NVIDIA rumored to be preparing GeForce RTX 4080/4070 SUPER cards

What you spend your money on is something I would respect. That said, I do not see a single game or program out there, unless you are making money from it, that justifies or even means that a 4090 at $1600 is worth it, at all. Games, back in the 2000's and maybe up to 2018 were often impressive and unique but, not really so much anymore.

The way it’s going you’re going to have to get a top tier card just to brute force some of the games coming out. Anything less will be a recipe for frustration.
 
The way it’s going you’re going to have to get a top tier card just to brute force some of the games coming out. Anything less will be a recipe for frustration.
Certainly some seem that way. Starfield comes to mind. No great visuals yet middling performance on top tier cards. Truly a terrific engine.
 
I don't really care about them releasing Supers of lower end cards, but it would kind of suck for me if they released a 4090 Super/Ti shortly after, because then my own purchase, which I just finally committed to, would lose some value. Knowing my luck, it'll happen though.

The thing that made me realize if I was itching for an upgrade, AMD just wasn't going to provide is looking at the 4080 vs 7900 XTX review earlier: https://hardforum.com/threads/techspot-7900xtx-vs-4080-revisited.2030904/

I was kind of rooting for the AMD chip to just be a much better value in that review, but... it didn't happen, their conclusion was that it was still sort of overpriced, even as low as its prices had gotten. So, no choice. I just really wanted a better GPU than my 3080 Ti for some reason unknown to me and the only thing that could provide it is a 4090.
Fortunately, the 4090 doesn't exactly leave a lot on the table with the 102, so at best Nvidia might be able to squeeze another 15% out of it, but would be doing so by cutting back on releases of the server L40 units, which are already in short supply and back-ordered to high hell and sell for way more than they could sell a 4090TI for. My pricing on an L40 sits around $10K CAD, has an 8-week lead time, and uses the full-featured ADA102, unless Nvidia puts something out on the market to replace that part then I can only see releasing a 4090TI as a net loss to themselves.
 
The way it’s going you’re going to have to get a top tier card just to brute force some of the games coming out. Anything less will be a recipe for frustration.
I would rather live in reality then, if that is the case. 😊 I could spend that money on things that are a lot more fun and also gets me out of the house.
 
Certainly some seem that way. Starfield comes to mind. No great visuals yet middling performance on top tier cards. Truly a terrific engine.

I think Hogwarts Legacy looked amazing with all of the RT settings jacked up, and with DLAA on a 4090 it looks quite phenomenal. On the other hand, what I don't understand is why I needed to upgrade my CPU from a 5950X to a 7800X3D to enjoy it. This is what's baffling me, is newer engines seem to scale with the CPU so hard at 3440x1440, even, that you can go up a GPU generation (or two)... by upgrading your CPU. I don't know how long this 7800X3D is going to actually last at this rate, if games continue using 2 threads for all of their RT processing...

Prices are creeping up for the 4090. So you should be good

Also No pressure from AMD on Nvidia to release a 4090 ti

The no pressure, I sort of understand, but technically is there much pressure for them to release the 4080 Ti/Super to begin with? The 4080 has been quantitatively proven to be competitive with AMD's 7900XTX in the topic I linked. I guess it will provide a middle line between the 4080 and 4090, but with how expensive they are to begin with, I don't know what good is going to come of this since I feel like most people near a 4080 Ti price ranger would just go for the 4090 anyway...

4090s are still plenty in stock, for what it's worth:
https://pcpartpicker.com/products/video-card/#sort=price&c=539&xcx=0

Although the only one at MSRP is pretty much PNY, and who actually wants PNY? I got my Suprim X Liquid for 1620 open box at Microcenter, and got it registered for warranty on MSI's web site, so I'm happy with that.
 
I think Hogwarts Legacy looked amazing with all of the RT settings jacked up, and with DLAA on a 4090 it looks quite phenomenal. On the other hand, what I don't understand is why I needed to upgrade my CPU from a 5950X to a 7800X3D to enjoy it. This is what's baffling me, is newer engines seem to scale with the CPU so hard at 3440x1440, even, that you can go up a GPU generation (or two)... by upgrading your CPU. I don't know how long this 7800X3D is going to actually last at this rate, if games continue using 2 threads for all of their RT processing...
This is more an engine thing, it is incredibly hard to build something consistent that automatically scales across cores and threads, so writing something that is designed for 6/6 or 8/8 as a hard target is much easier than trying to build something that can automatically scale across X-1 threads (one thread will always need to be the master). In a situation where you know exactly how many threads you are working with you can work to load balance that accordingly, but in the case where things are trying to do this automatically it leaves too much out of the developer's control and places that almost exclusively on the OS Thread Director, it then also changes how the engine must allocate memory and troubleshooting becomes a minor nightmare because the swing gets very extreme.

