Benchmarks – AMD Radeon R9 290X Versus NVIDIA GeForce GTX 780

NVIDIA is holding a press event in Montreal today and it looks like AMD is crashing that party. AMD set up shop at the hotel across the street and allowed the media to come over to run both the game titles BioShock: Infinite and Tomb Raider at 4K resolutions with the AMD Radeon R9 290X versus the NVIDIA GeForce GTX 780. This rather bold move came as a shock as NVIDIA flew us up here to attend their press event and here we had AMD asking us to come to their hotel to look at how their graphics card performs. As you could imagine NVIDIA wasn’t too happy with this marketing stunt and sat outside AMD conference room to deter their invited guests from entering. Since we were guests of NVIDIA we declined to visit with AMD here in Montreal, but AMD did send over pictures and results of the demos that they were running inside the room. Some press did cross into ‘enemy’ lines and they confirmed that the numbers that we were supplied with were very close to the ones they personally ran.
Here are the specs of the system AMD is running in Montreal:
  • Intel Core i7-3960X – 3.3Ghz
  • MSI X79A-GD65
  • 16GB DDR3-1600
  • Windows 7 64 w/ SP1
  • NVIDIA driver: 331.40
  • Future build of AMD CATALYST 13.11 beta (to be posted with launch)
PC Game Settings:
  • BioShock Infinite: 3840×2160 Ultra preset
  • Tomb Raider: 3840×2160 normal preset with TressFX off (UPDATE: AMD Says TressFX was off)


In Bioshock: Infinite with the Ultra preset the AMD Radeon R9 290X ran at an average of 44.22 FPS and the NVIDIA GeForce GTX 780 was at 39.63.This shows a significant 11.6% performance advantage for the new AMD Radeon R9 290X with the Hawaii GPU.
Tomb Raider showed the AMD Radeon R9 290X averaging 43.0 FPS and the NVIDIA GeForce GTX 780 was at 40.8 FPS. This would make the AMD Radeon R9 290X roughly 5.4% faster than the NVIDIA GeForce GTX 780 in Tomb Raider.
These legit benchmark numbers should help shed light on a number of things.  For one the AMD Radeon R9 290X appears to perform better than an NVIDIA GeForce GTX 780 at 4K resolutions, but pretty much everyone assumed or hoped that would be the case. It should be noted that these results are also on beta drivers and as time goes on we expect AMD to get some more performance from the brand new 28nm Hawaii GPU.
When it comes to pricing, the NVIDIA GeForce GTX 780 retails right now for around $659.99 online, so where will the AMD Radeon R9 290X be priced? AMD has traditionally priced their cards very competitively against NVIDIA, so will we see the AMD Radeon R9 290X in the $629-$649 price range?
AMD is also letting sites publish their own 4K benchmarks from Bioshock Infinite and Tomb Raider today, so expect to see some sites having numbers from their own unique test setup. We are here in Montreal and far from our test system, so expect to see our benchmark results when the full review goes live.
  • Shaq

    Wrong setup ATI with intel, use it with proper platform ATI -> AMD FX -> RADEON R3 memory and same setup for Nvidia with his proper platform Intel same resources so there you will notice the performance of the GPU and the real diff between ATI and Nvidia 😉

  • Sheraz wizzer

    just would like to go for amd cheap and better.

  • iostream

    this is a little sad but shut up and take my money AMD

  • jojetjoengsje

    #chair #sandniggersyunoarabs #crumble # jujojoenksje

  • Sowkloss

    Both of these games are optimized for AMD cards anyway? Not trying to be a fanboy or anything, although I guess nearly all modern games are optimized for either the Red or Green Corner…

  • Serpent of Darkness

    Going off the following website:

    Doing some number crunching on a theoretical performance output on the GTX Titan, the following could be true with some degree of error. Take into account, this is RX9-290 on “Quiet Mode.” These numbers should be considered with the idea of using the 2160p resolution…
    Bioshock Infinite: 41.79 FPS; 5.27%.
    BF3: 34.04 FPS; 8.37%.
    Crysis 3: 31.12 FPS; 3.52%.
    Hitman Absolution: 35.54 FPS; 1.13%.
    Skyrim: 56.35 FPS; 26.38%.
    Tomb Raider: 36.0 FPS; 11.94%.
    Far Cry 3: 31.62 FPS; 17.67%.
    All the percentages are the difference in performance between the RX9-290 and GTX Titan. All percentages show RX9-290 providing better performances based on the values given up above.

