RTX 5070 vs RTX 3080 specs comparison – how do they compare?
Table of Contents
The RTX 5070 and RTX 5070 Ti are heading our way, and out of all the 50-series cards so far, it looks like these two are being met with the most ‘mixed’ reception in the build up to their respective launches. In our poll, we noted that the 5070 was people’s last choice out of the line-up. And let’s not forget that finding one at MSRP may be challenging too due to potential stock issues and that the TI variant may not get a Founders Edition.
If you’re here though, you’re likely weighing up your options, and one card that will have come to mind would be the impressive RTX 3080. This card is still a great GPU for most people and delivers excellent performance at 4K, making it a worthy alternative to the RTX 5070 – and maybe it’ll be a bit easier to get a hold of one too.
🚀 Save Up to $1,200 on the Samsung Galaxy S25!
Pre-order now and save big with trade-in and Samsung credit. Limited time only!
*Includes trade-in value + $300 Samsung credit.
We’re going to take a look at these two cards side by side, with the information we have available to us right now. Will the card that is claimed to have RTX 4090 performance come out on top, or can the once popular 30-series GPU still throw down? Here’s a closer look.
Spec comparison
We’ve listed their specifications below to see the hardware difference between the two processors.
| Specifications | RTX 5070 | RTX 3080 |
| Cores | 6144 | 8960 / 8704 |
| Base clock speed | 2.16 GHz | 1.26 / 1.44 GHz |
| Boost clock speed | 2.51 GHz | 1.71 GHz |
| Memory | 12GB GDDR7 | 12 GB GDDR6X / 10 GB GDDR6X |
| Memory interface | 192-bit | 384-bit / 320-bit |
| Bandwidth | 672 GB/sec | 760.3 GB/s |
| TDP | 250W | 350 / 320 |
| Power connector | 2x PCIe 8-pin cables or300 W or greater PCIe Gen 5 cable | 2x PCIe 8-pin |
| Release date | February 2025 | September 2020 |
Core count and clock speed
The RTX 5070 has 6,144 CUDA cores, while the RTX 3080, 10, and 12GB variants have 8,960 and 8,704 CUDA cores, respectively. This is a difference of 2,816 and 2,560 cores, which puts the RTX 3080 on top in terms of hardware.
However, each RTX generation features newer RT and Tensor cores, which is the same for CUDA cores. So, even though it has fewer cores than the RTX 3080, it might be able to trade blows with it in raw performance.
In addition, it comes with enhanced AI technologies, which should give it the edge over the 3080. The clock speeds are also favorable for the RTX 5070 as it has a 2.16 and 2.51 GHz base and boost clock speed while the RTX 3080 10GB has a 1.26 GHz base clock, the 12GB has a 1.44 GHz base clock, and both share a 1.71 GHz boost clock speed.
One could argue that this difference can be bridged by overclocking the GPU; pushing the 3080 to cross the 2.00 GHz mark at the base would be too much, but probably not impossible under the right circumstances that would keep it from thermal throttling. However, for an everyday user, pulling that off would be difficult.
That being said, we can't ignore that the RTX 3080 has more cores to pull the heavyweight, and ultimately, I think we might see the RTX 5070 and 3080 clash in terms of performance.
Memory
A significant difference between these GPUs is the memory type and, for one variant, the memory capacity. The RTX 5070 has 12GB GDDR7 VRAM, while the RTX 3080 has 10 and 12GB GDDR6X VRAM.
While the difference is only 2GBs at max, the main factor is the memory type. GDDR6X has a maximum per-pin speed of 24GB/s, while GDDR7 takes it to the next level with 32GB/s. So, this means that compared to both 3080 variants, the RTX 5070's memory should give it the edge in performance as it can handle the same amount of data but at a much faster rate, meaning better performance and efficiency.
So, judging by this, we think the RTX 5070 should be able to compete with the RTX 3080.
Memory interface width
The memory interface width determines how much data can be transferred with each clock cycle. For example, the RTX 5070 has a 192-bit connection, meaning it can transfer 192 bits of data in each clock cycle at maximum; any more than that will cause a bottleneck.
So, if there is an influx of data but the memory interface width can't accommodate more, it'll slow down the GPU performance, but this is unlikely to happen. That being said, the RTX 3080 has a 384-bit / 320-bit memory interface width, meaning it can handle 384 or 320 bits of data in each cycle. This means it should have a better output and efficiency as more data can easily travel to and from the GPU.
Hardware differences like these most likely can't be overcome with software, as it won't help push more data where there is a physical limitation, so the RTX 3080 wins in this domain.
Performance
Now we don’t have performance figures for the RTX 5070 just yet, but we will once we put it to the test in our review. But we can share our benchmarks for the RTX 3080 10GB.
