News Zotac accidentally lists RTX 5090, RTX 5080, and RTX 5070 family weeks before launch — accidental listing seemingly confirms the RTX 5090 with 32GB...

[Admin]

Zotac accidentally added premature listings for the RTX 5090/D, 5080 and the 5070 family that were promptly taken down.

Zotac accidentally lists RTX 5090, RTX 5080, and RTX 5070 family weeks before launch — accidental listing seemingly confirms the RTX 5090 with 32GB... : Read more

 

[A Stoner]

I hope it is a massive increase in gaming performance and then I can feel bad about having an RTX 4090, but have to look forward to waiting for the 6000 release before I can legitimately argue I should upgrade.

 

[txfeinbergs]

Nah, don't feel bad. I will not be upgrading either from my 4090. What is the point? The only thing that would get me to upgrade is if they lock some great new DLSS technology behind the 5000x series that drastically improves games.

 

[CaptRiker]

have a regular 4080 (bought before super's come out)... not even interested in a 5080.. looks like a weak sidegrade... now if I can get a 5090 for $1599 (4090 msrp), I might be tempted, else ToTaL pass on the 5000 series.. F. NVIDIA!

 

[Notton]

Let me guess, RTX 5090 for the sweet sweet price of $4090

The 5070 being stuck at 12GB is annoying, but at least the Ti has 16GB.
There's a big gap between 5090 and 5080, I am guessing they'll launch a 24GB Ti model when enough 24bit GDDR7 is in stock?

 

[Giroro]

I hope it is a massive increase in gaming performance and then I can feel bad about having an RTX 4090, but have to look forward to waiting for the 6000 release before I can legitimately argue I should upgrade.

It probably won't be some massive increase in gaming, Nvidia is not likely to jump to much more advanced transistors, just to make the same basic thing, but bigger. Most RTX 4090s were bought as entry-level workstation cards for AI workloads and Nvidia knows it. A lot of people were paying $1800 for a gaming card to avoid paying $3500 for a workstation that performed close enough. So there's a good chance the RTX 5090 is not designed for gaming. Although the RTX 4090 also wasn't really meant for home gamers, but non-sponsored people still bought them apparently, so who knows.

Although there's always still a chance Nvidia severely nerfs the 5090's AI to stop from eating themselves out of their own market. They've always limited compute performance, started limiting hash rates for crypto back when people cared about that, so why not artificially nerf AI? The bump in memory would only be useful for AI, though.

I expect the price of the new cards to scale linearly with performance again, if not a bit worse. So maybe well get a card that games 15% better than a 4090, but at 20% higher prices and probably >20% higher power consumption. Unless they don't nerf AI, then the price jump could be a lot more. Nvidia wants their $3500 from these people.

 

[helper800]

I expect the price of the new cards to scale linearly with performance again, if not a bit worse. So maybe well get a card that games 15% better than a 4090, but at 20% higher prices and probably >20% higher power consumption. Unless they don't nerf AI, then the price jump could be a lot more. Nvidia wants their $3500 from these people.

If the 5090 only has 15% more gaming performance this would be a complete disaster of a launch...

 

[emike09]

I'd really like to see more than 16GB on the 5080 model. 20GB or 24GB would suffice for this upcoming generation. Several games are already maxing out 16GB, especially with tech like Path tracing. On my 4090, with the OS overhead, I've seen upwards of 21.5GB utilization in select games. Many sitting around 16-18GB utilization.

Who knows what the next 3 years of game development will bring - 16GB on the 5080 seems like just barely enough for the latest game releases and does not feel future proof. 5070 should be 16GB, and 5060 should be 12GB.

 

[SocDriver]

From everything I am seeing in leaks this generation is going to be a wash, especially if they do another price hike on them. The 5080 getting stuck with only 16GB is the biggest killer, I imagine that will be the bottleneck for the card, just not enough memory to push more performance.

Also it would be a disaster to leave the 5070 at 12GB of VRAM because if Intel has the same luck with performance leap in the B770 as they did with the B580 its going to crush a 12GB 5070.

 

[Eximo]

I suspect these are likely to come with more onboard cache as well. That may reduce the need for as much memory. 30 series high end has 6MB cache, 40 series topped out at 72MB. AMD at 96MB with the 7000 series.

