What Video Card to Buy - Late 2016

Linus Tech Tips ·Linus Tech Tips ·2017-05-06 · 1,859 words · ~9 min read
Floatplane YouTube

Transcript

JSON SRT VTT 154
0:00 over the last few months AMD and NVIDIA have released a slew of new graphics
0:05 cards promising better performance at in
0:08 some cases lower prices but which P
0:12 should you choose Polaris or Pascal and
0:16 how much should you spend let's try to answer those
0:28 questions tunnel bear is the simple VPN app that makes it easy to browse
0:32 privately and enjoy a more open internet to try tunnel bear for free check out
0:37 the link in the video description so we're going to start as
0:41 usual with the tldr if you're building a
0:44 budget box or upgrading an existing machine you can't go wrong with this
0:48 powercolor Radeon RX 460 2 gig video
0:51 card that we found at 110 bucks the
0:54 rx460 delivers smooth frame rates at full HD 1080p in popular titles like
1:00 League of Legends Dota 2 OverWatch and CSG go without having to compromise on
1:06 graphical quality settings stepping up to the $200 price point this EVGA GTX
1:13 10603 gig is a standout deal on the
1:16 green side of the fence with a susus strix rx4 74 gig holding down the fort
1:22 for team red these cards will perform on
1:25 average very similarly and pack enough horsepower for AAA games at high details
1:31 1080p with pcmr approved frame rates for
1:36 a little more oomph this Sapphire RX 484
1:39 gig costs 15% more than those but is
1:43 also about 15% more powerful making it an equally good value if your piggy bank
1:49 can handle the pain now there's an unusual void right now in the next price
1:53 range I would normally address $300 to
1:57 $350 it's populated almost EX exclusively by overpriced RX 480s and
2:03 GTX 1060s I mean they're good cards and
2:07 thanks to their extra horses and onboard memory they're capable of driving solid
2:11 gaming experiences at higher resolutions like 2560 x 1080 ultrawide and 2560
2:18 x440 but I wouldn't pay more than $260
2:21 to $270 for them given the availability
2:24 of this Gigabyte windforce 106 6 gig and
2:28 this Sapphire rx48 8 gig which leads us
2:32 finally then into Enthusiast class cards
2:35 that cost $400 or more where my pick is
2:39 this EVGA GTX 1070 Super Clock the FPS
2:44 per dollar number is a little lower than our other cards but the 1070s value is
2:49 also not bad it delivers 80% of the
2:53 performance of its bigger brother the GTX 1080 enough to run almost anything
2:58 at 4K with some tweaking but at a price that's closer to 65% as much thanks to
3:04 its lower cost gddr5 memory and cutdown
3:07 processor there are slightly cheaper GTX 1070s but I felt like the EVGA tax of 10
3:14 bucks was worthwhile in this case because there is more to life than
3:17 frames per second but that's a subject for the next portion of this video tldr
3:22 is over where I'll explain the methodology that I used to select these
3:27 cards this is my
3:31 spreadsheet as some of you understood
3:34 when I did my little rant on low-end video cards and why nobody should buy
3:38 them for gaming value by definition is
3:42 not determined by Price alone but by a
3:46 ratio of price to reward a meal that
3:49 costs half as much but contains a quarter as much food is cheap but it is
3:56 not a good value especially if you're still going to be hungry when you're
4:00 done so back to my admittedly imperfect
4:04 spreadsheet it was designed to help me use performance numbers from trusted
4:09 review sites around the web to calculate which cards will give me the best value
4:15 or the most FPS per dollar across a variety of games and resolutions the
4:20 more I add the more meaningful the final
4:23 average numbers become here's how it works I could use exact FPS values like
4:31 42 FPS if I was pulling all of my data
4:35 from one source where I knew that every game was tested with the same
4:39 methodology and the same settings but since no single review covers every
4:44 possible scenario I've opted instead to normalize to a last generation card
4:50 whose drivers are mature and that was well sampled making it a great point of
4:55 comparison for both modern flagships and mid-range Solutions the G gtx980
5:01 reference design so when I enter exact
5:04 FPS values for both the cardart I'm investigating and the gtx980 here I get
5:10 a relative performance number which I fill into my spreadsheet this allows me
5:15 to compare cards across different review sites or even cards that haven't ever
5:20 been directly benchmarked against each other with from some spot checking
5:25 anywhere from 2 to 3% margin of error
5:28 for cards that have never been benchmarked against the GTX 980 like the
5:32 750 Ti I used a card that has been
5:36 benched against both as an intermediary the rx460 though this method is not
5:42 ideal and I approximated a lot of other stuff as well only a small fraction of
5:47 the cards available for sale have ever even had a professional review done but
5:52 thanks to GPU boost 3.0 especially
5:55 actually were more limited by power limits imposed by by NVIDIA and AMD than
6:01 by Factory overclocks anyway these days
6:05 so I've approximated most non-reference cards to perform pretty similarly which
6:10 won't necessarily be true but as you can see here adjustments to Price have a far
6:17 larger impact on our value calculation than the kinds of singled digigit
