AMD Radeon R9 290X Press Sample Versus Retail

Jump To:

What is Causing The Radeon R9 290X Frequency Variance Issue

The AMD Radeon R9 290X and R9 290 have been getting a ton of attention lately due to a number of reports that the retail cards are performing differently than the press cards that the media sites received. We have been following these stories for the past few weeks and finally decided to look into the situation ourselves. We knew from the start that there was a performance difference between testing a ‘hot’ and ‘cold’ graphics card and that AMD’s new PowerTune technology was very temperature sensitive. The AMD Radeon R9 290 series is designed to run at 94C when gaming and the core clock is raised or lowered to maintain that temperature up to a set maximum clock speed. The new twist is that you’ll see performance differences on the same card at various room temperatures and the card will slow down as it heats up.  This caused mayhem for some review sites as they were testing from a ‘cold’ state and only doing one benchmark run and not averaging them. Luckily for us, we average our results and have been testing at regular operating temperatures for some time. The thing that caught us off-guard is that the retail cards were being shown to perform 10-20% slower than the press cards when run in quiet mode. AMD tried addressing this variance issue by releasing a new driver that increased the fan speed from 45% to 47% as they thought that it was due to heat. That moved helped fuel the fire that AMD cherry picked low-leakage cards to be sent to the press for the best performance numbers and the retail market has the typical high-leakage cards that run hotter and thus slower. Trying to normalize the fan speed is one way to go about it, but it also makes the cards louder.


The only way to figure out what is going on was to get a retail AMD Radeon R9 290X that was purchased from the retail market to see what is really going on. We did exactly that and have spent the past week trying to figure out if there is a significant performance variance between the cards.


Our retail card was purchased from Newegg and was a Sapphire Radeon R9 290X. The part number on this card is 100361SR and it can be purchased for $559.99 from Newegg.




The Sapphire Radeon R9 290X video card is based on the reference design, so Sapphire just stuck a few stickers on the card and called it a day.


One observation that we had on the Sapphire Radeon R9 290X retail card is that the stickers weren’t cut out in the right shape! This meant that dust, dirt and other gross stuff would adhere to the overhanging sides of the exposed sticky back. There is no excuse for this as it is the only thing Sapphire physically modified on this $560 video card.




Here is a shot of the AMD Radeon R9 290X reference card and the Sapphire Radeon R9 290X press card that we’ll be comparing to one another.

Jump To:
  • Alec Joy

    My 290x has a completely different bios from both of these

    can you post the retail BIOS somewhere? Mine is outdated compared to both.

  • Relayer

    So, if we set the card to quiet mode (cap the fan @ ~40%) and pump heat into the case, the card throttles. This isn’t news. It’s how the card is designed to operate. Interesting that we’ve never seen anyone go through such measures checking to see if there might be performance issues like this with any other card before this one.

    When we had press samples of GTX-680’s boosting to 1300MHz, did anyone run out and buy retail samples to see if they performed the same? Try capping a Titan or 780 @ 40% fan speed and put it into a heated case and see what happens. Oh, and don’t be too consistent in your technique either. Let the case get to 40°- 45° just for good measure.

    I can’t believe you guys sold out like this and purposely tried to show this card in the worse light possible.

  • Hmmm… the BIOS probably just pushes the GPU harder – increasing short term benchmark results but degrading its longetivity and reliability in the long term.

  • Guest

    The problem here is WHY a special BIOS was used on the press cards. Even if AMD wanted to boost the benchmark reviews and make the card seem better, there’s no reason why they should install the same BIOS on retail cards to give the consumer the same performance as well.

    I have a strange hunch that the new BIOS stresses the GPU harder and pushes the card more towards its limits, which makes the card perform better but reduces its longetivity and reliability in the long term. Reviewers just do their reviews and don’t notice the long-term effects.

    A leaked BIOS will probably surface soon. Let’s see what it really does.

