit - 12GB) - 5060 Ti (skimmed GB205 - 12GB)
5060 (full die GB206 - 128bit - hopefully with 3GB memory modules for still 12GB VRAM? - otherwise 8GB 🤮)
And 5050 Laptop (GB207 - 8GB entry - yay).
I expect this, just because it pleases my expectation OCD lol 🤣
more clickbait like the one of 40% ipc uplift in zen 5?
2 dies on gb203 rtx 5080 or no?
the 5090 will cost a liver
Cant wait to upgrade my 3060 12GB to a 5060 8GB ddr6 🤡 🤣
The huge generational uplift is reserved for the 3000$ card that will be a paper launch and distributed to tech youtubers and other shills and maybe to less than 1% of the gaming playerbase world wide.
$2,000 5090 stop making numbers up.
@@ZackSNetwork you made up that 2000 number and you are telling me to stop making up numbers 🤡
Just keep what you have for sure.
@@reichfrog2 I am gonna buy a 7900gre if I find one in my area around 500 euro before 8700xt is released.
@@lanelesic I wish I could just do that.
looks like AMD would compete with Intel now. 😅
Like yeah why not the 5080 have the up in BW like the 5090! 5080 Should be 384-bit and NOT 256-bit again! My 3070 was nerfed on the amount of RAM should be 3070 16GB like they did over seas. NIVIDA Stop Nerfing the GPUS . lol
xmen 97 was amazing brother
I just want 4090 just cuz lol have no use for it, i need help
Looks like another disappointing generation. What a shame.
You really think the 5080 will cost even more than last gen? Why would nvidia make the same mistake twice in a row ? The 4080 sold poorly at its ridiculous price and the 4080 super has done much better while being $200usd cheaper ( the 4080 super can found $300-400aud below msrp here in Australia) no one wants an 80 class card above $1,000....especially if the specs are accurate and it's going to be way beyond the 5090. It needs to be $899 imo. But I think they'll price it at $999. I can't imagine nvidia being dumb enough to complete ruin the 80 series class. Twice in a row would just about do that, it's in no many land above $1,000. Not powerful enough and not cheap enough and only 16gb, I'll be shocked If they even try to charge $1,200. It would be hilarious if they do 😂 annoying because I was somewhat looking forward to it before the leaked specs but it will be funny to see the reactions from media and consumers lol
If Navi 48 best chip have 256 bit bus and the same VRAM than 7900Xt and Leaks say performance from 7900XT to 7900XTX. The price is importent but can the RNDA 4 Raytracing power be the same like a 4070 Ti super or even 4080. For 599$ i am total in and pay my 700-750€ for Custom. Nvida way to greed for 2000 RTX to 4000 RTX and this strategies to give low amount of vram to force customers to buy every 2 years a new gpu no thx not with me.
No one notice rgt is a buff man? 💪
I noticed his t-shirt
5090 msrp probs $1800 with real price $2500 for 6 months
_clinches chest_
Soooo... GB207 is just an RTX 3050?
Don’t buy that shit, never buy low range nvidia. Go buy AMD.
Glad I downgraded several years ago from 4k back to 1440p, Don't care what they release with insane prices, I'll still get well over 80fps in every game I play.
5080 looks like another skip generation and they are not going to entice me to move up to 90 class with stupid pricing.
I heard 5080 Super / ti variant will be released later with more memory like 20gb, can only hope.
5080 $2299
The 5080 won't be that high, they have to make it buyable in China, so expect something close to the 4090D.
Anybody want my 4070ti ventus for 600 or 3080 12gb for 400
I don't think nvidia can charge more than $1k for the 5080 unless it has more than 16gb of vram somehow. I would expect the gpu to last until at least 2029 or 2030. With next gen consoles maybe around 2 years away I see vram usage jumping in the near future. I'm not spending more than $1k on a gpu then being forced to turn down textures in just 2 years cause next gen games and consoles will want more than 16gb.
