Gadgetory

we're back to test the 4 gigabyte vs. 8 gigabyte arguments for the RX 480 and I've got the box on the table because we have our two cards in crossfire testing this instant but today we're looking at 4 gigabytes vs. 8 gigabyte specifically we did this previously with the gtx 960 nvidia card testing 2 gigabytes versus 4 gigabytes the story has changed graphics cards now have more vram but games are also demanding more so that's what we're looking at before getting to the benchmark this content is brought to you by Origin PC and a new origin Kronos which is equipped with the RX 480 and is customizable and upgradable on their site including custom paint so getting into the coverage here first of all we used our reviewer copy of the rx 480 which performs identically to this I've validated it the one instance it is different and it is somewhat important one for debugging is we are able to flash our firmware to create a 4 gigabyte rx 480 so that's what I did go into this benchmark knowing that the one note that is relevant to that end is that the memory on 4 gigabyte cards that you can buy is 7 gigabits per second it is 8 gigabits per second on the 8 gigabyte card but when we flashed the firmware I checked and this is good on AMD for doing this the memory clock was actually slightly reduced and was comparable to what you can buy on the real market so those were variables that were taken into account and things look good for the comparison we are not doing thermals power or noise levels it will basically be the same as what you see in our our X 480 review check that if you want more what we're looking at is all games and a few things to expect here with 4 gigabyte vs. 8 gigabyte or really any disparity in the RAM for video cards the aspect of games that will stress that the most will be anti-aliasing which really heavily eats into vram potentially that's because taking multi taps or multiple samples per pixel to determine what color that pixel should be so if you have 8x MSAA also known as 8 tap its sampling each pixel 8 times that is a lot of additional sampling and that does impact B Ram another aspect would be the post effects side of things post FX and temporal effects will more heavily impact the processing side of the GPU than the vram so if we have a game that is more heavy on post effects and computational effects then we might actually not see that much of a difference between four gigabytes and eight gigabytes but let's run through the charts and then talk about why they look the way they do let's start with one of the games where we see a big difference in performance between four and eight gigabytes this is Assassin's Creed syndicate which I'm reviving to our bench specifically for the test I'm not even going to show the other video cards we're just looking at the RX 480 so that's all this video is about we've configured settings to very high with a four tab MSA and I've seen fifty-three FPS average on the eight gigabytes rx 480 which is a 14.5 percent increase over the 45 point 3 FPS average of the 4 gigabyte card the difference is actually very noticeable and it's not just because of the averages it's because of the low performance so as always with low-level GPU performance comparison it is important to look at 1% and 0.1% lo metrics or low frame times to get the full picture of things and the 8 gigabyte card although it runs at 53 FPS average it puts out v 32.3 FPS 0.1% lows and the 4 gigabyte card falls to just twenty five point seven fps 0.1% lows or a twenty five point six eight percent increase if you move to eight gigabytes so that is a twenty five point six eight percent change from four gigabytes to eight gigabytes on the 0.1% lows and this is noticeable in play where you will see more frame tearing on the four gigabyte card it gets choppier and that is actually something you feel as a player the fluidity just isn't quite as good as on eight gigabytes 1% lows are also separated by a percent difference of 25 so 25 percent difference this is also something we saw when testing the gtx 960 s on two gigabytes and four gigabytes where the cards were most heavily impacted on games like Assassin's Creed shadow of mordor will request all the vram you have available but requesting it and seeing quote-unquote utilization of GP Z doesn't mean that's actively engaging all that requested vram and tasking or instructing it we tested the game at 1080p 1440p and 4k with Ultra settings at 1080p we're seeing results of 78.3 FPS average 41 fps 1% low and 36 FPS 0.1% lows on the rx 488 gigabyte card drop into 4 gigabytes those numbers barely change there's a 0.3 fps difference and the bench pretty much perfectly executed every time we repeat the benchmark and we do repeat testing iterations to be sure so this is a measurable and real difference and it's accurate but it's still