Nvidia GeForce GTX 1070 Review: Titan X-like performance for a fraction of the price

I don’t care if all they did was rub banana juice all over the thing, it delivers by far the best cost per frame ratio of anything on the market and the power consumption is phenomenal. Why you would focus on where or how the performance is being achieve rather than the resulting performance is confusing to say the least.



This is a reference card and we reviewed it as such. The score is for the GPU, we have never reviewed a reference card before or focused on the reference card itself. If that were the case the 290 would never had scored 95pts.



A perfect score doesn’t mean a perfect product, a perfect product doesn’t exist. Unless Nvidia or AMD come out with a free graphics card that smokes everything on the market, that would probably be perfect. The score is subjective and it reflects the fact that it is the best card on the market at this time.

Everything is a factor when looking at a video card. Why do I have to tell a tech writer that heat generation and overclocking affect the selling price of a video card and it's respective score? I find it rather odd that there are no temperature tests done for the 1070 at all when other tech outlets have it. Did you guys just forget to do one of the biggest parts of a graphics card review? This is seriously the second major thing you guys left out. I had to call you out just to get the 4k benchmarks.


Why do the 980 and 970 get a lower score when they launched with excellent energy efficiency and as the fastest on the market? Maxwell is what gave pascal the base for it's speed and power efficiency.

https://www.techspot.com/review/885-nvidia-geforce-gtx-970-gtx-980/page9.html

Using the exact same criteria on the exact same website yet two different scores. Tell me what makes the 1070 better now than the 970 when it released?

So reference cards shouldn't be overclocked? That's a cop out given it's one of Nvidia's major selling points for the cards. Once again though, another contradiction because Techspot has done reviews focusing on a stock card and has overclocked it.

https://www.techspot.com/review/727-radeon-r9-290x/

So why don't you tell me the real reason you didn't overclock?
 
Last edited:
I figure it'll be either another 5000 series style victory or another 300 series flop. Hoping for the former, preparing for the latter given AMD's current track record.

Wild hypothesizing here, but wouldn't intel on AMD's upcoming releases be a pretty good explanation for the eminently reasonable prices on Nvidia's 10XX cards so far?

I, too think it unlikely that Polaris will beat Pascal (all indications point to low-midrange, rather than mid-high). We will find out in a month or so, I suppose, as information filters out from the Macau conference.

Here's hoping.
 
Overclocking may effect selling price in your book, but your book is not the book.

That's the point I was trying to make, everyone needs something different from this card. No criterion was stated. Reviews should cover all the bases as TechSpot has done in the past with reference cards.
 
Everything is a factor when looking at a video card. Why do I have to tell a tech writer that heat generation and overclocking affect the selling price of a video card and it's respective score? I find it rather odd that there are no temperature tests done for the 1070 at all when other tech outlets have it. Did you guys just forget to do one of the biggest parts of a graphics card review? This is seriously the second major thing you guys left out. I had to call you out just to get the 4k benchmarks.


Why do the 980 and 970 get a lower score when they launched with excellent energy efficiency and as the fastest on the market? Maxwell is what gave pascal the base for it's speed and power efficiency.

https://www.techspot.com/review/885-nvidia-geforce-gtx-970-gtx-980/page9.html

Using the exact same criteria on the exact same website yet two different scores. Tell me what makes the 1070 better now than the 970 when it released?

So reference cards shouldn't be overclocked? That's a cop out given it's one of Nvidia's major selling points for the cards. Once again though, another contradiction because Techspot has done reviews focusing on a stock card and has overclocked it.

https://www.techspot.com/review/727-radeon-r9-290x/

So why don't you tell me the real reason you didn't overclock?

"As far as we can tell, the 1070 Founders Edition card uses the exact same cooler as the 1080, and this would explain the slight improvement in thermal performance. With the fan spinning at near silent levels, the card typically operated at 76 degrees Celsius, though, on occasion, it would get as high as 79 degrees. Throughout the testing, a typical boost clock of 1782 MHz was seen, which is well above the quoted 1683 MHz for the boost clock."
 
Wild hypothesizing here, but wouldn't intel on AMD's upcoming releases be a pretty good explanation for the eminently reasonable prices on Nvidia's 10XX cards so far?

I, too think it unlikely that Polaris will beat Pascal (all indications point to low-midrange, rather than mid-high). We will find out in a month or so, I suppose, as information filters out from the Macau conference.

Here's hoping.

Yep. If Nvidia didn't know AMD was going to release their cards soon they would have definitely increased the prices, most any company would. Polaris is targeting everything put the highest cards. I'm willing to be the R9 490 will go a bit below the 1080 while the 480 will be around the 1070. Of course I also expect the AMD cards to be cheaper.
 
