1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Graphics R9 290X meet Eyefinity, Eyefinity 290X

Discussion in 'Hardware' started by xaser04, 27 May 2014.

  1. xaser04

    xaser04 Ba Ba Ba BANANA!

    Joined:
    27 Jun 2008
    Posts:
    2,551
    Likes Received:
    468
    Preface - I dun Derped!

    Hi all,

    This review has been a while coming and ultimately started with a completely different premise; initially the review was going to focus on how a 290X copes with Eyefinity when coupled with a budget processor, in this case a Athlon X4 (Richland) 760K. The reasoning behind this being that; at triple screen resolutions you are generally GPU bound rather than CPU bound. My conclusion was, I thought, going to summarise that you can get away with a budget cpu as long as you have enough GPU grunt to see you through. Boy how wrong could I have been!

    Unfortunately this whole review fell flat on its face the moment I started to bench Crysis 3. Without going into too much detail (I will be here all day otherwise), the game was completely unplayable. The CPU was pegged at 99% on all four cores yet the GPU utilisation varied wildly from 25-100% constantly. The end result was a 15-30fps stutter fest, so bad in fact that a 290X overclocked ended up performing slower than a HD7850.... :sigh:

    After much deliberation (and copious amounts of tea, mmmm I need a cuppa now actually - I ******* love tea! :D), I decided an upgrade was on the cards and thus finally I am able to put together a proper 290X review for your enjoyment.

    Welcome to the Jungle 290X - (VERY) Loud, Proud and ******* fast

    I am not going to go into much detail about the GPU itself as there are plenty of professional reviews that have done just that. I just want to give my opinion on what is a very capable card.

    Ah, what can be said about a reference 290X / 290 that hasn't already been said. Well let's get one, well two, things out of the way right away; yes it is intolerably loud above 60% fan speed, and yes it runs hot. Keep it at 60% and under (<50% ideally) and under though and the noise level is acceptable although in no way can it be considered silent. I have made a switch to gaming with noise cancelling headphones so the fan noise doesn't bother me, although it is still there. Overall I would say the cooler is adequate but could be a lot better.

    The rest of the card though, holy balls! Without getting bogged down in the performance numbers too soon I will simply say this card is ruddy brutal. The overall design of the Hawaii GPU is rather nice, lots of Shader cores coupled with a massive 512bit memory interface and more importantly a 64 ROP back-end leads to a card that can really push the pixels. This is especially important once you start ramping up the resolution. Arguably I would say this GPU is somewhat more balanced than the full fat GK110, although the overall performance is undeniably similar (well at 1080p anyway).

    The card I picked up is a reference XFX R9 290 which quite frankly is about as boring and generic as they come. Note that I say 290 there instead of 290X, this particular card is a full fat, no compromise unlocked 290. you may remember that a few cards around the 290 launch were capable of unlocking. Why this happened we can only speculate (I still think it was to meet demand for the 290). I am certainly not going to complain mind! :D

    A few pictures:

    [​IMG]

    [​IMG]
    (HD7850 vs R9 290X)

    [​IMG]

    Big but not that big (circa 28CM in length). Fits snuggly into a Corsair 250D with no room to spare :p.

    I need to get a few pictures of the complete system with the 290X in action.

    The bundle was the usual bland rubbish of a driver disc, power adaptors and installation guide. Nothing more to really add here.

    Overclocking, Under volting and all sorts of things ending in 'ing'

    Given the fact the 290X runs hot and I am running a mITX set-up I haven't really pushed the 290X at all when it comes to overclocking. Honestly given the performance numbers at stock there isn't much need too. With a little bit of tweaking I managed 1075 / 5500 at stock volts without any increase in temperature (at the same fan speed). I bought the card used and know it can do 1150/1600, but I can't recall at what voltage (or fan speed - seriously 100% makes the FX5800 sound like it is passively cooled).

    More interesting on this card was the ability to run at -37mv at stock clocks. This allows a slightly lower fan speed for the same operating temperature. Based on my testing this is 100% stable. I haven't tried pushing the volts lower but so far so good.
    In all of the benchmarks you see below the overclocked results are at 1075/5500 unless stated otherwise. Stock results are 1000/5000 which is bone stock 290X clocks.

