WEBVTT 00:00:07.640 --> 00:00:13.519 okay guys it's new video card review time again so we're going to be taking a 00:00:11.400 --> 00:00:19.680 look in this video at the 1080p performance of the allnew GeForce GTX 00:00:16.880 --> 00:00:24.680 670 from NVIDIA so you can see here I've actually got a variety of GTX 670s uh 00:00:22.359 --> 00:00:28.720 most of which I did unboxings of already so you can see I've got a few reference 00:00:26.320 --> 00:00:32.520 cards this one's Factory overclocked this one's not this is one a direct 00:00:30.439 --> 00:00:36.840 NVIDIA card I've got them set up with a three-way SLI Bridge so that's a bit of 00:00:34.040 --> 00:00:41.079 a hint of what one of my other videos is going to be about I've also got a non- 00:00:39.559 --> 00:00:45.039 reference card so this is with a dual slot cooler this one's going to be cool 00:00:43.120 --> 00:00:50.160 because it's going to go head-to-head against a sapphire hd7950 that has a 00:00:48.399 --> 00:00:54.559 similarly designed cooler so we can take a look at them you know which one has 00:00:52.160 --> 00:00:59.399 more coil wine or less coil wine which one has the the better cooling solution 00:00:56.760 --> 00:01:04.000 since galaxy and Sapphire are both uh key partn for NVIDIA and AMD 00:01:02.640 --> 00:01:11.159 respectively this will be a really interesting head-to-head because these are similarly priced products now in 00:01:07.960 --> 00:01:12.560 terms of the overall mix I'm not testing 00:01:11.159 --> 00:01:16.920 a ton of games today guys and I apologize for that but honestly I just 00:01:15.119 --> 00:01:20.079 am a little bit a little bit constrained in terms of what I can do right now so 00:01:18.799 --> 00:01:25.360 I'm going to be doing crisis 2 Battlefield 3 Skyrim and Witcher 2 and 00:01:23.079 --> 00:01:28.759 this is the selection of cards that I'm going to be taking a look at we're going 00:01:26.600 --> 00:01:31.159 to be looking at the 1080p performance and then I'm going to have a separate 00:01:29.680 --> 00:01:34.600 video where I'm also going to cover things like the physical power and noise 00:01:33.040 --> 00:01:41.479 I'm going to take a look at 3D Vision performance compared to the last generation GTX 570 that this new 670 00:01:39.439 --> 00:01:46.200 replaces we're going to do an overse guide seeing what kind of performance we 00:01:42.920 --> 00:01:47.560 can get compared to a 680 as well as oh 00:01:46.200 --> 00:01:52.200 yeah there it is there's that hint there for the SLI scaling video now before I 00:01:50.399 --> 00:01:56.039 go any further yes I covered this in the unboxings but I just want to give you 00:01:53.479 --> 00:02:01.799 guys a brief overview so the 670 is basically a slightly cut down version of 00:01:58.640 --> 00:02:04.960 the GTX 680 NVIDIA has made fewer Cuts 00:02:01.799 --> 00:02:08.200 than usual however because it still has 00:02:04.960 --> 00:02:12.520 the 256 memory bit memory interface with 00:02:08.200 --> 00:02:15.120 6 GHz gddr5 it only has some of the Cuda 00:02:12.520 --> 00:02:20.920 cores cut away and a slight reduction in clock speed giving you um again that 00:02:19.200 --> 00:02:24.560 that slight performance reduction now you can see the reference cards are 00:02:22.239 --> 00:02:29.000 actually very short pcbs so these are very power efficient these are very cool 00:02:26.800 --> 00:02:32.000 running they don't need beefy coolers and you can see they only have dual six 00:02:30.680 --> 00:02:36.720 pin connectors even though these are high-end graphics cards this is cool too 00:02:34.360 --> 00:02:40.239 this is some of the other sort of slight finishing touches that you see from the 00:02:38.280 --> 00:02:45.080 board