[Guru3d]Call of Duty: Infinite Warfare PC graphics benchmark review

[Glo.]

That sucks what memory bus does the GTX 1080 and Titan X have ?

GTX1080 256 but GDDR5 with 320 GB/s, and Titan X has 384 bit of GDDR5, with I think 448 GB/s(?) of bandwidth.

 

[littleg]

256 bit memory bus.

Just hoyed a few numbers around and the 1080 and 1070 both scale to around 60% fps at 4k vs 1080 whereas the 980ti and AMD cards are in the 65 - 67% range.

 

[Head1985]

The GTX 1070 is bandwidth limited for sure, with only 256 GB/s. The GTX 1080 has GDDR5X, which gives it 25% more bandwidth for reference clocked parts.

Higher resolution also increases pressure on the shaders, as there are more pixels to shade.

+3GPC and 1080 have 33% more SP.Bigger resolution=bigger gap.
1070 is pure 1080p card?

From computerbase GAP 1070 vs 1080:
1080P 20%
1440P 26%
4k 30%

 

[Carfax83]

+3GPC and 1080 have 33% more SP.Bigger resolution=bigger gap.
1070 is pure 1080p card?

I wouldn't say the 1070 is a pure 1080p card, as it really depends on the game. But overall, it does better at 1080p than it does at 1440p for sure. GTX 1080 excels at 1440p, and has fair performance at 4K.

 

[pcslookout]

I wouldn't be all happy yet.

The GTX 1080 is a pure 1440P card.

 

[Abwx]

Benchmarks from game.gpu, here . Very different results.

Because very different setting.

The highest image quality is in Computerbase.de test followed by Guru3D who are a step below, PCgameHardware is another level below and GameGPU is yet another level below the latter..

 

[Carfax83]

Because very different setting.

The highest image quality is in Computerbase.de test followed by Guru3D who are a step below, PCgameHardware is another level below and GameGPU is yet another level below the latter..

GameGPU says maximum quality setting according to Edge translator. PCgameshardware.de says:

"Max. detail except Point Shadows/Sun Shadows (buffering off);"

No idea what that means.

 

[frozentundra123456]

What story, i just mentioned those two resolutions because Fury X with it's 4GB of memory is beating Pascoal 8GB. Why do you feel the urge to defend NV at all costs?
Retarded arguments like these start wars here. Just stop

Well, if you consider it "retarded" and "defending nVidia" to want to look at all the data instead of just what fits one's agenda, then I plead guilty as charged. If you want to claim differences of 1 or 2 FPS, "faster" go right ahead, but I consider such differences within the margin of error, and as I said, the 5K results are barely above 30FPS, which I cant imagine anyone would consider acceptable for a fast paced shooter.

 

[PontiacGTX]

What's the deal with their 1070 cratering as the resolution increases?

Probably the bus width / ROPs/rasterizers are holding it back or R9 Fury X HBM improves its performance as the higher bandwidth is used

How come AMD GPUs with time appear to gain more and more performance?

Driver optimization or game optimizations

That sucks what memory bus does the GTX 1080 and Titan X have ?

the Titan X Pascal and GTX 1080 use GDDR5X 256 bit which has way higher clock speed then it has higher bandwidth

GTX1080 256 but GDDR5 with 320 GB/s, and Titan X has 384 bit of GDDR5, with I think 448 GB/s(?) of bandwidth.

GDDR5X and it allows higher clock speed

Just hoyed a few numbers around and the 1080 and 1070 both scale to around 60% fps at 4k vs 1080 whereas the 980ti and AMD cards are in the 65 - 67% range.

R9 280(X)HD7970/7950 384bit/32ROPs, R9 290(X) 512Bit/64ROPs, R9 Fury (X) R9 Nano HBM/64ROPs+Compression allow higher performance at higher resolution

but RX 480 isnt the best example compared with other AMD GPUs

 

[PhonakV30]

I remembered AMD said RX 480 will bring 390's Performance.now look at bench.What's story with RX 480 ?

 

[Gikaseixas]

want to look at all the data

Then look, there's several reviews, don't limit yourself by reading just my posts as they clearly don't always say what you like or support your agenda. Fact is Fury X does a little better than 1070 at higher resolutions in this title. What so unreasonable about it? Many others pointed out 1070 memory bandwidth limitations and that's what i wanted to point out too.

 

[Abwx]

GameGPU says maximum quality setting according to Edge translator. PCgameshardware.de says:
No idea what that means.

In Call of Duty: Infinite Warfare, there are no presets to set the graphics quality. Nevertheless, the graphics menu is quite extensive, so there is enough control to increase performance at the expense of quality. Only the player has to do this individually for each option, which is cumbersome and no longer up-to-date in 2016.



