gamegpuCall of Duty Black Ops III Beta Benchmark

[csbin]

http://gamegpu.ru/action-/-fps-/-tps/call-of-duty-black-ops-iii-beta-test-gpu.html

 

[csbin]

CPU Benchmark

 

[tential]

Gtx 980 pulled ahead of a r9 290x as resolution rose...
That makes sense

 

[DownTheSky]

You're turning this GW stuff to paranoia. There's nothing wrong with those results.

 

[monstercameron]

You're turning this GW stuff to paranoia. There's nothing wrong with those results.

It goes againts trends, amd should be ahead as resolution goes up.

 

[DarkKnightDude]

Not that it hardly matters. Its is beta to be somewhat fair. I ran this on my 770 and did get some horrible microstuttering.

 

[Arachnotronic]

Great stuff :thumbsup:

 

[Headfoot]

The big cache on the 5960x seems to make quite a large difference in this game, along with good multicore scaling

 

[Spjut]

Yet another game where it seems the i7 CPUs outperform their i5 brethren significantly. It feels fine to see my advice to go for i7 has been right.

 

[psolord]

970 and 780 have 100% scaling in SLI at 1080p. Pretty cool.

The 980 sli has more than 100% scaling at 1600p though. Secret sauce?

I am quite happy seeing the 970 SLI being faster than the 980Ti, since I have started entertaining the idea of gpu upgrade and I am between 980Ti or second 970, with my decision weighing on the second 970.

Now if only I didn't have that Unreal Engine 4 thorn on my side, the decision would be easy.

 

[Arachnotronic]

The big cache on the 5960x seems to make quite a large difference in this game, along with good multicore scaling

Wonder how well the 5775c, also known as "The Unicorn," would do in this game.

 

[biostud]

No cf scaling/profile

 

[RussianSensation]

Lower end and mid-range Keplers are falling apart based on this preliminary testing.

1080P
1. 950 is nearly as fast as a GTX680 (!) :sneaky:
2. HD7970Ghz = 280X is 53% faster than GTX680, and 42% faster than GTX770
3. GTX960 > GTX770

1440P
1. 950 is once again on the heels of the GTX680
2. HD7970Ghz = 280X is 55% faster than GTX680, and 45% faster than GTX770
3. GTX960 > GTX770
4. 780Ti loses to an R9 290/290X/970

Either there are some serious flaws in the Kepler architecture that are becoming exposed or this architecture is extremely susceptible to drivers which means unless NV optimizes Kepler for a modern title where the performance is poor, its performance tanks.

However, Kepler does well in games like Trine 3 so I am inclined to believe NV's lack of Kepler driver optimizations are just bombing Kepler architecture more than anything. There is just no way a GTX950 should be nearly as fast as a GTX680.

Whatever the reasons may be, Kepler GPUs generally cost more than AMD's counter-parts during that generation, and yet they continue to show very poor performance in modern titles. Sucks for Kepler owners but great for NV as they are just waiting for Kepler owners to jump to Maxwell or if not surely to 2016 Pascal. I bet once NV launches Pascal and shifts its driver focus on it, Kepler is basically done for because over the last 11-12 months it has already started to show major issues in performance, coincidentally after GTX970/980 launched.

Gtx 980 pulled ahead of a r9 290x as resolution rose...
That makes sense

Considering that's a reference 290X they are using and it's just 2 fps behind GTX980, and given that R9 290X cost $300-380 in many parts of the world for the 8 months period when GTX980 cost $550, that's a stellar showing from AMD. Kepler syndrome continues....

The $1000 Titan is just 5 fps ahead of the $300 R9 280X at 1600P. I still remember when the Titan/780 came out, people actually defended their prices and tried to sell us on their "future-proofness." Now, none of those individuals are man enough to admit they were wrong and I guess since most of them moved on to Maxwell, who cares about Kepler!

____

The editor, SVL, mentioned that they haven't been able to get a hold of Fury/R9 3xx cards in Russia which is why they haven't been able to test them at all.

 

[Spjut]

PCgameshardware.de also has a GPU benchmark
http://www.pcgameshardware.de/Call-...cials/BO3-Beta-Benchmarks-Windows-10-1169217/

 

[railven]

It goes againts trends, amd should be ahead as resolution goes up.

I don't even think this game has Gameworks or any Nvidia involvement.

Does it?

 

[biostud]

PCgameshardware.de also has a GPU benchmark
http://www.pcgameshardware.de/Call-...cials/BO3-Beta-Benchmarks-Windows-10-1169217/

The 390X beats the Fury non X

 

[monstercameron]

I don't even think this game has Gameworks or any Nvidia involvement.

Does it?

I don't know.

 

[railven]

The 390X beats the Fury non X

Quote from that article using Google Translate:

In addition, even customized graphics card driver is missing from both Nvidia and AMD. A detail: Unlike its predecessor Advanced Warfare is no Nvidia DLLs, see the Directory, so there is no explicit GameWorks used.

So it's not a Gameworks title.

 

[monstercameron]

The 390X beats the Fury non X

How odd. Maybe there is a bottleneck other than the shaders.

 

[monstercameron]

Quote from that article using Google Translate:



So it's not a Gameworks title.

That's a plus in my book. We can remove that element from any further debate.

 

[AnandThenMan]

The performance of Kepler is a disgrace there is no way the 280x should be beating it so handily. Maybe Kepler is highly driver dependent I don't know but the pattern of it tanking in newer titles continues.

That's a plus in my book. We can remove that element from any further debate.

Yes.

 

[railven]

How odd. Maybe there is a bottleneck other than the shaders.

Yeah, definitely driver issue. Article said neither side had a game driver, so there is performance on the table for both sides to snag.

@1080 390X > Fury by 10%
@1440 390X > Fury by 4%

So I don't doubt AMD can fix that.

 

[Hitman928]

According to what I can gather from the pcgamershardware translation, the engine is based off of the same that was used with Advanced Warfare but with the Nvidia effects stripped and replaced with their own as well as a few modern effects added. I haven't looked at the numbers, but it might be an interesting case study as to the effect of gameworks on visuals/performance when comparing the two games against their contemporaries. Should probably wait for drivers from Nvidia and AMD to really compare though.

 

[SimianR]

R9 290 hanging in there with the big boys (sorta) at almost 30 fps at 4K, hey not anywhere near playable but it's still pretty insane all things considered. :thumbsup: That card is the card to get if you have the power supply/cpu to push it. Kind of weird that they have it lower in the graph than the 970 even though it has the same minimums and slightly higher average fps. What's more impressive is if you look at how well the 290/290x are doing here, you can imagine where the 390/390x would land in these results.

 

[Samwell]

This game is eating vram like hell. That seems like on of the reasons for these results. Kepler with 2gb totally tanks. Maxwell is good in Full HD with 4Gb, at 2560 Vram is limiting it and it's going down a lot. 390X is also so good because of 8 gb. Normally you never have a difference of 20% between 290X and 390X. But seems GCN is more resistent when vram is full, as it's loosing way less than the Nv cards.