  • john

    ok this is for nvidia fans around: – Buy an exploding card! Buy a GTX780Ti Now! As long as they’re HOT! ROFL… And you bitch about 1% variance and high thermals… apparently none of the 2 can build a decent ref cards to save their lives… This whole GPU wars stats too look more and more ridiculous. I suppose we need to wait a bit for these cards to become really available and get non ref designs… Well at least AMD doesn’t blow up… it just hisses away like a jumbo jet ROFL….

  • HisDivineOrder

    AMD’s on a roll.

    TrueAudio – Soundstorm Reborn for the 2013!
    Mantle – Glide Reborn for 2013!
    Reference Coolers for R8/R9 series – Geforce FX5800 Reborn for 2013!
    R9 290/290X Variance – Quack Reborn for 2013!

    Sometimes the past needs to stay in the past, AMD. Especially those last three.

    • john

      so much ignorance in this post it’s not even worth the network bandwidth really…

      1) True Audio – Soundstorm… not by a longshot… True Audio is a minuscule DSP on the GPU die not on a chipset and can use the GPU grunt to compute real 3d sound unlike soundstorm which was basically just any audio processing unit…

      2) Mantle – Glide – Mantle is syntactically close to DirectX and represents a sort of upgrade to it by giving access to lower level functions by preserving all the rest… that’s about it. Glide was a interface built from ground up that the brilliant minds of 3DFX( now NVIDIA) have cooked up because there were NONE at the time. It took off for a while and died when the software provider that should have done it in the first place actually stepped in (Microsoft with directx). This always happens, software that fills a gap left by the company that should have done it in the first place usually end up like this. Mantle is set for the same fate it will die out once Microsoft includes that functionality in DirectX the advantage here is another, if they can pull multiple OS coverage this might actually be a unifying standard that would be : Open source, Supported on multiple OS’ and offer low level access and possibly controlled by an entity other then Microsoft… Mantle is what DirectX 11 should have been and what DX 12 may yet become…

      3) The blower on R9 is actually (much?) worse then that of the FX5800 series from what I remember. From what I’ve seen at 100% your computer might be advancing 1cm/minute, after a few hours of gaming you might be forced to get it back from the yard or nail it down to the floor.

      4) It seems that ALL NVIDIA fanboys, have reading comprehension problems: there is *NO* variance with the updated bios – read the review again! Then have a read on toms too maybe that clears it up for you (even though I find this review shorter and easier to understand – yes somehow you still don’t get it… I get it…)

  • GeRa


    Any news on the fan speed in rpm during the test with the two bios ?

  • tech90

    I don’t have this issue because I have watercooled my 290x. what idiot would use stock cooler I wonder. AMD didn’t cheat anything, improper cooling can cause this to any GPU,

    • Nathan Kirsch

      This performance difference between retail and press cards was only seen in Quiet mode with high ambient temperatures. How many gamers are buying a $550 to run in quiet mode anyway? Something is clearly going on, but how many people is this really impacting? That would be interesting to find out.

  • hellbindercda


    So, did AMD put a special BIOS on the press cards to influence the reviews or are the add-in board partners using an older version? We aren’t sure and will likely never know the answer to that. “””

    Um.. you are not sure? lol obviously they did. Its pretty much a modern day intentional benchmark cheating scandal.

    • john

      absolutely because they paid developers to compile benchmarks with fixed compilers….aaah damn… wrong company again… I keep getting mixed up with this…

      It’s not like the “special” bios is so special at all you can get it from your aib or even amd how is that a modern day “intentional benchmark cheating scandal” when you can get same performance with a simple software update???

      The problem is that AMD changed the bios and never sent a memo to the aibs to update the boards they had before launch that is the problem.

      You make really no sense at all! You can buy ANY card and if it’s not up to snuff you make a bios upgrade like you do any bios update and it will be up to par(As clearly shown by this review!!!) what’s the big deal about it??

