Upgraded GTX 970 to GTX 1080 (FE) - Initial Thoughts

14 replies
3602 views

Yesterday I received my EVGA GTX 1080 FE.
It was an upgrade from the EVGA GTX 970 (ACX 1.0) I had been using while I waited for Pascal.

Just prior to removing the 970 I decided to run Firemark Ultra, receiving a measly 2096 marks. I was hitting ~12 FPS in many of the scenes. Bleh.
The makers of 3DMark actually say somewhere in the 2500 range is acceptable for VR so I was below their recommendation.
Despite my woefully inadequate computer, I met the min specs according to Oculus and had enjoyed my time with PC, AC, ED, and most recently Chronos. It was good enough.

But now I have a GTX 1080.

My first impression is that everything just feels a little... smoother. Just like how 1.3 seemed to make everything crisper in a way that was difficult to nail down, the 1080 is just all around better than the 970. It FEELS faster and smoother.

However, to ensure this wasn't just confirmation bias I threw Firemark Ultra back on.
Over 5000. Frames in the 26-27 range. I'm happy. It made a pretty massive difference.

The rest of my system is an MSI m7 hosting an i5 6600k and 16gigs of g.skill 3200. Win 10 on a Samsung SSD, with a 1TB hybrid drive for storage.

It was absolutely worth the wait. I cranked the settings in ED up to max and nary a stutter. In the next few days I'll play around with debug SS and see what sort of difference I notice. Same with Project Cars, although I was just doing solo lapping at Infineon so I can't comment on performance with a dozen cars on the circuit with me.

A

AntDX3162

Up and Coming

Synthetic benchmarks don't matter. It's all about the real-world results. Overclockers like me were benching that stuff since Pentium 4 on phase change/peltier cooling.

What a bench that all CPU overclockers used was SuperPi 1MB but that died out. The thing is a 2600K OCed at 5GHz would set lower times than a 6700K at 4.5GHz but would get smoked on some real-world applications due to the optimizations that exist in hardware with the later generation.

Some games have more improvements than others but I think the most important improvements than frame rate are the frame draw response times, multiprojection to have objects not stretched when they are near the sides of the screen, and the electric bill power draw.

I got tired of stupid benching every single time I put something new in and installing new drivers. The best way to get a high score is to disable unnecessary services and turn off apps that aren't related to the bench. Also making the video card settings all low and force disabling quality settings to high performance. Force AF and AA off. That isn't a real default bench but it will still pass the test but the score will be higher. This had to be done every single time there was a change of hardware, clock speed, and/or memory timing.

The novices bench and think it matters to be honest. My friend got a 980 Ti from some slow PC, he listened to me (last year), and he benches all the time but not w/ the advanced how to get your score higher type technique.

facebook.com/AntDX316

A

AntDX3162

Up and Coming

Oh and you can get an ultra high cinebench score with 2x xeons of way back having like 20-core than a 6700K but get lower or equal rendering performance in Adobe after effects.

facebook.com/AntDX316

C

cybereality

Inspiring

Nice. Hope you're happy with the upgrade.

I'm not really concerned about the 3DMark score anymore. I had 3 GTX 980's before, and updated to 2 GTX 1080's. However, the 3DMark score was nearly the same. Which was a slight disappointment, until I actually tried some games and saw a major improvement.

Z

ZoomieAuthor

Up and Coming

I guess the bench was significant to me because I changed LITERALLY nothing else.
If it was just a score increase from 2000->5000 it wouldn't mean as much, but improving from 12FPS to 26FPS under some very demanding graphics was a significant jump. Likewise, I was able to firewall my sliders to the right on ED and PC without impacting performance.

I agree that benchmarks aren't the be-all and end-all of comparisons, but I don't think it's fair to entirely discount them like you suggest. In my case, I think the raw score and FPS improvements speak for themselves. I didn't resort to any trickery to improve my secondary score which (IMO) would invalidate the testing method anyhow.

Any sufficiently advanced technology is indistinguishable from magic. - Arthur C Clarke

nalex66

Honored Contributor

Nice. Any heat issues? Are you running with the stock fan speed settings or taking @Shadowmask72's recommendation and applying a more aggressive fan speed curve?

I've been really happy with the EVGA GTX 1080 SC that I scored last week. Great performance, nice and quiet, and dynamically boosts well above the rated boost clock (I'm seeing boost speeds in the 1930's).

DK2, CV1, Go, Quest, Quest 2, Quest 3. Try my game: Cyclops Island Demo

Z

ZoomieAuthor

Up and Coming

No heat issues, but I haven't really had much play time this week.
I'll give the card a good workout this weekend and report back.
Still not sure whether I'll use MSI or native EVGA controls. My mobo is an MSI m7 and I'm not terribly impressed by their management tools so far.

On a related note about SMP and LMP -

"And while Nvidia says SMP can yield “a 3x VR graphics performance improvement over previous generation GPUs”, the gains don’t matter if the tech isn’t built into VR games. So far we’ve only seen SMP built into a few tech demos, but it’s soon to get a potentially huge boost thanks to integration with Unreal Engine and Unity, the two biggest game engines and the primary tools used to create VR content.

Nvidia hasn’t said yet exactly when we’ll see SMP built into these two popular game engines, but the company notes that there’s already VR games being developed with the rendering tech, including Pool Nation VR, Everest VR, Obduction, Adrift, and Raw Data."

SOURCE

Any sufficiently advanced technology is indistinguishable from magic. - Arthur C Clarke

nalex66

Honored Contributor

I just read the the Nvidia funhouse demo is being released free on Steam. If that can be run without tracked controllers, then it'll probably be the first opportunity to test SMP in VR on our GTX 1080s.

DK2, CV1, Go, Quest, Quest 2, Quest 3. Try my game: Cyclops Island Demo

Z

ZoomieAuthor

Up and Coming

It's also being implemented in several games (listed at the end of my previous post).
Pool Nation VR, Everest VR, Obduction, Adrift, and Raw Data. Not sure why Pool Nation needs SMP, but I'll take the performance gain regardless :)

This is big news because unlike VR SLI which was supposed to be implemented with Maxwell cards, this means SMP and LMP are already being integrated into software. It's not just vapourware. This means Pascal series card owners will be able to enjoy optimization beyond just the hardware improvements from 10xx and DX12.

Any sufficiently advanced technology is indistinguishable from magic. - Arthur C Clarke