NVIDIA GeForce RTX 5080

Written by John Yan on 2/6/2025 for PC  

By now you have heard of Blackwell, NVIDIA’s newest GPU architecture. If you haven’t seen the review of the GeForce RTX 5090, you can check out my review here. While that product is the halo product, today we have what I consider the one that most high end gamers might consider as of this writing.

The NVIDIA GeForce RTX 5080 is the second of four cards announced at CES. Interestingly, NVIDIA priced the Founder’s Edition at $999, something that surprised me. Whereas the RTX 5090 got a $500 increase in price over the RTX 4090, the RTX 5080 got a $200 reduction in price over the RTX 4080. It matches the RTX 4080 Super pricing, which was the card that replaced the RTX 4080.

We won’t go over the architecture and features too much here as it’d just be a rehash of my GeForce RTX 5090 review. If you want to get a refresher, you can read it here. So let’s get to the heart of the review.

Spec wise, the GeForce RTX 5080, it seems to be a slight increase in various aspects over the GeForce RTX 4080 Super. The RTX 5080 has 10752 CUDA cores, a 5% increase over the RTX 4080 Super, but a generational increase in architecture. ROPs go from 112 to 128, Tensor Cores go from 320 to 336, and RT cores go from 80 to 84. Base clock for the GeForce RTX 5080 is 2295 MHz while the boost clock does get a slight uptick over the RTX 4080 with it being 2617 Mhz. If you compare the two side by side, the improvements is marginal.

VRAM, unfortunately, stays the same at an amount of 16GB but it is GDDR7. If you plan on trying to use this card at 4K, which is where all my tests are, you might run into some VRAM issues unless you dial down some settings. Clock speed for the memory sits at 1875 MHz with a bandwidth of 960.0 GB/s.

Here’s a table showing the differences:

	GeForce RTX 5080	GeForce RTX 4080 Super	GeForce RTX 4090
SM	84	80	128
CUDA Cores	10752	10240	16384
Tensor Cores	336 (5th gen)	320 (4th gen)	512 (4th gen)
RT Cores	84 (4th gen)	80 (3rd gen)	128 (3rd gen)
Texture Units	336	320	512
ROPs	128	112	176
Base Clock	2295 MHz	2295 MHz	2235 MHz
Boost Clock	2617 MHz	2550 MHz	2520 MHz

Physically, the card is the exact match of the GeForce RTX 5090 and uses almost the same great engineered cooler that that card has. A two slot cooled card, the blow through style cooler helps keep the card cool but the RTX 5080 doesn’t have the liquid metal that the RTX 5090 has and the heatsink itself doesn’t have the vapor chamber design. I’m guessing since this card has less TGP, NVIDIA opted to go with a less efficient cooling solution since it didn’t need to and it would also help reduce the cost of producing the RTX 5080 in order to meet the $1000 price point. As I said in my RTX 5090 review, I really do like the look of the card and if you’re one to show off the internals of your PC, the cooler design is top notch.

The power adapter is located in a much nicer location versus the previous generation of cards. NVIDIA has said they’ve fixed the issues of the connector melting its pins, but only time will tell on that one. The perpendicular angle connector makes routing a cable much, much easier in a case and cleaner to with this orientation.

Included with the card is a power adapter that’s built the same way as the RTX 5090’s power adapter. Unlike that one though, it’s got three 6-pin connectors instead of four. The RTX 5080’s TGP is 360W, a 20W bump over the RTX 4080 Super. It’s definitely a lot less than the RTX 5090’s 575W monster. So while there’s a small bump in TGP, you can get by with three 6-pin power connectors here. Like I said in my other review, I do like the more flexible connector and hopefully, this design helps alleviate any pressure that can come if you do need to bend the cable in your setup.

We’re going to compare the GeForce RTX 5080 with the GeForce RTX 4080 Super. All tests were run at 4K and multiple runs were averaged out, which you see in the graphs below. My test system consisted of:

AMD Ryzen 7 9800X3D
G.Skill Trident Z5 Neo RGB 64GB (2 x 32GB) DDR5-6000 PC5-48000 CL30
ASRock B65-E PG Riptide WiFi motherboard
NVIDIA GeForce RTX 5090
Samsung 990 Pro 4TB SSD
LG OLED42C2PUA 42"
NVIDIA Drivers 572.16

We'll start with rasterization results.

The uplift from an RTX 4080 Super is small in rasterization numbers and there's actually some titles where it's even a little slower, but here the price is essentially the same. On the surface, that's a little bit disappointing. I expected a lot more improvement over the RTX 4080 Super and I think we should have expected the RTX 5080 to best the RTX 4090 as that was how the cards in previous generations in this tier was. The main benefit it seems is access to the faster GDDR7 RAM for those that would take advantage of that memory in their workloads and access to Multi Frame Generation.

Let's take a look at DLSS performance without frame generation.

Again, while performance using DLSS is better than the RTX 4080 Super, there's not solid uplift we're used to when these cards are released.

Now, let's look at what happens when you turn on RTX to improve. Indiana Jones and the Great Circle didn't run on the RTX 4080 Super nor the RTX 5080 when Path Tracing was turned on so those results have been omitted.

As you can see, turning on Path Tracing really puts a big hit on even the 80 series of cards. But, we have a few features that can help. Let's turn on DLSS.

The performance is still going to be a little below what we want for smooth gameplay. That's where Frame Generation comes in, to help smooth out the animation. Let's see how it works in all scenarios such as without ray tracing turned on and with various DLSS options enabled.

Combining DLSS with Frame Generation, it does help a ton to smooth out gaming and with the newer Transformer model that's available in some games, you can go lower in terms of quality and still get great visuals. The GeForce RTX 5080 does support Multi Frame Generation, which can produce an extra 2 or 3 frames before it displays one rendered frame. It works well when the game's already running well and you want to use a high refresh rate monitor. Here's Cyberpunk 2077 and Star Wars Outlaws using this feature.

Finally, we have VR benchmarks and 3D Mark Time Spy Extreme.

VR wise, it looks like you'll get a nice boost in performance over the RTX 4080 Super. Getting above the 90FPS range will provide a smoother gameplay experience in VR.

When it comes down to it, the GeForce RTX 5080 is a GPU that is a little disappointing in terms of performance when compared to the past generation card. The uplift is small at times and if you aren't interested in the new Multi Frame Generation software, then there's really no reason to upgrade from Ada Lovelace. The 16GB of VRAM also makes it a card that will just be OK for some AI usage, but I would have preferred 24GB here. We've been used to the 80 series cards outperforming the 90 series card of the previous generation, but it's definitely not happening here.

I did run into some small stability issues that seem to have improved after a few driver updates. Firefox is my primary web browser and I've had a few times where it would crash with a NVIDIA error in the Event Viewer. I even forced my motherboard to use PCIE Gen4 and that didn't help before some recent driver updates. Putting in Ada Lovelace cards into the same system, I never had any of these errors pop up. As soon as I put in the GeForce RTX 5080 again, the errors appeared. I was able to send some info to NVIDIA so hopefully that helps.

Like the RTX 5090, I don’t think the RTX 5080 is for those who are coming from the previous generation. If you already own an RTX 4080 Super or even the RTX 4080, this isn’t the card for you. It’s for those who are in the 3000 series and lower who are looking for better RT performance and better DLSS performance. That is, if you can find the card at MSRP.