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

[Steve]

For now, the GTX 1080 goes unchallenged as both the best performing and best value high-end GPU, though at the price it obviously isn't for everyone. As attractive as the $600 GTX 1080 is, many enthusiasts are undoubtedly holding out for the much more affordable GTX 1070. Partner cards will start at $380, while the arguably less desirable Founders Edition card will retail for $450.

That presently puts the GTX 1070's introductory price at $50 more than its Maxwell-based predecessor. As a natural upgrade for GTX 970 owners, this premium will probably concern many of you, so we plan to determine if it hurts the card's appeal.

At 37% cheaper than the 1080, there are two major differences between the 1080 and the 1070, the most notable of which has been made to its core configuration. An entire graphics processing cluster has been disabled, reducing the streaming multiprocessor count to 15 - and thus the CUDA core count is 1920, 25% less than the GTX 1080's.

Read the complete review.

 

[davislane1]

Excellent review. Will definitely upgrade to the 1070 if it finds its way to EVGA's step-up program in the next few months.

 

[poohbear]

Wow, Nvidia looks like it hit a homerun with this card!! Still waiting to see what Polaris brings to the table before upgrading from a GTX 970....

 

[Evernessince]

"1506 MHz to 1726 MHz, which should put the boost clock at or above 1827 MHz, or only a 6% increase. However, the actual operating frequency was consistently above 2000 MHz so that is more like a 16% overclock."

Not really when the stock card gets the boost clock. Pascal can't overclock as well as Maxwell that's for sure.

The exclusion of 4k benchmarks is noticeable to cast the card in only a good light. From other reviews I've seen, once again, Nvidia's xx70 series card has worse and worse performance as resolution increases.

1070 is still only a 1080p / 1440p card just like the 970. 1070 didn't live up to the "faster than a titan x" claims Nvidia was making either.

The fact that TechSpot gives every video card Nvidia releases a 100 a joke and proves they are not being critical. I'm sure that AMD won't pull past Nvidia this gen but a perfect score is just pandering.

 

[Badelhas]

"1506 MHz to 1726 MHz, which should put the boost clock at or above 1827 MHz, or only a 6% increase. However, the actual operating frequency was consistently above 2000 MHz so that is more like a 16% overclock."

Not really when the stock card gets the boost clock. Pascal can't overclock as well as Maxwell that's for sure.

The exclusion of 4k benchmarks is noticeable to cast the card in only a good light. From other reviews I've seen, once again, Nvidia's xx70 series card has worse and worse performance as resolution increases.

1070 is still only a 1080p / 1440p card just like the 970. 1070 didn't live up to the "faster than a titan x" claims Nvidia was making either.

The fact that TechSpot gives every video card Nvidia releases a 100 a joke and proves they are not being critical. I'm sure that AMD won't pull past Nvidia this gen but a perfect score is just pandering.

I totally agree. And where are the VR benchmarks?

 

[dividebyzero]

Many thanks Steve for another very comprehensive review.
With regards the overclocking, were the results obtained at default settings? No increase in thermal headroom, stock fan profile etc?
I notice that some sites have found the card to OC quite well when the default settings are tossed for a more OC friendly profile (Guru3D, HardOCP, Hexus)

 

[Steve]

Many thanks Steve for another very comprehensive review.
With regards the overclocking, were the results obtained at default settings? No increase in thermal headroom, stock fan profile etc?
I notice that some sites have found the card to OC quite well when the default settings are tossed for a more OC friendly profile (Guru3D, HardOCP, Hexus)

I will look into the overclocking, I have only had 2 days with the card so far. So still testing, I just updated with 4K results for you guys.

 

[Evernessince]

I totally agree. And where are the VR benchmarks?

I would like to see those as well, especially considering the bulk of Nvidia's claims were in the VR space.

 

[Evernessince]

I will look into the overclocking, I have only had 2 days with the card so far. So still testing, I just updated with 4K results for you guys.

Thanks!

 

[Julio Franco]

4K benchmarks have been added.

We haven't settled on a proper way to do VR benchmarking, until then there's a simple rule of thumbs: the minimum you want to have for smooth performance is a GeForce GTX 970 or equivalent. A 980 Ti, equivalent or above will be ideal and the more room you get, the more future proof your choice is going to be since VR games run at a high resolution and they are bound to get more complex graphics compared to the current first wave of titles.

Nvidia claims it's further improved VR performance with Pascal but we have not tested this yet.

 

[Antykain]

Very nice to see how well this 1070 performs.. While I am still going to be upgrading to the GTX 1080 from my current EVGA GTX 780 Dual Classy, I've been looking forward to seeing just how well the GTX 1070 does. I was expecting the 1070 to have a bite more "oomph", but still a pretty amazing card in the mainstream seg.

