Nvidia RTX 4070 Founders Edition Design
We're still more than a little bit curious about the lack of an RTX 4070 Ti Founders Edition. We confirmed with Nvidia that there will not be a 4070 Ti FE, leaving a small hole in the RTX 40-series lineup for those that like reference cards. Perhaps it was a Nvidia's way of throwing a bone to its add-in board partners, or maybe it's just part of that whole RTX 4080 12GB rebranding nonsense from last year. Regardless, we have the RTX 4070 Founders Edition, which looks like the 30-series parts, only with a few differences.
If you check out the RTX 3070 Founders Edition photos, you'll notice that — unlike the RTX 3080 or RTX 4080 — there's no large "RTX 3070" logo on the card. The RTX 3070 Ti meanwhile looks like the other high-end offerings. Basically, Nvidia created a different design and aesthetic on the RTX 3070 and RTX 3060 Ti Founders Edition cards. Those two are also more compact designs, but we did appreciate the look of the higher tier reference models.
Nvidia has "fixed" this with the RTX 4070 Founders Edition, in that there are now two fans on different sides of the card, with a large blank area sporting the RTX 4070 logo. It's not just the RTX 3070 Ti / 3080 design repurposed, however, as the 4070 Founders Edition includes the larger fan used in the 40-series models and also comes with a smaller form factor than the 3080. It's actually just a touch larger (maybe 1mm) than the RTX 3070 FE, but it's also about 2mm thicker.
Also "fixed" is that the RTX 4070 Founders Edition comes with the same large clam box that you'll get with the 4080 and 4090 Founders Editions. Is that a good thing? Maybe, maybe not — less expensive packaging in exchange for an RGB-lit logo would have been better, we think.
The full dimensions of the RTX 4070 Founders Edition are 244x111x40 mm (compared to 243x111x38 mm on the RTX 3070 FE). It has two custom fans that are 91mm in diameter (versus 85mm on the 3070), and it weighs 1021g (compared to 1030g for the 3070). While we generally prefer the new aesthetic over the 3070 design, there are a few blemishes that are sticking around.
Like that completely pointless 16-pin power connector. Seriously, this is a 200W card, and while it might have required dual 8-pin (or 8-pin plus 6-pin) connectors to ensure there's ample room for overclocking, we'd rather have that than the rather clumsy 16-pin to dual 8-pin adapter. Even if it doesn't melt, it's just one more piece of cabling you'll have to deal with in your build. The PNY RTX 4070 incidentally opts for a single 8-pin connector, though it's also limited in overclocking options, so if you don't want the 16-pin adapter (and/or you don't have a new ATX 3.0 power supply), you might want to give that a look.
Most PCs aren't using much in the way of expansion cards beyond the GPU these days, so a relatively compact dual-slot design will work well in mini-ITX cases all the way up to full-size ATX builds.
I generally like the look of the current and previous generation Founders Edition cards. They're not flashy, particularly on the models that don't have any RGB like the RTX 4070 model. But when they're installed in a typical case with a window, you can see the nice "RTX 4070" (or whatever model) logo facing upward. Also, cooling performance seems to have improved quite a bit with the 40-series, or at least the second generation GDDR6X doesn't run anywhere near as hot as the first generation did on the 30-series.
As with pretty much every Nvidia Founders Edition GPU of the past several generations, you get the standard triple DisplayPort and single HDMI outputs. (Some might still miss the USB type-C VirtualLink from the 20-series, though.) It's an HDMI 2.1 port and three DisplayPort 1.4a outputs. Not to beat a dead horse, but while DP1.4a is technically worse than HDMI 2.1 and DisplayPort 2.1 (54 Gbps), we have tested a bunch of GPUs with a Samsung Odyssey Neo G8 that supports up to 4K and 240 Hz, thanks to Display Stream Compression (DSC). Until we start seeing DisplayPort 2.1 monitors that run at 4K and 480Hz, DisplayPort 1.4a should suffice.
Since this will be our reference RTX 4070, we're skipping disassembly and teardown — we need to guarantee the card keeps working properly, and ideally not change any of the cooling or other factors. We do know there are six GDDR6X chips inside, and during testing we saw peak temperatures of 70C on the memory — a far cry from the 100–104C we often saw on the RTX 30-series cards with GDDR6X!