Nvidia GeForce GTX 1070 8GB Pascal Performance Review (Archive)

Page 7 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.
I was testing my gigabyte G1 970sli on i5@ 4.8k and at 1440p ultra settings, hit 50 fps with nvidia hairworks off, ~40fps with hairoworks ON, I can get more out of it perhaps if I OC my GPUs, also haven't patched the game still v1.08. The FPS changes a lot from place to place, I can easily go over 60 fps depending where I go.

So yeah.. is it game breaking NO, would I want either of the new GPU's sure, but its not like I have too.

If someone is building new rig, hell yeah go for the new cards.
 
I assume you are talking about Wither 3 (Just finished Blood and Wine yesterday).

1. The game will depend much more upon the OC on your GFX cards than on the CPU.

2. My son averages about 60 fps on his OCd 970s (1440p IPS Acer Predator @ 144 HZ) .... can't say what, if any settings are different than yours.

3. I would definitely upgrade to 1.21

4. I agree on a new rig, it rarely pays not to get the latest and greatest (one exception was 780s after price drop just before 780 Ti came out.). That being said, I can never recommend buying right away after a new release.

a) None of the aftermarket boards are here and they will bring higher performance. The Giga G1 980 Ti managed 31+% fps increase w/ overclock ... the initial reference cards, just 9.4%. That's a 20% fps difference.

b) The initial release or "1st stepping" cards will be replaced with later steppings that contain bug fixes and tweaks. The initial EVGA SCs had a cooler whereby 1 of the 3 heat pipes missed the GPU.... MSI had a plastic take over the fans that, sometimes, when not very carefully removed, damaged the fans ... both were fixed shortly thereafter.

c) When supply is short, prices are higher

In short, buying early costs more to get less.
 
yah kinda hard to hit every game at every resolution. i'd take the general % increases and do some quick math on benchmarks from the game you find elsewhere. i have all the comparison %'s i could come up with over at the megathread so you don't have to find those if you wish to figure it out.

http://www.tomshardware.com/forum/id-3047729/nvidia-geforce-gtx-1000-series-megathread-faq-resources.html right in the op are % for a number of cards that come from a handful of sites averaged together over a bunch of games.
 


I agree . add to it the fake specs they fool us with today , 1 billion to one contrast (lol) and 1000 fake hz ... so you get lost when you want to compare with other products ... it is impossible to compare today from their fake specs , and must wait for review and real life measurement. in the past I never needed a review ..

in the past , contrast numbers were real and not hidden (300:1 , 800:1 , 1000:1 , 3000:1 ) now ? we see fake digits .

The same for other specs ... I think that the laws must force them to write the real specs and not hide behind fake made up specs.

 


Just the same thing that is wrong with all reference cards ... weak coolers, "adequate" as opposed to hi-end components on PCB, no heat sinks / thermal pads on some PCB components, substantially lower performance than non-reference models and ... this time .. you pay an extra $100 to get one 2 weeks before the non-reference cards are released.

This is for the 1080, expect no different for the 1070.

http://videocardz.com/60838/msi-geforce-gtx-1080-gaming-x-is-much-better-than-founders-edition

NVIDIA’s own reference design suffers from severe throttling just after few minutes. It probably wouldn’t be that bad if not the frequency spikes. While average clock is somewhere around officially stated boost clock, those spikes cause micro-stuttering, which negatively affects gaming experience.

Hardware.info:

Founders Edition suffers from a horrendous amount of throttling and it runs +- 150 MHz lower all the time.

Meanwhile, MSI GTX 1080 GAMING X generates almost a straight line for GPU frequency (~1910 MHz), with no spikes and rather constant sub-70 C temperature. This should mean that the gaming experience will be much better, and card should theoretically generate better results in most tests. Also according to H.I. this is also the best custom design they so far tested.


 
If you consider 1440p @ 60Hz to be a reference, 1080p @ 144Hz requires 1.35 times the processing power and bandwidth as the reference. Bumping 1440p to 144Hz requires 2.4 times as much. 2160p @ 60Hz requires 2.25 times more compute and the same resolution @ 144Hz requires 5.4 times as much power. 2880p requires 4 times as much power as 1440p @ 60Hz. Running that up to 144Hz will cost you a whopping 9.6 times as much power as a GTX1070 presently provides.
 


1440p is still less than 2% of gamers while 1080p continues to climb and is at 36.81% of gamers according to the Steam Hardware Survey for May 2016(Primary Display Resolution)

Source: Steam Hardware Survey - May 2016
 


1080p is a waste with either a 1080 or 1070. that's why it is not included. if you are only playing 1080p, then stick with a 970 or wait for the 1060 to show up whenever that is. it's not worth the time t test just to say "look at that big number". if you REALLY need them, look for 980ti/titan x 1080p numbers and add a few to that. oh wait there are few of them as well, since they are also WAY overkill for 1080p.

i disagree currently there are 180hz 1920x1080 monitors. So for a fraction of users that demand the absolute fastest response time nad need to hit huge 180fps numbers the 1080 and 1070 is useful
 


The card is longer than expected ... the GTX 970 was very short in comparison and fits in compact machines . I was disappointed at the 1070 length .. which is not needed , they seem just used GTX 1080 boards
 

The 1070 is a clocked down 1080 with one quarter of the chip disabled. There is no reason for board manufacturers to re-design the whole board to fit what is effectively the exact same chip in every other way that matters from a board layout perspective. Leaving some PCB footprints unpopulated due to the 1070's power power requirements (the only significant difference board-wise) is much cheaper than re-designing a whole PCB specifically for the 1070 to save maybe a square inch of board space.
 



it is a 150W GPU ... and chip is smaller 16nm ... so I dont see a reason for this card not to be the same length of the GTX 970 which was 145W TDP and larger chip size.

I was expecting the GTX 1080 being a short card given its low TDP

if you look carefully at the card , it is not dense .. allot of wasted area on the board

1070-front-bare.jpg

 

Looks dense enough to me. The GDDR5X chips cannot be packed much closer to the GPU due to the need for space to length-match the data and control lines. The only seemingly wasted space I see there is maybe two square inches in the VRM area which may be necessary as a heatsink for the FETs and possibly for extra board capacitance.

The Founders' Edition is a reference card and the main focus of reference cards is ease of design and fabrication to get the first products out quickly, not achieving the smallest and cheapest footprint possible.
 


That is true, but I controlled for 1440p because of multiple previous posts about 1080p @ 144Hz and the fact that the GTX 1070 is such a good fit for 1440p @ 60Hz. Something has to be set to 1.0. I made a choice.
 
Am I the only one thinking that the clock is the one, overriding spec that makes these cards faster than their predecessors. The core count decides a lot about how high you can set the features at a particular resolution, but there is no substitute for clock when you are trying to hit a frame rate. What am I missing?
 


2 Billion more transistors at a cost of only 15 more Watts is nothing to scoff at. 16nm FinFET has arrived so maybe next year, I can get one and dump my 770:bounce:
 
What are you basing this on? AFAIK, there's nothing inherently better about increasing frequency instead of increasing core count as far as gaming performance is concerned.

 
Status
Not open for further replies.