NVIDIA GeForce RTX 5080 Founders Edition

[Daven]

looks like about 12% performance over 4080 at 4K? Better than I was expecting, but still not what it should be. This morning I was guessing 8%.

It's 15% performance over the 4080 (100 divided by 87).

 

[10tothemin9volts]

As with the 5090, the 4080 Ti..I mean the 5080 scales unfortunately almost linearly in price to performance and power efficiency due to the same process node and NV not having put something secret out of their drawer:

• only a 9.89-11.11% power efficiency improvement over the 4080 (Super)
• 13.6 % Performance per $ improvement over the 4080 Super @4K resolution

Whats up with:

• Idle power consumption increased from 14W 4080/13W 4080 Super to 20W 5080
• V-Sync 60 Hz power consumption increased from 68W 4080/67W 4080 Super

@W1zzard I wonder whats the idle power consumption using an (OLED) TV at 4K, 8bit, 120Hz, 4:4:4 and HDMI VRR enabled.

And the DLSS 4 Multi Frame Generation feature has received a critical review:

And the DLSS 4 transformer based upscaling is available for all RTX cards, from the "new" RTX 50 to RTX 20 series, so also no reason to upgrade, not that I would have (if there was a 24GB VRAM version of the 5080 using 3GB per VRAM chip, I would seriously consider upgrading, though I really want a cheaper than the 5090 32GB VRAM consumer GPU: 5080 clamshell VRAM design (16GB * 2).

Only 16GB VRAM, but "is being pitched as a 4K graphics card":
16GB VRAM not enough in Indiana Jones And The Great Circle in 4K, Full RT and DLSS Quality. Then the reviewer explains that when the "Texture Pool Size" is set from Supreme to High, that the textures still look kind of the same, because they are streamed/adjusted when they are close to you:

Also, while the 16GB of VRAM "should be enough memory for most people", "it would have been nice to see the RTX 5080 adopt the 24GB featured on the RTX 4090, especially since this GPU is being pitched as a 4K graphics card".[45]

High VRAM density, 192-bit/256-bit (=much cheaper than big 512-bit chip), low TDP consumer GPU for local AI LLM self-hosting/inferencing:
Wish there was at least a 24GB VRAM (=3GB per VRAM module) or much better a 32GB VRAM clamshell (=VRAM modules on both sides of the PCB) for local LLM self-hosting/inferencing, so there would be at least one reason for me to upgrade. Maybe in next year's Refresh. But I have not much interest in a 384-bit, 2GB per VRAM chip, 24 VRAM version Refresh (it would be basiclally a 4090) (should NV decide to design such a chip 1 year later), as it's going to be 450W + xW: I want more VRAM and less TDP, e.g. a 192-bit or 256-bit consumer GPU, hense 3GB per VRAM chip and/or clamshell VRAM design.

 

[mxthunder]

On the tear down page - it says the lighting effect is static and cannot be changed or turned off. Is this different than the 30 and 40 series then? Many didnt know it, but you could turn off the LEDs, and change the color of the LEDs on the "X" part of the GPU (Just not the color of the "GeForce" logo)
icue and open RGB, as well as Gigabyte RGB fusion allowed for this control.

 

[JIWIL]

... well at least I no longer need to worry about that voice in the back of my head tempting me to upgrade even though I already decided not to.

 

[ZoneDymo]

Dear Lord huge oof, I knew anyone not buying the 90 series was a second rate citizen in nvidia's eyes but damn son

 

[Bwaze]

Worst generational performance uplift award?

If we wait a bit for the real world prices to materialize, not this fake MSRP, I think this could beat the GTX 1080 Ti -> RTX 2080 by a mile!

 

[Neo_Morpheus]

And he said 5070>4090.
What joke.

Thats what happens when there is no competition.

There was competition, but this happened:

And it will continue happening.

Personally, i dont blame AMD for cutting back.

 

[remekra]

GeForce RTX 508.3%.

 

[gffermari]

And now what??
We stay with the 4080s....?

I think we need to bench next gen games with neural features etc.
Normal raster or ''normal'' RT performance improves seem to be forgotten by nVidia.

 

[Hecate91]

Terrible generational performance uplift over the 4080, yet it still gets the "highly recommended" award. Also throwing mud at AMD on an Nvidia review? Oh come on. I hope AMD drops dgpus all together at this point and focuses on apus like strix halo, because no matter what at every chance reviewers bash on any competition.

 

[WonkoTheSaneUK]

I don't understand the idea of a modular card instead of everything being on a single PCB like traditional cards.