Nevertheless, the performance can be raised slightly without the graphics quality suffering significantly. If, for example, the "Shadow Map Resolution" is set to "High", the "Volumetric Light" is deactivated and the option "Detail Level" is reduced to High; this is not the case with the naked eye Low performance is a sensible measure..

https://translate.google.fr/translate?sl=de&tl=en&js=y&prev=_t&hl=fr&ie=UTF-8&u=https://www.computerbase.de/2016-11/call-of-duty-infinite-warfare-benchmark/&edit-text=

 

[Sweepr]

Looks like another game where you need 8GB to enjoy the highest texture quality.. Anyway, here are PCGameshardware.de's benches. I prefer using them, since they actually use parts that people will be buying, rather than some reference clocked nonsense:

4GB VRAM is already a problem for Fury / Fury X if you opt for the highest texture qualify. It's actually worse than 1060 3GB because Fiji-based cards are high-end, used to cost $550-650 just a year ago - and people have higher expectations at this segment. For those wondering why it does better at 4K than 1080P relative to other cards, apparently the game is (weirdly) using more memory at lower resolutions.

Spoiler

What's very surprising is that VRAM usage goes down at higher resolution; normally, we'd expect the opposite. I have no answer for as to why this is happening.

 

[dogen1]

Probably the bus width / ROPs/rasterizers are holding it back or R9 Fury X HBM improves its performance as the higher bandwidth is used

Most likely because higher resolutions shift the load away from geometry. The fury/x is able to better take advantage of it's large shader performance.

Higher resolution increases the ratio of ALU/TMU/ROP work to geometry work, but in general the ratios between ALU/TMU/ROP stay relatively constant.

 

[SlickR12345]

I'd like to see the people defending the 3GB 1060 now. It was just yesterday where they were defending 3GB of ram and saying its enough for all games and it will be enough, but we aren't even reached 2017 yet and every single new games that has come out in the past 2 months has used over 4GB of ram and on average about 6gb of vram.

 

[IllogicalGlory]

I'd like to see the people defending the 3GB 1060 now. It was just yesterday where they were defending 3GB of ram and saying its enough for all games and it will be enough, but we aren't even reached 2017 yet and every single new games that has come out in the past 2 months has used over 4GB of ram and on average about 6gb of vram.

I agree about the 3GB 1060, but this is one of those games that allocates as much VRAM as you have available. It does not imply that the game actually needs that much.

 

[Bacon1]

Few other reviews, FPS seems to be very dependent on where they test.

http://www.hardwareunboxed.com/call...e-benchmark-23-gpus-tested-at-1080p-1440p-4k/

 

[RussianSensation]

Another amazing win for the RX series graphics, this is the 5th clear cut and easy win for AMD, but the writing has been on the wall for quite some time now and we are going to be seeing more and more wins for the RX series, especially as DX12 titles become more prominent and DX12 is the primary development and not an afterthought.

Yup, foreign sites are already recommending RX 480 over GTX1060 6GB as RX 480 is leading in the latest titles.

Amazing how Kepler continues to fall apart. GTX780Ti is barely faster than R9 380X at Guru3D. Computerbase has R9 280X beating 770 by 25%, and smashing GTX960 by 35%. RX 480 is beating GTX1060 by 20%. The horribly underpowered and overpriced GTX1050Ti is trounced by $169 RX 470 by a whopping 62%!
https://www.computerbase.de/2016-11...rk/2/#diagramm-cod-infinite-warfare-1920-1080

Haha Nv paying for cutdown GTX1070 so much.Furyx is faster.If nv cut 1070 like normal card GTX1070 would still be faster by atleast 10%.
The gap is again mega Huge GTX1080 is 30% faster in 4k and 26% in 1440p than 1070.

GTX1070 is severely cut down but you have to take into account its price/performance. I bought 2x Asus Strix 1070s for less than a single Asus Strix 1080. GTX1070 SLI wipes the floor with GTX1080 in 1440p and 4K. It's misleading to talk about how GTX1080 is beating GTX1070 in 1440p and 4K while ignoring that for the price of a 1080 it's often possible to purchase 2x GTX1070s. Every top YouTuber I trust has said that GTX1080 does not provide a sufficiently faster level of performance in 1440p or 4K compared to 1070. If you want more performance, you go 1070 SLI, 1080 SLI or Titan XP. Perhaps the one advantage 1080 has over 1070 is 1080p 144Hz gaming but otherwise, for higher resolution gaming, 1070 SLI all day over 1080. The average performance delta in games is usually 22-24% in favour of the 1080, which is not enough when it's often possible to buy GTX1070 SLI for $700 or less....