  • Lominas

    290x is about the same as a TITAN, surpasses in some areas, falls behind in others… leave it at that….. More in-line with the 780, bench and price range too. TITAN is a GK110 compute card too, keep that in mind u sick fuks. 780 is pretty much the graphics only version of the GK110.

  • Nikos

    R290X is a faster from GTX780 but Titan is still is faster in most tests

  • Tequila_Mckngbrd

    Not looking good. If they have to let loose two gaming benchmarks, it’ll be ones where they perform best against the GTX 780.

    Unlike the GTX 680, the GTX 780 can unlock voltage. The HD 7970’s in Xfire was always faster fully overclocked than GTX 680’s in SLI. With the GTX 780’s now being able to overclock to their potential, the R9 290X had better be able to overclock like a monster.

    • Principle

      I am positive there will be benchmarks that actually show even more favor to the R9-290X. Bioshock was not favored to AMD before, but there are titles that even an HD7970GE keeps up with the 780.

  • Jason

    I love how amd uses an I7 to test the NEW R9 290X.
    Must of not got the numbers they wanted with a FX cpu.
    Big victory Amd!

    • gonchuki

      AMD’s drivers are still heavy on CPU usage, so a weaker CPU makes their card look worse than it actually should.
      I don’t know why review sites don’t talk more about this one, but these can easily be spotted on game benchmarks where they compare both AMD/Intel CPUs and AMD/Nvidia cards for scaling, and you can see that while using Intel CPU the AMD card ends up better or tied with Nvidia, switching to an AMD GPU puts the same card way behind on the AMD platform (finding it hard to google for one of these right now, you will have to trust my word this time.).

      • Tequila_Mckngbrd

        That’s the whole point he was getting at…

        • gonchuki

          yes, I was explaining the reasoning behind going for an i7 instead of an FX as his sarcastic remark didn’t infer that.

      • Jason

        So… for enthusiast gaming everyone running an AMD card should be running an intel chip-set? And why cant AMD run this card on the Scorpius platform and get all of the magical benefits of unlocking an AMD platform.

        • gonchuki

          Basically yes, until AMD gets better drivers that depend less on the CPU IPC (or memory bandwidth, both are inferior on current AMD platforms), or otherwise release a CPU with such characteristics. Kaveri/Steamroller will probably help in both departments, we are still not sure about how much.

  • PJ12

    by the way above mentioned games favors AMD more.

  • PJ12

    *Quite Mode* interesting.
    Hoping there will be game mode, which will be able to bump up fps

  • cum

    It goes to show, AMD FEAR Nvidia’s progress, they wouldn’t have bothered to crash anything anyway. We also know what happens to trolls as well. Heil Nvidia!

  • bodz

    I thought R9 290x is a Titan killer? Why compare it with just the 780 was it because it was their target price range? If that is so, NVDIA can always tune-down their prices 🙂

    • gonchuki

      “Quiet Mode” probably means Turbo Boost is disabled which (if the leaks are accurate) parks the card at 800mhz instead of reaching up to 1ghz (basically a 20% clock speed penalty). Since there’s enough bandwidth to throw a party, scaling the GPU clock will mean close to linear performance increase, and we all know the Titan is not even close to 20% faster than the 780 (the Titan is 15% faster at best, usually 5-10%).
      All in all, this means the R9 290x in quiet mode is already halfway between the 780 and Titan, now imagine it with 20% more clock speed.