    The Benchmarks

    I would like to start this section by mentioning that I am not a professional reviewer. I do this for fun and hopefully as a by-product it may helpful to some people. All of the components you see are bought personally and thus I am unable to compare directly like for like (as I have to sell one thing to buy another). In general the base systems are comparable enough that the underlying GPU performance is as close to like for like as I can get. I also have no bias when it comes to components. I simply buy what I would like to try out. If it performs poorly I will say it does and vice versa.

    Test Setup & Benchmarks

    The following is the test setup & card comparisons you will see in the benchmarks below:

    [​IMG]

    In all cases the latest available drivers were used.

    Unless otherwise stated the GPU clock speeds are as shown above.

    The benchmarks are as follows:

    BF3
    Crysis 3
    Sleeping Dogs
    Thief (290X only)
    3DMark11
    3DMark (13)
    1080p of above games


    Battlefield 3

    Mixed settings - 5 minute FRAPS run through of Operation Swordbreaker. Starting the recording the moment you leave the ATV.

    BF3 - Ultra with 4xMSAA

    [​IMG]

    [​IMG]

    Right out of the gate we can see what the 290X is capable of, some 27% and 50% quicker than a GTX780 and HD7970Ghz respectively. With a mild overclock it simply extends its lead even further. In this specific test the GTX780 OC (running at 1189/6800) represents roughly what a stock GTX780Ti would be capable of at the same settings. The 290X overclocked just owns the show here. Heavily clocked it would probably be pushing 60FPS.


    BF3 - Ultra No MSAA

    [​IMG]

    [​IMG]

    Another excellent showing from the 290X. This time the performance delta to the GTX780 is a little lower at ~18%. Overclocked vs overclocked the 780 closes the gap although the 78- itself is overclocked further in % terms than the 290X. My initial thought is that the extra ROPS on the 290X are providing the extra grunt when MSAA is applied - thus allowing it to extend its lead. This is also backed up by the much stronger minimums enjoyed when overclocked.
    Of interest here is that the 290X, like the GTX780 OC, is quicker than a pair of GTX670's in SLI. The 290X overclocked is actually quicker than the 670's overclocked as well (not shown). If I overclocked the 290X to say 1150/6000 I imagine it would breeze through the 80FPS mark quite easily here.


    BF3 - Custom Settings
    (all high bar shadows on medium)

    [​IMG]

    [​IMG]

    Nearly 100FPS at what amounts to high settings in game.... :jawdrop: Again we see a stock R9 290X matching the overclocked GTX780 (780Ti). The performance from the top cards here is truly stunning. Way back when I can remember playing BF3 in Eyefinity on a single HD6970. It managed around 35FPS at identical settings. The performance of the top end single cards really has come some way at this resolution.

    Overall the BF3 experience was fantastic with the 290X. Not a hint of displayport tearing (a major problem for me with the HD7970Ghz) and no signs of stutter or uneven frames (seat of the pants style anyway). The 290X out performs the GTX780 across all settings although the actual gameplay experience was very similar. The one exception being ultra settings with the addition of MSAA. The reference clocked (993 boost) GTX780 struggled a little here, whilst the 290X seemingly didn't care.


    Crysis 3

    "High" system spec with Very high textures, SMAAx1 - 5 minute FRAPS run through of Welcome to the Jungle, starting at the point you exit into the Dome.

    [​IMG]

    [​IMG]

    Crysis 3 paints a similar picture to BF3 with the stock 290X outpacing the GTX780 by a noticable margin. The overclocked GTX780 claws back the difference against the overclocked R9 290X although this is more down to the much larger OC I had on the 780 vs the 290X.

    Again as with BF3 the overall experience was similar with the 290X taken a small win due to its more consistent overall framerate. Overclocked the 780 is putting up one hell of a fight though!


    Sleeping Dogs

    Built in benchmark, Max settings split by AA type

    Normal AA
    [​IMG]

    High AA
    [​IMG]

    Extreme AA
    [​IMG]

    Sleeping Dogs appears to favour AMD hardware quite noticeably so it isn't surprising to see the 290X stomp all over the GTX780 here. Extreme AA is still just as pointless and unplayable as ever. The only thing of interest with this benchmark is the fact I needed to add a little bit of core voltage to get the 290X stable at the mild overclock I am running. Crysis 3 was absolutely fine yet this engine for whatever reason was not. Playability wise at "gameplay" (normal AA) settings the 780 and 290X are basically indistinguishable, whilst high AA starts to tip things in favour of the 290X.