Dealing with Thread scaling is a pretty complicated and hot topic for Engine Developers, and I would bet that Epic has something on the drawing board for it in Unreal Engine 6 in concert with Microsoft DirectX-Next work, but as things currently stand dynamic thread allocation for an engine is just not feasible for the development cycle.

You can do it for workloads like blender and rendering and all sorts of jobs because each frame is 100% independent of each other, the final result doesn't care if frame 18 finishes before frame 2 or frame 1 finishes after the very last one because the only thing that matters is the end result and it knows where each piece goes.
But in Games, there is too much real-time work going on and too many things that go wrong when you leave that to chance.
 
RTX Thuper

AKA 'The Champ'

1697749777482.png
 
I just want a smaller card with less power draw next generation. I still can't believe how big my 4080 is like the biggest widest card I ever owned. I like it but all that weight it's like three times as heavy as my Noctua Heatsink which is 2lbs.
 
I just want a smaller card with less power draw next generation. I still can't believe how big my 4080 is like the biggest widest card I ever owned. I like it but all that weight it's like three times as heavy as my Noctua Heatsink which is 2lbs.

I was impressed with my 3060ti size/performance//power - 4070 is just bigger size, but also much bigger performance in same power envelope - I've been impressed tech wise (save the price complaints, I have them too, who doesn't)
 
I just want a smaller card with less power draw next generation. I still can't believe how big my 4080 is like the biggest widest card I ever owned. I like it but all that weight it's like three times as heavy as my Noctua Heatsink which is 2lbs.
I dunno what the deal is with this generation. Most of the cards could be smaller. Seems like everyone is trying to keep their GPUs under 70c. Which is overkill.

IMO, AMD has it right, with their reference cards. Temps are great. They just need to R&D some fans which keep a better noise profile as they speed up. But the temps are so good, you can manually control the fans and still have good temps.
 
NVIDIA is tired of getting beat up on VRAM, and the "NVIDIA does not care about gamers anymore," narrative. Needs to make a positive PR move. So this will give them some positive press.

I am hearing Super to be the CES launch with a retail roll out right before Chinese New Year.

All this is while NVIDIA is cutting AIB GPU allocations for Q4.
 
NVIDIA is tired of getting beat up on VRAM, and the "NVIDIA does not care about gamers anymore," narrative. Needs to make a positive PR move. So this will give them some positive press.

I am hearing Super to be the CES launch with a retail roll out right before Chinese New Year.

All this is while NVIDIA is cutting AIB GPU allocations for Q4.
I was under the impression the cuts were mainly to the ADA104 and 106 supply as the 4070 Super will move to the 103 to combat the 7800xt and the 4060 Super goes to the 104 but its market value is still going to be questionable as all hell, and AIB’s still have more 4060 non Supers than they know what to do with because they are selling worse than even they expected them to.

I figured thought it was mostly to feed the OEM demand for mobile chips and their first party cards. Dell, HP, and Lenovo make their own GPU’s do they not?
 
I honestly can't imagine 4070 Ti owners being that happy, if its performance is anywhere close to the 4070 Ti...
I can't imagine Nvidia cares but I would like to think they do.

A newer faster card is always just one CES away, we all know this, buy the card you would be happy with now, because tomorrow it's obsolete.
 
I honestly can't imagine 4070 Ti owners being that happy, if its performance is anywhere close to the 4070 Ti...
The 4070ti doesn't magically become slower or anything.... They knew what they were buying when they bought it. New stuff comes out all the time anyway. If it didn't give someone what they wanted for the price, they probably shouldn't have bought it and should have waited for market adjustments in the first place :).
 
I can't imagine Nvidia cares but I would like to think they do.

A newer faster card is always just one CES away, we all know this, buy the card you would be happy with now, because tomorrow it's obsolete.
The 4070ti doesn't magically become slower or anything.... They knew what they were buying when they bought it. New stuff comes out all the time anyway. If it didn't give someone what they wanted for the price, they probably shouldn't have bought it and should have waited for market adjustments in the first place :).

