NVIDIA 5000 SERIES

[TNA]

The whole ‘retained value’ thing only really works if you are flipping gen-to-gen (selling a 40 series to get a 50 series), either before or around the time of release of a new card. Clearly it’s a legit strategy, with success depending on how well the new gen is hyped / received / looks on the spec sheet.

Once the 60 series is out and makes the 40 (+ 50?) series look like a pipsqueak(s), the resale value will be massively reduced.

Hence my confusion with some sentiments each upgrade cycle (not just this one) of “I will be keeping my current card - not worth the upgrade and my own card will retain its value” - if you are keeping the card, it will be totally outclassed and have a reduced resale value by the time you actually sell it… so then you won’t be getting any real benefit of ‘retained value’…? You can’t keep the card and get the benefit of optimum resale value - pick a lane

Yep. I have done this for two decades now. It works. Majority of the time anyway.

The 4090 will be worth around £500 when the 6000 series is out I recon

It's like who wants a 3090 today? 24GB VRAM though bro!!

Who cares. I would take a 5070 Ti with 16GB over a 3090 all day every day. Just like most people will rather have new 6070Ti over a crusty old 4090 when the 6000 series is out.

 

[Tinek]

Also agree. The 5090 certainly won't hold as well as the 4090 is at the moment if the performance increase is 45-50% on the new node. Having said that, 4nm to 3nm (I doubt it will be 2nm) isn't a massive step forward, so who knows.

I don't want to be a total downer, there's still time for Nvidia to improve things but one has to remember - the moment they release GPUs to the market, the next generation is already in internal testing (and not just design) phase usually, which means they already know what to expect from its performance and process. And so far the only thing I've heard from their VP of AI stuff was that it's very very hard to scale up GPU performance anymore and the only way forth is AI and even more MFG. Which is consistent with what Nvidia CEO repeated all the time about 5k series for months now. In other words, I don't expect much from 6k series currently and will not have any big hopes.

What newer process do these days is lower power use OR increase performance - not both anymore. So we could get faster but with even more power use or similar performance but less power used - unless they change the design of the chip a lot but they seem to have no real incentive to do it. Most of their r&d is focused on AI for a while now, whilst at the same time they are a monopoly with gaming GPUs. So, they can do whatever they want really and people will buy it anyway, because it's Nvidia. So what I expect them to do is to go for the best revenue, which means AI focus (Stargate project well keep them busy for many many years to come and with full pockets) and just do bare min. for gamers.

 

[TNA]

The man maths side of me is saying, 'buy a 5090 and sell your 4090, otherwise your 4090 will be worth £500 when the 6090 launches'. A moot point if supply is non-existent I suppose.

Haha. You know it will.

But your 5090 might be worth £900. One way or another huge depreciation. Only way you get most of your money back on a 5090 is to sell just before Jensen does a presentation saying 6060Ti = 5090 performance

Basically you need to be able to slum it out with a lower end card for a few months to not lose much.

 

[Absolutely Clueless]

I don't want to be a total downer, there's still time for Nvidia to improve things but one has to remember - the moment they release GPUs to the market, the next generation is already in internal testing (and not just design) phase usually, which means they already know what to expect from its performance and process. And so far the only thing I've heard from their VP of AI stuff was that it's very very hard to scale up GPU performance anymore and the only way forth is AI and even more MFG. Which is consistent with what Nvidia CEO repeated all the time about 5k series for months now. In other words, I don't expect much from 6k series currently and will not have any big hopes.

What newer process do these days is lower power use OR increase performance - not both anymore. So we could get faster but with even more power use or similar performance but less power used - unless they change the design of the chip a lot but they seem to have no real incentive to do it. Most of their r&d is focused on AI for a while now, whilst at the same time they are a monopoly with gaming GPUs. So, they can do whatever they want really and people will buy it anyway, because it's Nvidia.

Guess I need to buy a 5090 now because the 6090 won't be much better. Thanks, Tinek. Now where can I find an overpriced Astral card.... mhmm.

I'm just hoping chiplets introduces some big gains, but I also don't have big hopes tbh.

 

[Absolutely Clueless]

Haha. You know it will.

But your 5090 might be worth £900. One way or another huge depreciation. Only way you get most of your money back on a 5090 is to sell just before Jensen does a presentation saying 6060Ti = 5090 performance

Basically you need to be able to slum it out with a lower end card for a few months to not lose much.

Yep, that's the only way. Unfortunately, I don't have a spare GPU.

Maybe I keep my 4090, buy a 5090, and then use the 4090 in between selling the 5090 and buying a 6090. Gosh, nothing but the big brain moves over here.

