Rectify Gaming

RTX 2070 GPU Review: Is a $600 graphics card upgrade worth it?


   PC gaming is having that upturn within its cycle again, despite Nvidia’s best attempts to ruin it. It always happens and IS one of the biggest reasons we have premium consoles this generation, as I have covered before. As the performance limits top out on consoles, enthusiasts look for more and PC is where that sits almost exclusively if power and performance are your priorities.
 As the title suggests though, it comes at a cost with top level but not bleeding edge equipment carrying an inflated premium. But is a $600 RTX2070 really worth the cost? let alone a $1200 2080Ti? As it stands an overclocked 2070 is well within 1080 performance in many games and in excess on others, the gains to the Ti are there but the value makes that card much, much worse. The promise of things to come such as RTX features which are still almost non-existent besides EA’s partnership with Nividia allowing the Dice team to add ray traced reflections. The more promising from a wider and less demanding solution comes from DLSS ( Deep Learning Super Sampling ) which can deliver checkerboard like solutions that retain much of the image quality with a reduced performance impact, but again is mostly promises at this point.
  For now I will leave these possible benefits out of the equation and concentrate on what Nvidia have been drumming into its consumer base for years, new hardware = improvements. If you are interested I wrote a detailed article on my predictions for the new range and performance expectations on Rectifygaming before its launch, link on screen and below. Short version is they are a victim of their own success with the tech media largely slating the launch ( as predicted) and thus the market place echoing that. The $ to performance chart I created was interesting and shows that value wise none of them really add up, helped by Nvidia inflating the price to help shift 10 series stock returned or sat in warehouse, possible by the lack of competition, a point I have talked about for years.
 I decided to dip into the RTX range myself and managed to get a deal via some friends within the trade on this GeForce 2070, a brand I have sworn by and used for years to build and sell alike. This is the cheapest entry into top tier GPU’s and from my current Radeon 470 is a worthy increase which itself was a side step from my previous GTX970. The 470/570 cards fall into the £175 or $200 range. The 2070 is 3x that at $600 I managed to get mine for £430, which is a small discount but for this review I will keep to RRP figures as that is what most will pay.

THE MACHINE

  It is all tested on my Zen 2700 CPU oc at 3.9GHz, 16GB DDR4 @ 3200 and SSD. I have covered an in-depth review of the zen cpu and how CPU testing is very different to GPU testing, check that out in the links to learn more, but the 2700 is a perfect CPU to max out this card on almost anything here. From the offset then I am looking at 3x the performance delta from my 570 to this 2070 to draw even on bang for buck. I have my AMD OC by 200mhz on core and Ram which is a decent increase and is an obvious boost anyone will add for free. So we have the facts, specs and figures now all we need are the results, 12 games should be a good spread.

THE GAMES

 The tests are pretty simple, although benchmarks in PC games are notoriously misleading for final figures, they work here to represent a like for like test. On many titles I used a fixed real-time cut-scene, the best tests for GPU loads and exact opposite for CPU or a fixed run of repeatable game-play.
  I run everything at max where possible ( the 470 lacks the ram to go Uber on Wolf for example) with both a 1080 and then native 4K run. V-sync is off to show absolute max levels but of course be aware that V-sync will affect the performance levels when engaged. To be safe a title unlocked like this should average into the mid to high 60’s to expect a capped 60 to be realistic, 70’s is the ideal for headroom. The range covers some old and new titles with a couple DX12 and Vulcan examples included but bar 1 DX12 title most run best in DX11.

The Division

 This MMO is about to be superseded by the sequel but the core focus on consoles means this is very well threaded and efficient on CPU’s giving the GPU all the heavy lifting. Side by side at 1080 ultra we see a gulf between the 2 cards as expected, the 470 can deliver 58fps at 1080p, so a great 1080/60 card for sure. The 2070 can push this up to an 107fps average. A 184% improvement at HD resolutions is nice but makes the 2070 worse value than the 470.
  4K changes the results with the 8GB memory of the Nvidia being double that of the AMD card makes it a much better specced 4k card. Indeed we are 196% on average better now, the minimum is now 240% better. Still not quite matching the bang for buck of the 470 but certainly makes a 4k/60fps option possible with some cutbacks on certain areas, again not for this test.

 

AC Unity

 An oldy but still cracking game that stresses modern hardware and certainly flexes CPU. At 1080 the 2070 takes a commanding lead, some of this comes from the game favouring the API & architecture of team green. With an average 80fps it is easily maxing the game out here and almost 200% faster than the 470, so we have clawed back that value here at HD resolutions, 4K is even better. 4x the min performance and 3.4x average are the result with a locked 4K/30 possible here and higher averages if you want to run unlocked makes it the best result for Nvidia so far and gives us better value than the 470. Even  at 1440p for an extra test the Nvidia is 263% better,will it continue?

Watch_dogs 2

 Was something of an oddity, a much better game technically than the first but was less enjoyable or gripping for me. It certainly looks good with another huge open landscape which can never be fully tested in such a finite manor. That said it is clear the 470 struggles to hit 1080/60 at ultra settings, we are running without the temporal filter option that uses a checkerboard alike reconstruction technique such as Rainbow Six that uses the MSAA or EQAA hardware to reconstruct the final image from 2 lower frames. This can offer almost 50% improvement on performance with a much smaller IQ impact, but here we stay native. This gives use 195% increase at 1080 on average and 227% at 4K making the point clear in which card is a genuine 4K one.

