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

[Thinker_145]

For the last 2 generations Nvidia cards have had more OC headroom than AMD so yes using reference cards is a dodgy thing to do when comparing AMD and Nvidia.

However it becomes absolutely ludicrous when you talk about the Fury as the reference cards are essentially max OC cards. At minimum you must use the best AIB Nvidia cards to compare with Fury. I am surprised some of our otherwise very logical members just ignore this point.

 

[Bacon1]

However it becomes absolutely ludicrous when you talk about the Fury as the reference cards are essentially max OC cards. At minimum you must use the best AIB Nvidia cards to compare with Fury

What?

Fury OC's fine, people average 10%+ on theirs yet it is never oc'd in benchmarks unlike many 980 Tis that often are, especially on sites like Gamersnexus where they use a hybrid WC'd OC'd 980 Ti.

 

[Thinker_145]

What?

Fury OC's fine, people average 10%+ on theirs yet it is never oc'd in benchmarks unlike many 980 Tis that often are, especially on sites like Gamersnexus where they use a hybrid WC'd OC'd 980 Ti.

Please show me reviews achieving 10%+ like you claim? I just did a quick check on a few reviews and didn't see that. It seems like 5% on average to me. "Users" love to exaggerate their overclocks.

 

[Carfax83]

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..

.

Performance might be very scene dependent like what Arachnotronic stated. Computerbase.de has some outliers as well. Look at the GTX 970 and the GTX 980. The 980 is a whopping 42% faster than the 970, which is too high to be credible. Also, the 970 is only 10% faster than the GTX 770, which is unbelievable..

I'm starting to think that this game might be unbenchable due to the weirdness of these scores..

 

[Bacon1]

Please show me reviews achieving 10%+ like you claim? I just did a quick check on a few reviews and didn't see that. It seems like 5% on average to me. "Users" love to exaggerate their overclocks.

You do realize that 10% is only 1150 clocks right and only 1100 for Fury.

Most reviewers try to shove as much power as they can which is actually the wrong way to do it. The cards often OC better when undervolted.

Reviewers haven't tried OCing lately so not sure how I'm supposed to find any reviewers showing that and not just actual users that use the cards daily.

But I guess all those users claiming they can OC their 980 TI 30% no sweat are just exaggerating as well then.

 

[Samwell]

Performance might be very scene dependent like what Arachnotronic stated. Computerbase.de has some outliers as well. Look at the GTX 970 and the GTX 980. The 980 is a whopping 42% faster than the 970, which is too high to be credible. Also, the 970 is only 10% faster than the GTX 770, which is unbelievable..

I'm starting to think that this game might be unbenchable due to the weirdness of these scores..

I don't think the 970 has such a weird result. There are games which fill more than 4Gb ram, but still don't really need it. COD is different, it really needs ~6gb or it's reducing texture details even on same settings. So you can't compare 4Gb against 8Gb cards in this game, because 4 Gb Gpus are reducing texture details even on max details as pcgh states. For the 970 the game is pushing for 4Gb ram as it thinks this is the usable ram and that's killing the performance. In other games Nvs driver might push for 3,5Gb as the games don't really need more.

 

[Bacon1]

I'm starting to think that this game might be unbenchable due to the weirdness of these scores..

Guru3d said that in their conclusion:

Call of Duty Infinite Warfare will NOT very likele make it into our regular benchmark suite for videocards. In the past we run into this as well, but there are things odd and off. As such we recommend you to look at the performance benchmarks as indicative performance. At one point the result set would be 50 FPS, and on the second similar path and execution 42 FPS and another 56 FPS. This means that the results shown today are indicative, not a precise measurement. Another problem was that some levels average out at say 40 FPS, yet there are a handful of scenes where the framerate will crumble down to say 25 FPS or where running a luxurious 60 FPS all of the sudden. That is the reality of this game engine.

 

[Thinker_145]

You do realize that 10% is only 1150 clocks right and only 1100 for Fury.

Most reviewers try to shove as much power as they can which is actually the wrong way to do it. The cards often OC better when undervolted.

Reviewers haven't tried OCing lately so not sure how I'm supposed to find any reviewers showing that and not just actual users that use the cards daily.

But I guess all those users claiming they can OC their 980 TI 30% no sweat are just exaggerating as well then.

lol I am not talking about clocks but actual performance. Generally no 980Ti achieves an outright OC of 30% but when you OC an AIB card it can end up 30% faster than the reference version at stock.

Reviewers don't know how to overclock? Okay...

 

[raghu78]

The benchmarks seem to vary a lot based on where in the game the cards are being tested and the settings. The Rx 480 performs very well in all reviews, beating GTX 1060 by 10-20% in a few.