 

[TNA]

When I sold my 3090 because I had bought a 4090 at launch, I got 25% of the original value. Rtx4000 killed the 3090's value. However right now the highest offer I've received for my 4090 is 68% of its original value, so clearly rtx5000 has not had the same effect

We have confirmation he did not smash his 3090 after all gents.

Says he can smash it, doesn't, now whining on about the 5090 as he can't get one.

Not as minted as he says he is after all

 

[TNA]

PS. Doesn't matter anyway Jensen pretty much smashed it for him as he only got 25% back. Lol

 

[TNA]

Yep, that's the only way. Unfortunately, I don't have a spare GPU.

Maybe I keep my 4090, buy a 5090, and then use the 4090 in between selling the 5090 and buying a 6090. Gosh, nothing but the big brain moves over here.

Get a cheap 4070/4080 or something close to the time.

 

[Hostile_18]

If its just a few months a super cheap card will do. I've been playing FF7 Remake maxed at 4k and Returnal 1440p just fine since I sold my 4090 on a 3050.

 

[xaldub]

Also agree. The 5090 certainly won't hold as well as the 4090 is at the moment if the performance increase is 45-50% on the new node. Having said that, 4nm to 3nm (I doubt it will be 2nm) isn't a massive step forward, so who knows.

Even if by moving to 3nm we only get another 20-30% performance boost, the power consumption & thermals will be better than with a 5090. Mark my words, the resale value of a 5090 next gen will plummet ... this is what happened to the 3090 when the 4000 series launched.

 

[Twinz]

I can't help but think these stories are related:

Nvidia fixes Blackwell chip flaw with help from TSMC, mass production back on schedule

Overcoming this issue was crucial for Nvidia, as it aims to maintain its dominant position in the AI chip market. As demand for high-performance AI computing solutions...

www.techspot.com

Nvidia's data center Blackwell GPUs reportedly overheat, requiring rack redesigns and causing delays for customers

Nvidia redesigns NVL72 servers to tackle overheating problems.

www.tomshardware.com

NVIDIA has removed "Hot Spot" sensor data from GeForce RTX 50 GPUs - VideoCardz.com

GPU-Z won’t show hot spot temperature for RTX 5090 NVIDIA has decided to disable the sensor entirely. Performance aside, there are other interesting topics arising from today’s review coverage of RTX 5090 cards. NVIDIA has spoken a lot about its new dual-slot cooler design, which is intended to...

videocardz.com

 

[Tinek]

The man maths side of me is saying, 'buy a 5090 and sell your 4090, otherwise your 4090 will be worth £500 when the 6090 launches'. A moot point if supply is non-existent I suppose.

I went a bit deeper with math in my head, kind of like this: I sell 4090 for £1600 (£200 below purchase price), I need to add about £500 to get 5090 (likely more but let's be positive!). Resale value of 5090 will be possibly low, so I can get only £1k back when time comes. So that's -£1.7k when 6k series comes (assuming I want to upgrade to 6090 and need to pay 1k on top of 5090 resale again). Now, if I don't sell 4090 and then can sell it when 6090 comes, for just below low value of 5090 assuming £1k (£700 let's say), that's £1k down from my purchase price. Not £1.7k. Ergo, I'm actually losing much much less money by not swapping 4090 to 5090 but waiting for 6090. And I can invest that in ISA and actually need to pay even less when 6090 appears. With that 5090 looks like it would be a horrible choice to upgrade to.

 

[Glanza]

 

[xaldub]

I went a bit deeper with math in my head, kind of like this: I sell 4090 for £1600 (£200 below purchase price), I need to add about £500 to get 5090 (likely more but let's be positive!). Resale value of 5090 will be possibly low, so I can get only £1k back when time comes. So that's -£1.7k when 6k series comes (assuming I want to upgrade to 6090 and need to pay 1k on top of 5090 resale again). Now, if I don't sell 4090 and then can sell it when 6090 comes, for just below low value of 5090 assuming £1k (£700 let's say), that's £1k down from my purchase price. Not £1.7k. Ergo, I'm actually losing much much less money by not swapping 4090 to 5090 but waiting for 6090. And I can invest that in ISA and actually need to pay even less when 6090 appears. With that 5090 looks like it would be a horrible choice to upgrade to.

I did similar calculations and came to the conclusion that by waiting this gen out then picking up a 6090 I'll be approx 1k better off than if I sell my 4090, buy a 5090 FE then did the same when the 6090 is launched. My assumptions are : 6090 FE will be £2400 and resale value of a 5090 FE will be 1K at best ( personally I think it will be slightly less ).