      • Serpent of Darkness

        I have to agree. The hurdle, or amount of performance need to surpass the GTX 780 and Titan, is very small, and proof of this is shown in the leaked benchmarks on In addition, more proof of this is shown in NVidia’s reaction to the RX9-290 benchmark. I think they got scared, and with that, released information about G-Sync and GTX 780Ti. So they are probably looking at the worst case scenario for the GTX 780 and Titan on the competition front. The Turbo Frequency pushes the RX9-290 over GTX Titan and GTX 780. Enough to be a small percentage over, but it surpasses the current competition in performance. I suspect the GTX 780 TI will be the same Cuda Core count at the GTX 780, but with less leakage. So that will translate to a higher core frequencies and Overclock headroom. G-Sync is a nice niche, but it’s going to suck if it’s NVidia only. This could cause other Monitor producing companies to adapt the same technology if demand goes up. Seeing as how it’s will more than likely be NVidia only, NVidia will only milk the NVidia Consumer Base. Some Consumers on the AMD side may switch, but if PCIe Crossfire through the PCIe Bus is successful, it will buffer any exodus of consumers from one side switching to the other…

  • Palash Sarate

    no cooling fan for cpu with nvidia gtx

  • basroil

    anandtech’s benches are pretty far from AMD’s, showing the two cards just 1% off (probably within margin of error) in bioshock with everything maxed out. It seems the R290x is made to play games at lower quality settings, because it only shows significant leads when quality is turned down!

  • Gogo

    Go go go

  • john

    AMD lift the NDA damn it! 🙁 we’re dying out here! WE. NEED. BENCHMARKS!!! kapishe? no seriously wtf si quiet mode?? I want to see this card in rage mode!!

    • David Calloway

      LOL!!!!!! Right….Right…

  • papoulka

    With one exception this is the most civil, intelligent, and informative comment string I’ve come across in years. And I tend to read only the top sites.

  • you need to remove one of the ‘as time goes on’ phrases

    “It should be noted that these results are also on beta drivers and as time goes on we expect AMD to get some more performance from the brand new 28nm Hawaii GPU as time goes on.”

    • Nathan Kirsch

      yes, we do and that has been fixed!

  • snook

    Socius, big up to you for keeping level-heading while they questioned your statements. rare internet calm.

  • Serpent of Darkness

    I think the blurred out images on the first column, is showing 59 and 57 FPS. The image is twisted, but you can see parts of it if you zoom in. I’m assuming this is at 1.1 GHz Turbo Speed. I have no clue what a “Quiet Mode” is. Must be something new. I know Tomb Raider is heavily favored by AMD because of Directcompute. Having the RX9 beat the GTX Titan on Bioshock, is gratifying since it’s a AMD Evolved Game. The AMD 7000 Series was failing in performance with that game. Between both cards, on TB there’s a 39.47% performance difference in favor of 290. This percentage drops, still in favor of AMD, in “Quiet Mode.” 36.15% difference in performance between the RX9 and GTX Titan on Bioshock, in favor for AMD.

    • Serpent of Darkness

      I have them reversed. 36.15% for Tomb Raider. 39.47% is the performance difference for BioShock.

    • sCuBaSte

      It’s a stock GTX 780 not a Titan.

      • ActionKamen

        780 Out performs Titan.

        • Robert Boluyt

          GTX 780 does not outperform Titan. 780 is a cut down Titan.

        • Louis Preston Thornton

          780 out performs the titan in only gaming, rendering its a different story

        • William M

          Nooo….. just no. The 780 is a cut down Titan in every aspect. The Titan beats the 780 in EVERYTHING

        • sCuBaSte

          The Fastest GTX 780’s Out perform a Titan in gaming benchmarks etc. True Story! Google it! EVGA GTX 780 Classified, Zotac Amp edition etc etc etc

        • Robert Boluyt

          These are overclocked models. We’re talking stock clock speeds. The 780 is a cut down Titan.

        • Rick

          No, the titan has a locked clock speed and cannot be overclocked. The 780 HOF pulls ahead in every way.

        • Socius

          Titan doesn’t have a locked OC limit. It’s limited by power draw only at Max 1.212v stock. Swapping to water cooling and unlocking voltage and using a modified bios allows you to bypass all that. I can run mine at close to 1400MHz though only do it for benching and stick to 1.34v at 1350MHz for day to day. Memory also clocked at 7600MHz.

        • asskicker

          In a way it does just like all other GPU´s out there but it´s in bios which mean your Titan can be clocked even further if you unlock bios limitations as far as vcore etc goes right?

        • Socius

          I’m not sure I understand. Even prior to bios modification or water cooling I ran my Titan at around 1200MHz for benchmarking. I don’t know where you heard that the OC capability was locked. That is certainly not true. The only limit, as I mentioned, is the voltage at 1.212v. That’s the one you can use modifications to remove. Otherwise all else is free to clock as you please.