Partners Galaxy's gone ahead and put a slightly different back plate on 00:02:41.879 --> 00:02:46.879 theirs MSI has actually slightly 00:02:45.080 --> 00:02:54.319 adjusted the positioning of the slots compared to the reference one and 00:02:50.200 --> 00:02:55.800 uh yeah so let's get going shall we now 00:02:54.319 --> 00:02:59.599 there are a couple critical details before we really get started and that is 00:02:57.680 --> 00:03:04.239 test benches and the driver revisions that are are being used for so for the 00:03:01.159 --> 00:03:06.959 GTX 670 I'm using the publicly available 00:03:04.239 --> 00:03:13.519 beta driver 301.4c I'm using 00:03:10.200 --> 00:03:17.319 301.4c I'm using the latest 12.4 drivers 00:03:13.519 --> 00:03:19.280 my test bench is a 4.4 GHz 3930k from 00:03:17.319 --> 00:03:25.440 Intel so that is an overclocked processor with 16 gigs of Kingston 00:03:21.200 --> 00:03:29.239 hyperx RAM at DDR 3600 I'm using a 00:03:25.440 --> 00:03:32.680 Gigabyte uh what's it called x79 a UD 00:03:29.239 --> 00:03:34.200 x79 ud7 motherboard this is sort of like 00:03:32.680 --> 00:03:40.400 the spiritual successor to their previous generation OC board so that's 00:03:37.400 --> 00:03:42.720 uh going to allow me to run up to four 00:03:40.400 --> 00:03:47.159 graphics cards in SLI or Crossfire should the uh should the need arise and 00:03:45.200 --> 00:03:52.879 then for boot Drive using the Kronos Deluxe 120 gig and for the power supply 00:03:49.879 --> 00:03:55.799 the ZX 1250 wat 80 plus gold power 00:03:52.879 --> 00:03:59.920 supply from ocz will be dealing with that side of things for cooling I'm 00:03:57.519 --> 00:04:04.519 going to be using come on Focus please an h100 which will uh keep my processor 00:04:02.760 --> 00:04:08.439 at a reasonable temperature during all this testing for crisis 2 I'll be mixing 00:04:07.280 --> 00:04:13.799 things up a little bit I'm going to be using the ultra preset since all of the 00:04:10.720 --> 00:04:15.519 cards that I am using are very high-end 00:04:13.799 --> 00:04:19.720 cards I still hate motion blur so I'm keeping that on medium and I'm going to 00:04:17.400 --> 00:04:24.120 be using highres textures this time around once again to try and strain 00:04:21.880 --> 00:04:26.600 these cards as much as I can so I'm going to do this a little differently 00:04:25.199 --> 00:04:30.400 and I'm going to go one game at a time in terms of the analysis so here you can 00:04:28.240 --> 00:04:36.440 see the crisis 2 1080p P Ultra performance here is the GTX 670 which 00:04:33.479 --> 00:04:41.280 puts in a very very strong performance because while the average frames per 00:04:38.039 --> 00:04:44.479 second are below the 7970 check this out 00:04:41.280 --> 00:04:48.560 the actual minimum frame rate of the 670 00:04:44.479 --> 00:04:50.160 was above the hd7970 so that means that 00:04:48.560 --> 00:04:55.479 when you see a dip in the action you're going to see pretty much the same dip 00:04:51.800 --> 00:04:57.199 with a 670 as with a 7970 which is 00:04:55.479 --> 00:05:00.240 outstanding I mean remember too guys you know there's stuff that's sort of within 00:04:58.639 --> 00:05:05.560 margin of error here but uh in general that shows us that 00:05:02.560 --> 00:05:06.800 this is a very very strong card and when 00:05:05.560 --> 00:05:11.720 you compare it against its more similarly priced competitor the 7950 the 00:05:09.840 --> 00:05:16.000 difference becomes much more pronounced although bear in mind the 7950 is an 00:05:13.680 --> 00:05:20.759 outstanding overclocker which will allow it to uh put some distance between 00:05:17.720 --> 00:05:22.840 itself and the 7870 which at stock speed 00:05:20.759 --> 00:05:26.720 