NVIDIA GeForce GTX 1080 Put Through 3DMark

[HumanSmoke]

What the videocardz piece says as well is that the LN2 cooled 980ti FM results at similar clocks destroy the 1080. So perhaps Pascal isn't very good until we see daddy Pascal. Same as 980 Maxwell wasn't really much better than a 780ti - this is the same all over again. It's only because of it's high clock speed that it's improving...

That is at least something. A theoretical clock-for-clock "loss" to GM200 doesn't mean too much in the real world if 99.999% of 980Ti/Titan X users are capped at <1500-1550MHz. Like the previous generation, big die to small die isn't the natural upgrade path, but it does take the shine off the previous kings of the hill.
I think more interest (at least for me) will be
ROP count - although Maxwell wasn't particularly ROP limited- just as an indicator of what we might see with GP102/GP100
TMU throughput - something that has limited Maxwell to a degree. It still wouldn't surprise me to see the 980Ti/Titan X shade GP104 in texture fill and narrow any gap to the new Pascal at high resolutions/DSR/FSAA.
Board power limits and overclock headroom.

the sad part is nbadia is still going to sell a lot of these "best" gpus to their religious flock...
"hey look im part of that orderof10tacles secrect club stuff! is awesum! i pity u lame poor outsiders"

Because that is solely a Nvidia trait obviously...

Demand the best.......get a neutered Tonga....no hyperbole there then

 

[truth teller]

I am pretty sure Pascal has some architectural differences over Maxwell as well as being more efficient per clock.

does a pig with makeup stops being a pig?

 

[TissueBox]

Really surprised at the community here. The 1080 replaces the 980, and it is on average ~34% faster (assuming GTX 980 Ti @ 1190MHz performance levels). Similar to the 980 reveal, which was about 30% better than the 780 and just barely faster than the 780 Ti (5-10%).

That's pretty solid. 980 Ti owners should be looking at the 1080 Ti or Pascal Titan as their replacement.

I'll probably opt for SLI 1080s as it shifts down a price bracket when the 1080 Ti releases.

 

[idx]

ROPs only matter when you increase the amount of raserization; higher resolution, higher framerate, higher AA, higher resolution temporary framebuffers, etc. More Gflop/s increases the throughput and does of course increase ROP load if there are no other bottlenecks, but then someone first would need to increase the throughput. AMD is currently struggling a lot with their architecture, despite having up to 50% more GFlop/s than Nvidia, so the other bottlenecks will have to be addressed first.


As stated above, not true at all. Higher resolution increases the load on the ROPs, so that's irrelevant in this case. The problem with Fury X is the scheduler which is unable to feed the massive amount of cores.

I am sorry, I don't mean to look like arguing. but It seems that you dont understand my point.

The processing of pixel shaders and any other kind of shaders is already done by the time the data reaches the rasterisation part. And there where the rasterization of vector graphics happens AT the gpu ROPs. And thats why it is so important to have more ROPs in case if we have a big number of SPs and TMUs on the gpu.

Again I really apologize if I sounded like arguing. I am really not , I'm just sharing an info (talking from experience).

 

[Legacy-ZA]

So 2 cards maybe that actually stand for some better performance we actually need (and might I add for quite a few generations now) and those will be the highest priced cards?

Yeah good stuff.... anyone giving a damn about actual damn progress and actually something worth being called "next gen" would want a GTX1060 to be at the level of a GTX980Ti now and everything above it should destroy it.
But nope, guess another "Generation" of mediocre upgrades unless you pay 700 dollars, great, good stuff.

I mean http://www.techpowerup.com/reviews/Gigabyte/GTX_980_Ti_XtremeGaming/7.html

a current top of the line card, that this new pretty much close to top of the line card gets similar performance to, gets a mere 43 fps on 2560x1440....
You are buying a card that is incapable from the start.
Lets not play around people this is the "80" not the 70 not the 60 but the 80 line of cards, and this is what we can expect from a future card, not even being able to max a CURRENT game, let alone games in the near future?

I dont see the point.

I am so glad somebody else gets it.

 

[xvi]

I'd be curious to see what kind of mobile offerings this generation spawns. Performance per watt, please.

 

[N3M3515]

Yeah good stuff.... anyone giving a damn about actual damn progress and actually something worth being called "next gen" would want a GTX1060 to be at the level of a GTX980Ti now and everything above it should destroy it.
But nope, guess another "Generation" of mediocre upgrades unless you pay 700 dollars, great, good stuff.

This. MEDIOCRE upgrade. And yes, GTX1060 should be above every single gpu card.

14nm was supposed to a HUGE leap though, the 1080 non-ti should have smashed a 980 ti... 1080 ti 15% increase max if we lucky, it is sad honestly. all the hype saying 50% increase cause finally going to a new die shrink, same hype they said about skylake, and my 4.8ghz 2500k owns stock skylake in games still, lulz..