"As far as we can tell, the 1070 Founders Edition card uses the exact same cooler as the 1080, and this would explain the slight improvement in thermal performance. With the fan spinning at near silent levels, the card typically operated at 76 degrees Celsius, though, on occasion, it would get as high as 79 degrees. Throughout the testing, a typical boost clock of 1782 MHz was seen, which is well above the quoted 1683 MHz for the boost clock."

That makes the 1070 a bit more appealing. Any chance we can get temps along with perhaps overclocking in another article? I've heard the 1080's single 8-pin connector limits OC but that shouldn't be the case with the 1070.
 
I'm gonna be that guy and say I don't think AMD has anything to go up against these cards. If they did, why are they doing their marketing via Twitter and not screaming from the rooftops about their offerings? Either way, I don't like AMD simply because they are a component company and have been for a while. Meaning, I hate their drivers, the time it takes them to produce drivers and time it takes to fix drivers. Aside from selling you a card, they have nothing to offer over nVIDIA products. nVIDIA is primarily a software company with impressive GPU's, and it shows.

Also, AMD has said they are targeting the mainstream market with Polaris, so I can't see them alienating the Fury X and from what I've been reading, we're looking at 390Xish performance from their top Polaris GPU.

I will definitely be getting a GTX 1070 in time for Battlefield 1. Regardless what AMD comes out with.

Thanks for the review TS. Another job well done.
 
Last edited:
That makes the 1070 a bit more appealing. Any chance we can get temps along with perhaps overclocking in another article? I've heard the 1080's single 8-pin connector limits OC but that shouldn't be the case with the 1070.

https://www.techspot.com/review/1182-nvidia-geforce-gtx-1070/page6.html

https://www.techspot.com/review/1182-nvidia-geforce-gtx-1070/page7.html

I meant more like overclocking and the resulting temperatures.
 
Why do the 980 and 970 get a lower score when they launched with excellent energy efficiency and as the fastest on the market?
Duh, maybe because the paradigm isn't the same. The GTX 970 at launch sat between the R9 290 and R9 290X in performance. The GTX 970 was 11% cheaper than the 290, and 29% cheaper than the 290X. Likewise, the GTX 980 was around 19-20% more expensive than 290X for a gain of ~ 10% more performance.
This time around the GTX 1080 is offering 25% more performance for what could be the same basic cost. A Fury X costs $600-700, and vendors like EVGA have already made public pricing on custom models in the same price brackets - from ~$600 to a full custom FTW for $680 . Same price, 25% better stock performance, and much better OC headroom. The GTX 1070 is in even more rarified company - same basic performance as AMD's competition at a fraction of the cost. Now factor in that the GTX 1080/1070 absolutely destroy their current competition in performance per watt to a much, much, larger extent than the GTX 970/980 did to the 290/290X, and using the current scoring metric, 100% looks positively miserly in comparison
perfwatt_3840_2160.png

Tell me what makes the 1070 better now than the 970 when it released?
Done. But I'm guessing you switched off as soon as the comparisons didn't validate your argument.
So why don't you tell me the real reason you didn't overclock?
I'd take overclocking results as an indicator only at this point in time. I haven't seen any site actually fine tune a Pascal overclock as there isn't actually any overclocking software that allows GTX 1070/1080 past some broad strokes. Pascal has variable voltage ofsets to optimize voltage/frequency overclocking.Currently, EVGA Precision is buggy and MSI Afterburner doesn't currently offer variable V/F offsets - rather a more prosaic Maxwell "one-size-fits-all" blanket offset as per Maxwell. So, any current overclocking isn't going to be tuned for voltage/heat/frequency. Hopefully this will be resolved soon.
 
Last edited:
Impressive card. I also think AMD won't have anything up their sleeve looking at their history. Still wouldn't give the card a perfect tho.
 
Whats with the "for a fraction of the price"? I mean like... all the prices can be represented as a fraction... like ... 37/38 of the original price? or for 3.33 times higher price you can say 10/3 of the price :D

GTX 1080 is also a fraction of GTX 1070s price... 699/449 in fact ... a fraction of the price :D

There are still 7.2 billion transistors crammed inside, just 10% fewer than the most complex Maxwell GPUs.
Although technically yes, 1070 has 7.2B transistors, roughly only 75% of them are active ... thus why count the dead transistors? Just to show bigger numbers for no reason?
 