just 0.3 FPS so it is imperceptible as a user basically the same and there's no real advantage to an extra 4 gigabytes in this game regardless of resolution Mirror's Edge catalyst is processing and post FX intensive but it also used a lot of texture filtration techniques and using those techniques does impact B Ram pretty heavily for our upcoming crossfire benchmark this will be an interesting test case to see how post-processing impacts the crossfire setup but for now we're still looking at vram and Mirror's Edge catalyst has some unique performance results at 1080p Ultra we see the average FPS between 8 gigabytes and 4 gigabytes is about the same 70 4.3 FPS versus 72 0.3 FPS it's a two point seven three percent difference and even the minimums are good with mostly identical results between resolutions at 1440p Ultra we're still seeing gaps of 1 FPS maximally the difference between averages is 0.64 percent so negligible but then we look at 1080p hyper and this increases texture quality mesh quality other filtration effects and vram syncing settings and 1080 hyper produces dismal stuttering after a few minutes of play this isn't something you'll see reflected in a short benchmark period but if you're playing the game properly for a bit the vram begins saturating with resources and frame rates can drop hard just from this chart we see our average it looks like 40 FPS versus 53 FPS or a 27.96 percent difference but the 0.1% lows are even worse down to 18 FPS from 31.3 that's a fifty three point nine six percent difference these lows are noticeable as severe in frame rate output because the frame times have become inconsistent but there's more to it than that here's a results table different from our chart we ran multiple extra passes on Mirror's Edge specifically because of this issue and found that the first few sets of data were poor in performance but not completely unplayable after playing for some time and every single time thereafter we'd see drops to 12 fps 0.1% lows and in the 20s for the average frame rate making some really intense stuttering and slowing down the game time overall so by dropping to four gigabytes we've gone from a relatively playable fifty-three FPS average to an unplayable and fluctuating range of twenty six to forty seven FPS average GTA 5 was completely retested on the our X 480 cards for our bench after speaking with Andy's Scott Watson we received an unreleased in development driver that should resolve the previously mentioned GTA 5 stuttering issues from launch day the driver update is sixteen point seven point one and this is the only game in the test that runs with this driver set the rest are still on 16.6 dot two or the press drivers for release day we see an average FPS of 85.3 on the eight gigabyte card at 1080p with the four gigabyte card at 83 FPS that's an imperceptible but measurable difference 1% lows are also close only three point three five percent difference and 0.1% lows are fifty four point three FPS verses 51 FPS on the eight gigabyte and four gigabyte options respectively another imperceptible but measurable difference 4k in 1440p shows similar performance results leaving GTA 5 a title where it doesn't seem to matter too much whether you get a 4 gigabyte or 8 gigabyte are X 480 that said we're only testing for single GPUs here and crossfire may change the story black ops 3 is another title with mixed results like Mirror's Edge black ops has also been an optimized title for us in the past on these benches and one which has generally pushed Andy a bit further up the ranks than some other games on the bench at 1080p we see at the 8 gigabyte card running at 130 2.3 FPS average 105 FPS 1% lows and 93 FPS 0.1% low is all very tightly timed and suited for 120 Hertz gaming if you want that and you even gets 144 Hertz with some tweets moving to four gigabytes our average frame rate drops by four point one seven percent to 127 fps and 0.1% lows dropped by thirteen point nine eight percent 280 FPS increasing it to 1440p we see more of an impact to those 0.1% lows so previously the difference was actually pretty large but not something that you really perceive at 1080p unless you're really hardcore about 18144 Hertz and at 1440p the eight gigabyte RX 480 is now at 80 3.3 FPS average 60 7.3 1% low 60 1.3 0.1% lows which is incredibly tightly timed and the average is only slightly faster than the 4 gigabyte card at 80 fps but the 0.1% lows on the 4 gigabyte card are twenty eight point seven percent lower than eight gigabytes that's almost 30% lower percent change from eight to four gigabytes and that means occasional stutters become visible in gameplay and that is actually something you can feel as a player 4k produces similar results at 41 FPS average for 8 gigabytes 34 for 1% low is 32 for 0.1% lows but on the 4 gigabyte card that changes to