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

NVIDIA GeForce RTX 5080 Founders Edition

Nothing in stock here, and cheapest one is $1450cad so about $1640cad after tax. Over 300% more than I paid for my RX6800 ($500cad).

I wonder if Nvidia will have a $500cad (~$350usd) card this gen??
 
Nothing in stock here, and cheapest one is $1450cad so about $1640cad after tax. Over 300% more than I paid for my RX6800 ($500cad).

I wonder if Nvidia will have a $500cad (~$350usd) card this gen??
Even if they do deign to offer us a 5060 around that price, going by the 5080's improvement over the 4080, it would be slower than your 6800.
 
I'll take screen shots and post it in the site issues thread.
EDIT: Nevermind, it just showed up.

WOW! You got north of 3300mhz stable? Nice! So the 5080 at stock may have been less than great but they OC well!
 
Not all supers were good 2080 super and 4080 super was let down, hardly much of an improvement and I think the 5080 super will be the same with maybe 3-5% improvements
5080 is already using the full gb203, which was not true of the 4080 and ad103. If there is going to be anything slotted between the 5080 and 5090, it would have to be more of a 5080 Ti on a cutdown gb202 die. If they do that (and it's a big if imo), then I suspect you'll get something better than 3-5%.

Given no competition at that performance tier though...I'm just not sure they'd do that. Only if their yields on gb202 is so bad for 5090s that they can get enough 12-15k shader dies out of it and maybe cut down to a 384-bit 24GB card and call it a 5080 Ti.
 

12VHPWR Melty connector lol no thanks. I want 4x 8pin connectors.
Also isn't it time to move towards fully enclosed GPUs with internal USB4 80Gbps USB-C GPUs that can fit into a 5.25 bay or externally? 4 slot PCIe cards no longer make sense because of the weight. PCIe is also vastly more unreliable because of the pin count compared to USB-C. USB4 80Gbps USB-C is far more elegant. Use 4x quality USB-C 240W cables to increase power. Internal USBC 240W cables are a far more elegant modular and standardized connection for internal PSU connections too.
 
Last edited:
I'll take screen shots and post it in the site issues thread.
EDIT: Nevermind, it just showed up.

WOW! You got north of 3300mhz stable? Nice! So the 5080 at stock may have been less than great but they OC well!
JayZ did a video on overclocking a sample and he was impressed too.
 