In recent history of NV, every single x80 card was never truly worth its price. It's better to buy the x70 card or x70 SLI for more performance and just upgrade again next gen.

GTX480 never outlasted 480, 470 SLI > 480
GTX580 never outlasted 570, 570 SLI > 580
GTX680 never outlasted 670, 670 SLI > 680
GTX980 never outlasted 970, 970 SLI > 980
GTX1080 will not outlast 1070, 1070 SLI > 1080

It's even more in favour of 1070 SLI now because 2x 1070s make 2x the profits mining. That means 1070 SLI user will enjoy superior gaming experience to 1080 at 1440p and 4K for the next 2 years, be able to earn more $ towards next gen Volta/Vega GPUs. 1080 is just another overpriced mid-range x80 card.

You are right that GTX1070 is not a true x70 card by specs and the criticism against it is fair from that perspective. However, when GTX1070 often drops to $350-370 and GTX1080 is $600, the choice cannot be clearer for 1080p/1440p 60Hz gaming. Anyone who wants a top-end PC gaming rig and wants top performance is buying 1080 SLI, Titan XP or Titan XP SLI. 1070 SLI is easily the best bang for the buck at 1080p 144Hz, 4K 100-165Hz and 4K gaming compared to all of those setups. 1080 sits in no man's land and once this generation is done, no one will care about that card -- exactly the same history that happened to 580, 680, 780, 980 --> all overpriced and under-performing videocards.

When will nv release game ready drivers for 970? Both reviews show it at Tonga level of performance

Look at this forum -- the same people who recommended 950/960 over R9 380/380X/280X, 780/780Ti over 290/290X, GTX970/980 over 290X/390/390X/R9 295X2 have moved on to focus entirely on the Fury/Fury X because AMD's 2012-2015 cards are smashing 2012-2015 NV cards (except 980Ti) in all key segments. The irony is that I would bet the vast majority of people on this forum who bought Fury/Fury X have either benefited from a much cheaper FreeSync monitor and/or have been mining on Fury/Fury X for months, and/or have bought Fury/Fury X for a fraction of their MSRP (i.e., recent deals on $225-275 Fury for example, or $325 Fury X). In other words, almost no one here who owns Fury/X actually cares that Fury/Fury X is limited to 4GB of memory because almost no one here who actually paid $650 around launch and then not mined with it to pay for itself. I would bet Fury/Fury X owners are already lined up for a largely subsidized or nearly free Vega. They never bought the Fury X to keep for 5 years.

At least 1060 is actually a good card compared to the GTX660/660Ti/760/960 turds that came before.

The entire Kepler generation fell off a cliff as of November 2014. 980 looks awful as usual, typical NV marketing SKU to milk the loyalists as in modern games released in the last 12 months compared to R9 390/R9 290X it never shows the worth of its $550 MSRP. The only truly great card NV released from 2012-2015 was GTX980Ti. It's the only card in the entire Kepler and Maxwell line that continues to perform well and isn't falling off a cliff.

As far as current gen goes, the primary reason 1070/1080 look good is because they are competing against last generation's Fury/X videocards. Deep down, objective PC gamers know that 1070/1080 are just GTX660Ti and GTX680 -> aka GTX560/560Ti lineage and the true AMD/NV flagships will only show up in 2017. I am kinda over this generation because even if Vega 10 and GP102 are fast, they will cost $700-800 USD and in 2018 a mid-range $400-450 x70 Volta will likely obsolete both of them for close to half price, use way less energy, have better features, 16GB of G5X/GDDR6 memory, likely have far superior DX12 architecture, etc. It's going to be a repeat of 780Ti -> 970 and 980Ti -> 1070 all over again. Even AMD has Vega 10 230W rumoured to be shrunk to a 130-150W card in 2018.

 

[frozentundra123456]

I'd like to see the people defending the 3GB 1060 now. It was just yesterday where they were defending 3GB of ram and saying its enough for all games and it will be enough, but we aren't even reached 2017 yet and every single new games that has come out in the past 2 months has used over 4GB of ram and on average about 6gb of vram.

Looks to me like it plays the game quite well. 1080p max gets over 60FPS average and 54 FPS minimum. The 6gb model is 12/23 percent faster (ave/min) right in line with the price difference. Granted, the 470 is faster, but it is also as fast as the 1060 6gb, so it seems more like an architectural difference rather than simply vram. (These results from the hardwareunboxed tests just above.) Honestly, people love to bash this card, but in the vast majority of cases, the data simply does not back up the unrelenting criticism.

 

[Carfax83]

Yup, foreign sites are already recommending RX 480 over GTX1060 6GB as RX 480 is leading in the latest titles.

Foreign sites = one Russian YouTube channel?

