• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Best time to sell your used 4090s is now.

On you own risk , maybe you sell your 4090 now for 2.5k and then you wont find any 5090 under 4k :) , probably wont be this extreme but who knows? :D
In the world of bots, scalpers and ai. We have to be the anomaly to survive.
 
  • Like
Reactions: izy
I don't plan on selling my 4090 anytime soon nor do I plan on paying $1000 more than what I paid for it with the model replacing it. There seems to be room for a couple cards in between the 5080 and the 5090 and with little competition this series of card will probably be around for a while.
Yeah the specs are insane, 5070 to 5080 all squeezed together very close to each other on a one to one basis, then a humongous gap to the 5090.
 
Yeah the specs are insane, 5070 to 5080 all squeezed together very close to each other on a one to one basis, then a humongous gap to the 5090.

Honestly to me I wouldn't be surprised if AIB 5080s end up significantly more expensive than the 999 usd msrp. Wouldn't be surprised if most models start around 1200 and go up to 15-1600.
 
Honestly to me I wouldn't be surprised if AIB 5080s end up significantly more expensive than the 999 usd msrp. Wouldn't be surprised if most models start around 1200 and go up to 15-1600.
Those are the prices in EU. Reference cards go for ~1200-1250€. The problem is, are we going to get any reference cards at stock? Cause everything else that is not reference is over 1.5k. And with that huge gap between the 5080 and the 5090 in prices, I think nvidia is just allowing AIBS to milk the cows, they can easily sell 5080s for 1500-1700€ since the 5090 is going to be over 2.5k unlike last gen.
 
Video about possible power bottleneck for the Blackwell and Titan/5090ti in plain site.
View attachment 379810
Is it me or the increase of TDP indicates a less power efficient architecture?

Honestly to me I wouldn't be surprised if AIB 5080s end up significantly more expensive than the 999 usd msrp. Wouldn't be surprised if most models start around 1200 and go up to 15-1600.
Yeah I buy only FE now, also that FE isnt factory overclocked so no instability like I have had with previous AIB.
 
Is it me or the increase of TDP indicates a less power efficient architecture?


Yeah I buy only FE now, also that FE isnt factory overclocked so no instability like I have had with previous AIB.

I love the FE they are just really hard to get... I wasted 3 1/2 months trying to get a 4090 FE not becuase of the msrp but becuase it's my favorite card aesthetically.

Could be that it's less power efficient or that Nvidia pushed it way too hard to get it to it's performance level. A 4090 could be dropped to less than 350w and you'd still get 95% of the performance. We won't know till actual reviews hit and they do power scaling test on them. It'll be easier to test with the 5070-5080 becuase they have the same amount of vram.
 
I love the FE they are just really hard to get... I wasted 3 1/2 months trying to get a 4090 FE not becuase of the msrp but becuase it's my favorite card aesthetically.

Could be that it's less power efficient or that Nvidia pushed it way too hard to get it to it's performance level. A 4090 could be dropped to less than 350w and you'd still get 95% of the performance. We won't know till actual reviews hit and they do power scaling test on them. It'll be easier to test with the 5070-5080 becuase they have the same amount of vram.
I wish someone, anyone, would include power scaling results in the pre-release reviews.

I'm most interested in seeing performance at the 330-350W range.
 
I wish someone, anyone, would include power scaling results in the pre-release reviews.

I'm most interested in seeing performance at the 330-350W range.


That testing is probably beyond the scope of a standard review although i find it more interesting than OC results.

Maybe in a separate piece.


Nvidia does seem pretty confident in 5090 perfomance though. I can't remember the last time we got reviews a week before launch.
 
If the stuff I'm selling is in perfect working order and great cosmetic condition, I adopt a moral code: no more than 70% of what I initially paid. My objective here isn't to make a profit, just help fund my upgrade. Which is the entire point. Oftentimes I just pass my hardware to my brother and sell his old stuff. I'm likely going to end up giving him my RTX 4080 and selling his RTX 3070 when I upgrade to the 5090. That way I get a cushion on the purchase, he gets a massive upgrade for free, and everyone's happy.
i say sell it at a price so the buyer will not turn around and sell it for more. (or ‘research’ and find out if they will be reselling it in the future… no point in making it easy for the buyer to make money off of it.)

i
 
Video about possible power bottleneck for the Blackwell and Titan/5090ti in plain site.
View attachment 379810

If Nvidia really released a TITAN with a Full GB202 & 48GB VRAM (thanks to GGDR7 3GB chips), then it would definitely be sold $3000+

But imo unless RDNA 5/UDNA in 2026 is really kicking ass and close to a 5080 Ti/5090 performance, I doubt we will ever see a 5090 Ti with those specs.

LOL, I have two of these wonderful CPUs, the fastest CPUs money can buy -14900K and 13900KS, trying to sell one of them for quite some time now, and nobody for some reason wants to buy it. People are so dumb, you would not believe it... :laugh:
Yeah they're so good that you're trying to sell one... :roll:
The 9950X3D will just destroy them and won't need any microcode update ;)
 
If Nvidia really released a TITAN with a Full GB202 & 48GB VRAM (thanks to GGDR7 3GB chips), then it would definitely be sold $3000+

But imo unless RDNA 5/UDNA in 2026 is really kicking ass and close to a 5080 Ti/5090 performance, I doubt we will ever see a 5090 Ti with those specs.