I get that, but the 4070 Ti's existence was kind of a kick in the proverbial nuts to begin with, thanks to its 192 bit bus (which did show bottlenecking in 4k though) and only 12 GB of ram. People still ended up buying it because it was good enough. The 4070 Ti released in January of this year. Normally I think cards within the same gen might last a little longer than a year before becoming completely obsoleted by them releasing a version that would basically be better in every way possible. People ate that 4070 Ti price after giving the card a fair chance, and then they do this... They keep this up, no one's going to buy round 1 cards. They'll just wait until the Supers come around that fix the issues.

Which is fine by me, Nvidia needs to get kicked a bit. Maybe they should have worked harder at releasing a generation where the only card that was really worth it was the 4090...
 
I would say if Nvidia and AIBs reduce the stupidly big cooler sizes for cards that don’t need it, virtually all of them, so normal people with normal size cases have less issues. Increase ram sizes some and reduce cost, Nvidia would do extremely well.

Also, a 102 die that can be sold in China with upcoming if not implemented ban of 4090 sells there. I would think Nvidia with their very successful 102 die sells and production would have a stockpile built up of failed GPUs that could work in a 4080 Super, 20gb. 10 to 15% slower than a 4090.
 
I would say if Nvidia and AIBs reduce the stupidly big cooler sizes for cards that don’t need it, virtually all of them, so normal people with normal size cases have less issues. Increase ram sizes some and reduce cost, Nvidia would do extremely well.

Also, a 102 die that can be sold in China with upcoming if not implemented ban of 4090 sells there. I would think Nvidia with their very successful 102 die sells and production would have a stockpile built up of failed GPUs that could work in a 4080 Super, 20gb. 10 to 15% slower than a 4090.
I was under the impression it was already technically banned because its performance exceeded the limitations that were already in place, they just didn't realize when they were shipping it that it could exceed those thresholds.
 
I would think Nvidia with their very successful 102 die sells and production would have a stockpile built up of failed GPUs that could work in a 4080 Super, 20gb
This:
https://www.nvidia.com/en-us/design-visualization/rtx-5000/

Has only 70% of the AD102 chip enabled, the 4090 "only" 88.8% that is already a significantly cut down failed to be RTX 6000 or L40 something card, both are not that low for a 600mm affair, but specially the RTX 5000 if it was selling well could have took a lot of it.

Maybe they have a stock of 75- 80% chips, too good for RTX 5000, not 4090 caliber that could make a nice 4080TI.

I could see them keep the ridiculous power delivery-cooling system already ready and overclock the super quite a bit, give them the first +8-10% performance.
 
I honestly can't imagine 4070 Ti owners being that happy, if its performance is anywhere close to the 4070 Ti...

My assumption is it will be right in the middle of the 4070 and 4070ti in terms of performance. But will have 16GB of VRAM. Which would be odd to have a slower card with more VRAM. I suppose they will also have a 4070 Ti Super? Have there ever teen Tis and Super nomenclature in a single generation before?
 
My assumption is it will be right in the middle of the 4070 and 4070ti in terms of performance. But will have 16GB of VRAM. Which would be odd to have a slower card with more VRAM. I suppose they will also have a 4070 Ti Super? Have there ever teen Tis and Super nomenclature in a single generation before?

I'm actually kind of wondering if they would bump the price of the 4070 Ti down instead, and then put the 4070 Super at a slightly more expensive price tag so it comfortably sits in between the 4070 Ti and the 4080. That way its VRAM would make sense and it would really cut into 7900 XTX sales. For people looking a bit more premium, then they don't have to choose the 4080, they could get the 4080 Super... I guess?

Well, just something that they could possibly do.

i ordered a pny for my rendering computer. What is wrong with pny?

I've done a fair share of research on various brands before I got this Suprim X Liquid. Most things I've read about PNY and Zotac says that their low prices are deserved. PNY especially I have read very bad things about their warranties.

You can read various anecdotes around Reddit. ie:
https://www.reddit.com/r/nvidia/comments/r1eqg2/trying_to_rma_a_dead_pny_rtx_4000_and_getting/
Also their warranty does not transfer:
https://www.reddit.com/r/nvidia/comments/14g6zow/pny_4080_secondhand_warranty/
It's just their general reputation:
https://www.reddit.com/r/nvidia/comments/y0dyqq/pny_a_misunderstood_yet_good_gpu_brand/irrrych/

They seem fine for their Quadro line, but the consumer GPUs are reported to have slightly cheaped out cooling solutions (though on the 4090 and probably 4080, that's a non-factor since all of the coolers are already overkill).

