Nvidia GTX 1080 Gets Stripped, Benched and Overclocked

The flood gates have opened. The deluge has begun. Tech reviewers across the globe can breathe easy as the NDA gets lifted and the green light to publish finally begins to glow. The big news is that the new flagship card from Nvidia is without doubt going to cause a seismic shift in performance metrics, pushing way ahead of today’s Titan X and GTX 980 Ti cards while simultaneously being more efficient overall.

The new reference card, or Founders Edition GTX 1080 landed in the hands of ccokeman from OverclockersClub who has done a solid job of breaking down all the details for us. Like any good review we have a look at the PCB and the cooler design, plus benchmarks and a first look at how the card overclocks. Ccokeman managed to push the card to 2,050MHz using the reference cooler, a nice push of 18% over stock clocks of 1,733MHz.

“The best combination of results that I was able to run all of the game tests with came in at a GPU Boost 3.0 clock speed of 2050MHz on the Pascal core. In some game tests 2076MHz to 2088MHz was perfectly stable, but not in all games. This gives the end user about an extra 315MHz of headroom to play with that does indeed drive the performance to another level.”

“Micron's GDDR5X memory also plays well when you start clocking it up. I was able to push the 8GB of GDDR5X memory to a data rate of 11Gbps. This final step again helps with memory bandwidth to get you that last bit of performance out of the GTX 1080. Having good architecture and a great power delivery system help improve performance across the board by allowing stable overclocking at high levels.”

“The last piece of the puzzle is how well the Founders Edition cooling works when the screws are put to it. Keeping the card somewhere south of 70 °C goes a long way towards helping deliver added performance from overclocking.”

Check out the full and detailed review over on OverclockersClub.


11

Taiwan sdougal says:

Exciting times....

Netherlands rsnubje says:

And yet I wonder why no-one has yet uploaded any results, even with the FE version you could make some nice results.

Germany der8auer says:

rsnubje said: And yet I wonder why no-one has yet uploaded any results, even with the FE version you could make some nice results.


Because Overclocking is dead.

j/k was wondering the same.

Indonesia speed.fastest says:

Yep, still wondering, no one get sample here?

Belgium Massman says:

Still nothing on Futuremark either ... no idea

TaPaKaH says:

could it suck real bad? or not be compatible?

Germany der8auer says:

buildzoid said: I'm worried about there being a really bad cold bug. Clock for clock the 980Ti is faster. Also the stock PCB is garbage those FETs are rated to do 50A at 25C. So at realistic operating temps they might do about 35A and you see why the power limit on the card is so low.


yea I would also wait for custom PCBs. Stock PCB seems pretty bad indeed.

Czech Republic buildzoid says:

I'm worried about there being a really bad cold bug. Clock for clock the 980Ti is faster. Also the stock PCB is garbage those FETs are rated to do 50A at 25C. So at realistic operating temps they might do about 35A and you see why the power limit on the card is so low.

Indonesia Lucky_n00b says:

buildzoid said: I'm worried about there being a really bad cold bug. Clock for clock the 980Ti is faster. Also the stock PCB is garbage those FETs are rated to do 50A at 25C. So at realistic operating temps they might do about 35A and you see why the power limit on the card is so low.


I don't know about 'cold bug', but yeah I'm curious about temperature scaling on these 16nm FinFET.
About efficiency per clock, maybe that's because the 1080 is using less shader units than 980Ti? (1080 has 2560, 980Ti has 2816), but on the last gen, maxwell has better performance even using less shaders IIRC. So safe to say maxwell and Pascal have roughly the same performance on same shader unit count? (Pascal being faster because of the clocks).

And reference model PCB and cooler is just bad as previous gen. I wonder why NV choose to make it 'special' by naming it 'Founder's Edition', and adding 100 USD to it..

Romania Alex@ro says:

I hope to see 2500 on 'hwbot water" and maybe 2800-3000 on ln2.

Germany QAI says:

Lucky_n00b said: And reference model PCB and cooler is just bad as previous gen. I wonder why NV choose to make it 'special' by naming it 'Founder's Edition', and adding 100 USD to it..


Because this way they can sell their first few cards for premium price and start the hype train instead of doing a real paper launch due to low 16nm production :D

Please log in or register to comment.

Leave a Reply: (BBCODE allowed: [B], [QUOTE], [I], [URL], [IMG],...)