That's kind of disappointing, my 980ti at 1600~ scored 23400~ on gpu score. Maybe memory bandwidth limitation

That's kind of disappointing, my 980ti at 1600~ scored 23400~ on gpu score. Maybe memory bandwidth limitation

It`s a reference card with the ****ty VRM and Stock Cooler.

The pipelines are just a little longer to be able to reach the higher clocks compared to the 980 series. PIII to P4 sort of deal, but not such a big increase in pipeline length. iirc PIII was ~12 long, and P4 was ~18-20 long. It may clock higher, but it will yield less performance increase than on the 980 series, on a per clock basis. Its good for the future, since clocks are basically only going to get higher, as the process matures, and especially when the transition to 10nm comes along. I'm not sure the extra power plugs on the SuperClock versions etc are actually going to be able to deliver that power, or that the GPU can use it. @Hilbert : Did the max power consumption increase a lot between the stock values, and the best OC values ?

Nice video, your English is fine.

That's kind of disappointing, my 980ti at 1600~ scored 23400~ on gpu score. Maybe memory bandwidth limitation

Yeah, or just not so magic after all..

It`s a reference card with the ****ty VRM and Stock Cooler.

I think that's about it with these chips. It can do 2 - 2.2ghz max (~50% from base 1.6k), unless you HW bios override higher voltage control or volt mod.. Also it seems base frequency is not so OC'able only 90MHz OC, rest is extended boost 3.0 thing, what up to 350mhz boost lol 🤓

It`s a reference card with the ****ty VRM and Stock Cooler.

still it's running 2k on the core.

That's kind of disappointing, my 980ti at 1600~ scored 23400~ on gpu score. Maybe memory bandwidth limitation

probably not memory bandwidth since they claim much better memory compression,(that is if they are not lieing) The scores are consistent with pascal being more of a die shrink + minor improvements, seeing as the 980ti has 10% more cuda cores and 50% more rops, and the memory bandwidth being similar , the only big change is the higher clocks on the core for the most part, assuming the rops are not a weak point and that the core design is almost the same, a 980ti at 1600mhz would be roughly equivalent to a 1080 at 1760mhz, this may be enough to explain the similar performance depending on how the boost clock fluctuates (would be cool to see a graph of clocks over time from guru3d). If the firestrike score is reflective of the 1080 throttling down below the maximum boost clocks , then this explanation is sufficent But if this firestrike score is reflective of the 1080 not throttling down and sustaining high clocks 1850mhz+ , this is bad news as it would suggest a weakness, either rop limited or changes to the cuda core design. Thats just my 2 cents

I'm curious when someone gets a hold of a custom or ever reference one, to downclock it if possible to 1500-1550MHz and pit it against a 980Ti at the same clocks. I curious to see the outcome. Just for "scientific/educational" purposes.

Small preview of new feature, which will be available to 1070/1080 owners in beta 4: http://forums.guru3d.com/showpost.php?p=5280644&postcount=137

Or you could skip the 1080 and wait for big pascal and vega while enjoying a 980ti at 400-450bux :infinity:

Nice to hear your voice Hilbert, looking forward to OCing my future Pascal card with the updated MSI AB, which I couldn't live without anymore. 🙂

Nice to hear your voice Hilbert, looking forward to OCing my future Pascal card with the updated MSI AB, which I couldn't live without anymore. 🙂

Amen!! to that Msi AB is an awesome piece of software.

Something is not adding up. Something is not adding up for me. I have EVGA 980Ti, ASUS GTX 1080 FE and ASUS R9 295X2 and the scores published here looks mile away from what my scores were in 3d mark firestrike. the r9 295x2 beats the gtx 1080 by quite a lot the gtx 1080 beats 980Ti by a bit. The published GTX 1080 score is way higher than in real life and the R9 295x2 way lower than in RL. Can somebody offer reasonable explanation? https://fbcdn-sphotos-e-a.akamaihd.net/hphotos-ak-xpf1/v/t1.0-9/13346978_1024297997655562_1064258689889053595_n.jpg?oh=f069136f54e37c51e7a71a565dba3f42&oe=57CAA3E8&__gda__=1473073623_1a9adfdf6474e1ca8a624dc78d300501