 

[psolord]

Hello everybody. I did some benchmarks too, for anyone interested in some more down to earth system's performance.

I benchmarked the Black Sky:Under Attack mission. As usual, I did long runs, 14 min for this one, so I can expose every possible weakness of my systems.

The settings I used were fairly high. I only kept a few of them at bay.

Shadow maps at high instead of extra
Cache spot shadows disabled (needs too much vram)
Cache sun shadows disabled (needs too much vram)
Film grain disabled (I don't like it and reduces video quality)
AntiAliasing Filmic SMAA 1X
Tesselation near

All these have minimal impact on the visual quality but a huge impact on the performance.

Cod Infinite Warfare 1920x1080 High GTX 970 @1.5Ghz Core i5 2500k @4.8GHz - 82fps

CoD Infinite Warfare 1920x1080 High 7950 @1.1Ghz CORE i7-860 @4GHz - 67fps

Now I have to say that this is one of the smallest deltas I have seen between the 970 and the 7950, but there is a big difference if you take a look at the frametimes.









With these settings, the 970 was registering 3.7GBs of video ram usage in MSI Afterburner, while the 7950 was registering 2.93GBs. You can see that the 7950 system could be under quite some data swapping between ram and vram, hence the higher frametimes.

Overall the game looks and feels quite nice and was quite playable on both systems, although neither of them is new.

I was surprised to see the 2500k being hammered to death, even at 4.8Ghz, yikes.



The good news is that you can have 60fps vsynced, with the cpu at 4.3Ghz and the gpu at stock.

PS The game has a serious issue with FRAPS. Whenever FRAPS is running, you get jumpy mouse movement. You turn the mouse suddenly aiming for a 30 degree turn and you end up with 180 degree some times. Really weird, don't know why this is happening, but I am certain about this. That's why I play like a jerk in the runs from time to time. Does not affect frametimes however.

 

[guachi]

I'm most impressed with how well the AMD 8370 does compared to the Intel 5960X CPU. In the CPU tests using the 480, is the 8370 actually faster at 3840x2160?

I bought a 8350 (should be very similar to the 8370) new for $100 in August and it's within a few percent of a $1000 5960X.

That's crazy. 8370 + 480 + Freesync = crazy cheap and good gaming, apparently.

 

[SlickR12345]

Overclock for overclock is not the same. Nvidia 1060's can overclock by 20% or more, but they only gain 10% or so in performance at best. AMD's cards on the other hand can overclock by 10% or so and gain 10% performance.

Its a different architecture, Nvidia can overclock 1000mhz more and not even gain 10% performance, MHz doesn't equal performance! Its all about the architecture and how it uses that MHz.

 

[frozentundra123456]

I'm most impressed with how well the AMD 8370 does compared to the Intel 5960X CPU. In the CPU tests using the 480, is the 8370 actually faster at 3840x2160?

I bought a 8350 (should be very similar to the 8370) new for $100 in August and it's within a few percent of a $1000 5960X.

That's crazy. 8370 + 480 + Freesync = crazy cheap and good gaming, apparently.

Not sure how they got those cpu results. Game.gpu tested with a 1080 and Fury X. 5960 X at stock was 36/50 and 42/50 (ave/min) percent faster, respectively with the two cards. And of course, the 5960 X has a lot more overclocking headroom than the 8350. Even a skylake i3 was equal or faster in both ave and min FPS than the 8350 with either card. In fact, the skylake i3 was equal or faster in both cases, even compared to the 9590.

 

[Abwx]

And of course, the 5960 X has a lot more overclocking headroom than the 8350. .

And of course that you re posting without even reading the article, hence a pile of non sens..

The Guru3D comparison is at 4.3GHz for both CPUs, that s basically a 300W 5960X against a 150W 8350, and btw, so much for the overclocking headroom...

 

[Carfax83]

I don't think the 970 has such a weird result. There are games which fill more than 4Gb ram, but still don't really need it. COD is different, it really needs ~6gb or it's reducing texture details even on same settings. So you can't compare 4Gb against 8Gb cards in this game, because 4 Gb Gpus are reducing texture details even on max details as pcgh states. For the 970 the game is pushing for 4Gb ram as it thinks this is the usable ram and that's killing the performance. In other games Nvs driver might push for 3,5Gb as the games don't really need more.

What about the GTX 970 only being 10% faster than the GTX 770 then? And if the game actually needs 6GB of VRAM and is reducing texture details for GPUs with less VRAM, then the game is an unoptimized piece of turd. Apparently IW don't know how to program their engine to efficiently utilize the PC's non unified memory pools, versus the consoles which don't have disparate memory pools..

