Gadgetory


All Cool Mind-blowing Gadgets You Love in One Place

RX 470 CrossFire Benchmark vs. CF 480s, GTX 1070

2016-08-06
for no particular reason we decided to benchmark crossfire Rx for 70s for the full rx for 70 review using this card you can check our separate video on that on the channel what we got in the mail after the sapphire rx for 17 is a 4 gigabyte card was this MSI rx for 70 I think it's another gaming X model yes and this one is an 8 gigabyte card it's clocked slightly differently it's got more vram so these aren't a perfect match in terms of what you might set up for crossfire but we were able to do it anyway and I'll talk about how and why we set up the way we did so this test is mostly going to be focusing on the comparison between a single 470 and 2 4 70s we're not going to be going through crazy look at every single card on the bench type of thing they will be on there but check the full review for more information first off the card setup I ran the Sapphire card in the top slot so that for games that are not using explicit multi-gpu which is basically all games at this point except for dashes for those games this is the card that we'll be using used for all the vram so it's a 4 gigabyte limit we're basically meeting the least common denominator here so this card is only going to be as good as the Sapphire card despite the fact that the MSI one does have more vram we're not going to be tapping into it so that is one note and consideration to make they perform effectively the same as two of these because we're limited to the worst card but I do want to point that out they're not they're not actually the same card so that's something to note for dx12 things are different to ashes is a different benchmark we'll talk about that later first off let's look at power draw for power draw we're looking at the parent power draw in volt amps as measured at the wall this means that we're looking at power draw from the system level so we're comparing Delta's between configurations not between actual cards the system power draw with a single rx 474 gigabyte card from sapphire the Platinum Edition was measured at about 220 2 volt amps under full load with idle at about 80 volt amps the rx 470 crossfire setup was consuming at 404 volt amps load or eighty-three idol so definitely a significantly higher power draw nearly actually 2x the load power draw of a single card and this is in step with most multigp setups where you see the most valid arguments for why power draw matter for FPS benchmarking as always a link in the description below if you want to check this article with potentially extra charts but definitely extra test methodology explanation you can read about the drivers we use the system we use all that stuff but otherwise the system used is on the screen now and we use that for very specific reasons you can read in the article if you're curious let's look at Metro last light first Metro last light at 1080p with very high quality and high tessellation post the performance output of 109 FPS average on the crossfire RS 470 cards with a 70 point 7 fps 1% low and 64 FPS 0.1% low for comparison the RX 470 single card performs at 70 point 3 FPS average so we're about 30 to 40 lower and that's a difference of forty two point six percent that puts the RX 470 crossfire cards at the level of performance achieved by a gtx 980ti not too far from a 1070 and really close to the RX 4 80s in crossfire interestingly RX 4 70s in crossfire somewhat invalidate the argument for RX 4 80s in crossfire with this particular game and those four eight is word by the way tested with 8 gigabyte models and let's just look at more results before reaching a definitive conclusion though at 1440p for metro we're seeing an output of 81 FPS average for the crossfire for 70s with a 56.7 fps 1% low and forty six point seven FPS 0.1% low for a comparison the single RX 470 performs at 47 FPS there is a positive scaling and that's not always the case the crossfire for 80s are at 89 FPS average a difference of 9 percent against the RX 4 70s in crossfire and a single gtx 1070 is still outperforming the crossfire RX 4 70s that was a crossfire config does markedly outpace a gtx 1060 our previous Mirror's Edge catalyst multi-gpu test showed negative scaling for the crossfire rs4 80s and we're still seeing negative scaling with the crossfire 4 70s though not as fiercely as the original crossfire testing showed still it's better to disable crossfire for Mirror's Edge catalyst the crossfire Rx for 70s are posting 65 FPS average with 44 fps 1% lows and about 31 fps 0.1% lows for comparison the single rx 470 performed at 65 point 3 FPS average 44.7 FBS 1% low and 31 fps 0.1% lows so the configurations are effectively identical and this has improved at least over the original crossfire 480 benchmark or we saw really negative scaling but it's still not positive so you'd still be best to say when crossfire for optimal performance scallion is mostly the same or the lack thereof anyway at 1440p with the duel cards output in 41 FPS average 28 200 and a stuttery basically unplayable 19 fps 0.1 drusen below the single card pushes 42 29 and 19.7 so we're again effectively the same GTA 5 was a nightmare with crossfire with the 16 point 8.1 drivers these resolve the single our X 480 stuttering issues over 16 62 and we saw it still intense micro stutter that nukes performance with the crossfire 470 set up the same issue is not present on a single 470 when tested including the single Platinum Edition that we have here in our crossfire config but introducing crossfire does basically break the output and the issue by the way is what you've been looking at on the screen if you're curious and obviously paints a picture of an unplayable game this is also exactly why we test for 1% in 0.1% lows because they're an accurate depiction of those swings here's the raw data for a few test