Yeah they're so good that you're trying to sell one... :roll:
The 9950X3D will just destroy them and won't need any microcode update ;)
Yeah I doubt we will see a fuller fat gpu unless the ai market peaks in my guesstimate.
 
The Asus Tuf 4090 OC easily sells for $1,400. The cards were $1,850 plus tax when released on Oct 12, 2022.
 
I'm lucky i got mine used like new for only $1750 CAD, user who had it paid like 2K for it lol.
 
Yeah I doubt we will see a fuller fat gpu unless the ai market peaks in my guesstimate.
Yeah I doubt it either... Nvidia want to make huge margins with their AI chips so a TITAN even at $4000 wouldn't be a great value compared to $30K to $40K for each AI chip...

The Asus Tuf 4090 OC easily sells for $1,400. The cards were $1,850 plus tax when released on Oct 12, 2022.
If the 5090 are out of stock for weeks or even months then they will be sold between $2500 and $3000 and the 4090 available online will probably be sold around MSRP ($1600).

I'm lucky i got mine used like new for only $1750 CAD, user who had it paid like 2K for it lol.
I got my 4090 SUPRIM LIQUID X at MSRP ($1750) and even right now they're very hard to get and are sold $2000+
 
Jay already got his 5090 and fit it into a small form factor case. Looks amazing imo!

 
This isn't aging very well.
 
Jay already got his 5090 and fit it into a small form factor case. Looks amazing imo!

We already know it's a SFF GPU... what we wanna know are the Temperatures + Power Consumption lol
 
We already know it's a SFF GPU... what we wanna know are the Temperatures + Power Consumption lol
We are 7 days away from embargo. Which means they have a week or so worth of testing. I am personally curious on long benchmark runs outside a 5 minute benchmark to see if the card throttles itself or the system in long game sessions. Majority of reviews will probably be based on 5 minute benchmarks with an open test bench.
 
We are 7 days away from embargo. Which means they have a week or so worth of testing. I am personally curious on long benchmark runs outside a 5 minute benchmark to see if the card throttles itself or the system in long game sessions. Majority of reviews will probably be based on 5 minute benchmarks with an open test bench.
Open test bench or case shouldn't make a difference. You are not going to be using a 500w GPU on a case that can't actually handle it. Was in that situation with my 3090, it was scorching hot in the o11, migrated the system to a fractal torrent and the temps dropped by 11C on the GPU. If opening the side panel drops temperatures by more than 2-3C then your case can't handle the hardware you have in there.
 
Open test bench or case shouldn't make a difference. You are not going to be using a 500w GPU on a case that can't actually handle it. Was in that situation with my 3090, it was scorching hot in the o11, migrated the system to a fractal torrent and the temps dropped by 11C on the GPU. If opening the side panel drops temperatures by more than 2-3C then your case can't handle the hardware you have in there.
According to many believe that the 5090 fe in that tiny case will not throttle using full power 575 watts) and will pleasantly surprise those with personal experience and a thing or two about laws of physics.
I'm with you although the warm ambient air eventually will recycled into the system unless air conditioning is used in these cooler winter days. Unfortunately I don't believe I'll find my answer from scripted reviews so I'll have to do my own research. lol.
 
We already know it's a SFF GPU... what we wanna know are the Temperatures + Power Consumption lol
SFF is up to 2.5 slots and Nvidia limited the 5090 to only 2.0 slots. A lot of performance just left on the table and that single 16 pin will prevent it from hitting 3.0 GHz. At least come up with a 22 pin that can handle 900W. Now what's the point we're already hitting a wall.
 
I was super stoked for the 5090, but once I saw that it's only 30% better, I decided to keep my 4090 for two more years. Nvidia is holding back way too much this generation. The 6090 will be perfect for GTA 6 :D

Selling my 4090 would pay like 50% of the 5090 cost in my country.
 
SFF is up to 2.5 slots and Nvidia limited the 5090 to only 2.0 slots. A lot of performance just left on the table and that single 16 pin will prevent it from hitting 3.0 GHz. At least come up with a 22 pin that can handle 900W. Now what's the point we're already hitting a wall.
Yeah performance improvements are getting harder and harder without AI, hence Nvidia using Tensor Cores to do most of the work on RTX 50s. I think people need to understand that Moore's Law might really be dead (unless someone finds a "magic recipe" but yeah..).

The era of huge Performance gains & good Pricing died with the GTX 10s. We're only getting more and more expensive GPUs with AI in them now. The only reason the RTX 50s are cheaper (except the 5090) is because they're using a very similar process node (TSMC 4nm) with the same memory bus sizes (256-bit or 192-bit) that are not expensive to make.
I'm pretty sure Nvidia will raise their prices with the RTX 60s since they will be using a TSMC 3nm node (probably close to N3P or N3X).
 
If I decide to sell my 4090, what would be a fair price , Micro Center would do trade in of 1150 dollars , on eBay , would get a lot more , thought about selling here , not sur yet , no hurry in my case , but my concern is if a 5090Ti comes out in 6 months , do noy want to go thru a selling and buying again.
Screenshot 2025-06-26 115954  primemenber.png
 
Back
Top