This.

The 980 is still a high end card. No ifs or buts. It's €500. If the 1080 is about as much that too is a high end card.

This.

the 60 is low end gaming
the 70 is mid end gaming
the 80 is high end gaming
the Ti is a boosted version

Titan is overpriced
Thats how the nomenclature worked,
the GTX960Ti would not suddenly better then the 980 purely because of the Ti thing

This.

If it does not destroy the GTX980Ti its a fail in my book

Exactly!, this is more of an incremental upgrade, as if they where still at 28nm

Really surprised at the community here. The 1080 replaces the 980, and it is on average ~34% faster (assuming GTX 980 Ti @ 1190MHz performance levels). Similar to the 980 reveal, which was about 30% better than the 780 and just barely faster than the 780 Ti (5-10%).

That's pretty solid. 980 Ti owners should be looking at the 1080 Ti or Pascal Titan as their replacement.

I'll probably opt for SLI 1080s as it shifts down a price bracket when the 1080 Ti releases.

Pretty solid?, seriously?
So, you expect that a vcard on 14nm performs 10% more than one on 28nm? wtf???
The 1060 should be on par with 980Ti!, that is the way it always is when there is a node change and generational change, hell, they even skipped 20nm!!!!!!

Where is the huge jump in performance from 14nm? could anybody talk about that??
From my own perspective AMD and NVIDIA are screwing with us.

 

[efikkan]

I am sorry, I don't mean to look like arguing. but It seems that you dont understand my point.

That's OK

I completely understood what you said. Your first point was correct (when everything else scales you need more ROPs as well), your second was not (the one about Fiji being limited by ROPs in 4K).

 

[efikkan]

So, you expect that a vcard on 14nm performs 10% more than one on 28nm? wtf???
The 1060 should be on par with 980Ti!, that is the way it always is when there is a node change and generational change, hell, they even skipped 20nm!!!!!!

Where is the huge jump in performance from 14nm? could anybody talk about that??
From my own perspective AMD and NVIDIA are screwing with us.

Yes we can!
Well, I think your expectations are too high based on PR. What TSMC calls "16nm" FinFET and Samsung calls "14nm" FinFET would be classified as "20nm" FinFET if it were made by Intel...

 

[G33k2Fr34k]

The 980TI with an i7 4770K gets around 12,000 points in Fire Sttike Extreme. Correct me if I'm wrong, but isn't the GTX1080 slower with a score of ~9000 running on an i7 3770K?

 

[TissueBox]

Pretty solid?, seriously?
So, you expect that a vcard on 14nm performs 10% more than one on 28nm? wtf???
The 1060 should be on par with 980Ti!, that is the way it always is when there is a node change and generational change, hell, they even skipped 20nm!!!!!!

Where is the huge jump in performance from 14nm? could anybody talk about that??
From my own perspective AMD and NVIDIA are screwing with us.

It roughly follows the same increase in performance per generation - Kepler to Kepler refresh (20%-30%), and Kepler refresh to Maxwell (20-40%). As I recall, the switch to 16nm offers a 40% improvement in performance compared to 28nm, OR use 50% less power.

It performs ~34% better than the similar card on their previous generation, so yes, I would say it's solid (Or perhaps I should use expected) and in line with what they've been doing.

 

[yogurt_21]

So ...
1. It pretty much matches a factory clocked 980 Ti performance wise
2. The commas are missing from the memory, core clock, and memory clock compared with regular entries
http://www.3dmark.com/fs/6704407
http://www.3dmark.com/fs/6264721
3. There's a 1290 MHZ difference between the 2 core clocks
4. The memory bus speed is over twice that of a 980 ti


I call BS. These are 980 Ti's with edited graphics card specs on the entry.

 

[efikkan]

It roughly follows the same increase in performance per generation - Kepler to Kepler refresh (20%-30%), and Kepler refresh to Maxwell (20-40%). As I recall, the switch to 16nm offers a 40% improvement in performance compared to 28nm, OR use 50% less power.

It performs ~35% better than the similar card on their previous generation, so yes, I would say it's solid (Or perhaps I should use expected) and in line what they've been doing.

Yes, and accounting for the significant performance increase in the Pascal archtecture a ~400mm² should be able to achieve up to 70% or so over GM204, but since this is a new node GP104 is only ~333mm². So ~35% performance gain over GTX 980 is exactly to be expected.

GV100(2018) is scheduled to be made on 10nm FinFET, but it wouldn't surprise me if the successor of GP104 will be still made on 16nm FinFET, but with a slightly bigger chip ~400mm², and with the architectural improvements we'll get another 30% once again...</speculation>

 

[PP Mguire]

oh right 980 midrange, add a "T" and a "i" and its top end.
Makes sense.....

GM204 midrange, GM200 high end. P104 midrange, P100 next year will be high end. They've done the same release strategy since the 680. I could bet money on them doing the same with Volta in 2018.

