KitGuru KitGuru.net – Tech News | Hardware News | Hardware Reviews | IOS | Mobile | Gaming | Graphics Cards
A month on from the launch of Nvidia's flagship RTX 4090, today we have put the RTX 4080 through its paces. Built on the new AD103 silicon, RTX 4080 is notably cut-down against the RTX 4090, and that shows when looking at overall performance.
At 4K for instance, the RTX 4080 is on average 24% slower than the RTX 4090, with very consistent performance scaling across the 12 games we tested. That still means the RTX 4080 is the second-fastest graphics card we have ever tested – it's certainly no slouch for 4K gaming, averaging 101FPS across our test suite.
It may not be as large of a performance uplift as some were expecting however. Versus the RTX 3090 Ti, the AD103-based GPU is 17% faster on average, and versus the RTX 3080 10GB, it offers a 46% uplift, with a 30% lead over AMD's RDNA 2 flagship – the RX 6950 XT.
Interestingly, we also noted a number of instances where the RTX 4080 would scale better at 1440p than it does at 4K. Nine of the twelve games we tested saw the 4080 perform better, in relative terms, against the 3090 Ti at QHD. The differences aren't huge, as it is on average 17% faster at 4K, but 19% faster at 1440p, but it was a repeatable trend.
The explanation for this seems to be a combination of factors, including less mature drivers for the 40-series, but I'd wager memory bandwidth is the main one. By cutting the memory interface down to 256-bit, even with fast GDDR6X memory, total bandwidth of 716.8GB/s is a good deal slower than the 1008GB/s speeds offered by the 3090 Ti, which becomes a significant factor for 4K gaming.
Ray tracing performance numbers also scale in a similar fashion to raster performance, with the RTX 4080 between 25-30% slower than the RTX 4090, but still holding a decent lead over the RTX 3090 Ti. By and large this means you will be able to crank up settings to the max, but there are exceptions – Cyberpunk 2077, for instance, averaged 28FPS at 4K using the Ultra ray tracing preset, so DLSS would still be required to get things up to a playable frame rate.
Speaking of DLSS, I encourage you to check out the video review where we take a look at the new DLSS 3 Frame Generation feature. It certainly provides an immediate benefit of improving the visual smoothness of supported games, though you do have to be happy to contend with some visible artifacts as a result of the new frames being generated entirely by AI. That is something a lot of people could live with given the boost to frame rates, but the impact on latency is more of a problem for me. Frame Generation marks a significant paradigm shift where increasing frame rates don't lead to a decrease in latency, so depending on the games you play, it won't be an ideal solution. It's still early days however, but we will keep a close eye on the technology as it develops.
In terms of the Founders Edition, here Nvidia is using the exact same cooler as found on the RTX 4090. This has an immediate benefit of providing exceptionally low temperatures and low noise levels, simply due to the fact it was designed to cool a 450W GPU but is now faced with the much easier task of taming the 320W RTX 4080. That being said, I can't help but feel the decision to re-use the 4090 cooler results in an unnecessarily large RTX 4080. We've shown how much bigger it is compared to the RTX 3080 Ti Founders Edition, and that's a 350W card, so it does feel somewhat overkill in that regard.
Speaking of power, we even found the 4080 would consistently draw less power than its rated TGP. Using our in-depth methodology where we test power draw of each GPU across twelve games at three different resolutions (starting on page 28 of this review), at most we saw the 4080 drawing 303W, but even at 4K it averaged 280W across our testing. In discussions with Nvidia about this, we can only conclude that Ada is significantly less power-limited than Ampere, instead being either voltage or even clock-limited.
Of course, lower power draw is a very good thing indeed, and that has clear implications for overall efficiency. In simple terms, the RTX 4080 is the most efficient graphics card we've ever tested, offering 10% better performance per Watt than the RTX 4090 at 1440p. At the same resolution, it is 60% more efficient than the next-closest challenger, AMD's RX 6900 XT. This is a fantastic hallmark of the Ada architecture and should get better still as we move down the stack.
In an objective sense then, the RTX 4080 is an excellent graphics card. It stomps through 1440p and 4K gaming, it's remarkably power efficient and its lower-than-expected power draw makes it significantly easier to cool.
And yet, our final conclusion is not so straightforward, for one main reason – value. As shown earlier in this review, the RTX 4080 offers basically equivalent value to the RTX 4090, being 24% slower and 25% cheaper, which means we are looking at linear value scaling between the two 40-series cards.
That doesn't add up, however. The RTX 4090 has been talked about as the card where value for money doesn't matter – it's the GPU for those who want the best of the best and don't care what it costs them. So how does it make sense to offer the RTX 4080, which is a slower card at a lower price point, but still end up with the same value proposition as the RTX 4090? It doesn't make sense.
The comparison to the situation with the RTX 3080 is also telling. That GPU launched back in 2020 at £649, coming in just 10-15% slower than the RTX 3090 but at roughly half the price. With the RTX 4080 however, not only is it further behind the 4090 than the RTX 3080 was, it scales linearly in terms of value, resulting in a pretty unappealing cost per frame situation.
We can't forget that this is still an extremely expensive graphics card. At £1269, this is hardly a card for the mass market. Part of me wonders who would be prepared to spend this much but not step up to the RTX 4090, especially when considering the price premium that AIBs will add to the baseline figure. I'm sure there will be some people looking at the 4080 who couldn't justify spending the extra on the 4090, but considering the performance differential, dropping another few hundred quid doesn't seem unreasonable to me.
I'm not usually one to get caught up in nomenclature either, but with the RTX 4080 at this new, incredibly lofty price point, compared to generations of previous xx80 SKUs which launched in the £600-650 region, it is concerning to see how rapidly things are shifting upwards. With the RTX 4080 coming in 46% faster than the RTX 3080 at 4K, but for nearly double the price, there are many who simply wouldn't consider this GPU in the first place, and I can't blame them. It is also worrying to think what generational improvements – if any – will be on offer once a 40-series card in the £600 price bracket finally arrives.
Of course, the elephant in the room is AMD's RDNA 3, with the 7900 XTX set to hit the market at $1000 next month. We can’t speak to how it will compare to the 40-series just yet, but anyone considering a 4080 would have to be mad not to wait and see what AMD can bring to the table. I personally wouldn’t be surprised to see an RTX 4080 Ti launched before too long if the 7900 XTX looks as competitive as AMD claim.
The Nvidia RTX 4080 has an MSRP of £1269. Retail availability is slated for 2pm tomorrow, November 16.
Discuss on our Facebook page HERE.
Pros
- Very fast for 1440p and 4K gaming.
- Power draw is lower than the rated TGP.
- The most efficient graphics card we've tested.
- Runs very cool and with low noise levels.
- Overclocked pretty well, particularly the GDDR6X memory.
- DLSS 3 is a promising new technology.
Cons
- Doesn't offer any value benefit versus the RTX 4090, which doesn't make sense.
- Many spending this much money may want to step up to a 4090 considering the significant performance differential.
- Vast increase in price over previous xx80 SKUs.
- The cooler feels unnecessarily large considering the power requirements.
KitGuru says: The RTX 4080 is a very capable graphics card, but the value proposition just doesn't add up.
