NVIDIA GeForce RTX 5080 Founders Edition

[Darc Requiem]

Honestly that's the biggest dissappointment AMD ditched the high end at the wrong time Nvidia just phoned in this generation my guess is somthing didn't work out with 3N likely because the B100/200 were rushed to keep margins super high....

Yeah I've expressing the same thought. The performance crown was there for the taking. AMD once again misses an open opportunity.

 

[Darmok N Jalad]

Seeing the poor uplift for this, I'm baffled by AMD's decision to go and sit in a corner. Just redoing a slightly larger 7900 XTX on N4 would have probably been enough to beat this.

I think AMD tried something bold with RDNA3 with the MCM approach, but it failed to produce the desired results, it was too expensive, or it couldn't scale like they hoped. I'm guessing the chiplet 7900XT/XTX-level of RDNA4 was scrapped, and so all we get is the monolithic version. Basically, imagine if Zen and the chiplet concept had failed to compete and scale--they would have had to go backward in order to go forward and recover.

 

[Vya Domus]

I get what you're saying, but they have the IP to convincingly beat the 5080 and probably even the 4090. Doing so would have been beneficial for the rest of the lineup.

It just isn't a good time to pull "one last hurray" type of thing, their main customer base is of those who already bought AMD which is very tiny.

If you want to have a chance to displace some of Nvidia's market share you need to wait for them to pull a couple more of these "4080 ti super mega multi frame GDDR7 neural AI tops 1000$" so their customers become even more indifferent towards Nvidia. Ryzen was successful in part also because people didn't want to pay for a 4 core CPU with 5% better clocks for the 5th time.

 

[W1zzard]

Ps. I only check out TPU review to see "highly recommend" badge as usual - this time for 2 year 13% generation uplift
I get it by now....job is a job and all these websites depend on manufactures $...

alright, i'll feed the troll. what would you recommend to me if i had $800 to $1200 to spend?

 

[Vya Domus]

alright, i'll feed the troll. what would you recommend to me if i had $800 to $1200 to spend?

We know there isn't really anything else in that price range but I still think it's silly to consider a positive aspect the fact that Nvidia didn't up the price considering that there would be almost nothing to back up that price hike.

 

[TheinsanegamerN]

Seeing the poor uplift for this, I'm baffled by AMD's decision to go and sit in a corner. Just redoing a slightly larger 7900 XTX on N4 would have probably been enough to beat this.

Either that or designing a larger monolithic chip. Either way, there's opportunity here.

I think AMD tried something bold with RDNA3 with the MCM approach, but it failed to produce the desired results, it was too expensive, or it couldn't scale like they hoped. I'm guessing the chiplet 7900XT/XTX-level of RDNA4 was scrapped, and so all we get is the monolithic version. Basically, imagine if Zen and the chiplet concept had failed to compete and scale--they would have had to go backward in order to go forward and recover.

The other baffling thing with rDNA3 was the shaders. They doubled the shader count per CU for, apparently, 0benefit. The RT performance per CU isnt any better than rDNA2, but it DOES bloat the die sizes up for no reason.

Honestly they would have been better off just taking rDNA2 and expanding it on the newer nodes instead.

 

[Krit]

I got the 4060 Ti 16GB on sale for not significantly more than the 8GB is now going for so I think it was a fair deal. Like they say there's no bad cards, only bad prices. It will last a few years at least for 1080p gaming at a decent level, it's very efficient, cool and quiet and that's all I want.

That's great you are lucky! But do not advertise nvidia for free! That's how this all BS works in favor to nvidia. You are basically supporting higher prices by doing this.

 

[cellar door]

alright, i'll feed the troll. what would you recommend to me if i had $800 to $1200 to spend?

Not giving Nvidia any awards for a card that doesn't deserve any would be the right thing to do.

 

[PixelTech]

I bet you managed to sell your 4090 for a disgusting amount of money though! Use that wisely once the dust has settled and the scalpers have moved on.

Unfortunately not, got my money back though.

 

[Shakallia]

Highly recommended

 

[oxrufiioxo]

alright, i'll feed the troll. what would you recommend to me if i had $800 to $1200 to spend?

Still that makes it sound like it gets the highly recommended badge by default like the only prerequisite is having no competition.

 

[Chrispy_]

Loses the same or even more performance % with RT on. New RT cores my *ss...
Despite more then 2x AI TOPS performance, only 0,1 sec gains in LLM test? It should be faster than 4090 in AI apps according to the stats.
View attachment 382345

This entire generation is a scam...

Yeah, actual IPC appears to be zero.

 

[W1zzard]

Still that makes it sound like it gets the highly recommended badge by default like the only prerequisite is having no competition.

there is competition, 4080, 4080 super, 4090, 5090, 7900 xtx, 7900 xt, ps5, xbox, buy $1k weed and get high in that order (imo). i would buy 5080

 

[AnotherReader]

I think AMD tried something bold with RDNA3 with the MCM approach, but it failed to produce the desired results, it was too expensive, or it couldn't scale like they hoped. I'm guessing the chiplet 7900XT/XTX-level of RDNA4 was scrapped, and so all we get is the monolithic version. Basically, imagine if Zen and the chiplet concept had failed to compete and scale--they would have had to go backward in order to go forward and recover.