        • asskicker

          Sorry I´m not a native English speaker so probably my fault I did not make sense, English is far down on list of languages that I speak but learning more and more every day now.

          All GPU bios have limitations in them by default that´s about what I mean, or call them predetermined values if you like after all it´s the software that controls the hardware on low level and have to have it but like you say yes one can always flash to a modified or different bios. When OC a card it´s going to need more and more juice the further you push it and if stock cards have no limitations in them many noobs would burn up their cards no doubt about it, you seam to be under the impression I´m arguing against you and I´m not just pointing out the obvious.

          As far as performance goes when talking Titan vs R9 920x we simply do not know yet but in a few days we will, nor do we know if R9 290x will become a beast of a OC card or not. Some on here are pure fanbois with nothing to add that much anyone can see and it´s very sad to see people reach those ridiculous levels cause any intelligent person will buy what´s best for her/him and most likely look at the cost as well and then we have some standing on knees in worship of a bloody company and defend like it was their mom in a court room, lol

        • Richard Amable

          The 780 HOF edition due to the overclock beats a Titan non-OC in gaming. Not in a stock to stock comparison.

        • Carl Hill

          The 780 can only defeat the Titan IF the Titan is NOT overclocked…

        • tristen mahle

          It really dosnt the 690 out performs the titain in game rendering the 690 is the best for gaming

        • Socius

          Unfortunately the 690 has a limited bus, higher thermals, and only 2gb VRAM. And since SLI causes increased micro stutter, many people go with a single Titan for the smoothest gameplay. And this way there is no issue with SLI profiles and etc etc…I used to have 2 gtx 690’s.

        • FrankVVV

          LOL, what?

        • lou

          pretty sure everything is reference so this isn’t at Titan level performance…..

      • asskicker

        ..of course it is my friend, that´s what the R9 290X is aimed at according to AMD hence all the speculative talk about $600-650 price range

        • sCuBaSte

          I know 😀 Was simply pointing it out to a previous poster.

        • asskicker

          Yeah I get it buddy, my English does not always come out like it should either hehehe..

          What I think about it? Well imho it´s Nvidia feeling the heat and they want to stay on top so they simply start to push their marketing machine once again to not allow AMD to get to much attention? I see no real reason for that card unless price stay at GTX780 levels and GTX780 itself fall down a bit and still it would be slightly unnecessary + What about Titan LE then?

          But just my personal view of it, I like it when these 2 companies go to war cause it will benefit us the customers! I will buy from any company and always did but sure I would love to see AMD have as big share of the market as Nvidia when it comes to dedicated GPU´s but even better with a few new companies in there a well?

          just my 2 cents though

  • Suspicious


    Jesus fuck. This card probably electrocutes 6/10 people who use it or something, why the cloak-and-dagger:ing otherwise?

    • Robert Boluyt

      They don’t want to tip their full hand to Nvidia just yet.

      • Trey

        Nvidia already knows. A tremendous chunk of their company depends on accurate, timely information.

        • Sly Cooper

          *industrial espionage

        • Suspicious

          Their “full hand”? The card wont trump much of anything before mantle. Weve seen TWO confirmed benchmarks on TWO AMD-gaming-evolved titles in 4k resolutions. I am dissapoint.

          The only efficient part of their strategy is that nobody is buying 780s because we dont know what performance were looking at 😛

        • Raijin

          they may be amd games but 290x don’t run at full power.notice that it says “Quiet Mode”?the full mode or what ever is called isn’t shown.i bet its more than 50

        • Suspicious

          Obviously companies only run their products in a low-noise mode when showing its competetiveness. Oftentimes race cards lose races because the driver is hard on eco driving.

          Get a grip, if it was in quietmode then that was because they needed it to be or it didnt make any difference.

  • Monir Khader

    i think it will stay on par with GTX TITAN since it beats the 780…

    • Socius

      Really depends on how much it overclocks. I have my titans voltage unlocked running memory at 7600MHz (stock is 6000MHz) and GPU clock at almost 1400MHz (stock is 876MHz). So how well the 290 can overclock will be important as my Titan is essentially running at 1.6x the speed of a regular Titan. So a benchmark that would normally get a 45fps average on a stock titan would instead get 72fps. And even more important will be micro stutter and amd’s general issue with crossfire. Since most 4k display users will be running a multi-gpu setup, this becomes an important question.

