Gadgetory

welcome back to harbor unbox for today's video I want to discuss some of the results from my day one valued coverage as well as share a few opinions so two days ago now I rushed at my vega 56 coverage having only received the card the day prior to the release that is so it wasn't really ideal but that's what we had to work with anyway I was very excited to discover what Vega was all about so I pulled my first all-nighter of the year and tested 25 games the days prior to the release or the days prior to receiving the Vega 56 sample I spent testing six other GPUs for the comparison so we had fresh data for 25 games at 1080p 1440p and 4k overall I found that Vega 56 was on par with the geforce gtx 1070 MSI's custom-designed gaming x mall no lesson i'll touch on the reference versus custom card subject a bit later in the video getting back to the testing there was one result that stood out like a sore thumb and that was dirt for here Vega 56 was an unexplainably 32 percent faster than the gtx 1070 despite providing a full disclaimer and warning viewers that these results are massive outliers and until they can be confirmed take them with a grain of salt but yeah despite that a few viewers got very upset called the benchmarks fakin or that usual angry pitchfork stuff anyway as all the reviews went live I began combing the internet looking for other dirt for Vega results eventually I stumbled upon a hard OC peer review from longtime tech veteran Carl Bennett unfortunately Kyle didn't test Vega 56 but he did compare Vega 64 alongside the geforce gtx 1080 and 1070 so that's really going to tell me all i need to know the problem for me at least was that Kyle's results were basically the complete opposite to what I was showing here Vega 64 was 10% slower from the GC X 1080 quite different to the 32 percent gain Vega 56 had over the gtx 1070 in my video my heart sank and I immediately thought I've screwed up big time I was also really confused and I couldn't work out what I might have done wrong as I spent over an hour confirming all the dirt for results three times anyway at this point it was 2:00 a.m. on Tuesday morning just a few hours after the reviews went live so I went to bed and had a bit of a think I 100% trust and all the results the guys over at hard OCP throw up so I knew there had to be something more to it the next day I woke up with the intention of looking into this further and it was then that I know as Carl was testing with eight times msaa enables an extremely aggressive anti-aliasing method there's certainly nothing wrong with that but I tested using CMAA so could there be something in that conservative morphological anti-aliasing is position between fxaa and SMA in terms of computation cost so it's another high-performance alternative to traditional multi sample anti-aliasing otherwise known as MSA a the technique was originally developed by Intel for using Codemasters grid 2 video game the obvious thing to do here was retest all the graphics cards using 8 x msaa and then for good measure disable anti-aliasing altogether so let's start with the disabled AAA results here we see Vega 56 is indeed much fast in the gtx 10:17 29% faster in fact i also added the founders edition model which was 2% sold in the MSI gaming X version with a a disabled anyway it's quite clear that Vega 56 absolutely crushes the competition in dirt for as it wastes the GTX 1080 and was even able to beat the 1080 T is one percent low result something else worth noting here though is that with a a disabled a few X is also five percent faster than MSI's gtx 1070 gaming X moving to the CMAA results shown a few days ago we find similar margins with anti-aliasing disabled or with CMA enabled again we do see much the same margins though the fury X now slips behind the GTX 1070 now time for the 8 x msaa results and wow things have changed quite a bit previously using CMAA vega 56 is one percent low result was 38 percent higher than that of the custom 1070 now though it's 6 percent lower though it's still 5 percent faster for the average that's a far cry from what we saw previously it's been quite shocking to see how much of an impact to these various AAA modes can have on the margins whichever way you slice it though Vega 56 does well in dirt for it just does incredibly well when using CMAA although I am yet to test with lower MSA a-levels it looks like Vega does do well at two times and even four times modes based on some testing done by PC games Hardware they found using four times MSA a that Vega 56 was 20% fast and the GTX 1070 so it seems like Vega really only loses its efficiency in this title when using eight times MSA a anyway I just wanted to clarify the results shown in Monday's video as they did look very out of place that said we know already that the RS 580 beats the GTX 1060 by a handy margin in dirt for at least when using CMAA so the results probably shouldn't have been that surprising in fact as I was wrapping this up I did actually go back and have my dirt floor released a results and here I was using four times MSA a and as you can see the RX 580 is 20% faster than the GTX 1060 and not much slower than the gtx 1070 moving on to another topic i would like to discuss a few viewers claimed that my results were often that i've done something wrong at vega 56 is indeed faster than the gtx 1070 and every other review showed this AMD themselves were also quite surprised with my findings and said vega 56 should have won now I'm not getting my nose out of joint over these comments they actually weren't that many of them and I did openly admit that my testing was a rushed though I am confident in the numbers and I wouldn't have risked my reputation by publishing them if I wasn't but looking around the net there are certainly mixed results and opinions as there often are so I thought it was worth looking into honestly though looking at many of the big trusted tech sites I felt like the results I was showing were quite similar at least the margins were anyway Anandtech for example showed the same or at least very similar margins in the games that overlapped they tested nine games and of those nine games I looked at six of them dead overlapping titles included dawn of war three total war Warhammer Deus Ex mankind divided battlefield one doom and Ghost Recon wildlands they also tested with ashes of the singularity f1 2016 and Grand Theft Auto 5 now I dropped ashes of the singularity because it's a pretty rubbish GPU benchmark at least in my opinion I like it for CPU testing but not so much for GPU testing plus very few of you seem to