Geforce GTX 980 Ti vs GTX 1070 vs GTX 1080 Overclocked Performance

[Qwertilot]

Doubt AMD were a big factor actually - one upmanship is fun but not much direct competition right now of course.

Their timetable simply works much better with a launch round now. They seem to want to stagger things, so there's the 1070/80, 1050/60, mobile pascal etc all to launch to be in place in time to go into lots of systems selling round Xmas.

Can't be impossible that they're going to try and sneak the titan in just in time for Xmas.

It looks a very relaxed timetable but slip everything a month and it'd start to get a little tough.

 

[Mopetar]

I think a more respectable decision would have been to postpone launch an additional month and build up additional stock to satisfy demand at a lower price.

Why do that when you can release early, get more publicity for being first, and rake in extra money while doing so?

Even if AMD had come out with much better base-level Polaris performance than they showed off, Nvidia still wouldn't lose anything for moving first. It's not like you can go out and buy a competing new generation AMD GPU if the 1070/80 are out of stock and most people will just shrug their shoulders and wait with the only people who spend much energy caring beyond that being those who are firmly embedded in one camp or the other.

The people who will buy early for higher prices are either NV fans that don't care about a so called tax or are GPU enthusiasts who will pay for premium performance regardless of company or price. Everything NV is doing is good business sense.

 

[moonbogg]

I think the performance is more impressive than the GPU itself. Sure, I'm glad to see the new process shrink and they clearly had to do some redesign, but the numbers of pretty much everything in the chip design except clock speeds are lower than the 980Ti. That isn't going to last through the lifetime of of the 14/16nm process. We will probably surpass the 980Ti specs with 50% even though starting _lower_ with the initial 1080.

I hope to GAWD its at least 60%. It damn well better be.

 

[BlueWeasel]

Just for my own curiosity, I took the overclocked results at 1080p and 1440p and compared the new cards to the 980Ti. These are the same benchmarks that Sweepr linked to, just reformatted to show relative percentages of the FPS.

The green icons represent a greater than 5% improvement, yellow dash is 5% or less change, and the red arrow represents a decrease.

 

[Sweepr]

Just for my own curiosity, I took the overclocked results at 1080p and 1440p and compared the new cards to the 980Ti.

The green icons represent a greater than 5% improvement, yellow dash is 5% or less change, and the red arrow represents a decrease.

Nice job! Added to first post.

About the claims that Geforce GTX 980 Ti was throttling in the OP article, looks like it wasn't:

Testing for the maximum clock speed consisted of looping Unigine Heaven 4.0 for thirty minutes each to see where the clock speeds failed when pushed. If the clock speed adjustment failed, then the clock speeds and tests were re-run until they passed a full hour of testing.

Temperature testing will be accomplished by loading the video card to 100% using Unigine Heaven Benchmark Version 4.0, with MSI Afterburner overclocking utility for temperature monitoring. I will be using a resolution of 1920 x 1080 using 8x AA and a five-run sequence to run the test, ensuring that the maximum thermal threshold is reached. The fan speed will be left in the control of the driver package and video card BIOS for the stock load test, with the fan moved to 100% to see the best possible cooling scenario for the overclocked load test. The idle test will involve a twenty-minute cooldown, with the fan speeds left on automatic in the stock speed testing and bumped up to 100% when running overclocked.

 

[BlueWeasel]

Ashes of the Singularity and 3D Mark results have been added below. I have no idea what resolution was being used for 3DM (as the article does not say), so those results are under 1920x1080.

 

[2blzd]

1080 FE is a premium card (+$100), using premium components including vapor chamber cooling, hence should not be compared with reference 980Ti.

See how easily this kind of argument can be turned around?

PS: wouldn't it be more appropriate to to keep all cards from throttling, since this is an oc test?

FE's are not premium cards and are not using premium components. The PCBs on the FE's are junk compared to what the AIBs are providing. THe only thing you can pass as 'premium' on the FE is the price, and maybe the vapor chamber, ohhhhhh!

 

[CakeMonster]

Great work! I feel this is giving a better (but ofc far from perfect) idea of the chip design than the non-OC numbers since it pushes both chips equally. My conclusion is that 16/14nm will improve a LOT in the coming years, and that NV can easily pull a ~30% improvement "high-end" release each year for the next 4 years.