Last edited:
Whats with the "for a fraction of the price"? I mean like... all the prices can be represented as a fraction... like ... 37/38 of the original price? or for 3.33 times higher price you can say 10/3 of the price :D

GTX 1080 is also a fraction of GTX 1070s price... 699/449 in fact ... a fraction of the price :D
Do you not understand what an idiom is?
Although technically yes, 1070 has 7.2B transistors, roughly only 75% of them are active ... thus why count the dead transistors? Just to show bigger numbers for no reason?
Duh! Because the die has that many transistors, and that is the accepted way of describing a ICs trans count. Absolutely no one - except you it would seem - recalculates transistor counts for salvage parts any more than they would quote a smaller die area because parts of it are inactive.
You must lead one helluva sad life if you devote time to posting this kind of rubbish.


shaking_head_breaking_bad.gif
 
Do you not understand what an idiom is?

Duh! Because the die has that many transistors, and that is the accepted way of describing a ICs trans count. Absolutely no one - except you it would seem - recalculates transistor counts for salvage parts any more than they would quote a smaller die area because parts of it are inactive.
You must lead one helluva sad life if you devote time to posting this kind of rubbish.


shaking_head_breaking_bad.gif

Well said, it wasn't worth my time to address that "rubbish" . Still I appreciate your effort nonetheless and Mr. Ehrmantraut summed it up perfectly.
 
I will read through the comments, but, I read the bit about the best Banff for buck is the 1080, which would be the case if Nvidia and their claim that the 1080 had the power of two 980ti's which from videos I've seen, they have only like 10% increase on the 980ti, which quite frankly, isn't worth £635.

So I'll wait for the next iteration. New *tock* version of skylake, and the 1180 may be worth the upgrade
 
I will read through the comments, but, I read the bit about the best Banff for buck is the 1080, which would be the case if Nvidia and their claim that the 1080 had the power of two 980ti's which from videos I've seen, they have only like 10% increase on the 980ti, which quite frankly, isn't worth £635.

So I'll wait for the next iteration. New *tock* version of skylake, and the 1180 may be worth the upgrade

FYI Nvidia never claimed that a single GTX 1080 would beat two 980 Ti SLI cards, they said it would match/beat 980 SLI cards, so not the faster Ti version.
 
Marvelous card, though I won't buy this solely based on the last 2-3 years of Nvidia's shadey practices. Awaiting to see what AMD's Polaris brings to the table.
That's always a good option but don't condemn nVidia alone for shady business practices, AMD is no better, in fact their business practice is often more questionable. All these monolithic corporations are tarred with the same brush.
 
Long time AMD graphics supporter here, but I have to admit, this is starting to look line AMD vs. Intel for processors. AMD has done o.k. so far, but if this keeps up I'll be buying Nvidia
 
The 970 came out 1 year and 8 months ago. The PC master race is expensive. If I would have kept my old amd card as a backup, I could have sold my 970 last month for 300$ on ebay.

I do find one thing funny, Techspot used to do benchmarks which would show the 970 usually at least 60 fps on 1080p. Now with the max quality benchmarks it has dropped down quite a bit. Nudging us to buy the 1070.

will edit with details soon

It looks like they tweak the benchmarks.

The division on release Directx 11 ultra quality was just fine then, 970 56 fps.
https://www.techspot.com/review/1148-tom-clancys-the-division-benchmarks/page2.html

The division on the 1070 review, now they turn on SMAA ultra to make the 970 look like a piece of junk. 49 FPS.
https://www.techspot.com/review/1182-nvidia-geforce-gtx-1070/page5.html

I notice because I check every techspot benchmark review and look at where the 970 places in the 1080p range. I only looked up the division but if memory serves me the other games on this list over a month old will be no different.
 
Last edited:
I have a GTX 760 so upgrading to the 1070 is a no-brainer for me. Not feeling the founders Edition though, so I will wait for other vendors to release theirs particularly MSI.
 
Numbers numbers numbers. All I see is numbers on a page telling me that this one is better than that one. When I buy a video card I buy it for what it does not the numbers quoted by some guy on the internet. Hundreds of pages of reviews in which most people just skip to the conclusion to see how good the new card is. You should buy a graphics card to suite the needs of exactly what you are doing, people have no idea of the immense work that goes into the making of a video card. Video cards are multi tasking units. You could fill volumes of information on just how many applications there are that involve the power of the GPU. No one answers this question: How can I see the difference between 1920 x 1080 and 4k while watching the video review on "fill in the blank with monitor or tv or whatever". The reason I buy a new video card is not because my old card "wore out", but that the new monitor I bought needs a better card so that it will work, plain and simple. A hint to all video card 'testers', you should test all types of games, then test video editing systems, cad/cam, photo editing etc. Only then will you be able to give the buying public a real hands on evaluation of the new card.
 
Back