      • Dominic

        Do you have to be watercooled to get those kinds of numbers? I have my GTX 780 overclocked a little with the stock cooler.

        • Socius

          Yes. Triple-radiator loop. The vrm will blow if you go too high. Though the 780 generally does overclock better than the Titan due to a much better vrm used. I’ve heard if people getting 1200-1300 on air with the 780. For how long and what dangers exist if you do it, I don’t know.

        • HughhHoney

          What voltage are you running on your Titan?

          I have my 780 at 1333/7240 on air. Voltage is 1220mv with llc disabled. Can’t go much higher without temp being a problem.

        • Socius

          Definitely envious of your 780’s OC ability. I need 1.34 for 1350mhz. I only go higher for benching. Not 24/7. My temps range from 40c-48c at those clocks.

        • HughhHoney

          Nice. I’d like to put my 780 under water, but I don’t really feel like dealing with it right now. Gotta love temps in the 40s…

        • Nnngghh

          My 7950 goes to 1500 with air

      • 124234

        micro stutter is gone with 290 series since there’s no longer a crossfire bridge needed which caused the problems.

        • Socius

          Micro stutter isn’t caused by the cf/SLI bridge. It’s due to multiple cards having to render one scene. The same data passed through CF previously is still going to be sent but through pcie this time. Amd has done much better of a job with micro stutter as of late and it’s be interesting to see if they up their game yet again with these new cards.

        • Serpent of Darkness

          This isn’t completely true. CFx causes the individual GPUs to produce Frames, one at a time, in a series. So GPU 1 produces frame 1, GPU 2 produces frame 2, GPU 1 produces Frame 3, GPU 2 produces Frame 4, etc… That’s how it is different from SLI. SLI uses link or tiny rectangles on your screen, and divides the work amongst the GPU to display an image or fame at the same time. Those images come in sync with one another. This reduces the load by each GPU on the same image. CFx basically pushes 2 or more GPUs at full load, pushing fames almost at the same time. Problem is that they don’t scale, and you can’t push 3 or more frames over a smaller interval of time fast enough. The current issues with CFx, right now, is that the frames overlap one another. Frames 2 and 3 might overlap one another, or Fames A and B, 40 sec later, overlapped. The overlapping causes the GPU to say “hey, we screwed up, we need to re-draw this frame.” As a result, a Runt Frame is produced, and time to produce additional frames goes up. Hense, Frame “Pacing” software is on the AMD Beta 13.8, 13.10, and 13.11 Drivers.

        • wcafricanus

          Good info, knowledge acquired:)

        • i dont understand why people keep thinking no bridge = better CF, the amd slide said ‘no penalty versus bridge’, meaning the same as bridge

          i find it interesting that there has been talk about how 3 cards, both sli & cf fix microstuttering

        • Serpent of Darkness

          The AMD CFx/NVidia SLI Bridges could have an operating limit. There could be a limit to the amount of transfers between each Buffer Card with the Primary Card. The point is this. The 50 Giga-Transfers per second, through the PCIe Bus is more than likely going to be regulated by the AMD Driver. So there could be a possibility that his number will go up. Transfer rate between the GPU and PCIe Bus could increase because PCIe 3.0 x16 has so much bandwidth. If Hardware specs possibly allowed it, it could probably go up. Increase rate of transfer between the GPU, the PCIe Bus, and the CPU, will probably result in increased performance for AMD and CFx. In addition, it could also mean an increase in performance between the Primary and Buffer Cards for AMD. If this is true, Runt Frames could be less of an issue because Frame Rates between in RX9-290 in CFx, will push out frames faster. There’s no physical bridge holding it back because the bridge has a limit to it’s operation. If a Physical Bridge is already producing or transferring at the same rate as a PCIe-Based CrossfireX, then there probably won’t be a change. There’s the possibility that the CPU could have some form of latency, or there won’t be a diminishing performance between the GPU, PCIe Bus, and CPU.