      It’s an internal AMD PR scandal, I would really fire the guys that handled this or hire some because it almost looks like there is not PR at all!(No statements no press release, no answers to reviewers, NADA!) –

      That is the actual problem, YOU the client are not affected by this as long as you know how to update your gpu bios or have a geek friend around who can do it for you I think even retailers could help you with it… (The same is true for any MB’s, you sometimes can’t run certain GPU’s, CPU’s, memory or even gbit or sound correctly until you make a bios update – was there EVER a “scandal” of “cheating” about it??? NO!)

    • Nathan Kirsch

      There are many revisions and some of the AIB’s could have released with an older ROM or something like that. If AMD wanted to cheat, I doubt they’d do it in quiet mode in high heat situations. That makes no sense to me, but there is a difference there. I am curious if all retail brands are like this or if it is limited to just a couple partners.

    • Mechaboswerftw

      Relevant question here. Why in the world would they intentionally cripple the bios of their card over the review versions? They wouldn’t. That would only lead to bad PR. They obviously manufactured all the cards in advance and just did a bios update for the review cards which means all of the cards have the same or similar potential.

  • john

    @55f939740ca6294fc44ae3024f2803af:disqus @f4d4ddf07d3e3ed531adf4648b33bff6:disqus @basroil:disqus … really… you do realise your’re bitching about a software glitch not… not whatever you monkeys think it is…

    I would really like to see you formulate a false advertising claim on this… would be so Al Bundy’esque…

    YES! Yes! stay away from AMD… because AMD managed to get chips to disconnect from the main board while in use or flash a woden card as the real product… oh wait that was NVIDIA. And you’re bitching about AMD having a software glitch (that can be fixed in minutes with a bios update – offered as far as I know even by AIBs nowadays) … REALLY!!!?!??! geez…

    • hellbindercda

      no fan boy they cheated plain and simple. this was not a simple software glitch. this is intentionally misleading people knowingly and trying to get away with it.
      but you are right that NVidia has a much greater resume of cheating from the early and mid 2000’s.

      • john

        how is it NOT a software problem when with the press release bios copyed to the retail card the variance was 1%… wtf are you smoking??? Do you even know what a bios actually is? Or do you think it’s the chip on your motherboard that has bios written on it??

        Most probably the press release got a much newer bios then the aib’s got a month before when they got the chips and built the cards…

        Anything that can be fixed without RMAing the card IS a software fix… so tell me what was the hardware issue? And what exactly was misleading? Misleading were all the reviewers that benchmarked a quiet mode card against NVIDIA boost cards – that is misleading. Having your GPU & VRM fall of the card is hardware failure and claiming that the card is ment to work at those temps – that is misleading…

        AMD has made a mess both with the ref cooler blunder, with the quiet/uber/bios mode confusion & also with bf4 confision – they have a hugely bad PR, they need to fix that asap… but this is completely blown out of proportion! They need to fix this pr debacle and fast! And they definitely need to do something about the thermals & that turbojet of a fan they have… They can beat NVIDIA with a 30% smaller chip with slower memory but they can’t build a decent fan/cooling solution to save their lives… wtf is wrong with these guys! Pick up the phone and talk to powercollor or any other aib that has decent cooling solutions!!! NOW!!!!

        • zaxxon

          BIOS is NOT SOFTWARE you idiot. How can you not understand this ? How many people do you think actually upgrade the BIOS on their cards ? It’s not a driver.
          It’s NOT SOFTWARE. Get it through your thich head.

        • john

          if a bios is not a software what is it cause sure as hell it’s not hardware and the fact you can update it means it’s software so go bark a different tree you retard…

          I never said it’s ok I just said you nvidia fb’s are just blowing it out of proportion trying to mislead people into thinking these chips & cards are golden chips and cards which they obviously are not – get it??