Fans can't blow air through a solid PCB

 

[Macro Device]

Worst generational performance uplift award?

No. RTX 2000 series provided even less uplift per dollar at the time.

RTX 4080 was launched for 1200 USD (=1270 USD in today's money, a.k.a. 127% of 5080's price).
RTX 5080 is about 115% as fast as 4080.

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

 

[dir_d]

Its very efficient, they could have upped the core clock more to get the boost up and still been as efficient as the 4000 series. Very disappointing especially knowing that you will not be able to get the card at MSRP due to shortages.

 

[Neo_Morpheus]

Yeah, it's funny how MSRP models are always only around the first time around and never make a comeback, pffff. As I said, scumbags, all of em.

And dont forget, that original msrp will be used to benefit one brand and crucify the other down the line.

 

[rattlehead99]

So - 40% better performance / $ gen on gen. That's pretty good. I remember vividly a couple of days ago people saying that "this is impossible blabla".

The rtx 4080 Super was 1000$ when new and this is 8-10% faster for the same money.

 

[ThomasK]

You won't get these at MSRP, so what's the point on listing this as a pro?

 

[Neo_Morpheus]

Terrible generational performance uplift over the 4080, yet it still gets the "highly recommended" award. Also throwing mud at AMD on an Nvidia review? Oh come on. I hope AMD drops dgpus all together at this point and focuses on apus like strix halo, because no matter what at every chance reviewers bash on any competition.

Dont forget having dlss as a pro, instead of being a Con, since its not a platform agnostic tech.

 

[JustBenching]

The rtx 4080 Super was 1000$ when new and this is 8-10% faster for the same money.

This is replacing the 4080 not last years super.

 

[dgianstefani]

Winner here is people who weren't on Ada already. Upgrading gen-gen has always been a bad decision. No price increase despite inflation is nice, considering the lack of competition. I'm also surprised they squeezed 10% better efficiency despite same node. Perks of power efficient GDDR7 vs GDDR6X I guess.

 

[mxthunder]

Technically yes, but its maxed out the GB203 silicon so there probably wont be a 5080 "super" unless they cut down GB202 die, and if thats the case it would more than likely be a 5080Ti not a 5080 super.
Really have to look at it as the 4080 super is what the 4080 should have been.
Fully enabled AD103 for $1000 vs fully enabled GB203 for $1000. naming convention aside. so its really only a 13% improvement.

This is replacing the 4080 not last years super.

 

[Hecate91]

And dont forget, that original msrp will be used to benefit one brand and crucify the other down the line.

Even though there's no way the 5080 would be available for MSRP. Also the 5080 getting praised for not being more than the 4080 super, even though the 4080 was a bad value and the 4080 super should have been the 4080.

Dear Lord huge oof, I knew anyone not buying the 90 series was a second rate citizen in nvidia's eyes but damn son

That is how Nvidia has treated customers who don't buy the xx90 card since at least the 30 series.

 

[JustBenching]

Winner here is people who weren't on Ada already. Upgrading gen-gen has always been a bad decision. No price increase despite inflation is nice, considering the lack of competition. I'm also surprised they squeezed 10% better efficiency despite same node. Perks of power efficient GDDR7 vs GDDR6X I guess.

Now all we need is a bit of competition so we can buy these cards for cheaper. Looking to replace my 3060ti soon, this and the 5070ti are the candidates.

 

[Ruru]

Even though there's no way the 5080 would be available for MSRP. Also the 5080 getting praised for not being more than the 4080 super, even though the 4080 was a bad value and the 4080 super should have been the 4080.

That is how Nvidia has treated customers who don't buy the xx90 card since at least the 30 series.

At least 3080/3080 Ti was a cut-down big chip, not an overpriced mid-sized chip with 256bit bus.

 

[Darmok N Jalad]

Winner here is people who weren't on Ada already. Upgrading gen-gen has always been a bad decision. No price increase despite inflation is nice, considering the lack of competition. I'm also surprised they squeezed 10% better efficiency despite same node. Perks of power efficient GDDR7 vs GDDR6X I guess.

View attachment 382318

Yeah, at best you have to jump up a product line as well for the gain to be worth it. A 4070 super/Ti owner might consider this, but it's not a big enough gain for a 4080 owner. Even then, though, you still aren't gaining any more VRAM.

 

[JustBenching]

At least 3080/3080 Ti was a cut-down big chip, not an overpriced mid-sized chip with 256bit bus.

Is it overpriced though? This is the fastest GPU below 999$.