Also just because more than 16GB is loaded into the GPU, doesn't mean it is using all of that data. Just that it hadn't needed to offload anything to make room. That is still a good thing, like not needing to swap things from disk to memory. So some shorter load times or the occasional avoidance of a little stutter/frame time variance.

The real solution is to just drop the settings down a notch or two. If you want top of the line, buy it, don't complain that the cheaper product has less.

 

[purposelycryptic]

The 5090 is almost twice as powerful as the 5080, with twice the RAM, yet the 5070 Ti looks to only be around 20% less powerful than the 5080, with the same amount of RAM...

Feels like they really should have done more to differentiate the 5080 as a high-tier card, especially since the 5070 Ti looks to be releasing around the same time - as it stands, it almost feels like more like a 5070 Ti Super Deluxe w/ Cheese than a card worthy of having its own tier in the Nvidia hierarchy.

An extra 8GB of RAM would probably be enough, but really, given the spec delta, the 5080 probably should have been a cut-down 5090 with 24GB, the current 5080 relabeled to 5070 Ti Super, (or whatever), and the rest kept as is.

Maybe their yields were just too good?

 

[Tonet666]

I hope it is a massive increase in gaming performance and then I can feel bad about having an RTX 4090, but have to look forward to waiting for the 6000 release before I can legitimately argue I should upgrade.

You can be certain of one thing, massive increase in price. 🤣

 

[aberkae]

If the 5090 only has 15% more gaming performance this would be a complete disaster of a launch...

This would be atypical. Even if we look at the weakest generational jump going from 1080ti to 2080ti which used dedicated silicon for rtx Tensor cores it still had a bigger jump in rasterization. I'm expecting a minimum of 40% in rasterization and hopefully double in rt or at least a 4k 60 fps patch patchracing without *upscaling or frame generation.
We can probably expect brute forced propriety exclusive features dedicated for Blackwell as well.
Current rumors 50 to 70% delta gains over 4090 for the 5090 and lower gains the lower you go down the stack.

 

[artk2219]

They're probably leaving that gap in between the 5080 and 5090 for a few reasons.

1. Yields, that is a big chip, and im sure theyre seeing some not so great yield rates currently, this lets them market more of them as "RTX 5080's". This also lets them retain the better binned chips in between for AI accelerators.

2. Competition from AMD and Intel, if the RX 8800 and Arc B770 (if it survives Intels chopping block) are competitive withe the RTX 5080, all they need to do is release a "super" or "TI" SKU and they're good.

3. Its Nvidia. They've been fleecing people for years and everyone loves them for it. If people are unhappy with how the RTX 5080 performs, they will yell, complain, buy it anyway, and then cheer when the super or ti comes out for marginally more.

 

[Eximo]

They're probably leaving that gap in between the 5080 and 5090 for a few reasons.

1. Yields, that is a big chip, and im sure theyre seeing some not so great yield rates currently, this lets them market more of them as "RTX 5080's". This also lets them retain the better binned chips in between for AI accelerators.

2. Competition from AMD and Intel, if the RX 8800 and Arc A770 (if it survives Intels chopping block) are competitive withe the RTX 5080, all they need to do is release a "super" or "TI" SKU and they're good.

3. Its Nvidia. They've been fleecing people for years and everyone loves them for it. If people are unhappy with how the RTX 5080 performs, they will yell, complain, buy it anyway, and then cheer when the super or ti comes out for marginally more.

B770 I presume.

4080 and 4080 Super are very close. So the binning was minor, they even used the same cuda count for the mobile 4090 which is a little different than usual. Doesn't seem like they had alternate plans, just waited until yields were better. Also the 4080 Super was a welcome price reduction of $200. The others simply replaced the existing tier of cards and the older cards dropped in price.

For Ada they were actually using cut down 4090 to make 'second' tier Quadro with a 256bit bus.

So not sure the 5080 doesn't represent the same thinking. Bin and save, wait for yields, then launch a slightly improved version.

 

[AnimalisticDisskord]

You can be certain of one thing, massive increase in price. 🤣

If it drops the price of the 4080 super, that would be just fine. The 4080 super already allows you to play any game at max settings with 4K. What would spending $4,000 on a card really do for you?