 

[dogen1]

Performance might be very scene dependent like what Arachnotronic stated.

I wonder if that has to do with the depth prepass the game does(confirmed by a dev). Doubling cost of geometry could result in more performance fluctuations. Doom, and some other games, do the same thing though..

 

[Carfax83]

Overclock for overclock is not the same. Nvidia 1060's can overclock by 20% or more, but they only gain 10% or so in performance at best. AMD's cards on the other hand can overclock by 10% or so and gain 10% performance.

Its a different architecture, Nvidia can overclock 1000mhz more and not even gain 10% performance, MHz doesn't equal performance! Its all about the architecture and how it uses that MHz.

I think you have it backwards man. NVidia gains much more from overclocking versus AMD:

 

[SlickR12345]

I think you have it backwards man. NVidia gains much more from overclocking versus AMD:

Really? I'm supposed to take the word of some random unknown youtuber as fact? As far as I know it might be you with your Nvidia bias. AMD's overclock is smaller, but gives better performance Mhz for Mhz.

In Fact RX 480 8Gb has won the last 5 AAA games released by big margins over the 1060 6Gb, when overclocked it wins by even bigger margins and this is not even using full DX12 engine for any of the titles, most come with DX11 engines with tacked on DX12 support.

 

[happy medium]

In Fact RX 480 8Gb has won the last 5 AAA games released by big margins over the 1060 6Gb, when overclocked it wins by even bigger margins

Please show me these benchmarks with and without overclocks .
First what are the last 5 AAA games?
Thanks.

 

[Bacon1]

I think you have it backwards man. NVidia gains much more from overclocking versus AMD:

Thats using reference cards as he made sure to point out in the video. You should probably take this to a different thread since its completely off topic. Thats showing 8% avg gain on OC'd 480 (power limited reference) vs 10% gain on 1060 OC.

 

[SlickR12345]

Please show me these benchmarks with and without overclocks .
First what are the last 5 AAA games?
Thanks.

Forza 3
Battlefield 1 (destroys the 1060 6gb in DX12, even rx470 is faster)
COD: Infinite Warfare
Titanfall 2
Doom

 

[ultima_trev]

Expect that BF1's DX12 path is broken and has very inconsistent frame times. Luckily RX 480 is close enough to GTX 1060 in this title at 1080P (and slightly better at 1440P) in the DX11 path that it's essentially a tie.

While I haven't seen benches for Forza 3; It is definitely true that Doom (w/ Vulkan), Titanfall 2 and Infinite Warfare sees RX 470 / RX 480 punching well above their weight. Hopefully in a years time with enough driver updates Polaris will be the dominant sub $300 graphics architecture.

 

[Carfax83]

Thats using reference cards as he made sure to point out in the video. You should probably take this to a different thread since its completely off topic. Thats showing 8% avg gain on OC'd 480 (power limited reference) vs 10% gain on 1060 OC.

It doesn't matter whether it's reference or not. The clock speed he reached is typical of the RX 480, reference or otherwise..

TechPowerUp Asus Strix RX 480
Computerbase.de Asus Strix RX 480
eTeknix Sapphire Nitro RX 480
KitGuru Sapphire Nitro RX 480

The Sapphire Nitro seems to overclock the best, but Kitguru was only able to get 1390 out of it, not much more than the reference models..

Video review of XFX Radeon RX 480 Black edition. Seems like it throttles in heavy games like the Witcher 3 unless you raise the power target substantially..

 

[Samwell]

What about the GTX 970 only being 10% faster than the GTX 770 then? And if the game actually needs 6GB of VRAM and is reducing texture details for GPUs with less VRAM, then the game is an unoptimized piece of turd. Apparently IW don't know how to program their engine to efficiently utilize the PC's non unified memory pools, versus the consoles which don't have disparate memory pools..

970 to 770 is the same as i said. COD uses 4GB with 970 and the slow 512mb are killing the performance. Using just 770 with 2gb is alright, the game uses lower texture quality.
Why is reducing details with lower vram bad? Actually its good, as you get max details in some scenes and middle in others versus always middle details. Also this is not a new technology. CoD is using this since Advanced Warfare, Mirrors Edga has it, Hitman has it. Only Advantage is that you can deactivate it in Hitman and Mirrors Edge, but more and more Engines will use it in future. Actually this is tech coming from consoles to us. Next step is adaptiv resolution depending on framerate which consoles are using. Everything good tech, as long as you can also deactivate it if you want. It's just important to always look how websites are testing, whether they deactivate it, i'm thinking many testers don't care and then test with different image qualities.

 

[moonbogg]

No SLI support! Huzzah!