I’ve chosen CS:GO and Doom Eternal to illustrate its performance, but ideally we would have gone for something like Cyberpunk 2077 if we could have! As you can see, the card could achieve 327 fps on CS: GO at 4K, and 99fps on Doom Eternal in 4K with Ray Tracing. By any standards that’s impressive.
The only comparison we can make to the 5070 at this time is to use the card which it claims to be similar to from a performance perspective – the 4090.
The closest comparison here would be with CS: GO, which we tested both at max settings. The 4090 achieved 377 fps at 4K, a 15% increase over the 3080.
It will be interesting to see how the 5070 performs without DLSS 4 and MFG on, but if it does perform like a 4090 (at least on this game), I think we won’t be seeing a huge difference between the 3080 and 5070 at 4K. Time will tell, and we’ll update this article with performance benchmarks when we get them.
DLSS and Reflex
DLSS and Reflex are two technologies that both GPUs share and work to deliver a smoother and better gaming experience. However, the RTX 5070 has the edge with DLSS 4, which will be available for previous-generation cards and Multi-Frame Generation.
This most likely gives the RTX 5070 a performance good enough for Nvidia to claim that it can deliver the same performance as the RTX 4090. Currently, the RTX 3080 has DLSS 3.5, which, in Nvidia's video, increased the FPS in Cyberpunk 2077 from 27 FPS to 141 FPS.
However, DLSS 4 increased it to 243 FPS, more than 100 FPS from DLSS 3.5. This technology renders frames at a low resolution, which doesn't put much stress on the GPU, and then upscales them to a higher resolution using AI.
As for Reflex, it is a technology that reduces the system latency for better responsiveness. However, on launch, all the RTX 50-series GPUs will feature Reflex 2, and like DLSS 4, the other cards will get it later on. Reflex 2 introduced Frame Warp, which further decreases the latency by warping the pixels; let me explain what this means.
When you input any instruction while playing the game, such as moving your character, interacting with a door, or moving your mouse around, it takes time for that action to be displayed on the monitor. While that time frame is so small that it doesn't affect regular users, it can spell victory or defeat for competitive gamers.
So, in Reflex 2, when you perform an action, the CPU calculates the next frame that'll be displayed based on that. The pixels are shifted in advance to reflect that subsequent frame, and to ensure that there isn't any discrepancy, it waits till the last moment for the GPU to render the most recent frame to use as a reference.
However, this can cause issues such as tearing or holes in the frames. To counter that, Reflex 2 has something called inpainting, which uses prior frames, color, and depth data to fill these holes so that all you see are complete frames with a lowered latency. Nvidia has an excellent video that explains and showcases this technology working in-game, and we definitely recommend checking it out.
RTX 5070 vs. RTX 3080 – price comparison
The RTX 3080, in the 10 GB and 12GB variants, launched at $699 and $799, respectively. The RTX 5070, in contrast, has an MSRP of $549. However, if you're looking for a new RTX 3080, you might run into some trouble unless you're willing to purchase renewed options from Amazon.
We found this EVGA GeForce RTX 3080 FTW3 Ultra Gaming on Amazon for $648.99, but there doesn't seem to be any stock for the RTX 3080 at Best Buy. On the other hand, the RTX 3080 Ti is more readily available in new condition on Amazon but has a $800 – $900 price tag.
Newegg also has more options for the RTX 3080 Ti, and the one we found was an MSI RTX 3080 VENTUS 3X 10G for $588.99, but it is refurbished. However, all hope isn't lost, as the Nvidia Marketplace has a Zotac RTX 3080 TrinOC available for $1,115.54, which for a four year old card is just absurd.
So, we recommend going for the RTX 5070 as it features better technology, has the latest generation cores, and has 12GB VRAM, the same as a high-end RTX 3080 but with GDDR7. To top it off, the founders edition costs $549, and while third-party variants cost more, you can go for a PNY, ASRock, ASUS TUF, or Gigabyte GPU that are relatively less expensive compared to ASUS ROG Strix or MSI variants.
Early impression
After taking a close look at both GPUs, it seems like the RTX 5070 will offer the same performance as the RTX 3080, meaning solid FPS at 4K; however, DLSS 4 and Multi-Frame Generation will give it a significant edge in gaming, which the RTX 3080 won't be able to compete with.
Besides that, investing in an RTX 3080 doesn't make sense as a new card costs $1,000+, and even renewing for refurbished options costs more than the RTX 3080. If you were to spend more than $1,000 on a card, we'd recommend you go for the RX 7900 XTX, which can compete with the RTX 4090 in gaming.
However, if you can wait, we recommend doing that for the RTX 5070 as real-world performance benchmarks will also give you a clear idea of whether the RTX 5070 is worth investing in or should you go for some other option. We'll also be testing it in-house so check back with us for the best and latest.
About the Author
