Gadgetory

the past four years has seen little change when it comes to GPUs the arrival of the Radeon HD 7970 for AMD was a big deal back in January of 2012 as we saw the birth of the gcn 1.0 architecture and crucially the first step to the 28 nanometer design process the 7970 was a huge step forward from the 6970 before it at the time was on average 40% faster likewise for Nvidia the GeForce GTX 680 marked the transition down from 40 nanometers to 28 nanometers assuring and improved them performance and power consumption upon release the GTX 680 was faster less power-hungry and quieter than the 7970 which would see AMD release an overclocked 7970 called the gigahertz edition three months later since then the 7970 was rebadged as radiant r9 280x in 2013 whilst AMD pushed their design further to create the 290 X which was later rebadged as the 390 X with twice as much VRAM finally AMD reached the limits of what was possible with the design on the 28 nanometer process when they created the monstrous fury acts with 4096 SPU's in 2015 and video 2 has been constrained to the 28 nanometer process though there have been a little more productive in my opinion the Kepler architecture was mostly dumped in late 2014 to make way for the new Maxwell GPUs which are more efficient and more powerful one might think that this change to a more efficient architecture was a good thing for NVIDIA fans but for many of the opposite seem true the problem started when gamers began to notice a rather large discrepancy in performance between Kepler and Maxwell most noticeably in modern game works titles such as The Witcher 3 some went as far to say in video is deliberately gimping performance on the previous generation GPUs in an effort to force the consumer to upgrade I think that theory has mostly been debunked now and instead it seems much more likely than a video simply forgotten their older generation GPUs when it comes to performance optimizations focusing their efforts solely on Maxwell given in videos in the business of making money about selling new products and not keeping second-hand shoppers happy this makes sense from a business perspective however if they were deliberately and decapping the older products that would be another issue but if they're simply forgetting about them then that's the pitfall of buying their products but is it can you really expect your ultra expensive shiny new NVIDIA GPU to be near useless once the next generation lands I've seen plenty of articles and videos that compare Kepler based GPUs using new and old drivers to Maxwell with mixed results yes there are instances where Kepler GPUs such as the GTX 780 4 well behind the new GTX 970 in some modern games but it is hard to pin that entirely on the driver after all Maxwell is a more efficient architecture and under the right conditions considerably more efficient in fact if the GTX 970 was merely a rebadged 780 with more vram up bloody Hal I've gone and done it I mentioned vram when talking about GTX 970 where was I run if the GTX 970 was merely a rebadged 780 with more vram than you wouldn't expect or accept discrepancies in performance from playing modern titles anyway let's just get to the point very soon consumers will have access to the next generation AMD and NVIDIA GPUs for many longevity will play a key role in their purchasing decisions I wanted to find out the truth and I didn't want to do it by comparing two different Nvidia architectures instead I'll be taking the GeForce GTX 680 and comparing it to the Radeon HD 7970 for testing I pick nine games from around the time the GTX 680 in HD 7970 were released in 2012 all of which were popular choices for benchmarking these GPUs so we can safely assume both cancel well optimized following that I then went ahead and benchmark no more recently released games most of them very recently such as rise of the Tomb Raider black ops 3 Far Cry primal and the division for example once the dust settles I'll be looking at how the two compared in the older games compared to the newer games to see if there's any performance discrepancies both graphics cards will be tested using the default AMD and NVIDIA specifications and please note we are using the original 7970 not the overclocked gigahertz edition card I'm not particularly interested in which card is faster but rather the performance margins that separate them and if they change over time first up let's look at the older games from around 2012 please note I will be focusing on the 1080p performance have also included the 1440p results for those interested the 7970 was 7% faster in Batman Arkham City when comparing the average frame rate this time the GTX 680 was just 3% faster in battlefield 3 now when testing with Crysis 2 we find the 7970 is now 3% faster though it did suffer from a considerably lower minimum frame rate the 7970 stays on top for dirt showdown with a 4% lead the GTX 680 hits the lead once again winning by a 