    Thief

    Built in benchmark - Max settings with no SSAA

    [​IMG]

    Note - there are no comparison cards included here as I only have the 290X to test.
    I haven't seen any results for this game running in Eyefinity therefore I can only give my own impressions of running the benchmark. That said the only thing I can really say is Mantle seemed a little smoother than DX11 across multiple runs. I don't know why the Stock Mantle run had much better minimums than all other runs but there is a momentary pause at the beginning which happened on every run bar that one. The average is not out of whack thus I am putting it down to benchmark gremlins. Mantle is showing small gains here even in a heavily GPU limited scenario. This bodes well for future releases.

    At this point I should note that the little 760K performed admirably here with Mantle enabled. It basically matched the DX11 results you see above.

    I have less interest in playing this game than I do listening to a Miley Cyrus "song" (currently listening to Fear Factory to give you some context), thus I am not too fussed if these numbers don't really agree with other results.


    Test Drive Unlimited 2

    This is where the results for the TDU2 benchmarks should have been. As you might be able to tell there isn't anything here bar a picture of a confused minion.

    [​IMG]

    The game ******* refused to run. :wallbash: I have reinstalled it twice and still nothing. It may be a driver problem but it ran "ok" before I changed out the CPU (same 14.4 drivers used both times). Overall I would like to think that the 290X would perform in line with the GTX780 here.


    1080p Benchmarks

    These are just a bit of fun as I don't game like this anymore.

    [​IMG]
    Performance across all of these games is much closer between the GTX780 and 290X at 1080p than it is at 5888x1080. This is pretty much in line with any mainstream review you can find on the internet. Depending on the game each card will win and lose pretty much equally. Generally speaking the 290X is quicker than the GTX780 at stock whilst both are equally matched overclocked (albeit with a great % overclock on the 780).
    There is one outlier though and that is Crysis 3 with 8xMSAA. I don't really know what happened there. The SMAAx4 results are in line with expectations yet the 8xMSAA results - where I would expect the 290X to pull ahead if anything - are heavily in favour of the GTX780. Ultimately the framerates are a little low for my liking here but the difference is there.


    Synthetics

    I have no real interest in synthetic benchmarks but I have run these to give some comparison for anyone interested.

    3DMark11
    [​IMG]
    There are no overclocked numbers for CPU lines for the 290X as my 4770S is locked. It is running at 3.9Ghz on all four cores though. Performance is in line with expectations.


    3DMark (13)

    [​IMG]
    I have only listed the 290X overclocked results here as the card was making such a racket that I was afraid it was going to gain sentience and ask WTF it had done wrong to suffer such pain. The score seems in line with what I have seen on-line thus everything seems in order.


    Conclusion & summing up of things that can be summed

    Brutally quick, horrendously noisy and great fun to play around with are the three things I would say which best describe the 290X (and by extension the the 290). The Eyefinity performance is astounding for a single card. It bests the GTX780 in virtually everything I have thrown at it and based on my conservative overclock has plenty left in the tank to extend its lead even further.
    This shouldn't be read as taking anything away from the GTX780 (GK110)though as that card still performs admirably, especially clocked up to basically 780Ti levels of performance. The 780 (GK110) runs a lot cooler, a **** ton quieter and will consume less power. The noise and temperature can of course be sorted by a custom cooled 290X/290 but as with anything you need to look at all of the variables.
    Finally to conclude I would say if you want to run triple screens and only have the budget/space for a single card (or only want to run a single card) then you can't really go wrong with a 290/290X. The 780 is a perfectly viable option but will cost a bit more initially.

    Positives
    + Quickest card I have ever tested in four years of Triple screen gaming
    + Cheaper than its direct competition
    + Custom coolers remove the noise / heat issues
    + It is red

    Negatives
    - Reference cooler is noisy
    - Reference card will run hot if you want to keep the noise levels down
    - It isn't blue, I prefer blue!

    Now I am off to boil the kettle for another cuppa. Did I mention I like tea?! :thumb:

    Oh and if I have to listen to the same non skip-able opening section of Operation Swordbreaker again I will go quite quite mad.
     

Share This Page