passes just the crossfire cards and the RX 470 card on the screen now you can see the variance with the crossfire configuration is tremendous and the lows are dipping really hard when they do dip and just to reiterate the point we sometimes have folks comment and challenge the use of the word unplayable when referring to high average frame rates for example this output even at a 93 FPS average is completely unplayable but you've got to look beyond the average because if look Jessa the average it looks actually pretty good looking at the low metrics we can understand why and it's the stutter and micro stutter in this case that we can show in the video capture regardless the normal chart shows that 1080p GTA 5 operates at about 97 FPS average with 1% lows of 39 fps and 0.1% lows of 24 the single rx 470 pushes 77 average but has lows of 54 and 48 genuinely playable with the single card here's the interesting thing with its high average GTA 5 would actually show some reasonable scaling if and the order to work out the drivers to properly support crossfire which will probably happen in the next update but performance is there it's just the frame throughput is to stutter II we were not able to get dooms at Vulcan update running on the crossfire cards this time the screen just went black has never really been well supported anyway but the crossfire cards did technically run with OpenGL I say technically because performance is mostly identical to a single card with very slight negative scaling for the low values though not in a substantial way or generally noticeable fashion right about 75 to 76 FPS average for both the single and dual GPU setups the same is true for 1440p by the way with scaling at near zero and a frame rate of 52 51 FPS average the lows are pretty similar but favoring the single GPU the division also has negative scaling with crossfire and we're looking at about 64 FPS average for the CF RX 4 70s with lows in the 30s while the single card outputs 71 FPS average and a 48 fps 0.1% lows and that's the biggest improvement here and that's about the same as what we saw with the rx for 80s and crossfire for positive scaling we have to look at games like Metro last light and black ops and GTA 5 though GTA did have that micro stutter is so you really can't count it right now you could also look at shadow of mordor and some of the other games but we didn't retest that title here and so we reach the same conclusion as previously with multi-gpu setups just generally they're not that reliable for a wide berth of games if you're playing a very specific game that does scale well one game that you're playing regularly then it's worth considering if it's a good value proposition versus a more powerful single card but even then you're still looking at the power argument the noise arguments not so bad depending on which cards you buy but the power argument certainly is and we've shown that it's nearly a 2x increase 1.8 X if you're just doing it linear anyway but 1.8 X over the original power draw so 220 to bold amps apparent power draw for a single card configuration versus the 400 plus load with the two cards that's certainly an argument worth considering and at least some instances maybe not everyone cares but really it's a it's a small market segment where multigp is for gaming it makes sense it's the same thing we've said friend video this is an exclusive to crossfire in AMD it's the same for SLI with SLI ten seventies we reach the ultimate conclusion of it's really not worth it you're better off buying a single card in some production applications you could make this argument because maybe if you're doing something OpenCL accelerated in this case for cuda accelerated an invidious case the argument could be made for software like blender or premier if it supports multi GPU render then you might actually see a benefit in that application and you could potentially get benefit in a few of the games you play as well just go into it knowing that you're occasionally the minimum going to be disabling crossfire the most interesting takeaway here if we're going to take something away is that the crossfire Rx for 70s in some games perform actually in most games perform really close in terms of frame rate output to the Rx for 80s so by performing as closely as they do it almost really invalidates the argument of crossfire rx for 80s we didn't recommend those either just like we didn't recommend as I said any of the other multi-gpu setups for gaming general use cases but if you were to buy one and you're looking at crossfire for ATS I would say also consider the option of four 70s because you drop your price a little bit if you really want that set up it's let's let's kind of just assume for sake of ease that the price of these is somewhere around $190 I know it's all over the map right now but let's assume it settles there for a semi-decent card you're looking at something like three hundred eighty bucks for two of them and if you do two for 80s at least a gigabyte models anyway then you're looking at something like five hundred of course if you do the four gigabyte model then you encounter the same thing we talked about in our 470 review which is the pricing is so close in some instances that is it's just it's really weird the market is strange right now and you can watch our previous videos for that but basically the price argument is potentially here for these cards if you really want to do crossfire if not don't because it's not worth it get a single card instead that's all there is to say so patreon like the post roll video as always link in the description below for more information test methodology all that stuff subscribe so you don't miss more content I'll see you all next time I'll see you all next time good night
We are a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for us to earn fees by linking to Amazon.com and affiliated sites.