Compute & Synthetics

Shifting gears, we'll look at the compute and synthetic aspects of the RTX 2070. Though it has its own GPU in the form of TU106, the hardware resources at hand are similar in progression to what we've seen in TU102 and TU104.

Starting off with GEMM tests, the RTX 2070's tensor cores are pulled into action with half-precision matrix multiplication, though using binaries originally compiled for Volta. Because Turing is backwards compatible and in the same compute capability family as Volta (sm_75 compared to Volta's sm_70), the benchmark continues to work out-of-the-box, though without any Turing optimizations.

Compute: General Matrix Multiply Half Precision (HGEMM)Compute: General Matrix Multiply Single Precision (SGEMM)

At reference specifications, peak theoretical tensor throughput is around 107.6 TFLOPS for the RTX 2080 Ti, 80.5 TFLOPS for the RTX 2080, and 59.7 TFLOPS for the RTX 2070. Unlike the 89% efficiency with the Titan V's 97.5 TFLOPS, the RTX cards are essentially at half that level, with around 47%, 48%, and 45% efficiency for the RTX 2080 Ti, 2080, and 2070 respectively. A Turing-optimized binary should bring that up, though it is possible that the GeForce RTX cards may not be designed for efficient tensor FP16 operations as opposed to the INT dot-product acceleration. After all, the GeForce RTX cards are for consumers and ostensibly intended for inferencing rather than training, which is the reasoning for the new INT support in Turing tensor cores.

In terms of SGEMM efficiency though, the RTX 2070 is hitting a ridiculous 97% of its touted 7.5 TFLOPS, though to be fair the reference specifications here are done manually rather with a reference vBIOS. The other two GeForce RTX cards are at similar 90+% levels of efficiency, though a GEMM test like this is specifically designed for maximum utilization.

Compute: CompuBench 2.0 - Level Set Segmentation 256

Compute: CompuBench 2.0 - N-Body Simulation 1024KCompute: CompuBench 2.0 - Optical Flow

 

Compute: Folding @ Home Single Precision

Compute: Geekbench 4 - GPU Compute - Total Score

The breakdown of the GB4 subscores seems to reveal a similar uplift like we spotted with the Titan V, which had scored in excess of 509,000 points. We'll have to investigate further but Turing and Volta are clearly accelerating some of these workloads beyond what was capable in Pascal and Maxwell.

Synthetic: TessMark, Image Set 4, 64x Tessellation

Given that TU106 has 75% of the hardware resources of TU104, the tessellation performance is in line with expectrations. For reference, we noted earlier that the Titan V scored 703 while the Titan Xp scored 604.

Synthetic: Beyond3D Suite - Pixel Fillrate

Synthetic: Beyond3D Suite - Integer Texture Fillrate (INT8)

Total War: Warhammer II Power, Temperature, and Noise
POST A COMMENT

121 Comments

View All Comments

  • webdoctors - Friday, October 19, 2018 - link

    You should shop around. I bought a new EVGA 1070TI from Amazon last weekend for ~$270.

    The regular 1070 is also that price now on Amazon website.

    The 1070TI definitely seems the sweet spot for upgrading right now...
    Reply
  • philehidiot - Tuesday, October 16, 2018 - link

    Have to agree here. I'd like to see the 1070Ti for completeness but Anandtech may not have the card to hand (you simply can't always keep stock of / put your hands on every card you want), might be planning on adding it later or just thought it'd be easy enough to extrapolate where it will fall. As for value for money, I did a massive table (in pen for some reason) which calculated aggregate benchmarks for a load of cards (1080Ti, 1080, 1070Ti, 1070, Vega 64) that I was interested in and then calculated a performance per Pound (UK here) for each one. Aside from the 1080Ti (which was far better value in terms of performance per Pound), the rest had practically identical value rankings aside from the AMD card which was slightly lower. I actually ended up buying a Vega 64 as the price briefly dropped by £110 just before Turing was released which made it an absolute bargain and, seeing the performance here, I'm relatively delighted that for once the GPU market has not screwed me over. I'm guessing the prices dropped as people were expecting Turing to be amazing and now they're back up as it's just pretty meh all over and very overpriced. My old card went for £100 (I don't normally sell them, but give them away to a friend) and so my cost to upgrade was ~£300 which I'm very happy with given that I would never EVER consider spending 2080 / 2080Ti money on a GPU. The worst part for Nvidia is that I am their target market - I'm a gamer who spends a lot of money on his PC to play the latest games at decent quality settings and I'm also a professional who isn't exactly poorly off. Could I afford a 2080Ti? Easily. Would I EVER buy one? No. The price is insulting and so is the marketing. I'd have never have considered AMD before they released all that insulting marketing rubbish. They gave AMD a sale. Reply
  • Spunjji - Wednesday, October 17, 2018 - link

    I know at least one other person with the exact same story. It's irritating for folks like us, but as long as the market continues to swallow it they won't miss our sales! Reply
  • hansmuff - Tuesday, October 16, 2018 - link

    A 1070 Ti is way too close to the 2070 in perf, why bother at this point when you don't know if the new features bring anything to the table with the 2070?

    A lot of people have 1070's, non-Ti, and have more reason to wonder if this a significant enough upgrade. I think Anandtech did the right thing here.
    Reply
  • JoeyJoJo123 - Tuesday, October 16, 2018 - link

    Nah. 1070Ti is as low as ~$400. 1080 is as low as ~$440.

    http://gpu.userbenchmark.com/Compare/Nvidia-GTX-10...

    Game performance differentials range from 6% to as much as 22%, but i'd say for 10% more on the price, you get a tiny bit more than 10% of the performance uptick. Hence, 1080 is a better overall value in terms of price/performance to 1070 Ti.
    Reply
  • Death666Angel - Tuesday, October 16, 2018 - link

    Not according to Techpowerup and German prices. The cheapest 1070ti is 430€, the cheapest 1080 is 480€. The relative performance towards a 2080Ti is 62%/55%/50% for the 1070Ti in 1080p/1440p/2160p and 66%/58%/52% for the 1080 in the same resolutions. That means the 1080 delivers only 95%/94%/93% of the performance/€ compared to the 1070Ti. Things are so close though as to warrant a real inspection when seeing a deal for one card or another. Reply
  • Vayra - Monday, October 22, 2018 - link

    While true, once overclocked, the 1080 can stretch its legs a bit more due to the faster GDDR5X, making the lead a bit larger, and its additional shaders also benefit a bit more from a 'similar' clock to a 1070ti. They're almost a perfect match if you consider perf/dollar, and in that case the 1080 is and was always the better choice, because higher absolute performance should usually result in a worse perf/dollar number. Reply
  • Marlin1975 - Tuesday, October 16, 2018 - link

    I'll take a $329 RTX2080. Comparison chart is off.

    Performance is not bad, but the price is. Until the price comes down the 1080 seems like a much better buy, esp with rebates on some now.
    Reply
  • Koenig168 - Tuesday, October 16, 2018 - link

    Paying for RTX which we cannot use ATM and not sure we want, bearing in mind the performance hit. Reply
  • Antoine. - Tuesday, October 16, 2018 - link

    Why do you compare in your final words the RTX2070 to the GTX1070? Just because they have the same marketing endname does not mean it's in the best interest of your readers to compare these two.
    The only relevant guide here is the pricing of course! RTX2070 have the same pricetag as GTX1080TI. Comparing it to GTX1070TI is a stretch but why not. But the 1070?! a 400-dollar card?
    Reply

Log in

Don't have an account? Sign up now