With the release of the GeForce RTX 2080 barely behind them – and the RTX 2080 Ti technically still in front of them – NVIDIA has announced the release date for their next RTX card: the 2070. The previously revealed card, which was scheduled for an ambiguous October, has been locked in for October 17th, where the Founders Edition version will hit the streets at $599.

In line with the general performance progression for the GeForce RTX family, the RTX 2070 is slated to deliver around 75% of the RTX 2080’s performance. The exact performance depending on how each game scales with the smaller collection of resources. The RTX 2070 has about 75% of the shading/texturing/tensor hardware as the RTX 2080, however it has the same ROP count and the same 256-bit memory bus. So we expect that the biggest performance differences are going to be in SM-bound scenarios – now including ray tracing – while pixel-bound scenarios that rely mostly on ROP throughput should take a much smaller hit.

NVIDIA GeForce x70 Specification Comparison
  RTX 2070
Founder's Edition
RTX 2070 GTX 1070 GTX 970
CUDA Cores 2304 2304 1920 1664
ROPs 64 64 64 64
Core Clock 1410MHz 1410MHz 1506MHz 1050MHz
Boost Clock 1710MHz 1620MHz 1683MHz 1178MHz
Memory Clock 14Gbps GDDR6 14Gbps GDDR6 8Gbps GDDR5 7Gbps GDDR5
Memory Bus Width 256-bit 256-bit 256-bit 256-bit
VRAM 8GB 8GB 8GB 4GB
Single Precision Perf. 7.9 TFLOPs 7.5 TFLOPs 6.5 TFLOPs 3.9 TFLOPs
"RTX-OPS" 45T 45T N/A N/A
TDP 185W 175W 150W 145W
GPU TU106 TU106 GP104 GM204
Transistor Count 10.8B 10.8B 7.2B 5.2B
Architecture Turing Turing Pascal Maxwell
Manufacturing Process TSMC 12nm "FFN" TSMC 12nm "FFN" TSMC 16nm TSMC 28nm
Launch Date 10/17/2018 N/A 06/10/2016 09/18/2014
Launch Price $599 $499 MSRP: $379
Founders $449
$329

Overall, the RTX 2070 is a bit of a wildcard. Rather than being a cut-down version of the GPU used in the 2080, as has been the case for NVIDIA product stacks for most of the last decade, the RTX 2070 gets its own GPU: the TU106. A third GPU in as many cards has more upfront costs, as NVIDIA has to tape out and manufacture another die, however the 454mm2 GPU allows them to min-max costs by not having to use the larger TU104 to fill what will be greater demand for the cheaper card. Still, like the rest of the RTX 20 series, relatively speaking this is a very large die for this product segment.

Consequently, while NVIDIA is officially setting the MSRP for baseline RTX 2070 cards at $499, we don’t expect to actually see them at that price any time soon. NVIDIA’s own Founders Edition card will carry a $100 premium, pushing it to $599, and we expect NVIDIA’s board partners to follow suit. This will price the RTX 2070 well ahead of current GTX 1080 cards, so it will be interesting to see where the new card fits in the bigger picture.

Source: NVIDIA

POST A COMMENT

54 Comments

View All Comments

  • Spoelie - Wednesday, September 26, 2018 - link

    If Ford called the next version of the Mustang the F150, would you consider it for your next truck purchase?

    You guys are getting caught up in names, when they have no practical meaning except for relative positioning in a stack.

    What matters is what you get in your hands, what it does at what price & performance.
    Reply
  • 29a - Wednesday, September 26, 2018 - link

    We're caught up in names because they should have meaning. When I buy an F150 I expect it to be in the 1/2 ton category, the F250 3/4 ton category and the F350 in the 1 ton category. Been like that forever. Reply
  • 29a - Wednesday, September 26, 2018 - link

    Ford is not going to call the Mustang an F150. F150 means a particular class of truck. Calling a Mustang a F150 would only serve to cause confusion. Reply
  • damianrobertjones - Wednesday, September 26, 2018 - link

    Brackets can be defined by whatever makes sense. Price, model name, spec, colour etc.

    But, yes, I do regard a 2070 to be the step from a 1070. Just as a 2080 is the step from a 1080.
    Reply
  • 29a - Wednesday, September 26, 2018 - link

    "If it's named 2070, then it's the successor to 1070, simple as that."

    I totally agree, you are 100% right.
    Reply
  • sorten - Tuesday, September 25, 2018 - link

    $500 for a second tier card. Ouch!

    I'll be keeping an eye on 1070 prices during the holidays.
    Reply
  • TallestJon96 - Tuesday, September 25, 2018 - link

    While $500 for a "106" gpu is still outrageous, this is probably the only card that actually improves in value of its pascal competitor. Its priced like a GTX 1080, but it should have performance between a 1080 and a 1080 ti.

    That's assuming it ever sells at $500 though. I have a feeling we'll be seeing them at $550-$600

    As an aside, I hope to see the 1070 ti tweaked and re branded as the "GTX 2060" for around $350 early next year.
    Reply
  • Alistair - Wednesday, September 26, 2018 - link

    I agree a 1070 ti re-release would make a great 2060 card. Not hopeful though. Reply
  • a5cent - Wednesday, September 26, 2018 - link

    While it might not be below nVidia to do such a thing, rebranding a 1070ti as a 20x0 series card isn't a great idea. If the difference was, as in previous generations, just about rasterization&shading perf/$ it would be fine. However, the 20x0 series is primarily about RTX and Tensor cores. nVidia sacrificed a LOT of shading & rasterization performance (relative to die size) to include those capabilities.

    Neither as nVidia nor as a consumer do you want to see any product in the 20x0 series omitting those features, particularly not in the more affordable/mass market SKUs. Omitting them in the most widespread SKUs would significantly reduce the chances of RTX and Tensor cores becoming widely supported in software!

    AMD being non-competitive in the "accelerated ray intersection testing game" is bad enough. We don't need nVidia shooting themselves in the foot, by also hampering adoption.

    Differentiating only the suffix, i.e. RTX VS GTX isn't enough. Any GPU that lacks RTX support and tensor cores must be clearly and obviously relegated to the "legacy tech" category.
    Reply
  • Zoolook - Wednesday, September 26, 2018 - link

    It's too early to tell, if AMD is non-competitive, MS, Nvidia and AMD has worked on DXR for a long time, it's not a surprise + AMD has a lot more compute in their cards so I wouldn't be surprised if Vega is twice as fast as Pascal in DXR, however they're not likely to touch dedicated hardware as in the RTX cards.
    That's were Navi comes in, unfortunately we likely have to wait until next year too see what they have come up with.
    I think the big problem with AMD GPU is that they have spent all their industrialization resources on the custom customers MS and Sony for a long time, that's why they are unable to compete with Nvidia.
    However they are starting to make money again and hopefully that means they can increase resources for the GPU side and increase their release cadence somewhat and get back into competing with Ngreedia again :)
    Reply

Log in

Don't have an account? Sign up now