Today Nvidia has announced three new graphics cards – RTX 2060, 2070 and 2080 SUPER. RTX 2080 SUPER isn't coming until later this month, but today we have reviews of both the 2060 and 2070 SUPER editions. This review is focused solely on the latter card, so if you want to read about 2060 SUPER, head over to our full review of that card over here.
What we're focused on here, however, is what RTX 2070 SUPER can bring to the table. Nvidia tells us it is faster than a GTX 1080 Ti, so could it be the new mid-range king?
Unlike the RTX 2060 SUPER – which is actually closer to the RTX 2070 than its namesake, the 2060, in terms of specs – Nvidia's RTX 2070 SUPER is more of a true ‘2070 Ti' card, as it slots neatly in between the 2070 and 2080. That said, as RTX 2070 is a full implementation of the TU106 GPU, Nvidia had to take TU104 – the basis for RTX 2080 – and cut it down to 36 SMs, and 2560 CUDA cores, to make RTX 2070 SUPER.
Priced at £475 here in the UK, it marks a £16 price increase over the original RTX 2070 MSRP, or 3%. The MSRP in the US, however, is the same for the 2070 SUPER as the original 2070, at $499 – so the slight increase for those in the UK is simply due to changing dollar-to-pound conversion rates. In any case, £475 is the baseline MSRP, so custom cards will be coming in with higher prices.
GPU | RTX 2080 (FE) | RTX 2070 SUPER | RTX 2070 (FE) | RTX 2060 SUPER | RTX 2060 (FE) |
SMs | 46 | 40 | 36 | 34 | 30 |
CUDA Cores | 2944 | 2560 | 2304 | 2176 | 1920 |
Tensor Cores | 368 | 320 | 288 | 272 | 240 |
Tensor FLOPS | 85 | 72 | 63 | 57.4 | 51.6 |
RT Cores | 46 | 40 | 36 | 34 | 30 |
Texture Units | 184 | 160 | 144 | 136 | 120 |
ROPs | 64 | 64 | 64 | 64 | 48 |
Rays Cast | 8 Giga Rays | 7 Giga Rays | 6 Giga Rays | 6 Giga Rays | 5 Giga Rays |
GPU Boost Clock | 1800 MHz | 1770 MHz | 1710 MHz | 1650 MHz | 1680 MHz |
Memory Data Rate | 14 Gbps | 14 Gbps | 14 Gbps | 14 Gbps | 14 Gbps |
Total Video Memory | 8GB GDDR6 | 8GB GDDR6 | 8GB GDDR6 | 8GB GDDR6 | 6GB GDDR6 |
Memory Interface | 256-bit | 256-bit | 256-bit | 256-bit | 192-bit |
Memory Bandwidth | 448 GB/sec | 448 GB/sec | 448 GB/sec | 448 GB/sec | 336.1 GB/sec |
TDP | 225W | 215W | 185W | 175W | 160W |
On top of the fact that RTX 2070 SUPER uses the TU104 GPU, it's also interesting that Nvidia's specs claim it has ‘5 or 6' Graphics Processing Clusters (GPCs). This is noteworthy because, for TU104, each GPC houses 4 Texture Processing Clusters (TPCs) – and the specs are very clear that RTX 2070 SUPER has 20 TPCs. So basic maths says the GPU should have 5 GPCs. So why the ‘5 or 6'?
Well, it turns out not all RTX 2070 SUPERs are made quite the same. They will all have 40 SMs (two SMs per TPC), but these may be spread across 5 or 6 GPCs – i.e. some GPCs will not be fully populated with 4 TPCs if you get a 2070 SUPER with 6 GPCs. Nvidia's Rick Napier, Senior Technical Product Manager, told us:
‘We [Nvidia] sweep in a variety of ways, and as a result some RTX 2070 GPUs have 5 GPCs, and some contain 6. The performance difference between 5 GPC vs 6 GPC GPUs is negligible. Ultimately because we’ve outfitted the GPU with 20 TPCs and higher clocks, the RTX 2070 SUPER is on average 16% faster than RTX 2070.'
So while it may not mean much to the everyday gamer when looking at a FPS counter, I think it is interesting to note how Nvidia is achieving its 40 SMs for the RTX 2070 SUPER.
Elsewhere, we can note the relatively high 1770MHz rated boost clock. Nvidia told us that RTX 2070 SUPER is not a factory overclocked card, so 1770MHz is reference speed for this card. It will be interesting to see how far custom cards will push the clock speed.
Lastly, just like both the RTX 2070 and 2080, RTX 2070 SUPER still has 8GB GDDR6 memory clocked at 1750MHz (14000MHz effective).