AMD Radeon RX Vega 56 versus NVIDIA GeForce GTX 1070

Jump To:

Radeon RX VEGA 56 versus GeForce GTX 1070

On August 14th, AMD announced three new graphics cards based on the ‘Vega’ GPU architecture: Radeon RX Vega 64 Liquid Cooled Edition, Radeon RX Vega 64 with air cooling and the Radeon RX Vega 56. The Radeon RX Vega 56 came a little sooner than we expected and we frankly weren’t prepared for it when the card arrived. AMD allowed reviews of the Vega 56 to be published on August 14th and actually asked reviewers to focus on it, but it won’t be available to purchase until August 28th. When the AMD Radeon RX Vega 56 becomes available in the retail channel on August 28th it will command a $399 MSRP and AMD expects it to be the clear leader at that price point.

All of the Radeon RX Vega graphics cards support HDMI 4K60 and DisplayPort 1.4, Radeon FreeSync displays, Ultra HD, ultrawide, and single-cable 5K resolutions.  So, if you are looking to pick up a new display and want some flexibility and an affordable adaptive sync solution the Vega lineup looks pretty solid on paper. The Vega 56 at $399 has 56 Compute Units with 3,584 stream processors running at up to 1471MHz for 10.5 TFLOPS of peak single-precision compute performance. Vega 56 also comes tacked with 8GB of HBM2 memory running at 1.60 GHz (effective) and that gives this model 410 GB/s of memory bandwidth. When it is comes to board power you are looking at 210 Watts for this model. Over 10 TFLOPS of compute and over 400 GB/s of memory bandwidth for $399 seems like a steal for gamers.

The only problem for AMD is the NVIDIA GeForce GTX 1070 8GB GDDR5 series graphics card that was released in June 2016 for $449. NVIDIA has reduced the MSRP on the GeForce GTX 1070 over the past year and the MSRP on this model is $379. The NVIDIA GeForce GTX 1070 has 1,920 CUDA cores running at up to 1683 MHz or 6.5 TFLOPS of peak single-precision compute performance. The 8GB of GDDDR5 memory on the GTX 1070 is clocked at 8,000 MHz (effective) and runs on a 256-bit bus for 256 GB/s of memory bandwidth. NVIDIA was ableto get all that performance in the GTX 1070 and managed on keeping it at a 150W TDP rating.

RX Vega 56 GeForce GTX 1070
GPU Vega 10 GP104 Pascal
Transistor Count 12.5 Billion  7.2 Billion
GPU Cores 3584  1920
Base Clock 1156 MHz  1506 MHz
Boost Clock 1471 MHz  1708 MHz
Texture Units 224  120
ROP Units 64  64
Memory 8 GB 8 GB
Memory Clock 1600 MHz 8000 MHz
Memory Interface 2048-Bit HMB2 256-bit GDDR5
Memory Bandwidth 410 GB/s 256 GB/s
TDP 210 Watts 150 Watts
Peak Compute 10.5 TFLOPS 6.5 TFLOPS
Process Node 14nm 16nm
MSRP (Aug 2017) $399 $379

Looking at the numbers on paper makes it look like a clear win for AMD, but is it going to be? Could the tale of the tape be wrong?

AMD Radeon RX Vega 56 versus NVIDIA GeForce GTX 1070

To figure out if the Radeon RX Vega 56 or GeForce GTX 1070 is the best GPU for under $399 we grabbed  both cards and ran them through a series of benchmarks at 1080P, 1440P and 4K display resolutions. We have several GeForce GTX 1070 video card models here to pick from, so we actually randomly drew brands from a hat and the Gigabyte GeForce GTX 1070 G1 Gaming graphics card (part number GV-N1070G1 GAMING-8GD) was the sacrificial lamb from the group. This card has a base clock of 1594MHz and a boost clock of 1784MHz, so it has about a 100MHz factory overclock on it. The MSRP on this card is $429.99, but good luck finding any high-end gaming graphics card at its suggest retail price thanks to the cryptocurrency mining boom!

AMD Radeon RX Vega 56 versus NVIDIA GeForce GTX 1070

The AMD Radeon RX Vega 56 measures 10.5-inches in length versus the 11-inch long Gigabyte GeForce GTX 1070 G1 Gaming. When it comes to GPU cooler designs the Vega 56 model has a blower style fan and the GTX 1070 has the WINDFORCE 3X cooling system. We hate to compare a custom board partner card to a reference design, but lets face it not many are buying a GeForce GTX 1070 Founders Edition today.

AMD Radeon RX Vega 56 versus NVIDIA GeForce GTX 1070

To get these cards running properly in your gaming PC you’ll need to hook up two 8-pin PCIe power connectors on the Vega 56 whereas there is just one needed on the GTX 1070 model that we are comparing it to today. Both cards have backplates and lighting on the tops of the card.

