First Benchmarks: Nvidia GeForce GTX 1070 Mobile Pascal

[Guest]

Nvidia claimed that the new mobile Pascal line has the same performance as its desktop counterparts. We took a quick look at the mobile GeForce GTX 1070 to see if that was true.

First Benchmarks: Nvidia GeForce GTX 1070 Mobile Pascal : Read more

 

[Kenneth Barker]

Amazing

 

[bit_user]

Please use a more similar CPU, in the full benchmarks. Even if you have to underclock a desktop CPU.

Also, I'm really looking forward to power, battery life, and noise measurements. How is battery life & performance, on battery? On AC power, does gaming or VR make the laptop sound like a hairdryer?

 

[thor220]

These aren't benchmarks, they're garbage. Comparing a 1070 in a desktop with an i3 of unnoted model to a laptop with a skylake i7.

As a review site I no longer use Tom's because frankly it's obvious bottlenecks like this that show either obvious negligence and / or favoritism. It's on a higher level of stupidity than adding project cars to your benchmarking suite.

 

[Tron_knows]

Battery life? Really? Who plays games using the battery on a gaming laptop with highend graphic cards? Use the Intel graphics card if your worried the battery life.

 

[nitrium]

Who in their right mind pairs an i3 with a GTX1070? How do these comparisons mean anything? No one here will learn anything useful from this test.

 

[-Fran-]

You all missed the point of using the i3 it seems...

To "simulate" the same CPU class of performance as in the notebook you have to cap the CPU power in the desktop side to make a more "apples to apples" comparison. Unless the notebook has a full fledged i5 or i7, a mobile Skylake is around the same as a desktop i3 or i5 at best.

I don't know why 3DMark punishes so much the core count in the CPU score, but you all can see in the aggregate it doesn't count that much (as it should be).

Cheers!

 

[hannibal]

I have phenom2 with 1070. More than double more speed than old 560ti... So i3 would be huge upgrade to me ;-)

 

[ern88]

Pretty impressive that's for sure.

 

[Brian_R170]

I'm more interested in a mini-PC with these GPUs. I imagine something the size of a Skull Canyon NUC that's maybe twice as thick to support the GPU, GDDR, and additional cooling.

 

[mortsmi7]

Mobile i7 is two cores, four threads. How does that not compare to a desktop i3?

 

[-Fran-]

Mobile i7 is two cores, four threads. How does that not compare to a desktop i3?

That is "U" i7s. The "HQ" models are 4C/8T versions.

Cheers!

 

[thor220]

You all missed the point of using the i3 it seems...

To "simulate" the same CPU class of performance as in the notebook you have to cap the CPU power in the desktop side to make a more "apples to apples" comparison. Unless the notebook has a full fledged i5 or i7, a mobile Skylake is around the same as a desktop i3 or i5 at best.

I don't know why 3DMark punishes so much the core count in the CPU score, but you all can see in the aggregate it doesn't count that much (as it should be).

Cheers!

Couple of problems with that. First, you have to assume both are the same generation. The i3's generation is undisclosed. Second, the processor in the notebook is a quad-core that boosts up to 3.6 Ghz. Note only does it have double the cores of the desktop processor they are comparing it to but it also operates at a much higher frequency than the downclocked i3.

http://www.notebookcheck.net/Asus-ROG-G752VS-Notebook-Preview.171608.0.html

So yeah, Tom's has no excuse on this one.

 

[-Fran-]

Couple of problems with that. First, you have to assume both are the same generation. The i3's generation is undisclosed. Second, the processor in the notebook is a quad-core that boosts up to 3.6 Ghz. Note only does it have double the cores of the desktop processor they are comparing it to but it also operates at a much higher frequency than the downclocked i3.

http://www.notebookcheck.net/Asus-ROG-G752VS-Notebook-Preview.171608.0.html

So yeah, Tom's has no excuse on this one.

The i7 in the notebook used is a 45W TDP CPU. That 3.6Ghz is a best case scenario when no other cores are taxed and the TDP is not reached with a base clock of 2.7Ghz. The desktop i3 has a TDP of *at least* 50W and a base clock of 3Ghz+. Assuming stock cooling used, then it should be pegged at it's non-throttled speed using 2 cores and the 4 threads.