Forza Horizon 4

 The epic racing game hit Xbox and Pc this year with some more stunning views and technical feats. 1080/60 on the 470 is possible with cutbacks and the Turn 10 dynamic engine options which are incredible, but to keep tests even these are off with equal ultra options chosen. At 1080 the Rx is 53fps behind on average, some 213% which is still not better value for money but better. 4K expands this to 261% which manages almost locked 60 at 3840×2160. Another win for the RTX card obviously but still a little short of the 3x target.

Doom

 This continues with Doom, the Rx is a capable card here for 1440/60 on high settings but the extra options can lower the rates and we see another big gaps here. At 1080p the 470 can deliver an average 72fps the 2070 manages over twice that at 161 with the average being 224% faster. Unsurprisingly 4K is even bigger with it rising to 254% over the cheaper card. Again though all this does is keep us within 2x increase for a 3 fold price premium, the card is better but value from it is worse.

Rise of the Tomb Raider

 Rise is one title bound by Single CPU core speed at times, something we see in the 1080 tests as the 2070 manages a 182% increase on average to 120fps yet GPU load is never much more than 60% leaving performance on the table. This is the lowest result so far for the RTX but 4K resolution resolves much of that loading the GPU so the CPU is not stressed as much giving us 225% increase for a 54fps average which seems to be the common level thus far.

Shadow of the Tomb Raider

 A modern game and an outlier with the DX12 option being the only way to play this game from both cards. Comparing AMD at 1080 between DX11 and 12 we see a 16% improvement on average between the 2 API’s. Nvidia benefit even more with a 52% lead over DX11 meaning if you have the game play using this API for sure. Using DX12 then we see 1080p 184% faster for green and 4K once again boosts this to 195% which is great but below the 2x increase we had been seeing elsewhere at this point.

Batman:Arkham Knight

Shows the Rx to be capable of 1080/60 while the RTX can achieve similar at 4K with another 216% increase at uhd levels dropping to 136% at 1080 with it again having GPU fill-rate headroom to spare which simply goes to waste. That is a point to touch on at the end and has been highlighted here by many of my tests.

The Witcher 3

 CDPR’s monster RPG success has worked on all formats, better with Nvidia the results show that 1080 and 4K are around 240% faster on average using a fixed real-time cut-scene that stresses the engine. Hairworks are off here on all as they still tank the frame-rate far beyond any worthwhile visual advantage.

Wolfenstein II

  Now we come to machine games Idtech6 beauty that really takes advantage of Vulcan and modern rendering techniques, it really is a showcase on tight, optimised code and how to deliver. Using the fixed real-time cutscene at the start of chapter 2 we can get almost perfect examples and this gives us the largest gap yet by far. Almost 300% on average faster on the 2070 with 4K still giving us an 81fps average to the tune of 444% over the 470.
   This game represents the absolute best you can hope for when switching cards here and I am sure Nvidia worked hard on this alongside Machine Games with it being an AMD linked title it would not be missed just how good the Nvidia is here. Some of this may come from the specific reduced shading used on Nvidia cards here that reduce the complexity and fill rate of moving pixels taking into account motion vectors and depth but it is a worthy addition and secures the best result and a clear victory on straight performance, that was never in question though, The value was and after all these tests how do we stand?

SUMMARY

 With all the results complete including DX12 in Shadow of the Tomb Raider showing a solid increase in bottom and top line performance that I hope will continue. It is the best DX12 result I have seen thus far that has a dual rendering API within the shipped game and I hope bodes well for future DX12 titles, it also demonstrates this benefits both Nvidia and AMD cards alike with it being the only option to use in this game. 16% increase on average for the 470 along with higher peaks and lows means not only do you see better limits but the consistency Is better. Nvidia fair better with 50% increase making DX11 redundant here for these newer cards, it may be a different story for older gen cards mind.
 The fact is the results speak for themselves, the target was 300% increase or more would make the 2070 a bang for buck replacement for the 470 but we see a lower 216% average across the games both 1080, 1440 and 4K. This leaves you with a clearly more powerful and better performing card which is worse value for money than the AMD card. Fine so long as you are aware your upgrade is not linear, this value is better if you only factor in 4K at 238% increase, then the 2070 makes more sense. The fact is though you are paying $8.33 per frame on the 470 and that rises to $10.71 on the Nvidia card a reduction in bang for buck of 22%. My chart from my article shows how this sits within the 10 and 20 series range of cards.
With 8GB Of Ram and feeding a UHD screen it is truly capable of 4K30 ultra on almost all titles here and likely most others with 4k60 being possible on others again at ultra. The biggest gain comes from the choice to cut back on shadow maps, motion blur, depth of field, volumetrics or ambient occlusion to see 4K/60 on a great deal more. Realistically this is rarely an option in the 470 hamstrung by memory and bandwidth it is the perfect 1080 or 1440 card with 4K being possible with heavy cutbacks on titles but this will diminish as new games launch.
   The 2070 is the cheapest entry in to top flight PC GPU’s and will offer a substantial improvement to anyone upgrading from this or other equal level cards. If you have a 1080P screen it makes much less sense right now as some games cannot max it out at this resolution. But at 4K it makes a great deal more sense with it being one of the best value 4K cards you can buy, well aside the 1080 or Ti cards which at this point make this card redundant.
  I started this article with a statement and for value the 2070 is not as good a deal as the 400 or 500 series of AMD cards, it will cost you more for less return on your spend no matter your current card the facts are here. If you have a 1080 or Ti it makes no sense at all as it will be a side step.If you have a 4K screen or want to improve your VR titles then it really offers a valid and currently good choice for those below the Vega or 1080 cards just don’t expect a high return on your investment.
You can watch the video on this topic here. 
Share Everywhere!