Nvidia’s RTX 4080 graphics card with 16GB of VRAM – the high-end model, as opposed to the 12GB version – was spotted in a supposed 3DMark benchmark, giving us an idea of its gaming performance.
Now, we should be clear upfront that it is just that – a performance concept, more on that later – and we should be careful to assume that it is also genuine. It could also be faked as VideoCardz (opens in new tab), who flagged the benchmark, points out that there are doubts about how a working driver with support for the GPU was obtained. (Unlike the upcoming RTX 4090, the RTX 4080 graphics cards don’t come out until November).
So there are some massive question marks around this, and the source, Chiphell (opens in new tab) forum in China, it’s not the most reliable either, although there have been some genuine leaks in the past.
In the graphics scores (at 4K resolution) the RTX 4080 supposedly succeeded in 3DMark, which was 17,465 in FireStrike Extreme and 13,977 in TimeSpy Extreme. Looking at the benchmark scores for the RTX 3080 (12GB), this makes the RTX 4080 about 50% faster in both metrics.
An additional 3DMark Port Royal result was also highlighted, with the RTX 4080 hitting 17,607, which is around 45% faster, so again pretty much the same ballpark.
As a side note, the images provided show the specs with the RTX 4080 clocked at 2.5GHz boost.
Analysis: Let’s wait and see – but price concerns certainly abound
At the risk of sounding like a broken disk, as we always remind people with these kinds of leaks, it’s just a synthetic benchmark – what we really need to judge a GPU is testing in real games, and a wide variety of them (as Results may vary quite a bit with different games, and therefore graphics settings).
In fact, there is a gaming benchmark provided by this leak in the form of Shadow of the Tomb Raider, where the RTX 4080 achieved 128 frames per second at 4K resolution. While we have no benchmark for this – and none of the graphics detail – at first glance, this seems a pretty decent result.
While some people online have been a little disappointed with the 3DMark scores here, they’re in line with what we really expected – and what we also have to remember is that it’s possible that clumsy drivers are at play. When official release driver support comes for the RTX 4080, the GPU will definitely be faster somehow. And of course that’s before we start talking about other benefits like DLSS 3 which seems to be a huge boost for games that support the technology (by the way, the RTX 3000 GPUs won’t have this, although it may eventually come to the Ampere series ).
Of course, this whole leak could be debunked, but there’s some decent footage and images here, so if it’s fake, someone’s screwed up enough (mind you, it wouldn’t be the first time this happened). Well, let’s be cautious about this leak, and if it’s true, equally cautious about the possible performance the RTX 4080 achieves here, as the GPU will be faster at boot (or something would be really, really wrong ).
Probably the reason there has been some negative reaction to this leak in some quarters is looking at that gross 50% yield gain and comparing it to the large perceived price increase. But actually, if we look at the cheapest RTX 3080 12GB models right now (the GPU used in the performance comparisons above), the RTX 4080 is about a 60% relative price increase for that 50% performance boost.
This means that the Lovelace GPU is more expensive in terms of performance per dollar, but not by much. And hopefully, as mentioned, when the full speed of the released RTX 4080 is realized, it can be almost on par as an overall value proposition in terms of performance.
Which is still not great, mind you, as graphics cards are still overpriced in our book, and volumes of others for that matter – and of course there’s also the reality that the RTX 4080 16GB can be more expensive than MSRP when first breaks cover. We seriously hope not, though, but these days, it’s almost unrealistic to expect the suggested pricing to be maintained at GPU launch. Initial stock levels will, of course, be a big determining factor here…