7% margin in hard reset this time the GTX 680 was 7% faster in Max Payne 3 man I had some fun with that game all those years ago Metro 2033 was taken by the 7970 by a rather decisive 17% margin using total war Shogun 2 we found that 7970 was 13% faster finally rounding out the old-school gaming is another favorite stalker Call of Pripyat this test resulted in a dead heat with the cards tied at 65 FPS each now for the newer titles these are all modern titles that I use regularly to bench my new GPUs now mostly selected at random the GTX 680 was 3% faster in battlefield 3 and surprise surprise was again 3% faster when testing the fourth and current installment these are the most shocking results seen so far here the 7970 was found to be 33% faster than the GTX 680 could this be a sign of what's to come yet another strong win for the 7970 this time when testing f1 2015 here it's able to get 14% ahead of the GTX 680 the Far Cry primal results are a bit mixed the 7970 was on average 2 percent faster despite being 15% slower when comparing the minimum frame rate again this time when testing Just Cause 3 we find the 7970 was faster on average but much slower when comparing the minimum frame rate Star Wars Battlefront provides another big win for the 7970 as was found to be 16% faster when comparing the average frame rate the 7970 stayed ahead in the division leading the GTX 680 by 9% margin it was neck-and-neck in the new Tomb Raider game with both GPUs delivering the same 39 FPS that said the 7970 was 3% faster when comparing the minimum frame rate last but certainly not least is the Witcher 3 Wild Hunt unsurprisingly the GTX 680 takes the win here it was 8% faster when comparing the average frame rates some of you might have noticed how poorly the GTX 680 performed in some of the modern titles when compared to the 7970 the 1440p resolution in games such as Far Cry primal Just Cause 3 the division and rise of the Tomb Raider although I didn't discuss the 1440p results as these cards are better suited for 1080p gaming I believe there's a simple explanation for this and it has nothing to do with display drivers instead I think this rapid decline in performance is likely down to the limit of VRAM buffer of the 680 by default the 680 was paired with only a 2 gigabyte vram buffer while the 7970 was afforded a little more breathing room here with his 3 gigabyte buffer compounding this issue is the 256-bit wide memory bus which limited the 680 to 192 gigabytes per second whereas the 79 70s 384 bit bias enabled bandwidth of 264 gigabytes per second it's almost 40% increase in bandwidth means the 7970 can chew through its buffered data faster now that we have all the numbers let's find out what the overall results are looking to the old games we find that both cards delivered virtually the same performance on average the geforce gtx 680 averaged 69 FPS making it just 1% slower than the radeon HD 7970 which averaged 70 FPS for those wondering the minimums were much the same though the GTX 680 did come out on top here by a single frame 45 FPS verses 44 fps jumping to the new games we'd find similar results though this time the GTX 680 was 5 and a half percent slower than the 7970 51 FPS versus 54 fps meanwhile the minimum frame rates averaged out to be the same at 40 FPS each so the margins are far from significant despite growing 5 times larger in favor of the 79 70 for the newer games looking back to the older games the GTX 680 was slower in 5 of the 9 tested for the new games that were slower in seven of the nine games tested so take away from that what you will with the margins being as close as they are it's hard to draw a solid conclusion I'm not sure I'd be willing to claim the AMD GPUs do in fact age better the small percentage change that we are seeing here can very well be down to the fact that for the most part AMD has been able to get away with very minor changes in design since debuting the gcn 1.0 architecture back in 2012 not just that but today the current generation r7 370 series still uses the original GCM 1.0 architecture so you have to imagine AMD is still optimizing for it and video on the other hand is little incentive to keep optimizing for Kepler and I'm sure the four-year-old geforce gtx 680 is no longer on the driver teams radar unless something goes horribly wrong this doesn't do anyone who invested a thousand dollars in the geforce gtx titan two years ago any favors but I'm sure anyone with that much money to throw around has since moved on or plans to shortly once these next-gen cards land and we said all that we see no real evidence that in videos older GPUs have slid significantly when compared to the AMD counterparts so this all seems like a bit of a non-issue thanks for tuning in to another episode of hardware unboxed I'm your host Matt as always and if you have any questions please leave them in the comments section below have a great day and I'll see you guys next time