 

[Steve]

Many thanks Steve for another very comprehensive review.
With regards the overclocking, were the results obtained at default settings? No increase in thermal headroom, stock fan profile etc?
I notice that some sites have found the card to OC quite well when the default settings are tossed for a more OC friendly profile (Guru3D, HardOCP, Hexus)

Looking at Guru3D didn't we get pretty much the same overclock?

EDIT: Their memory overclock was much better though.

 

[Cryio]

What's with the AMD Battlefront results? 390X faster than FuryX?

 

[Cryio]

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.

 

[Evernessince]

What's with the AMD Battlefront results? 390X faster than FuryX?

I saw that too and though that was odd. Could just be because the game doesn't handle HBM well. Otherwise the two cards are very similar.

 

[Satish Mallya]

The fact that TechSpot gives every video card Nvidia releases a 100 a joke and proves they are not being critical. I'm sure that AMD won't pull past Nvidia this gen but a perfect score is just pandering.

The 1070/1080 are far and away the best out there, in terms of both price and performance - and they will very likely stay that way until Vega shows up - I.e. Q4 2016, if rumors are to be believed.

As someone who bleeds red (Radeon, not Hemoglobin), and has not owned an Nvidia card since ~2004 (GeForce 2 MX), it pains me to say this, but they deserve it this time around.

Additionally, these could have been released priced very close to the previous gen flagships - they weren't.

 

[Puiu]

I'm actually pleasantly surprised.

 

[Evernessince]

The 1070/1080 are far and away the best out there, in terms of both price and performance - and they will very likely stay that way until Vega shows up - I.e. Q4 2016, if rumors are to be believed.

As someone who bleeds red (Radeon, not Hemoglobin), and has not owned an Nvidia card since ~2004 (GeForce 2 MX), it pains me to say this, but they deserve it this time around.

Additionally, these could have been released priced very close to the previous gen flagships - they weren't.

First, these aren't flagship cards. Should we be thanking Nvidia for releasing reference, er "founder" edition cards for $100 more than what they used to be? We are still waiting on Nvidia's big card.

They deserve a 100? All of the performance increase you are seeing is coming from the nm shrink. Over that, the cards do have issues. 1080 had thermal throttling and neither overclocks as well as maxwell. What exactly earns pascal a higher score than maxwell when one it's obvious a lesser piggyback of the other simply riding a die shrink.

I don't care what side you are on but giving anything a perfect score means it has not issues to speak of. These cards do. 4k scaling is still poor just like maxwell, overclocking isn't great, DX 12 performance is still in the air (but looks to still be an issue so far), and neither the 1080 or 1070 provide what gamers wanted in a true 4k video card. What's the point of getting a 1070 if the 970 can handle 1440p just fine? 1070 is not fast enough for 144 Hz 1440p nor 4k so it's in a middle-spot where it doesn't do much good. The 970 overclocks amazingly as well.

 

[Evernessince]

I'm actually pleasantly surprised.

If you look at what the rumors were saying about how the card was cutdown, yeah the performance is pretty nice.

 

[darkzelda]

I'm impressed by how efficient this new series is. AMD better come up with something impressive so I can finally return to red team.

 

[davislane1]

It's a flawed card! You can't give a flawed card a perfect score!

I'd give it 100. Seems like a good value. But then, scoring systems are terrible by nature. Useless.

What does a card do? What doesn't it do? That's all that matters. Whether that totals an 85 or a 100 is irrelevant.

It's a $400 quasi-Titan.

 

[Steve]

They deserve a 100? All of the performance increase you are seeing is coming from the nm shrink.

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 achieved rather than the resulting performance is confusing to say the least.

Over that, the cards do have issues. 1080 had thermal throttling and neither overclocks as well as maxwell. What exactly earns pascal a higher score than maxwell when one it's obvious a lesser piggyback of the other simply riding a die shrink.

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.

I don't care what side you are on but giving anything a perfect score means it has not issues to speak of.

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.

 

[Theinsanegamer]

And I see people are already complaining that a great card got a 100 score again. Typical.

The 1070 looks like an amazing card. Unless AMD pulls a rabbit out of their hat with polaris, the 1070 is my next chip. Goodbye 770's, you will be missed.

 

[Steve]

And I see people are already complaining that a great card got a 100 score again. Typical.

The 1070 looks like an amazing card. Unless AMD pulls a rabbit out of their hat with polaris, the 1070 is my next chip. Goodbye 770's, you will be missed.

Stay tuned, I think AMD's Polaris could really be something special.

 

[Theinsanegamer]

And I see people are already complaining that a great card got a 100 score again. Typical.

The 1070 looks like an amazing card. Unless AMD pulls a rabbit out of their hat with polaris, the 1070 is my next chip. Goodbye 770's, you will be missed.

Stay tuned, I think AMD's Polaris could really be something special.

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.

A zen+polaris rig would be fun to build.