actually play that game or interested in it I mean while I dropped GTA 5 at least from that video because it heavily favors the Green Team and it's not really technically impressive anymore especially from a GPU standpoint unless you're moderate but we're probably getting a bit carried away by that point admittedly the game is still hugely popular but it's also quite old and I wanted to favor some more recently released titles such as hell-blade then we have f1 2016 and with limited time on my hands for testing I dropped that game because it's a bit of a pain to get accurate results sometimes due to a few issues with the game that required a delete config files after every GPU or Hardware change not a big deal but I didn't have time to mess around with a game that might give inaccurate results anyway Anandtech tested those nine titles which is absolutely fine however their conclusion was quite different to mine they found Vega 56 to be on average 8% fast on the gtx 1070 whereas i found it to be 2% faster the reason for this being that they did only test 9 games and of those 9 games seven of them were very favorable to AMD in fact if you take my results from the six overlapping games and then add their figures for GTA 5 f1 2016 ashes of singularity so the games I didn't test we find the same 8% margin they reported which is about as consistent as it gets so if I tested the 9 games they did I would have also found Vega 56 to be 8% faster and I probably would have walked away much more pleased with what I was seeing games such as quake champions Mass Effect Andromeda watchdogs to overwatch hell-blade Dishonored 2 player unknowns battlegrounds and Crysis 3 all during the way the Green Team I ended up with what I believed to be a more balanced test basically in video 1 11 titles AMD 112 while we saw a tie-in just to titles so you can see why results from one review to another 10 vary quite a bit as another example if an antic dropped GTA 5 they would have found Vega 56 to be 11% faster and now the margins are really starting to come quite significant so this is why we test with so many games at harbor unboxed and I do my best to include new popular titles such as player unknowns battlegrounds when I can years ago I used to test with about a dozen titles then I move to about say 16 no 18 that sort of became the norm and now I find myself usually testing with over 20 games to try and give you guys a more complete picture anyway I don't mean to single out an antique here those guys do insanely awesome work and the numbers are always spot-on it just comes down to the spreader games tested now while about five of the 1500 Plus comments I received on my Vega 56 video claim the results were wrong or fake because other reviewers showed better numbers many more upset that I tested using a custom GeForce GTX 970 ideally I would have liked to test with a gtx 1070 founder's edition as well as a custom board partner models such as the msi gtx 1070 game ex which I did use however due to limited time and we are always managing how much time we have to invest in these videos I could really only test one so I went with the custom msi gtx 1070 gaming ex for a few reasons I'll try to explain those reasons as best I can but I certainly don't expect to change the minds of those that disagree with my choice and even then I understand their point of view why if I said a compromise had to be made actually before I get into it I'm going to take Paul's explanation I'm sure you guys know Paul from Paul's hardware but if you don't check it out Paul is one of my absolute favorite tech tubers now Paul use the galaxy gtx 1070 XOC model despite the fact that AMD would have much rather he used the founders Edition the reasons he didn't use the EFI model in his words is because there are many more custom overclocked gtx 1070 s available on the market may offer better performance while sticking to the MSRP and this is certainly true he also noted the 1070 has been out for a year now and that's also something I touched on if the gtx 1070 was released a month or maybe even two months ago then sure i'd test the efi model because it would be somewhat relevant especially if customer clocked versions came in at a higher price however as it stands the efi versions actually cost more than the custom board partner models and for good measure it's slightly slower while running much hotter and louder since day one I've recommended highly recommended my viewers avoid the founders Asian models like the plague and the and I got my hands on the custom blog partner cards I stopped testing the FE models entirely much too invidious dismay now if AMD themselves agreed that the reference Vega 56 is limiting the GPUs performance which they certainly haven't suggested then why release the card in the first place it's not exactly putting your best foot forward is it rather why not do what they did with the rx 518 have the board partners take care of the cards design speaking of the board partners I did talk with multiple different partners and they all said that custom value cards are nowhere to be seen and don't expect to have models anytime soon two different sources also believe that for now AMD is writing so low so that's quite interesting whether or not this is true I I simply don't know guys but it does sound like it will be some time before we see the availability of custom-designed vega cards and if that is true that means those researching Vega performance right now with the intention of buying in the next week or two will likely be buying a reference Vega graphics card in my opinion it would be misleading to show that consumer gtx 1070 found edition performance when in fact they were most certainly be buying a custom gtx 1070 board partner card if they weren't to buy Vega anyway if custom bag of 56 models do come out in the next few weeks or months and they do offer 10% or maybe even more performance over the reference card then of course I will retest all 25 games possibly even more and give you guys up-to-date consumer advice for now though it just seems like common sense at least to me to compare products that consumers will actually be purchasing and I'm going to leave it at that you certainly don't have to agree with me but at least you now know the reason why I tested the way I did whether they agree with that or not anyway that's going to do it for this one there will be plenty more Vega testing coming up on the channel like what I do with rising I'll track Vegas progress closely as new drivers games and custom cards are released I'm your host Steve see you next time guys you