 

[RussianSensation]

Great work! I feel this is giving a better (but ofc far from perfect) idea of the chip design than the non-OC numbers since it pushes both chips equally. My conclusion is that 16/14nm will improve a LOT in the coming years, and that NV can easily pull a ~30% improvement "high-end" release each year for the next 4 years.

Of course they can. GV104 can grow to 400-450mm2, use 14Gbps GDDR5X 384-bit, or even adopt HBM2 by 2018. If they release a GP102 ~ 450-475mm2 GP102 384-bit with GDDR5X, then a 600-610mm2 1-1.5TB/sec HBM2 Big Volta with a new architecture can increase performance from 1080 -> 2080 by 50-70%, and the same for GP102->GV100/102. That means ~ 30-33% per annum increases will continue in 2017, 2018 and 2019. Then, they'll try to move to a new node in 2020.

What matters the most are monitor (4K 120-144Hz HDR, 5K HDR, 1440p 200Hz HDR) and next gen games. Already now a 2016 1080 is worthless for 1080p 60Hz but we still have a solid 3-4 years left in the PS4/XB1 cycle. Something needs to change or GPU hardware is advancing far faster than games can keep up while current gen consoles are A LOT further behind modern GPUs than PS360 consoles were just 2.5 years out. I don't know what the resolution is but I would have much rather preferred for Sony and MS to release consoles 6-10X more powerful than PS4 by 2019 than these rumoured stop-gap Neo and Scorpio consoles that won't magically prevent 90-100+ million PS4/XB1/NX consoles (in 2017) from enticing developers to target anemic level or CPU/GPU hardware. I smell horribly optimized PC ports because soon the brute force of GPUs will overcome shoddy console to PC port programming.

 

[msbytes]

FE's are not premium cards and are not using premium components. The PCBs on the FE's are junk compared to what the AIBs are providing. THe only thing you can pass as 'premium' on the FE is the price, and maybe the vapor chamber, ohhhhhh!

I don't agree with the PCBs being junk. Evga is using ref PCBs below the FTW card. They just added there own cooler on them.

 

[msbytes]

Just for my own curiosity, I took the overclocked results at 1080p and 1440p and compared the new cards to the 980Ti. These are the same benchmarks that Sweepr linked to, just reformatted to show relative percentages of the FPS.

The green icons represent a greater than 5% improvement, yellow dash is 5% or less change, and the red arrow represents a decrease.

Great Job!!!! Really helpful.

 

[Belkov]

Nice comparison.

 

[dark zero]

Personally I wouldn't buy it, seeing how custom models like ASUS's Strix perform in comparison.

Agreed with you this time. To not to say that the golden chips are not out there.

 

[tviceman]

Thanks for that graph msbytes!

 

[BlueWeasel]

Thanks for that graph msbytes!

You're welcome.

 

[JDG1980]

Do the Founders Edition cards use binned chips? Many people assumed that AIB cards would OC better because of improved cooling, and they may indeed be better able to hold an overclock without throttling. However, TPU's reviews of two 1080 AIB cards indicate they actually don't OC as high as the FE.

http://www.techpowerup.com/reviews/Gigabyte/GTX_1080_G1_Gaming/27.html
2050 MHz maximum OC for both Gigabyte and MSI cards. Founders Edition hits 2114 MHz.

 

[Kenmitch]

Do the Founders Edition cards use binned chips? Many people assumed that AIB cards would OC better because of improved cooling, and they may indeed be better able to hold an overclock without throttling. However, TPU's reviews of two 1080 AIB cards indicate they actually don't OC as high as the FE.

http://www.techpowerup.com/reviews/Gigabyte/GTX_1080_G1_Gaming/27.html
2050 MHz maximum OC for both Gigabyte and MSI cards. Founders Edition hits 2114 MHz.

Silicone lottery plays the largest role in OC's.

Think it was stated that the FE cards don't get binned chips.

Was his FE a retail purchase or was it a review sample? Guess it would be possible to bin the FE cards. Not really sure if it would be worth the hassle in the end. Not like a few MHz is going to make or break the overall performance.

 

[Tempered81]

You're welcome.

haha

 

[tviceman]

You're welcome.

haha

Whoops! Sorry Blue Weasel. Thank you. Graphs are great. Easy to read and helpful.