I will concede one thing honoring what you mention though: it would be nice to have had CPU graphs as well to show how they stacked, but it wasn't the main purpose of the article to show that.

Maybe with a full fledged review of the mobile versions we might get to see it.

Cheers!

 

[bit_user]

You all missed the point of using the i3 it seems...

To "simulate" the same CPU class of performance as in the notebook you have to cap the CPU power in the desktop side to make a more "apples to apples" comparison. Unless the notebook has a full fledged i5 or i7, a mobile Skylake is around the same as a desktop i3 or i5 at best.

That's ridiculous. A better simulation would be to match the core count and clock speed (approximately). Anything else is useless, since some things are well-threaded, while others are more sensitive to single-thread performance.

 

[bit_user]

I'm more interested in a mini-PC with these GPUs. I imagine something the size of a Skull Canyon NUC that's maybe twice as thick to support the GPU, GDDR, and additional cooling.

It's the same GPU! Really!

Just save yourself some money and buy a big mini-ITX case that can handle the desktop version.

 

[poweruser_24]

"The competition this round is one of our test benches, running an Intel Core i7-6320 and 32GB of memory. "

Really?, and perhaps if you managed to link to said i7 that would be good.

"The test bench was running only an i3, which explains the lower CPU and overall scores"
Ahhhh, so not an i7 then.


As others have pointed out... Why? If this is a GPU comparison, then surely the CPU should not be a factor. If the tests you are running are bottlenecked by the CPU on the laptop, then so be it; let it show in the results, because it's not like I can easily replace/select the desired laptop CPU to prevent that. There are enough differences in the other motherboard componentry and even layout, that some of these comparisons are never going to be like for like, so there is no need to have an obviously challenged desktop in the mix.

Even if the laptop is "slower" and you explain it away as the CPU differences, that is fine, because at least I can make the informed choice of form factor affecting performance.

If you wanted to have like for like, then at least choose a desktop CPU with the same core count. If you cant get one with the same GHz/TDP, then throttle it, but there should still be an unthrottled desktop comparison, so we can see the difference, or lack of.

 

[poweruser_24]

"The competition this round is one of our test benches, running an Intel Core i7-6320 and 32GB of memory. "

Really?, and perhaps if you managed to link to said i7 that would be good.

"The test bench was running only an i3, which explains the lower CPU and overall scores"
Ahhhh, so not an i7 then.


As others have pointed out... Why? If this is a GPU comparison, then surely the CPU should not be a factor. If the tests you are running are bottlenecked by the CPU on the laptop, then so be it; let it show in the results, because it's not like I can easily replace/select the desired laptop CPU to prevent that. There are enough differences in the other motherboard componentry and even layout, that some of these comparisons are never going to be like for like, so there is no need to have an obviously challenged desktop in the mix.

Even if the laptop is "slower" and you explain it away as the CPU differences, that is fine, because at least I can make the informed choice of form factor affecting performance.

If you wanted to have like for like, then at least choose a desktop CPU with the same core count. If you cant get one with the same GHz/TDP, then throttle it, but there should still be an unthrottled desktop comparison, so we can see the difference, or lack of.

 

[-Fran-]

That's ridiculous. A better simulation would be to match the core count and clock speed (approximately). Anything else is useless, since some things are well-threaded, while others are more sensitive to single-thread performance.

It would have, yes. To your second point, I don't think the software tested here scaled to 4 cores solidly.

Cheers!

 

[Makes Nonsense]

Hello. Can you share 1070 mobile bios please? If u share it, thanks in advance.

 

[mapesdhs]

I have phenom2 with 1070. More than double more speed than old 560ti... So i3 would be huge upgrade to me ;-)

Indeed you'd see a sizeable speedup just by moving the 560 Ti onto a better platform, but switchiing to an i3 would be unwise. ;D

As for the mobile 1070 (notice how we're all effectively giving it an M monicker in discussions anyway), I'm just annoyed that so many sites have posted results without a 970M included for comparison. Sure it's great that the mobile 1070 beats the old 980M, but how does it fare against a 970M? I should imagine many more people bought the latter in the past given the large price jump from 970M to 980M.