First NVIDIA GeForce GTX 1080 3DMark Benchmarks

[ondooy]

First NVIDIA GeForce GTX 1080 3DMark Benchmarks

It’s about time we share first benchmarks of GeForce GTX 1080, new flagship graphics card that will be unveiled tomorrow by NVIDIA.

NVIDIA GeForce GTX 1080, 8GB RAM GDDR5X at 10 GHz

I think the most interesting detail about GTX 1080 is the new memory. NVIDIA has finally used 8GB ram for its flagship card. It’s no longer an exclusive to mobile solutions. Additionally, the GDDR5X modules are clocked at 2500 MHz (which is shown as 5000+ MHz in 3DMark). However the effective clock is 10000 MHz, which means the bandwidth is somewhere around 320 GB/s (assuming it’s 256-bit wide).

On the GPU side we have a huge improvement in terms of frequencies. It is said that GTX 1080 can boost up to 1.8 GHz, with base clock around 1.6 GHz. At the time of writing we are not able to confirm the exact reference clock. For such reason I decided to avoid making comparison charts, so this post will essentially tell you what GTX 1080 is capable of and nothing more.

 

[Gattberserk]

Its just a GTX 980 on GHz steroid. It about 60-70% faster than a GTX 980 (25% above stock 980 Ti) which is reflected in the same amount of clock speed boost.


1.6-1.8GHz for 1080 vs 1.1-1.2GHz on 980.


which lead me to suspect GP104 has 2048 FP32 cores with 1024 Compute unit. Giving a total of 3072 units pack in a 320mm sq die (which is still larger than half the size of GP100, or 610mm)

 

[ondooy]

aaaaaaaaaaaaaaaaaaaaaaaaaaand

Spoiler

980Ti is still faster...

 

[royfrosty]

Woah not bad. 2x fury X also cannot beat fire strike extreme. Lol

 

[Gattberserk]

aaaaaaaaaaaaaaaaaaaaaaaaaaand

Spoiler

980Ti is still faster...

980 ti is not faster. The most extreme overclocked 980 ti is 1.55ghz which is as fast as the 1080. If the stock result for 1080 is true, den the 1080 is definitely 30% faster than a stock 980ti. How much 1080 can overclock i do not know.

This beg another question, why is the ipc performance so low. Given the size of 320mm sq its possible to pack in 8 billions transistors, same as gm200.

Which lead me to believe 1/3 of the die space is wasted on useless dp compute which is going arrive disabled anyway (in geforce). Damn nvidia! Milking at its best!

 

[enenyi]

By the way, this is my GTX 980 Ti Lightning at stock speeds:

I don't have GTX 980 Ti Kingpin, so cant benchmark it.

But I agree that the GTX 1080 is a GTX 980 on steroids. If Nvidia would allow Maxwell to overclock above the voltage thresholds (without going LN2), if I can match my GTX 980 Ti clock-for-clock with the GTX 1080, u might still see a 15-20% lead of GTX 980 Ti against GTX 1080.

But this looks promising for the GP100 chips, the Pascal Titan and the GTX 1080 Ti. Hopefully AMD can come out with something that will force Nvidia to either drop pricings on the higher-end offerings or face mass defections to OpenCL computing.

 

[Gattberserk]

So effectively Pascal bring nothing new to the table except die shrink.


All this faster speed and lower consumption has been mentioned by jumping to 16nm from 28nm. TSMC claim a healthy 40% speed boost (Which is the same jump from Tesla M40 to P100).



Nothing new. Nvidia is selling its Maxwell with just the benefit of 16nm.


DP unit is added in but, we know its going to be disabled anyway.

 

[Vincym]

The price of the card better reflect on the small improvement GTX1080 brings...

 

[Tommorrow]

So effectively Pascal bring nothing new to the table except die shrink.

All this faster speed and lower consumption has been mentioned by jumping to 16nm from 28nm. TSMC claim a healthy 40% speed boost (Which is the same jump from Tesla M40 to P100).

Nothing new. Nvidia is selling its Maxwell with just the benefit of 16nm.

DP unit is added in but, we know its going to be disabled anyway.

The only thing that really matters is whether we can get higher FPS in our games. Nvidia can talk all day about whatever fancy features but they're all useless if FPS remain the same.

 

[Tommorrow]

Anyway, apparently there's going to be a livestream of the May 6th event.

https://twitter.com/NVIDIAGeForce/status/728222996138168320

 

[MoneyFace =p]

Blaise Pascal's gonna jump out of his coffin to wack Huang Jen Hsun!!

 

[kawaii_boy]

Order of 10 .....

Almost expected a 15-25% improvement over previous generation GPU. I'm was hoping 2x or 200% improve over GTX970 with GTX1070.

So current GTX 970/980 still got value

 

[ondooy]

time to sell your 980Ti's....
buying 980Ti's in bulk @ 99 each...

 

[morgankell]

anyone seen any power consumption reviews?
Im suspicious that we will end up with a situation

a) at mainsteam and performance level
same performance for lower power consumption (1060 versus 970; 1070 versus 980) at minimal cost reduction (16 nm not cheap despite die shrink)

b) at high-end performance
marginal performance increase at similar cost (1080 versus 980TI)... same power consumption

c) no ultra performance parts (GP100) for consumer until AMD coughs out Vega...

quite sad for a process-shrink unfortunately...

 

[darkarn]

Blaise Pascal's gonna jump out of his coffin to wack Huang Jen Hsun!!

Or someone will go cosplay as Blaise Pascal for that event...

Ok wait, what am I talking about again lolol

 

[wwenze]

Wait... people still believe TSMC 16nm is a die shrink?

http://pc.watch.impress.co.jp/img/pcw/docs/670/675/html/6.jpg.html

 

[White dawn]

When 980 ti price drop?

Sent from Pillars of Creation using GAGT

 

[Entry Level PC]

tan gu gu for dx12 gaming benchmark.

 

[Vulpix]

Looks pretty good if you ask me.

 

[FinalTidus]

As expected very disappointing result.

Barely on par or slightly faster than a GTX 980Ti.

But not unexpected considering that the GTX 980 was barely faster than the GTX 780Ti.

Lets hope it's just a driver issue and that driver optimization will kick up the performance a notch or 2.

Meanwhile my GTX 970 will have a longer than expected lifespan. Provided it doesn't KO on me.