 

[truth teller]

1. It pretty much matches a factory clocked 980 Ti performance wise

as expected, i hope this doesnt come as a surprise to you

4. The memory bus speed is over twice that of a 980 ti

again, as expected for gddr5x, not that its gonna do any better than the previous gddr5
does you car go faster just because you are in a racing circuit?

I call BS.

dont, you will be sad once the cards are launched and definitive reviews are posted

 

[TheHunter]

Numbers mean nothing to me if I don't have anything to compare them with. Can someone also post numbers for current gen cards using similar system?

You can search the FM database........................

Anyway, its about spot on/a bit faster than a factory overclocked 980Ti and 5820K.

This is my factory OC 980Ti @ ~ 1418-1443MHz boost, vram stock 1806Mhz (346gb/s)




so only 300points more, or 1-2fps difference.. Right.
And the way that GTX1080 @ 1800mhz is clocked I bet it won't have much room left, probably 2000MHz limit.



3dmark firestrike extreme, same 1418MHz, vram stock 1805Mhz

Score 8 679 with NVIDIA GeForce GTX 980 Ti(1x) and Intel Core i7-4770K
Graphics Score
9 340
Physics Score
13 411
Combined Score
4 214

http://www.3dmark.com/fs/8357945

Ok here its 600-700points difference, but that's minimal..



Imo overall its nothing special, I suspected it will be like that.. Slightly Oc'ed 980ti perf. at best.

I think sometimes it will be slower when that extra 16ROPS and 20-30GB/s count.

Personally I dont regret anything, bought this 980Ti 2 months ago for 420€, sold my old 780GTX for 250€ (originally payed 320€) and its a bomb. For me its only big Pascal or next-nextgen Volta and AMD Navi that's really interesting. Think I'll just skip Pascal all together.

 

[lemkeant]

I'm pretty surprised regarding the arguments here. There's no way Nvidia or AMD are going to blow their load on the 1st gen 14/16nm cards. Why? As we saw with 28nm, this process is going to need to last at least a few years (with Moore's law being mostly dead).

The next set of cards will be good enough to beat the previous cards but thats it. The next series from there will repeat the same trend and a few years from now, we'll have the 1380ti and R9 590X still built on the same 14/16 nm process, just larger and more power hungry

 

[cedrac18]

Hmmm is no one expecting better performance when the drivers mature? Not sure why is everyone is flipping out already when these results are not even from proper review.

 

[ensabrenoir]

......same pro/con arguments....just a new year........its all just business.....Nvidia like intel like everyone else should be, is out to make consistent profit. year after year, model after model....business 101.

 

[truth teller]

Hmmm is no one expecting better performance when the drivers mature? Not sure why is everyone is flipping out already when these results are not even from proper review.

if these "new" cards could have their performance increased by driver tweaks, then that meant that maxwell cards can also have that the performance increase. drivers for that generation are pretty mature by now, it will _not_ happen
why are people still thinking this is an entirely new gpu? jesus christ folks, you have numbers right in front of you...

 

[dj-electric]

GTX 1070 (350$) slightly stronger than a GTX 970 (330$)? wtf?
GTX 960 (200$) slightly stronger than a GTX 760 (200$)? wtf?
HD 6870 (239$) slightly stronger than an HD 5850 (229$)? wtf?
HD 5850 (260$) slightly stronger than an HD 4890 (250$)? wtf?
HD 4850 (200$) slightly stronger than an 8800 GT (160$)? wtf?

C'mon guys, we've been there many many times. its a small step in performance with another in efficiency.
These are all examples of replacements and competitors with an MSRP higher than their replacement - that's how it starts. sometimes it doesn't, but mostly it does.

 

[Caring1]

again, as expected for gddr5x, not that its gonna do any better than the previous gddr5
does you car go faster just because you are in a racing circuit?

Fuck yeah! No silly restrictions to slow me down.

 

[Prima.Vera]

To be honest, after all the hype and giggles regarding the new generation, I was hoping that at least the 1070 to be ~5% faster than the 980Ti, just like the previous gen was with the pre-previous gen. Seeing that the 1080 barely beats a stock 980Ti, is so disappointing...

 

[rtwjunkie]

The 980 is still a high end card. No ifs or buts. It's €500. If the 1080 is about as much that too is a high end card.

Um no....it's an upper-midrange, using the GM204 chip which happens to NOT be Nvidia's high end Maxwell chip.

Cost has nothing to do with whether a GPU is high-end, and everything to do with it just being expensive.

 

[Prima.Vera]

Um no....it's an upper-midrange, using the GM204 chip which happens to NOT be Nvidia's high end Maxwell chip.

I always thought that from nVidia, upper-midrange means *70 series, lower-midrange *60 series, with the <midrange> gap being filled by AMD...