        • what about the dual gpu cards? there are no bridges or mobo slots to pass through, just the PLX chip, so logically wouldnt they perform better than CF/SLI at same clocks? (they perform the same)

          i would also think the bridges were made to simply fill in the needed bandwidth that the slots didnt have until pcie3 specs, would be strange to make a custom proprietary bridge that doesnt have enough bandwidth

        • Mills

          Youre right, the bridge isnt a bottleneck. But at the same time, your statement about dual gpu cards performing the same as their single gpu cfx/sli counterparts is slightly off. Most benchmarks will show slight variances in their average performance.

        • variance as in… better or worse average performance?

        • Serpent of Darkness

          Runt Frames is only with CrossfireX Cards. A single 7970 or RX9-280 or RX9-290 won’t produce Runt Frames because the single card takes up 100% of the work. In addition, PCIe for the 290 will be through the PCIe Bus aka through the lane. So PCIe 3.0 x16 will become useful for this purpose. The only question is how much better is 50 Giga-transfers per sec, through the PCI Bus, is CrossfireX going to improve in performance versus a physical bridge connector, and is there going to be a significant Frame Time Latency between the buffer cards (290 cards 2, 3, and 4).

      • John Smith

        I have a very hard time believing you, because Nvidia does a very good job at locking the voltages on these cards.

        So other than hard-modding your Titan I don’t see anyway you could’ve unlocked the voltage on it.

        In any case the R9 290X when overclocked does beat an OCed GTX Titan.

        • Socius

          Unlocking voltage isn’t hard. You can follow the thread on You can also see a picture of my rig on the front page of

          The OC’s you linked don’t show too end overclocks for either the Titan or the 290x. I’d like to see how they stack up because that will determine the dominant enthusiast card. Should be interesting to see.

        • sirus

          Very few people would void their $2000 warranty on the Titans just to Volt it up. I agree that we should wait for hwbot listing etc. to see what cards are better.

        • Socius

          You’d be surprised how many of us do it. Anyone going tri or quad SLI titans wants ridiculous levels of performance that is tapped by unlocking voltage. Check out the thread on for proof.

        • Principle

          You might be surprised how few of you mess with their two or three $1000 cards, compared to the number of them sold, which is already much lower than the number of overclocked 780s. So I am more than willing to place a bet that there are more 780s out there running faster than Titans for less money. You might be king, but there are a lot less kings than there are noble men, so what really matters.

        • Socius

          I could be wrong but my assumption would be that anyone who wants max stock performance would get a binned higher end 780 instead of a Titan considering stock clock differences, and the price difference. Purchasing a Titan, given the option of said 780, with the intention to keep it at stock clocks just doesn’t make a lot of sense. That’s not to say that some people won’t do that because they are uninformed and automatically think the Titan is the highest of the high end cards, but with the Titan Nvidia specifically targeted people who apparently have a little too much money on their hands. Those of us who have the money and also the experience, don’t have to worry about risks of oc’ing. I guess my perspective may be skewed as I generally deal with enthusiasts who are informed or are looking to expand their knowledge and partake in some of the less common overclock that is possible as a result.

      • john

        you know performance doesn’t scale with frequency especially at ridiculously high oc’s as yours. I don’t even want to know how big your total heat dissipation capacity is or how big that thing is for that matter 😛 The idea is if you’re running at 1.6x speed the performance increase will most likely not be linear.

        • Socius

          It is quite close to linear if you have enough CPU power to keep up. Fortunately I have my 3770k at 5.2GHz for 24/7 use and 5.3GHz for benching.

        • john

          well kudos to you… you’re a better overclocker then I ever was 🙂 I always had problems with artefacts (like green dots or sticky shadows) and the sorts when I pushed the gpu too much – It had good frames yes but the artefacts were just too annoying so I dropped the speed in favor of quality to play games. From my experience the higher I punched the voltage & frequency the more frequent the artefacts occurred. Just my experience/luck I guess 🙁

        • Socius

          Artifacts are almost Always caused by your VRAM and not GPU clock. Increasing voltage can help get higher stable clocks. But your thermals will likely increase and that is what causes the artifacts you speak of. Try lowering your VRAM and just play with the gpu clock. Shouldn’t get artifacts.