have well put it put on a pretty good show for this uh for this performance 00:05:24.560 --> 00:05:31.479 review as well for Battlefield 3 I'll be doing my usual two-minute run through in 00:05:28.759 --> 00:05:37.240 the comrades level in the parade and I will also be using Battlefield 3 to 00:05:34.000 --> 00:05:40.759 measure the maximum uh Power draw as 00:05:37.240 --> 00:05:42.880 well as maximum noise and load 00:05:40.759 --> 00:05:47.560 temperatures of the gpus now yes I realize these are not the maximum loads 00:05:45.080 --> 00:05:51.280 that I could create with this system but I'm looking for a more realistic 00:05:49.400 --> 00:05:55.560 scenario so what I'm going to do is I find a particular van in the parking lot 00:05:53.240 --> 00:05:58.639 when I'm done my run through and I shoot it until it starts smoking and some 00:05:57.120 --> 00:06:03.199 effects get going and that's how I've been able to generate a pretty good Peak 00:06:01.120 --> 00:06:08.680 power draw and then I'm going to sit there and stare at it and get all of 00:06:06.160 --> 00:06:13.319 those stats before I close the game also for Battlefield 3 we will be 00:06:11.199 --> 00:06:18.639 running everything on Ultra in Battlefield 3 the GTX 670 sits 00:06:16.880 --> 00:06:22.599 very comfortably with its competition the 7950 hold on let me just move that 00:06:20.560 --> 00:06:27.520 cursor so you can see that in terms of the average frame rate it is quite a bit 00:06:24.520 --> 00:06:29.720 higher than even than the 7950 and even 00:06:27.520 --> 00:06:34.199 is competitive with the 7970 although when the action gets intense you'll see 00:06:31.680 --> 00:06:38.560 a dip more like a 7950 as opposed to like a 00:06:36.199 --> 00:06:43.080 7970 and it sits sort of exactly where you'd expect it to be behind its younger 00:06:40.440 --> 00:06:47.560 brother the GTX 680 I think it should be noted that this is a huge generational 00:06:45.160 --> 00:06:52.440 leap in performance compared to the card that it's replacing I mean these are 00:06:49.960 --> 00:06:55.840 staggering numbers and you'll continue to see that throughout the rest of this 00:06:54.000 --> 00:06:59.759 performance evaluation of this card once again the 7870 puts on a pretty good 00:06:58.000 --> 00:07:05.000 show considering its price point though as well for Skyrim I'll be doing my 00:07:02.720 --> 00:07:10.479 usual two laps around the town of white run and I will be using Ultra presets 00:07:08.599 --> 00:07:15.960 which means everything you know all the view distances are at Max and everything 00:07:13.560 --> 00:07:22.240 is pretty Ultra and fabulous also I am using the 00:07:19.039 --> 00:07:24.479 stock highres texture pack now 00:07:22.240 --> 00:07:28.080 Battlefield 3 was a bit of an anomaly for me bear in mind guys I mentioned 00:07:26.400 --> 00:07:35.039 this before but I'm using a different driver on the GTX 6 7 than I am on the 00:07:30.919 --> 00:07:37.039 GTX 680 so I would expect that once the 00:07:35.039 --> 00:07:42.160 GTX 680 catches up in terms of the driver revisions that we're going to see 00:07:39.599 --> 00:07:46.159 um a similar performance boost however what this basically means is that at 00:07:44.280 --> 00:07:51.720 launch for all intents and purposes the GTX 670 is the fastest card that I 00:07:48.479 --> 00:07:54.560 tested in Battlefield 3 which is no uh 00:07:51.720 --> 00:08:01.840 no small matter and I mean you can see how much ahead it is of even the hd7970 00:07:58.919 --> 00:08:05.720 so if you play Battlefield 3 the GTX 670 is going to be one of the cards on your 00:08:03.280 --> 00:08:09.879 short list I think and once again staggering performance improvement over 00:08:07.520 --> 00:08:14.759 the last generation GTX 570 almost a doubling of minimum 00:08:11.800 --> 00:08:19.280 FPS The Witcher 2 is so demanding that I won't be altering my usual test setup so 00:08:17.680 --> 00:08:25.199 you guys can see here if my camera would ever Focus exactly what I am using to 00:08:23.120 --> 00:08:30.720 test these cards bearing in mind of course this is at 1080P and I'm doing my 00:08:27.919 --> 00:08:33.360 usual run through um which if you guys have seen it before then you know which 00:08:32.039 --> 00:08:40.440 run through I'm talking about and if you haven't then you don't but it's pretty good there's some combat there's some 00:08:36.320 --> 00:08:42.399 trees there's some buildings good stuff 00:08:40.440 --> 00:08:46.040 The Witcher is such a demanding game and the settings that I'm running it at are 00:08:43.880 --> 00:08:50.120 just totally unreasonable even for modern high-end cards but it still gives 00:08:48.399 --> 00:08:55.600 you an idea of where everything stands so in terms of average FPS the GTX 670 00:08:52.959 --> 00:08:59.200 Falls right in with the 7950 about where it's supposed to be we saw a bit of a 00:08:57.440 --> 00:09:04.200 dip but something to note about the witch is that it tends to just kind of 00:09:01.120 --> 00:09:07.160 randomly hitch up and and the the frames 00:09:04.200 --> 00:09:11.519 just sort of tank unexpectedly sometimes so I don't know how much you can really 00:09:08.880 --> 00:09:14.800 read into these minimum FPS I think that would explain why they're all kind of 00:09:13.360 --> 00:09:20.680 around the same level you know you can see the 7970 dips below even a 00:09:17.600 --> 00:09:22.120 7950 um so so sort of take take all of 00:09:20.680 --> 00:09:26.000 that with a grain of salt for this one I would be more interested in the average 00:09:23.839 --> 00:09:30.720 FPS and once again we see a sizable improvement over the previous generation 00:09:28.279 --> 00:09:37.480 and about an exp Ed gap between the GTX 670 and the GTX 680 so just a couple 00:09:33.959 --> 00:09:39.920 subjective things um the GTX 680 00:09:37.480 --> 00:09:44.680 exhibits a tiny bit of coil wine better than the reference cards from the red 00:09:41.839 --> 00:09:50.279 team that I have in my testing suite and uh better than the GTX 680 that I have 00:09:47.360 --> 00:09:55.240 um but both the 680 and the 670 uh do have a little do make a little bit more 00:09:51.680 --> 00:09:57.000 noise than the last generation cards um 00:09:55.240 --> 00:09:59.880 the fan has a bit of a had a bit of a grind has a bit of a grinding noise to 00:09:58.440 --> 00:10:05.079 it you can check out my other video on the noise temperatures and uh and power 00:10:02.920 --> 00:10:09.959 consumption of this card it is extremely efficient so power consumption wise it 00:10:07.200 --> 00:10:14.079 is very similar to a GTX 570 but it offers dramatically better performance 00:10:11.839 --> 00:10:18.560 and part of this maybe the smaller PCB I don't know but a huge part of it is 00:10:16.240 --> 00:10:22.360 definitely the Kepler GPU that it's got going on in there remember this card is 00:10:20.519 --> 00:10:28.839 also available in 4 gig configurations and if you're looking at running a giant 00:10:23.760 --> 00:10:30.959 resolution like a 2560x1440 or 2560 x600 00:10:28.839 --> 00:10:34.640 monitor or three monitors in surround I would definitely look at picking up the 00:10:33.000 --> 00:10:40.200 4 Gig version so thank you for checking out my performance review of the GTX 670 00:10:38.560 --> 00:10:43.480 from NVIDIA and don't forget to subscribe to lineus Tech tips for more 00:10:41.560 --> 00:10:46.480 unboxings reviews and other computer videos