And its true value as a 70 class tier GPU is between $600-700 mind you. Nvidia is the first scalper and retailers the second ones. The last ones are sadly, the only ones called such.
Artificial scarcity is the oldest scam. "Sorry guys the demand is just too high and we can't make them fast enough." Make up some nonsense about no memory or something. I bet Nvidia is sitting on a mountain of 5090s and 5080s and are just releasing in small amounts so that scalpers (nvidia themselves) can buy up stock and then inflate prices. It doesn't need to be like this. The number of units sold doesn't increase if you keep supply artificially low. The profit margin isn't higher if you artificially affect supply like this. Nvidia could sell directly 1 per customer if they wanted to. They don't want to. So they create this market. I have a review 5090 which i got for free and I'm not convinced it is worth it. 20% real world improvement slop. the 4x frame generation is nice, but its not worth the money. My 3090 and 4090 works just as well. The demand is artificially high, and the supply is artificially low. Wishing Tenstorrent (Jim Keller) would get into the frame generation market and just blow the competition out of the water. We need a massive disruptor like deepseek that just instantly devalues any of the competition. I think we are on the cusp of this. The moment that AI TOPS start being the only thing that matters with full frame generation, where nothing is rendered at all, with 8X, or 16x Frame Generation, it will give a lot of other companies the ability to get in on the action. I'm hoping AMD and Intel 4x frame generation when it releases gives us a twofold increase on current non 50 series cards. I believe the next iteration of Unreal Engine 6 with full frame generation will make something like this happen. Creating high performance AI TOPS cards is a lot more doable than creating cutting edge rasterization cards. I'm not interested in whether it is personally worth it for me, or whether I can get a card, since I mostly play old isometric rpgs (modern games are trash, for me), I can easily afford this card, even at the scalping rates, but rather what I am more concerned about is I want more people to be able to afford high performance gaming. Its become a snobby mess. 2000 usd on a gpu every year. These cards shouldn't be more than $300 for halo cards each regardless of inflation. How can any teenager get into this hobby without their parents massively subsidizing it? So this hobby is now only for old guys with money? You know how stale and cringe that is? 5090s smell like mothballs and should come with retirement home discounts for us old guys, since no one who is young is interested. The nvidia brand isn't relevant at all with young people since they cannot afford it. Think about this. Get the young people interested and higher volume of products = higher profits, not low volume, small high profit like you currently have it. And all the people without just end up not playing anything? They're killing PC gaming. Don't be surprised if PC gaming dies in the process. You did it to yourself Nvidia. Destroying your own market for short term profit and turning PC gaming into an old man hobby. Prune juice vibes. RTX stands for ReTirement eXperience. Your hair is gray, Grandpa Jensen. Think about this. We need to get the next generation of players interested in Enthusiast PC gaming or the hobby will die. It is very close to RTX retirement. Young people play phone games. They view PCs as archaic devices for Old guys with money. Think about this image and how you are perpetuating it. Everything relevant is going to be console and mobile gaming. Enthusiast PC gaming is a tiny fraction of the gaming market as a result. You did this because few normal people can afford to get into it. AMD refusing to make a high end GPU is the writing on the wall. How many 4090s did Nvidia sell? I bet its a drop in the bucket and will be a break even drop in the bucket on 5090 as well. I suspect that Nvidia will also get out of the Enthusiast PC GPU market and only focus on mid range, handheld and datacenter.
 
Artificial scarcity is the oldest scam. "Sorry guys the demand is just too high and we can't make them fast enough." Make up some nonsense about no memory or something. I bet Nvidia is sitting on a mountain of 5090s and 5080s and are just releasing in small amounts so that scalpers (nvidia themselves) can buy up stock and then inflate prices.

I doubt it. I think these cards are and will genuinely remain scarce - I think Nvidia knows there is a limited market for them even at MSRP, content creators more and more rely on online tools rather than complicate their lives with home brewing Stable Diffusion and other tools that get outdated quickly…

And as for wealthy gamers - there are still RTX 5080 and 5090 cards available both locally in stores and across Europe. Sure, for scalper +50% prices on top of already expensive product, but considering how almost non-existent stock was at release, I’m sure not every wealthy gamers already got their card?

Nvidia clearly focussed on server AI this generation. If you watched CES announcement of Blackwell, they spent large amount of time describing some vague future of AI acceleration of everything and a kitchen sink, without any clear dates and tools on how this will work on current generation. AI is the future - and the future isn’t now.

Right now they’re focussed on server hardware. And they decided to do a very limited runs on GeForce toys.

I‘m waiting on who will be the first to successfully sell Nvidia FX 5900 for $2000+. :p

IMG_0121.jpeg
 
Last edited:
Artificial scarcity is the oldest scam. "Sorry guys the demand is just too high and we can't make them fast enough.".
It definitely helps, know a store that got something like 200 5080s at launch and were delivered 3 5090's o_O, of course the news is the 5090s are the ones on fire with sales. 200 is low as well, 3 is just no words for it.
 
i did some 5080 undervolt OC and its Avg 3-4% faster in native VS 4090 Stock.

i have 2x 5080 i did tested both almost same results,i also have one 5090 and its great in 5K vs 4090

4090 OC +5-6% so 5080 OC is 1-2% slower when we compare them OC vs OC

But if ppls can buy 5080 someday for good prices,its very close to 4090 if u OC it,and cheaper.
 
i did some 5080 undervolt OC and its Avg 3-4% faster in native VS 4090 Stock.

i have 2x 5080 i did tested both almost same results,i also have one 5090 and its great in 5K vs 4090

4090 OC +5-6% so 5080 OC is 1-2% slower when we compare them OC vs OC

But if ppls can buy 5080 someday for good prices,its very close to 4090 if u OC it,and cheaper.
Doubt it. Wanna test it on cp2077?
 