When it comes to video outputs for display connectivity both cards have three standard sized DisplayPort connectors and one full sized HDMI port. The GeForce GTX 1070 does have a DVI output though that isn’t seen on the AMD Radeon RX Vega 56 reference design.

How about we spare you the chit-chat and let’s do this! Let’s take a look at the test system and then see the benchmarks!

Print
Jump To:
  • forextor

    Impressive review… you even included mining comparison!.. I am looking forward for comprehensive reviews like this in the future..

  • Meow Meow

    AMD should have priced BOTH cards 100$ USD LOWER and this would have been a much better discussion and review…

    Although I guess it doesnt matter what theyre priced if Joe Blow consumer cant obtain one.

    • Meno Mam

      One important piece of information that has been repeatedly presented wrongly in the review. 1070’s MSRP is $349, NOT $379. Miner inflated prices aside, this puts vega 56 in even worse light

  • Meow Meow

    They are limiting cards and still theres no stock.

    Big miners have contacts within the industry to procure cards before release…
    Also Coach, right now, probably not worth it. Unless you get free electric… Or get creative and mine an odd coin that may shoot up in value.

  • Coach

    I do not know why sellers will not limit the card to one per customer to start. That may help. At this point, the return on Ether is way down, is it not? I mean, even if you buy 4 Vegas in quad crossfire, would it be worth the electric to mine?

    • Nathan Kirsch

      The chart below is based one Radeon RX 480 video card at stock settings. It still is profitable to mine, but you are only making a couple bucks a day per card after electric expenses. As you can see the amount you make is generally dropping. https://uploads.disquscdn.com/images/6aaef01d878d256c93e1bb1ea7c62f6d1a7d0f6c4ce0692bc5ef7101bece9a11.png

      • Ted

        This is a truly outstanding contribution to the article, Nate. And if electrical holds steady while performance drops, we got to rethink a lot about latest-gen options in GPU mining.

      • Rui Da Silva

        Hi there, can you let me know what site did you use to create this chart?

  • Coach

    I wonder how the aftermarket versions of Vega will fair. The comparison is with a Gigabyte G1 card with Windforce cooling (3 fans) versus a stock reference model. We know they do not cool as well. If you compared a Gigabyte version of Vega with 3 fans, you would see more comparable sound levels and likely a very slight bump in fps in some games, as usual for a non-reference card. When they come out in another month or so, a more direct comparison of Gigabyte cards or another brand would be interesting to see.

    • Coach

      LOL, I jumped the gun posting this after reading page 13, page 14 calls me out!!!! 😀

      • Coach

        At least I wasn’t mean like some people. 🙂 It was a “LEGIT” question.

        • Nathan Kirsch

          Coach, Your comment wasn’t mean it was truthful and makes a valid point. When custom AIB cards come out hopefully next month I’ll of course review them and compare them. I had people call me mean things, tell me I don’t know what I’m doing, that this is a waste of space and even link to ‘better’ sites on this comment section. Funny they linked to sites that are pay to play and have no idea what is going on behind the scenes. Some sites get cards days before others, so to have someone come on here and bash my coverage and then link to another site with better more in-depth coverage does strike a nerve. When a site gets paid for something and gets it earlier than the rest… I do expect it to be more in-depth and better. Heck, it should be as they were getting paid for it. I haven’t deleted a post on LR in years, but that got me. If anyone wants to see something tested and I have the ability to do it then I’ll do it. I do it for free and have always done that for people in the comment section and the forums. It’s one thing to ask someone to try a different test setup if they could and another thing to tell someone they are garbage and bash them. That isn’t right.

          end rant… sorry 🙁

    • Coach

      However, I still want to see a Gigabyte Vega 56 with Windforce tested. 😉

      • Nathan Kirsch

        We are still around a month away from custom AIB cards coming out. Heck, we are still 10 days away from the Vega 56 reference card being released. I so want to see the custom cooled cards as well. Mining performance on the water cooled cards look good, so it looks like lower Voltage and temps might actually help this series! I don’t have any water cooled models here to try it on and AIB’s haven’t even shipped cards to the US yet.

        • Ted

          I believe said card is out today

  • Norman

    Finally a review that compares Vega to the newest GTX cards ….
    Interesting read, bit dissapointed at Vega results and availibility.
    Lets see what the price does compared to 1070 when it does come available over here.

  • AMDominant

    Weak, run of the mill shill “testing” here. Of course FPS, thermals, and noise are going to favor an AIB triple fan vs a reference blower. Thanks for wasting space.

    • Nathan Kirsch

      Glad to see AMD Team Red is still kicking. Please go take a look at Newegg and count the blower style GTX 1070 cards versus the others and let me know what you came up with. No waste of time here just facts. If you don’t like the temperature/noise results then don’the look at those pages. The FPS results are what they are. AMD has had Vega running in the labs since 2016, so I think they were able to properly cool it.

      • Conor

        Deleting comments that call you out, nice one.