I think the failure of RDNA3 to be a better competitor to Ada is unfairly laid on the MCM approach though it didn't help with power usage. RDNA3 had two misses: slightly inferior node to Nvidia (even 5% can be enough in this contest) and missing its frequency targets. Judging from AMD's slides, RDNA 3 should have clocked about 15% higher. RDNA 4 leaks indicate that they may have fixed the second issue. As for costs, the packaging for the MCM approach is more expensive, because of the higher bandwidth between the MCDs and the GCD.

 

[oxrufiioxo]

there is competition, 4080, 4080 super, 4090, 5090, 7900 xtx, 7900 xt, ps5, xbox, buy $1k weed and get high in that order (imo). i would buy 5080

Me too but I wouldn't feel good after waiting 12-24 months for it....

I don't even smoke weed and it at least made me debate internally if that was the better option

 

[Chrispy_]

Very interesting the 4080 Super has 88% of the 5080's performance at 4k, but 90% at 4k ray tracing. I would have expected the ray tracing performance to have a bigger gap than the regular performance, due to improved RT cores and more memory bandwidth from GDDR7.

Unless of course the "new" RT cores are just marketing scams like "5070=4090"

 

[RGAFL]

lmao, sure buddy. Go back to school, the grownups are talking.

Sure, I need to go back to school. When the big push is for DLSS 4 to show the big frame improvements with MFG when their not coming from the actual hardware apart from small percentage increases.

I worked with these things over 6 months ago, I said what to expect. It turned out roughly right. 10-20% short of a 4090 game dependent. When Nvidia themselves are comparing a 5070 to a 4090 with all sorts of shenanigans and people are swallowing it.

But I need to go back to school.

 

[chispy]

I feel a bit disappointed on this gpu :/ , i feel like this card does not deserve the 5080 naming and should have been better calling it at best 5070Ti , that's my personal opinion. Don't like it ...

 

[LastDudeALive]

Seeing the poor uplift for this, I'm baffled by AMD's decision to go and sit in a corner. Just redoing a slightly larger 7900 XTX on N4 would have probably been enough to beat this.

The RX 7000 vs RTX 40 competition showed that nobody wants to buy AMD even if they offer better performance for cheaper. Ray tracing, DLSS, frame gen, CUDA, and AI are features that sell GPUs, like it or not. There is absolutely no way AMD could "beat" the 5080 by making something that's slightly better and slightly cheaper.

 

[Chrispy_]

I think the failure of RDNA3 to be a better competitor to Ada is unfairly laid on the MCM approach though it didn't help with power usage. RDNA3 had two misses: slightly inferior node to Nvidia (even 5% can be enough in this contest) and missing its frequency targets. Judging from AMD's slides, RDNA 3 should have clocked about 15% higher. RDNA 4 leaks indicate that they may have fixed the second issue. As for costs, the packaging for the MCM approach is more expensive, because of the higher bandwidth between the MCDs and the GCD.

From an interview with AMD at (and immediately after) the 7900XTX's launch event, engineer(s) basically explained that they'd missed their internal clock targets due to instabilities found with the MCM design.

So as much as it pains me to see chiplets fail for GPUs right now, I think the blame for RDNA3's disappointing launch can be squarely blamed on chiplets.

I'm just hoping AMD haven't given up on the idea completely. Large dies on cutting-edge, highly contested process nodes is really hurting performance/$ for us consumers.

 

[JustBenching]

127% * 115% = 146%. We surprisingly received 46% more value for our money. Not impressive but it's definitely not bad considering all cicrumstances.

I said the same thing but people disagreed cause...I don't know, they can't do math or something? 46% increased per per dollar, by the market leader - gen on gen. People 2 weeks ago were literally attacking me when I suggested that the competitor should release a GPU at teh same price for 40% more performance gen on gen....they said that's impossible

 

[TheinsanegamerN]

The RX 7000 vs RTX 40 competition showed that nobody wants to buy AMD even if they offer better performance for cheaper. Ray tracing, DLSS, frame gen, CUDA, and AI are features that sell GPUs, like it or not. There is absolutely no way AMD could "beat" the 5080 by making something that's slightly better and slightly cheaper.

Ok, so which AMD gpus offer "better" performance then nvidia for a cheaper price?

 

[AnotherReader]

Either that or designing a larger monolithic chip. Either way, there's opportunity here.

The other baffling thing with rDNA3 was the shaders. They doubled the shader count per CU for, apparently, 0benefit. The RT performance per CU isnt any better than rDNA2, but it DOES bloat the die sizes up for no reason.

Honestly they would have been better off just taking rDNA2 and expanding it on the newer nodes instead.

That has been AMD's problem for ages; even when they have good IP, they don't take advantage of it by going for the kill. With the 4870 and 5870, if they had gone for big die sizes like Nvidia, they would have held the halo for a couple of years. A bigger 6900 XT would have been unmatched until Ada. As for dual issue, it works, but the compiler gets limited benefit from it, and it saves die area which matters in this age of expensive nodes.

 

[scooze]

Apparently, 4080 and 4090 were ahead of their time, Nvidia offered too much, these were cards from the future. True, they asked for a lot for such an opportunity. And now welcome to the present.

 

[Krit]

Apparently, 4080 and 4090 were ahead of their time, Nvidia offered too much, these were cards from the future. True, they asked for a lot for such an opportunity. And now welcome to the present.

Wrong GTX 1080 Ti was way better than that p/p wise not even in the same universe!