    • NZSol

      Do you understand what a prop is?..I do think is funny the AMD PR crew on here trying to bring up benchmarking issues from 13 yrs ago!..LOL..While this maybe a great GPU, its a epic fail of a card and only the most die hard of fans could ever defend this peice of shite!

      • john

        I agree with you it’s a failed card because of a failed cooler which mostly any aib will thow out and use theyr own… calling it an epic fail exagerated… I would call it an epic GPU with a bad refference design… and even worse PR… funny how you think I’m AMD PR crew…

    • zaxxon

      It’s obviously impossible for you too understand the significance of this. The cards people willa actually buy are slower than the press samples. This was either done because of AMD’s incompetence or it’s intentional and therefore cheating.
      Even a fanboy such as yourself should be able to understand this. But obviously not…

      • john

        I never said the card is great, it’s a bad refference design and a bad cooler… but calling it cheating is blown waaaay out of proportion considering the actual cheating done by other companies in the industry that never inflamed any of you like ever… so chill, it’s a PR mess and AMD will probably deliver a driver update &| bios update soon to fix the powertune problems

  • Nvidia wins again


    • john

      yess it does… 5% performance over R9-290x in uber mode… and costs 100+$ more by using a rev B silicon slab being 30% bigger and less economical to fab and being essentially a golden GF110 chip otherwise sold for 2-4K$, using a 2ghz higher clocked memory but 1gb less of it, marketing a failed console and bundling old games from only one producer.

      wow just, wow…


  • Quake2

    Can you put an aftermarket cooler on it, and test power consumption? I think it would really benefit the card looking at your temperature/power graphs.

    • Nathan Kirsch

      Great idea and I have a water cooler coming in the near future and can do an article focused on that when it arrices!


    AMD scumbags….
    The Retail version of the cards are slower from the samples (and you know why AMD did it)
    Someone should sue AMD for this trickery

    • john

      I’d like you to formulate this claim you inept little monkey… Look, I’ll make it so simple even a retard like you gets it… Ready? … here it goes:

      Press card + press bios = Retail card + press bios. That means that any card you buy can get to the performance of press cards using the newer bios!

      Got it?

      Now go back and play with your legoes…

      • zaxxon

        You seem to have an answer for everything. Yet you only show your own ignorance every single time.

        The fact is that the retail cards are SLOWER than the press sample cards. Not everyone is going to flash a different firmware and most people will never even know about this.

        So yes, AMD either cheated or screwed up. Either way the customer loses. It’s obvious to anyone except fanboys like you.

        • john

          yes they are slower out of the box because of a mixup with bioses not because the card & | chip(read hardware) are “special” in any way, as clearly shown by this very review… so your argument is bs… and it’s not a firmware it’s a bios… most anybody can flash the bios nowadays there loads of softwares doing this and most retailers will do it for you if you have no clue… so you’re just trying to mislead others by claiming that the hardware is somehow different on the retail sku which clearly is NOT the case so… stfu…

        • ok, but what if there was some big problem that required everyone to update their bios, so the companies provide a 1 click exe to do it for you? (protip: it has happened before on specific brands & specific models)

          nobody loses if it’s just 1 click to fix, not to mention the drivers can just increase the fan… as can the user… speaking of retarded users, what kind of person wont notice unusual stuttering performance & look into it!?

          if anything all these upcoming custom cooler cards are going to perform real nice while the stock coolers will throttle, now some user is going to get a stock one cuz it’s cheaper in store & not get the numbers they saw in some custom review, is that anyone’s fault???

          it’s time customers stop being lazy asses & do some basic research in the options & manual of their product (i’d rather not have the quiet bios either)

  • johndoe

    hey AMD I got an idea, why not uncap those fan speed and run this card at 100%, obviously your fanboys won’t care about noise because – o wait, they use headphones! yea! that’s a good one! almost got me, haha.

    • john

      I do game with headphones… always i could’t give a fuck how loud the fan is as long as it stays quiet in office tasks so yeah. I’m one of the many guys that don’t want to disturb the neighbors or the others living in the house… as pc games are mostly single user/pc why do i need to bother everybody else with shooting and explosions or monster roars sounds???