 

[Absolutely Clueless]

I went a bit deeper with math in my head, kind of like this: I sell 4090 for £1600 (£200 below purchase price), I need to add about £500 to get 5090 (likely more but let's be positive!). Resale value of 5090 will be possibly low, so I can get only £1k back when time comes. So that's -£1.7k when 6k series comes (assuming I want to upgrade to 6090 and need to pay 1k on top of 5090 resale again). Now, if I don't sell 4090 and then can sell it when 6090 comes, for just below low value of 5090 assuming £1k (£700 let's say), that's £1k down from my purchase price. Not £1.7k. Ergo, I'm actually losing much much less money by not swapping 4090 to 5090 but waiting for 6090. And I can invest that in ISA and actually need to pay even less when 6090 appears. With that 5090 looks like it would be a horrible choice to upgrade to.

My version of it is below. I was much more conservative on the 4090 side, perhaps unfairly so. It also doesn't take into account 5090 having no supply for what I expect will be MONTHS.

If I sell my 4090 for £1300 (being conservative. Paid £1600), and buy a 5090 for £2300 and sell it for £1200 (random guess on its value in 2 years), The total cost of ownership for those two cards is £1400.

If I keep my 4090 and it drops to £500 in 2 years, the total cost of ownership is £1100. So £300 more for a 5090 (aside from the fact i'd need £400 on a new case and PSU, but i would need that for a the 6090 anyway).

 

[Hostile_18]

I did similar calculations and came to the conclusion that by waiting this gen out then picking up a 6090 I'll be approx 1k better off than if I sell my 4090, buy a 5090 FE then did the same when the 6090 is launched. My assumptions are : 6090 FE will be £2400 and resale value of a 5090 FE will be 1K at best ( personally I think it will be slightly less ).

I think your under estimating the 5090 resale value substantially. It will still have 32gb of vram, still be amazing for AI and almost certainly the 60 series will have months of supply issues and expensive on launch. Nvidia never really gives performance for free anymore and AMD are too far behind to be a threat at the top end.

 

[Octoplex7]

I don't want to be a total downer, there's still time for Nvidia to improve things but one has to remember - the moment they release GPUs to the market, the next generation is already in internal testing (and not just design) phase usually, which means they already know what to expect from its performance and process. And so far the only thing I've heard from their VP of AI stuff was that it's very very hard to scale up GPU performance anymore and the only way forth is AI and even more MFG. Which is consistent with what Nvidia CEO repeated all the time about 5k series for months now. In other words, I don't expect much from 6k series currently and will not have any big hopes.

What newer process do these days is lower power use OR increase performance - not both anymore. So we could get faster but with even more power use or similar performance but less power used - unless they change the design of the chip a lot but they seem to have no real incentive to do it. Most of their r&d is focused on AI for a while now, whilst at the same time they are a monopoly with gaming GPUs. So, they can do whatever they want really and people will buy it anyway, because it's Nvidia. So what I expect them to do is to go for the best revenue, which means AI focus (Stargate project well keep them busy for many many years to come and with full pockets) and just do bare min. for gamers.

Without getting in to the whole Moore's law being dead or not debate, Nvidia have been telegraphing this for a while: they are no longer a graphics company, they are an AI company. Essentially their message to their gaming customers is: you need to get used to our new products being about improvements in AI (measured according to increases in TOPS etc). Our more important client base wants AI, so we are going to focus on persuading them to regularly update their Nvidia products by giving them more AI. We will find ways of using software to persuade gamers that 5070 = 4090 and so on.

 

[waspy88]

For the first time in a long time.. since the 780ti I think, I am going to sit this one out. The heat on the FE around the memory bridge and the ridiculous stock situation on the AIBs make this not a great launch for me. I am also sure the 4090 raster improvements were better than the 3090 by a fair margin. It feels like they have a 5090ti waiting to come out at some point.. if not I will just eek more usage out of my 4090. With us getting DLSS 4.0 anyway we get a lot of the benefits of the 5 series, plus their tech makes native 4k meaningless. Can someone convince me otherwise please.. I don't want to be like cr*ck addict on launch day refreshing to buy one at the last minute

 

[Gazanimal]

I’ll be upgrading to a 5000 series as and when the availability is good as I think release day will be a washout for most.

Upgrading from a 4000 series doesn’t make too much financial sense but if you want the extra performance then enthusiasts will pay it.

I’m still rocking a 3080Ti, but it will be a good leap for me to go to the 5000 series for my work and play.

Although depreciation is generally a problem when buying goods.

 

[waspy88]

Yeah going from a 3090 or 3080 looks to be a 200% increase at 4K + RT with most games, you'll love it!