a similar average and 1% low number but a decrease of 60% on the 0.1% low metrics and now of course you might say well this isn't really what I want to play on for black ops anyway 4k at 40 FPS but if you're doing setting tweaks then it actually might become relevant we tested more games than this like the division ashes of singularity and others and you'll find those results in our article linked below as we found previously for the 2 gigabyte versus 4 gigabyte gtx 960 testing the differences are present you can actually see them in almost every game there is a visible difference in numbers but not necessarily in gameplay to the user a 0.3 FPS difference is not something you care about neither is the difference we saw in a lot of the games like shadow of Mordor maybe 1 to 3 FPS stuff like that is really frankly irrelevant as a user but there are games where it is actually a massive difference we saw the 60% gap and 0.1% lows for black ops at 4k not really a super common use case but if you look at more common use cases we can kind of push that one aside look at something like 1440p in black ops and we were still seeing almost 30% reduction in those numbers and that remained true for other games as well like Mirror's Edge catalyst and Assassin's Creed syndicate another game that has proven to be very vram dependent in previous tests so it depends on what you're doing there are other considerations that matter as well for example putting two four gigabyte rx4 80s and crossfire versus two eight gigabyte ones could have pretty different results than what we're seeing here on large anyway we'll see a lot of the same kind of concepts but it may be exaggerated if you do crossfire cards and that's something we're working on testing as soon as possible but the reason you would see that difference would mostly be because you're increasing sort of the the computational potential and throughput or maximum potential throughput of the cards by doubling them of course not a doubling exactly of power but we're keeping still lower vram and we might be exceeding what the the cards are capable of in terms of VRAM capacity so we'll look at that as it comes up as for whether or not eight gigabytes is worth it for the our X 480 that really depends it's an extra $40 on reference from four to eight gigabytes 200 to 240 dollars and if you're playing games like black ops or Mirror's Edge with high settings meet when I say that I mean hyper settings or just running the higher texture and filtration effects or you're playing games like assassin's creed that series has historically had this sort of performance output on these cards it is definitely better to get the eight gigabyte card if you are playing the other games and I'll read the off the list here we've got ashes of singularity Talos principle the division Metro last light shadow of Mordor to some extent GTA 5 those games see very little impact from the vram change the impact might be a few percent at most but it's always a couple FPS maximally and that's not super noticeable so those types of games you can kind of ignore it you should definitely buy eight gigabytes for games that are text or heavy anti-aliasing heavy if or if that's just something you really like or texture filtration but otherwise four gigabytes for the RX where D is an acceptable way to save $40 if you're kind of in that lower price category but crossfire maybe an instance where we need to look more closely at performance and see if that changes and we will be doing that soon one thing for certain these games that produce the disparity and vram performance are consistent in their disparity so we've seen that a few times now for multiple cards and it's also time to move away from two gigabyte cards I know there's not a lot of them left out there but the the use case is for two gigabyte cards have closed especially with the price reduction on four gigabytes on the whole I would recommend two gigabytes for a really cheap home theater PC something that's not meant to really do a lot of gaming maybe if it is it's more casual or older games but generally definitely kind of start pushing towards that four gigabyte value because games are drawing more VRAM there's a lot of reasons for that that we can discuss and future ask G ends or something but as always thank you for watching that pretty much recaps the difference is totally up to you whether you want to buy four or eight but hopefully that data helps you make that decision patreon link the postal video if you want to help us out by supporting the channel and paying for all this because I did buy this out of pocket thanks for watching and subscribing I'll see you all next time you let's run through the charts and then talk about why they look the way they do with benchmarking music okay so that was enough time to look at all 20 charts