    • Surfx

      Well if someone like me who doesn’t use casing with a huge window and has sound damper installed it would’t be a issue for me.
      Honestly I have nobody to show of too other then myself and I usually invite frens to my house .

      On the other hand I use water cooled system, so noise wouldn’t be a issue.

  • johndoe

    hey AMD I got an idea, why not just uncap those fan speed and run them all at 100%, obviously your fanboys won’t care about noise because -o wait, they use headphones. yea good excuse. almost got me there.

    • Daniel M. Nielsen

      The internet, a place where troll roll rampant. Accusing people of being so called “fanboys” while obviously one them selves. Nice job bro, nice job indeed bro.

      Anyhow thanks for the article, very interesting.

  • Captain America

    folks, and this is the reason why I stay away from AMD

    • hellbindercda

      and did you stay away from Nvidia when they were cheating at benchmarks for several years during the 2000’s? and cheating on image quality for performance gain? and had vacuum cleaner sounding fans?
      don’t be a mindless fan boy hypocrite.

      • zaxxon

        It’s obvious who’s being a fanboy here. Nvidia had VC fans for a single generation. In general AMD cards are much louder. In fact Nvidia did an excellent job with the 7 generation coolers. AMD just stuck the usual blower and called it a day.

        But I wouldn’t expect a fanboy like you to let a tiny detail like the truth get in his way.

        • Ry

          And this happens for a single generation and you go crazy on it. Why would a company use a boosted bios for press release benchmarking if the performance difference between those cards and retail cards could be identified so easily? Kinda shooting yourself in the foot yeah?
          You’d honestly think they would believe nobody would spot a difference? Get over it, there’s no hidden agenda here, because doing this is just dumb as fuck.

  • GeRa


    Sorry but why do not you log the tpm fan speed during the test since it’s a very important ?

    Is it the same with the two bios ?

    • Nathan Kirsch

      I have all the logs still and fan speeds were logged. I’ll go back and take a closer look at that. When I had over 200 logs saved it started to be too much info, so I reduced it all down to one game title and the info I thought was most relevant.

      • GeRa

        Thanks, i’m waiting for your feedback !

  • basroil

    Press samples running faster due to bios changes is quite a big issue if it’s true (i.e. no confounding factors), it is equivalent to false advertising in many states and countries (due to fraudulent means of describing their product).

    • Nathan Kirsch

      Or it could just be that a handful of companies released cards with an earlier build of the BIOS. Regardless, if this is just something that can be resolved with a BIOS update, it would be great news for everyone.

    • guess what, the specs say up to X mhz, no where does anything state you get Y performance, which has always been the case with 3rd party benchmarks as they use different computer parts than your own

      nothing is falsely advertised, you can control the fan or use another cooler yourself….

  • john

    Ty LR!! Finally a well informed and detailed review that cleanly explains the variance clearly… for any idiot… like sa’s ashraf aesra to understand and hopefully stop his missinformed articles…

    I find it a bit cleaner then the toms article… but you also had much more time to fiddle with the card 🙂

    Now i want a non ref cooler and eventually rev b chips, it’s been more then a month since launch there should be a rev b chip already sitting in amd labs by now…

    • Nathan Kirsch

      Thanks for the positive feedback! I too want to see what the AIB’s are cooking up. I have one of these on the way and will try to water cool a 290X to see what happens to clocks then –

      • john

        Can’t wait for it! but no memory cooling at all? Well I get it that it runs at a mere 5ghz but still…

        Non ref cooling … care to say where the package is from? – I promise I won’t tell anybody! I think everybody on this comm list wants to know! (even the NVIDIA FBs)

  • where are the voltage numbers? fan rpm readings? that’s what a comparison is… you cant just look at the performance results as a good custom cooler will beat every single reference, even the magical press one, making the real performance of the card better than its original design

    it would be strange for the 2 bioses to have identical fan speeds, identical voltage numbers, identical mhz (i mean i guess if there’s some secret stuff in them, but first we have to deal with the 2 variables i mentioned, on top of that, the end user can control them to gain performance or reduce noise manually)

  • Strider

    That’s quite a significant discovery, and while I may be waiting on non-reference cards before I retire my 7970, it’s still good to know that the source of this difference is the BIOS.

