GTX 1080 Founder's Edition Only Slightly Behind AIB GPU's

AngryGoldfish

Old N Gold
I noticed something interesting that I thought I'd share. Apologies if it's been discussed elsewhere already.

It seems that the difference in performance between the Founders Edition and the new AIB cards is actually smaller than Maxwell; it's only around 1-3% from what I can see. I've looked at a few reviews so far from Guru3D, KitGuru, and of course OC3D and the FPS gains are surprisingly small.

So seemingly what you are getting with ASUS, MSI, EVGA, etc is better looks (subjective) and superior thermals and overclocking support, which is where the extra performance comes in, because the stock overclocks aren't as impressive as they were compared to Maxwell.

To illustrate that (I won't post the images as it may not be appropriate) here is a review by Guru3D of The Witcher III at 1440p.

NVidia 980ti (reference) - 61 FPS
Gigabyte G1 Gaming 980ti - 71 FPS (10 FPS gain)

NVidia 1080 (Founders Edition) - 82 FPS
MSI GAMING X 8G 1080 - 85 FPS - (3 FPS gain)

Another comparison, this time of GTA V at 1440P from KitGuru with their ASUS Strix 1080 review:

NVidia 980ti (reference) - 72 FPS
ASUS Strix 980ti - 79 FPS (7 FPS gain)

NVidia 1080 (Founders Edition) - 96 FPS
ASUS Strix 1080 - 98 FPS - (2 FPS gain)

And this is Tom's review just posted of the ASUS Strix 1080 testing Hitman Absolution at 1440p:

NVidia 980ti (reference) - 48 FPS
ASUS Strix 980ti - 53.4 FPS - (5.4 FPS gain)

NVidia 1080 (Founders Edition) - 67.4 FPS
ASUS Strix 1080 - 68.4 FPS - (1 FPS gain)

These were just randomly picked, but the 980ti wins every time. The same can be said for the 970 and 980. The Founders Edition clearly holds its own with these AIB cards. The differences are almost negligible. This is surprising considering the complaints against the 1080 and the stock cooler. Maybe the AIB partners expect you to overclock yourself. That would be odd, but quite possible, right?
 
Guess the stock cooler is not so bad, and that GPU Boost 3.0 is really good....

Most likely, biggest gains are found in synthetic benchmarks where better coolers make a difference. But not so many enjoys benchmarks as much as games so ...

The truly interesting AIB card will truly exploit the die shrink / improvement in efficiency and be made with much simpler power and cooling componenets and therefor also lower cost and end-user price.
 
I haven't really seen a review where AIB cards aren't clocking that much higher than the FE. Could either be due to 16nm not letting it go further, or simply Pascal, or power phase/8 pin limited, or even just early batches that were rushed out and had worse yields.. Aka worse cores on the market until later on. Could be many things.

But yeah I did notice the same thing you did. Just didn't have time to make a thread about it. I'm really not impressed with Pascal. It's basically Maxwell on a better process and smaller transistors sonit just clocks higher. My friend(who owns a 970) called it Paswell. Meaning hes passing until next gen and it's basically Maxwell. I can't disagree. It's amazing how efficient they are, but it's expected with such a large advance in process and nodes.
 
Last edited:
I haven't really seen a review where AIB cards aren't clocking that much higher than the FE. Could either be due to 16nm not letting it go further, or simply Pascal, or power phase/8 pin limited, or even just early batches that were rushed out and had worse yields.. Aka worse cores on the market until later on. Could be many things.

But yeah I did notice the same thing you did. Just didn't have time to make a thread about it. I'm really not impressed with Pascal. It's basically Maxwell on a better process and smaller transistors sonit just clocks higher. My friend(who owns a 970) called it Paswell. Meaning hes passing until next gen and it's basically Maxwell. I can't disagree. It's amazing how efficient they are, but it's expected with such a large advance in process and nodes.
Hmmm, yeah, true.

Tom's ASUS Strix 1080 boosted to 2050Mhz out of the box, but couldn't overclock much more than that. What did Tom's Founder's Edition boost to out of the box? I can't find it anywhere on his review. Whilst it isn't a fair comparison, Guru3D's sample of the FE boosted to 1759Mhz. That's a huge difference between 2050Mhz and 1759Mhz for only 1-3% gains. So when GPU Boost 3.0 boosts to such a high core clock over base, the performance gains are minimal. If I'm reading that right, I'm disappointed. Maxwell brought better results than that. Still a very powerful card that can be run in SLI with a 750W PSU, but I think Maxwell was better. The 1070 isn't even as interesting as the 970 was, in my opinion.
 
It could be that since the design have been called "more CGN like" they have more in common with AMD with their architecture. AMD is the same with overclocking scaling gains not that great.
 
My FE was struggling beyond 2100 much like the Strix did.


GPUBoost3 really is very good at ragging itself.
 
It could be that since the design have been called "more CGN like" they have more in common with AMD with their architecture. AMD is the same with overclocking scaling gains not that great.
I hadn't heard that.

My FE was struggling beyond 2100 much like the Strix did.


GPUBoost3 really is very good at ragging itself.
Yeah, maybe we'll have to wait for HOF/Classified to see the 1080 open up. I'm surprised your Strix overclocked so little in relation to past generations. Obviously 2.1Ghz is a massively high frequency, but it's not translating to as high numbers, which is really where it counts I suppose.
 
Back
Top