You're underestimating NVIDIA here... they can easily charge $1,600 for a 5080 if they wanted to.
4080 originally costed $1,200.
@@laszlozsurka8991 I don't see that happening. Nobody will buy it. At that point you just buy a 4090. Or a used 4090.
they don't have to change the lineup if they charge $1200. 5090>4090=5080 but less ram>4080 super(no card competing with another). the 5080 is probably a 4090 in performance but less ram and $400 cheaper. personally the 5080 looks like a bad card no matter what because of its ram, If it beats the 4090 in path tracing I will revise my opinion as that speaks to a lot of improvement and optimization so that it can run on less ram.
@@TV-xv1le If NVIDIA puts out a $2,200 5090... there will be people who buy it. Remember the 3090 Ti? That GPU costed $2,000 yet people still bought it.
Plus AMD is not even competing in the high end so what else you're gonna buy anyway if you want a solid 4K setup? The 5080 will be your only choice so expect NVIDIA to take full advantage of that.
Sure you can buy a 4090 used but by the time the 4090 comes down below $1,600... the 5080 will be long launched.
Considering they're giving you gddr7 when no one else is and nothing from amd or intel will even come close to competing they can more or less charge what they want. Esp if it has a big RT uplift
Milking this topic for views is so sad
Bruh these “Leaks” and “Rumors” videos are getting very old!!
Can Amd Stop launching the XTX first? or atleaste the name, like were going to have the same problems as RX7000s where there is a 7900"gre" a 7900xt and a 7900xtx Like wtf AMD? Just call it a 7800 7800xt and 7900 and then release your top end XT later if you really want to compete with the top end of the stack with nvidia. They keep doing this gpu number class mindshare marketing stupidity that isnt working
Is your Name Pol, Pool or Poool?
NVIDIA: lets give them what they preach
Tech influencer: how much do you think 5090 will be ?
Absolute morons: 3000 usd
NVIDIA: done !
Given the size gap between the GB202 and GB203, there is so much room for cutdown GB202 to deliver a 448-bit 5090, a 480-bit 5090ti, a 384-bit config, and a 320-bit config before dropping into the GB203. Given the risks of defects that can lead to such drastic cut downs, it is safe for Nvidia to have both 5090 and 5080 line ups to be handled by GB202. This opens GB203 for 5070, GB205 for 5060ti, GB206 for 5060, and GB207 for 5050.
My thoughts:
GB202 - 448 bit -> 5090: $2,200
GB202 - 320 bit -> 5080: $1,500 (RTX 4090 replacement)
GB203 - 224-bit -> 5070: $1,000 (RTX 4080 replacement)
GB205 - 192-bit -> 5060ti: $700 (RTX 4070 replacement)
GB206 - 128-bit -> 5060: $500 (RTX 4060 replacement)
GB207 - 128-bit -> 5050: $300 (Budget tier, budget gaming laptop)
Honestly, Nvidia would be better setting the GB206 as the 5050 and turn GB2070 into the 5040 because there are plenty of low numbers.
I am looking at speculated VRAM bandwidth, and if the RTX 5080 will be an RTX 4090 replacement, it needs to be like a GB202-250 cutdown just for bandwidth. And as far as prices, I feel optimistic.
This guy just repeats what kopite7kimi leaks each time and add lots of fluff, hype and BS.
Dead on arrival. If this ends up being true I will absolutely under no circumstances buy another nvidia product ever again.
Skip this video if you're poor, cuz Jensen doesn't do business with poor peasants.
5080...... $1600?
With those performances ? Not a chance in hell regardless of Nvidia's greed . I wouldn't pay over 1k for that card and even that seems too much for it , its a joke . It will likely be 1200 bucks .
@@blabla-fw9ix
I agree it's not worth that. Not even close. I'm just thinking of Nvidia price gouging.
this is getting worse then old women gossiping in the neighborhood 😂