    With this discovery, I wonder if we will see the press sample BIOS version hitting the web in the near future. I am sure some users would like to get their hands on it to try, while they wait for the next update AMD is referring to gets released.

    When the BIOS update hits, I suspect you will do a similar comparison? Thanks for the article, it was a great read, and very informative.

    • me

      If you’re already running a 7970, I would suggest to skip this gen altogether. Too small improvement to waste money on.

      • Steve Smith

        what i think, press cards AMD expects to be run in open air test benches which card will have better cooling cause colder air then what would be found inside a computer case. So on press cards had more aggressive bios cause that fact, problem is it was found out.

        • Strider

          To be honest, they should have just shipped with the press BIOS. I mean, who buys a card like this to drop it in a case with crap airflow? I know some may do it, but most wont. There is not much difference between open air and a properly ventilated high-end PC case, so it can handle it.

          As far as cherry picking, you know darn good and well they all do it, with every new GPU line. So to me, this is not a shocker.

          The BIOS tweaks, like I said, they should either have just included it, or they should have made it available as a “high performance” version you can download and apply yourself.

      • Strider

        To be honest, the 7970 may be a great card, and it can easily handle any games out there today at 1080. Especially considering the fact I have mine running at 1.1GHz core and 1575MHz VRAM.

        Even then, the 290X is still more powerful, and I am a sucker for the latest and greatest when it comes to my GPU. Since I prefer AMD the 290X is going to be my new card.

        I know the 7970 fully supports Mantle, but I also want True Audio, I really want to experience that. I am an FPS player at heart and in-game audio is something I tend to be a stickler on. This sort of positional audio is something I have been interested in for a long time.

        So while I don’t “need” to upgrade, I want too. Of course, I am also a hardcore geek, so that has a lot to do with it as well. lol

        Now I just need those non-reference cards … if they make me wait too long I might just say to hell with it and buy a reference card and drop a water block on it. =/

    • Chris Wells

      Why not just buy a 280x and crossfire?

      • Strider

        I am considering crossfire, I have not had the need to do that since the old 4870 days since I have only been running a single 1080 display for gaming. However that’s going to be changing, I will be moving to 3x 1080 displays in Eyefinity.

        So I am actually considering running dual 290’s, instead of a single 290X. The reason I don’t just get another 7970 or a 280X and run them in Crossfire is the 290’s are faster, better at higher resolutions, and they have True Audio support. All things I am interested in.

        However I will only go down that path if I can get 2x non-reference 290’s I want, or I may have to spend more than I had intended and go with liquid cooling. Not sure on that yet. I have time, don’t plan on doing any of this until after the holidays.

  • Borat

    Does this apply to r9 290 non X eddition too ?

    • Nathan Kirsch

      I’m not sure, but it is likely. I haven’t looked into that sku nearly as much as the 290X.

      • Steve Smith

        290 cards make so much heat that they slow down after 10-20min when heat soaks in to heat sink. Even after you close the game takes almost 10+min for it to cool down. I don’t know about everyone else but when i start playing a game i don’t take for 20min then done, i will game for 1-2hour on avg min at a time.

        • john

          valid points you got there… AMD cooling solution is garbage… The chip is great, 290 beating sometimes the 290x just goes to show how much more potential this chip has when paired with a decent cooling solution. Sad to see the 290x dragged through all this slum just because they couldn’t be arsed to do a decent cooler that doesn’t blasts off in order to keep the card cool… 😐