What issue?
I think he was mentioning the issues Roman had with his 5080FE, detailed here. In short, he received a dud straight from nvidia. And he's not sure but suspects there might be other units affected as well. I think another reviewer had the same issue, he mentions something along those lines in his review. Great QC but hey, what can you do when you send reviewers the cards a week after production.

At least this one's caught during review because it's black or white. Unlike the terrible memory temperatures issue on 3090 FE's which was only caught weeks later.
 
Doubt it. Wanna test it on cp2077?
im not sure about Cp2077 5080 vs 4090 but i dont think there is much differences if gaming whitout RT enabled and Oc vs Oc

5090 and using samsung G95sc there is bigger differences vs 4090, i think maybe because huge memory bandwidth
 
I'm not sure believe that. Not calling you a lair as you likely read that elsewhere, only that 90% on such a new lith process is very unlikely.
I found several websites saying that TSMC 3nm yields were around 85-86% so that is pretty huge! Apple have been using TSMC 3nn since they released the A17 Pro in September 2023 so it's been a while, it's not really surprising.
 
Ouch:


Lots of NVIDIA GeForce RTX 5090 & 5090D GPUs Are Getting Bricked, Possibly Due To Driver, BIOS or PCIe Issueshttps://wccftech.com/nvidia-geforce-rtx-5090-5090d-gpus-getting-bricked-possibly-driver-bios-pcie-issues/

Der8auer also posted a video in which he prophesied problems - neither Nvidia nor AIB partners had apparently have any time to properly test the end product...
Picture someone who discovers this. All that anguish to secure a scalped high end card at a massive mark up and then the thing does something like this. Imagine the frustration. The more common a card model is, usually the more issues have been ironed out. These high end cards are like massively expensive and finicky sensitive supercars that you have to baby. Endless power issues, underspecced connectors melting, pcb cracking near the PCIe connector because of weight, and buying privilege and no stock. The run of the mill car is infinitely more reliable. Just get a popular 4060/4070 model and it works. You could make the case with that, that mid range high volume sales cards are almost always more reliable. Low volume, high end is asking for trouble.
 
I found several websites saying that TSMC 3nm yields were around 85-86% so that is pretty huge!
Ok, but how is that 86% broken down? Is that fully functional? Is it binned down? If they don't state the break-down then those numbers can not be taken too seriously.
 
Ok, but how is that 86% broken down? Is that fully functional? Is it binned down? If they don't state the break-down then those numbers can not be taken too seriously.
These yield figures always annoy me. What kind of chip are they using as a test vehicle? What's its die size; does it have any redundancy (easy with large last level caches)? A defect rate is more useful, but TSMC sometimes takes time to share those.
 
Ok, but how is that 86% broken down? Is that fully functional? Is it binned down? If they don't state the break-down then those numbers can not be taken too seriously.
The guy on High Yields says that yields with a full GB202 enabled are probably about 56% or more, so the 5090 with only 170SM out of 192SM definitely has a lot more yields... Maybe 75-80% would be a good number knowing that TSMC usually have great yields 1 year+ after release.

Ouch:


Lots of NVIDIA GeForce RTX 5090 & 5090D GPUs Are Getting Bricked, Possibly Due To Driver, BIOS or PCIe Issueshttps://wccftech.com/nvidia-geforce-rtx-5090-5090d-gpus-getting-bricked-possibly-driver-bios-pcie-issues/

Der8auer also posted a video in which he prophesied problems - neither Nvidia nor AIB partners had apparently have any time to properly test the end product...
RTX 2080 Ti, 3090, 4090 and now 5090 all had issues at release! Seems like a constant pattern at Nvidia...
 
Last edited:
Back
Top