The horribly underpowered and overpriced GTX1050Ti is trounced by $169 RX 470 by a whopping 62%!

The RX 470 should trounce the GTX 1050 Ti, since the latter is a sub 100w gaming card and the former uses nearly twice as much power.

GTX1070 is severely cut down but you have to take into account its price/performance. I bought 2x Asus Strix 1070s for less than a single Asus Strix 1080. GTX1070 SLI wipes the floor with GTX1080 in 1440p and 4K. It's misleading to talk about how GTX1080 is beating GTX1070 in 1440p and 4K while ignoring that for the price of a 1080 it's often possible to purchase 2x GTX1070s. Every top YouTuber I trust has said that GTX1080 does not provide a sufficiently faster level of performance in 1440p or 4K compared to 1070. If you want more performance, you go 1070 SLI, 1080 SLI or Titan XP. Perhaps the one advantage 1080 has over 1070 is 1080p 144Hz gaming but otherwise, for higher resolution gaming, 1070 SLI all day over 1080. The average performance delta in games is usually 22-24% in favour of the 1080, which is not enough when it's often possible to buy GTX1070 SLI for $700 or less....

So your logic is that you can buy a second GTX 1070 to make up the ground to a single GTX 1080. Yeah, that makes a lot of sense. How many times have you used SLI before?

Apparently not enough, as you totally swept under the rug all of the issues and difficulties with running multi GPU..

1080 is just another overpriced mid-range x80 card.

A midrange card with 9 Tflops (roughly 10 Tflops for aftermarket) that can handle 2K with ease? Sign me up! It's exciting times we live in when a midrange card from NVidia can do all of that, when AMD's best high end cards fall flat on their asses and can't even run the highest quality textures..

Look at this forum -- the same people who recommended 950/960 over R9 380/380X/280X, 780/780Ti over 290/290X, GTX970/980 over 290X/390/390X/R9 295X2 have moved on to focus entirely on the Fury/Fury X because AMD's 2012-2015 cards are smashing 2012-2015 NV cards (except 980Ti) in all key segments. The irony is that I would bet the vast majority of people on this forum who bought Fury/Fury X have either benefited from a much cheaper FreeSync monitor and/or have been mining on Fury/Fury X for months, and/or have bought Fury/Fury X for a fraction of their MSRP (i.e., recent deals on $225-275 Fury for example, or $325 Fury X). In other words, almost no one here who owns Fury/X actually cares that Fury/Fury X is limited to 4GB of memory because almost no one here who actually paid $650 around launch and then not mined with it to pay for itself. I would bet Fury/Fury X owners are already lined up for a largely subsidized or nearly free Vega. They never bought the Fury X to keep for 5 years.

And how long did it take the Fury lineup to actually deliver the performance that they were supposed to, rather than underperforming? And this is mostly thanks to the console effect more than anything..

And when you look at actual AiB cards, you know, the stuff that people actually buy, NVidia actually does very well. The 780 Ti in fact is nearly as fast as the Fury X at 1080p, and the GTX 980 is faster at 1080p.

I know you like to always focus on reference models, since that helps your argument

As far as current gen goes, the primary reason 1070/1080 look good is because they are competing against last generation's Fury/X videocards..

Maxwell is killing the Fury line up, much less Pascal! The only time when Fury looks good is when you use reference clocked models. When you use aftermarket models, then NVidia clearly has the edge due to superior architecture efficiency combined with much higher clock speeds..

 

[Arachnotronic]

Looks very scene dependent.

Anyway, game is great but 91FPS cap in multi-player is disappointing.

 

[happy medium]

my low end dud crap overclocked gtx960 will give me 44 fps minimum and 50 average? I may have to try this game. is it any good? Oh that's max settings?
I should hit 60fps with very high settings which look just as good.

 

[Arachnotronic]

my low end dud crap overclocked gtx960 will give me 40 fps minimum and 50 average? I may have to try this game. is it any good?

It's pretty good, I like it.

 

[tviceman]

Maxwell is killing the Fury line up, much less Pascal! The only time when Fury looks good is when you use reference clocked models. When you use aftermarket models, then NVidia clearly has the edge due to superior architecture efficiency combined with much higher clock speeds..

The 780 TI is beating the 390X too.... just thought I'd mention that since he said Kepler was falling apart.

 

[Abwx]

The 780 TI is beating the 390X too.... just thought I'd mention that since he said Kepler was falling apart.

Does it.?..

Because PCGH has the results that are the most apart of all other reviews, how is it that the 1060/390X comparison FI is so different from what Computerbase.de got.?

https://www.computerbase.de/2016-11...rk/2/#diagramm-cod-infinite-warfare-1920-1080

So from 20% faster at CPTB it becomes 10% or more slower at PCHG..

.