So no it has nothing to do with branding. I would not buy ASUS, PNY, or Zotac if I could help it. The only time I would is if their prices were so much lower comparatively that I could buy the replacement plan from Microcenter, along with the GPU price, for a total price that would bring it roughly up to parity with MSI or maybe Gigabyte. Granted, I've heard Gigabyte has also gone downhill lately (what with the clip breaking due to their extremely flimsy PCB in that area--arguably user error, but one could argue that it's prone to failure, too), but they did take care of me in the past so I'm slightly partial to them. Thing is, I would reckon that most GPUs just tend to work and keep working. The people with issues to begin with are likely in the minority unless the chip just has systematic design failures. The only reason you buy a certain brand is for the warranty assurance. And comparatively, MSI right now (to me, I haven't had to use it) seems to have a warranty that might be the most reliable. It's also transferrable since you can make it start from date of manufacturing, which most companies don't even seem to do. The only thing better seems to be FE if you're buying Nvidia, but you're stuck with an FE cooler (which on the 4090 is fine though) and it's also not transferrable afaik.
 
Nothing, never had an issue with any of their workstation cards.
Same, and have a client with just under a billion cards. Not flashy. They work.

To be on point, I really don't need a super. I need the existing things to cost a lot less, for home use.
 
Although the only one at MSRP is pretty much PNY, and who actually wants PNY? I got my Suprim X Liquid for 1620 open box at Microcenter, and got it registered for warranty on MSI's web site, so I'm happy with that.
The current PNY stuff is really good. They aren't the cheap brand they used to be. Techpowerup said this about their 4080 Verto XLR8:

It's been a while since I've tested a graphics card from PNY, we all mostly know them from the professional-grade NVIDIA graphics cards that they produce. It seems they are making a push for the consumer market—no doubt to capture EVGA's market share that's available now. I have to say I'm seriously impressed by what PNY has delivered here. Their cooling solution is top-notch and can easily compete with the offerings from ASUS, MSI and Zotac. Having the most powerful cooler won't help any bit if your fan settings suck. Here too, PNY has done a fantastic job. Their card is actually the quietest RTX 4080 that I've tested today. Yup, even quieter than the premium offerings from MSI and ASUS. With just 26 dBA under full load, the card is pretty much inaudible in any setup that has other source of noise, like the CPU cooler or watercooling radiator fans. It seems PNY understands that offering great noise levels sells GPUs. MSI realized this many years ago, hence their stellar climb from value-manufacturer to tier one hardware company started. While other vendors include a dual BIOS feature with their cards, I see no reason why this is needed on the PNY RTX 4080. The card runs extremely quiet, and runs at super low temperatures of 64°C—what else could you wish for?


And they said similarly good things about their 4070 ti:

PNY's factory overclocked RTX 4070 Ti is based on the NVIDIA reference PCB design, PNY just added circuitry for their RGB lighting effects. The cooler on the other hand is a fully custom design, and PNY did amazing work here, just like on their other GeForce 40 cards. Our apples-to-apples heatsink comparison test that measures only the heatsink's capabilities reveals that PNY is offering the most powerful cooler of all the cards tested today. They also paired it with fantastic fan settings that are incredibly quiet. Just 28 dBA under full load is whisper quiet—even in a quiet room—very impressive. At this noise level, the PNY RTX 4070 Ti OC is the second-quietest card, beating all the other famous big name brands! Temperatures are really good, too, just 63°C is comfortably cool. What I really like is that PNY isn't hiding those great fan settings behind a manually-activated dual BIOS "quiet" profile like some other vendors do. As expected from all modern graphics cards, all GeForce RTX 4070 Ti cards come with the idle-fan-stop capability that shuts off the fans when not gaming.
 
Same, and have a client with just under a billion cards. Not flashy. They work.

To be on point, I really don't need a super. I need the existing things to cost a lot less, for home use.
If rumours are true we may be pleasantly surprised in Feb, again IF… but if they are true 4070 Super will be on par with the 7800xt and priced comparably to it, the remainder of the stack will be priced to clear.
Nvidia does not want stockpiles of the 4000 series around when the 5000 hits and they are upset at the AIB’s unwillingness to lower their margins to pre covid levels. So many of the Super variants may be FE only models and Nvidia is going to force them to drop those prices by putting the supers at the same prices as the existing TI variants while being faster than them.
 
Back
Top