 

[palladin9479]

Will they include the additional dedicated PSU you need to power them?

 

[Elusive Ruse]

God damn the 80 class is dead, there’s a damn chasm between the 5080 and the 5090.

 

[TeamRed2024]

I expect the price of the new cards to scale linearly with performance again, if not a bit worse. So maybe well get a card that games 15% better than a 4090, but at 20% higher prices and probably >20% higher power consumption. Unless they don't nerf AI, then the price jump could be a lot more. Nvidia wants their $3500 from these people.

If the 5090 only has 15% more gaming performance this would be a complete disaster of a launch...

... it would make not upgrading from the 4090 a really easy decision too.

 

[P.Amini]

The real solution is to just drop the settings down a notch or two. If you want top of the line, buy it, don't complain that the cheaper product has less.

Your opinion is against most people's opinion.

 

[valthuer]

... it would make not upgrading from the 4090 a really easy decision too.

But if it's a 40-50% generational performance increase?

 

[artk2219]

B770 I presume.

4080 and 4080 Super are very close. So the binning was minor, they even used the same cuda count for the mobile 4090 which is a little different than usual. Doesn't seem like they had alternate plans, just waited until yields were better. Also the 4080 Super was a welcome price reduction of $200. The others simply replaced the existing tier of cards and the older cards dropped in price.

For Ada they were actually using cut down 4090 to make 'second' tier Quadro with a 256bit bus.

So not sure the 5080 doesn't represent the same thinking. Bin and save, wait for yields, then launch a slightly improved version.

Correct, thanks for catching that, I've edited the original post.

The 1660 super wasnt a minor uplift in performance, but you're correct in that it wasn't really architecture related, the faster memory really helped out there.

I do think its mainly yields on why theres a chasm between the 5080 and 5090, but its definitely not the only reason why that space would exist.

 

[Mama Changa]

have a regular 4080 (bought before super's come out)... not even interested in a 5080.. looks like a weak sidegrade... now if I can get a 5090 for $1599 (4090 msrp), I might be tempted, else ToTaL pass on the 5000 series.. F. NVIDIA!

How is it a sidegrade. 5080 expected to beat 4090 in raster and RT at lot lower price. You could wait for the inevitable 24GB Super version though.

 

[thestryker]

Also just because more than 16GB is loaded into the GPU, doesn't mean it is using all of that data. Just that it hadn't needed to offload anything to make room. That is still a good thing, like not needing to swap things from disk to memory. So some shorter load times or the occasional avoidance of a little stutter/frame time variance.

The real solution is to just drop the settings down a notch or two. If you want top of the line, buy it, don't complain that the cheaper product has less.

Very much disagree with your "solution" to insufficient VRAM. If the only thing holding back performance is a lack of VRAM then the card doesn't have enough. The capacity of VRAM should match the capability of the GPU.

There's an increasing number of instances of VRAM capacity being the limiting factor which from a consumer side should be something not stood for. If you're buying a $200 or less card sure expect limitations, but on something that level it's unlikely to be able to run things at a playable frame rate which would also use up all of the VRAM. Intel has already put the B580 forward saying 12GB at $250 can be done and nobody should be willing to accept less.

4080 and 4080 Super are very close. So the binning was minor, they even used the same cuda count for the mobile 4090 which is a little different than usual. Doesn't seem like they had alternate plans, just waited until yields were better. Also the 4080 Super was a welcome price reduction of $200. The others simply replaced the existing tier of cards and the older cards dropped in price.

Realistically the 4080 Super was just a price reduction due to backlash.

 

[thestryker]

I'd really like to see more than 16GB on the 5080 model. 20GB or 24GB would suffice for this upcoming generation. Several games are already maxing out 16GB, especially with tech like Path tracing. On my 4090, with the OS overhead, I've seen upwards of 21.5GB utilization in select games. Many sitting around 16-18GB utilization.

Who knows what the next 3 years of game development will bring - 16GB on the 5080 seems like just barely enough for the latest game releases and does not feel future proof. 5070 should be 16GB, and 5060 should be 12GB.

Plenty of room for a 5080 Ti or Super to use 24Gb instead of 16Gb memory IC which gives 24GB capacity whether or not they increase the core count.