6:21 performance differences that you'll typically see from one GTX 1080 to
6:26 another so yes I know I said it already but it merits repetition here so the
6:30 entire comment section isn't full of people pointing it out this method is
6:34 imperfect and would require either many
6:38 many days of additional manual data entry if I wanted to go beyond relying
6:42 so heavily on Guru 3D for example whose reviews are linked in the Excel sheet by
6:46 the way or ideally some kind of fancy
6:49 custom statistical analysis software and it also ignores a lot of factors here
6:54 are the ones I could think of in no particular order multi-GPU you want to
6:58 shop for an SLI Crossfire setup they would each need to be added individually
7:02 for every pair or more of cards have fun I'm not doing that two characteristics
7:08 like board length display connectors or
7:11 the required physical plugs that might affect physical compatibility with your
7:16 system three warranty there is a huge
7:20 range of policies not just in terms of the coverage period but in terms of
7:25 whether resale overclocking Andor
7:28 physical modif ification is permitted and this is critical for folks who plan
7:32 to water cool four cooler style and
7:36 quality Aesthetics aside though some people care a lot about that for some
7:41 configs a rear exhaust blower design
7:44 might be better to avoid recycling hot air to cool other components while in a
7:49 better ventilated case an open air Cooler's additional performance for the
7:54 GPU itself will result in tangibly better performance in games consult your
8:00 case manufacturer if you're not sure oh and double ball bearing fans are
8:05 probably worth a couple bucks for a card you plan to keep for a long time five
8:10 frame buffer size now generally speaking
8:13 bigger frame buffer cards end up underpowered before the extra RAM is
8:18 necessary but there are some exceptions someone who is really into Skyrim
8:23 texture mods will definitely want to grab an 8 gig rx480 versus a 4 gig one
8:29 regardless of what the performance per dollar metric looks like because vram is
8:34 one of those things that doesn't affect performance at all until you run out of
8:39 it and the card is swapping textures out to system memory and at that point
8:43 everything goes south faster than all the Canadian birds and seniors do in the
8:46 winter time six driver improvements over
8:50 time and the features that are difficult
8:53 or impossible to quantify Shadow play and freesync are examples of things that
8:58 I care about but maybe someone somewhere cares about true audio or the automatic
9:03 game settings optimization in GeForce experience your mileage may vary here
9:07 and that can definitely affect your decision especially when you're looking at Red Team Versus Green Team seven
9:14 sanity checks this GTX 950 at $130 could
9:20 look like a great value for 4K gaming
9:23 with its similar FPS per dollar number but even though this 1080 Founders
9:28 edition cost $700 H it's 47 FPS versus 5 FPS might be
9:36 a factor here for you if you actually want to play games so Consulting actual
9:42 reviews for raw performance numbers is still part of the process leading us to
9:47 eight and perhaps the most important actual availability just because you can
9:52 Google some super cool card if no retailers in your region carry it then
9:56 you're kind of out of luck that is to say unless the retailer ships worldwide
10:02 like Mass drop though it does cost a bit extra if you're not familiar with mass
10:06 drop I'm going to run you through the basic concept Mass drop takes a whole
10:09 bunch of people who want to buy a thing and then a whole bunch of Manufacturers
10:13 who want to sell a lot of things and they put them together the community
10:17 says Hey we'd love to buy that product X
10:21 but it's like expensive if a whole bunch of us agreed to buy it do you think the
10:24 price could drop a little bit you know volume discount and mass drop goes to
10:28 the authorized distributors or even the manufacturers directly and negotiates a
10:33 better price if a bunch of people want to buy or in certain cases they even
10:39 have manufacturers produce special products just for them like the k7xx
10:45 headphones that we featuring today and this is the same ones that I reviewed
10:50 last year so they were configured by mast drop and manufactured by AKG
10:55 they're open-backed headphones with large cushiony ear cups that feature a
10:58 flat flat wire voice coil a genuine leather headband memory foam ear pads
11:03 with velour coverings and a 2-year warranty covered by mass drop and what
11:08 kind of a price do you get well the more buy the better it gets because that's
11:13 how Mass drop works so check out the link in the video description so thanks
11:18 for watching guys if this video sucked you know what to do but if it was awesome get subscribed hit that like
11:22 button or check out the link to where to buy the cards that we recommended in the
11:26 video description also linked in the description is our merch store which has
11:29 cool shirts like this one and our community Forum which you should totally join now that you're done doing all that
11:34 stuff you're probably wondering what to watch next so click that little button in the top right corner to check out our
11:39 latest video over on channel super
11:42 fun