[GameGPU] BF Hardline benchmarks

[Hitman928]

yikes, what happened to the titan x frametimes? Has any other site tested frame times to confirm?

 

[96Firebird]

Whats also interesting is the 290X DirectX results being so different in singleplayer vs multiplayer. It is fine in MP, but horrible in SP.

 

[Hitman928]

Whats also interesting is the 290X DirectX results being so different in singleplayer vs multiplayer. It is fine in MP, but horrible in SP.

Yep, that is weird, although it only happens under dx, mantle is fine on the 290x single player. Reading the translation it seems that their benchmark runs happen when driving in a vehicle at fast speeds. It would be interesting to see frame times of more regular game play as well.

 

[artivix]

Review found several fps lower in retail version compared to beta and R9 290x and 290 surprising held position against GTX 970, with 290x performance within 10% of GTX 980 in 1980x1080.

http://www.techspot.com/review/979-battlefield-hardline-benchmarks/page4.html

 

[Head1985]

I hate techspot reviews GTX970 is so slow in their benchmarks.
They must have something wrong with their GTX970.
20Fps slower than GTX980 in 1080P really?
Also GTX970 only 25%Faster than 7970Ghz?
Others webs shows +- 10FPS(15% performance) differences in 970 vs 980 and 40+% in 7970Ghz vs 970.

Btw pcgameshardware updated frametimes for GTX970 and other cards for singleplayer

 

[ocre]

yikes, what happened to the titan x frametimes? Has any other site tested frame times to confirm?

Obviously it is the drivers. Those spikes look bad but the scale is zoomed in pretty far. Its only 8-20, which isn't a big deal on its own.

It is when we look at the awesome 290x frame times. Those frames are so steady, it is stunning.

I think maxwell has nvidia struggling with frame times. Just looking at the 980, its not terrible but obviously the 290x has a better showing here. This is not unusual. I have this theory (and I may get a bashing for it) that maxwell bus has nvidia struggling at times. They are able to keep the cores fed almost perfect but not completely and entirely. These holes and gaps may be what we see in the frame time spikes. One way to test out this would be to overclock the ram as high as possible. If the spikes reduce, then it would be pretty clear. But it could be a caching issue, which may not improve much just by overclocking the ram.

This theory of mine is just in my head. And it doesn't mean that nvidia driver teams are powerless. Never underestimate talented software teams. Looking at early 980 launch frame times, there has been improvement. There is no reason to think that nvidia driver teams cannot further improve. They can and will.
But then I think about all the info we have seen on dx12 with nvidia maxwell cards looking pretty amassing. I wonder how much of nvidia's talent is working strictly on dx12. I don't believe for a second that AMD cards will be so much weaker in dx12, I believe it is just that nvidia has put more resources into their dx12 performance. Considering dx12 isn't even out yet, that there arent any games on this API, who knows when there will be.............I wonder how much that matters at this point.

But back to the quote....
looking at the titanX frame times,
I would say that nvidia has room to improve the performance with their drivers. There is no reason the titanX should be worse, it should be at least as good as the 980. So I expect it will be sooner or later.

Edit:
Now that I see the single player frame times, I am confused. The charts on the first page of the thread are multiplayer frame times. not just that, it says 64 players!!!
There are now 64 reasons as to why the 980/titanX may have more peaks than a 290x. It also seems logically that these frame times were recorded the same time as the average frame rates they use in the charts. So, perhaps the nvidia cards experienced a much more brutal MP.

There is something up. How could the 290x single player be so different than the MP? Does the game scale down things (such as graphics) for MP? Just by looking, I would wonder if the game reviewer didn't mix up his data or something. Its really odd looking

 

[Headfoot]

I hate techspot reviews GTX970 is so slow in their benchmarks.

It's really a terrible reason to "hate techspot reviews" because they don't line up with how fast you want the 970 to be... Gotta put up some evidence otherwise it's your word against the reviewers...

The techspot review is done on lightly populated multiplayer servers, the pcgameshardware is done in single player.

All this illustrates is that BF Hardline single vs MP is different and needs to be analyzed separately. It in no way shows the techspot review is "bad" because the GTX 970 doesn't win like you want it to

Maybe you can call the techspot review bad because lightly-populated servers are poor measurements of real world gameplay, and that would probably be correct, but calling it bad because it doesn't give you the results you want to see is ludicrous.

 

[Head1985]

It's really a terrible reason to "hate techspot reviews" because they don't line up with how fast you want the 970 to be... Gotta put up some evidence otherwise it's your word against the reviewers...

The techspot review is done on lightly populated multiplayer servers, the pcgameshardware is done in single player.

All this illustrates is that BF Hardline single vs MP is different and needs to be analyzed separately. It in no way shows the techspot review is "bad" because the GTX 970 doesn't win like you want it to

Maybe you can call the techspot review bad because lightly-populated servers are poor measurements of real world gameplay, and that would probably be correct, but calling it bad because it doesn't give you the results you want to see is ludicrous.

Thier results are wierd.Compare them to all other websites...
I posted here two differend websites benchmarks with pretty much same results.

Cpu benchmarks are also wierd same with other games like Dying light
http://www.techspot.com/review/956-dying-light-benchmarks/page5.html
51Fps on celeron 2.7Ghz yeah sure...they tested inside building...Thats just bad.
When i test it on i5 2500K@5GHZ i have like 50Fps just like here
http://pclab.pl/art61159-9.html
and here

I simply dont trust techspot, because wierd results.

 

[raghu78]

Thier results are wierd.Compare them to all other websites...
I posted here two differend websites benchmarks with pretty much same results.

Cpu benchmarks are also wierd same with other games like Dying light
http://www.techspot.com/review/956-dying-light-benchmarks/page5.html
51Fps on celeron 2.7Ghz yeah sure...they tested inside building...Thats just bad.
When i test it on i5 2500K@5GHZ i have like 50Fps just like here
http://pclab.pl/art61159-9.html
and here

I simply dont trust techspot, because wierd results.

Let me show other websites which show R9 290X much faster than GTX 970 especially at 1440p.

http://gamegpu.ru/action-/-fps-/-tps/battlefield-hardline-test-gpu.html

pclab.pl is known to be extremely Nvidia biased. :thumbsdown:

 

[RussianSensation]

I hate techspot reviews GTX970 is so slow in their benchmarks.
They must have something wrong with their GTX970.
20Fps slower than GTX980 in 1080P really?
Also GTX970 only 25%Faster than 7970Ghz?
Others webs shows +- 10FPS(15% performance) differences in 970 vs 980 and 40+% in 7970Ghz vs 970.

What makes you think the 40%+ faster is accurate and 25% faster is wrong?

Let's see.

970 vs. 7970Ghz
TPU - 1080P: 32%, 1440P: 26%, 4K: 25%

Sweclockers - 1080P: 26%, 1440P: 26% (over 280X which is slower than a 7970Ghz)

Computerbase: 1080P - 28% against 280X, 23% at 1600P.

If anything, 970 is about 25-28% faster than a 7970Ghz, nowhere near the 40% you are quoting. Both GameGPU and Techspot show that 7970Ghz is a lot closer to a 970 in BF: Hardline than the 40% you are quoting and this is consistent with the average performance increase 970 has over the 7970Ghz. Also, 980 leads 970 by 15-19% depending on the review and TechSpot shows 980 outperforming the 970 by about 20%, which is in line with expectations.

pclab.pl is known to be extremely Nvidia biased. :thumbsdown:

I can hardly remember a single time / game where AMD led NV significantly on that site. Their results almost never align with GameGPU, TechSpot, TechPowerup, Sweclockers, Computerbase, AnandTech. For as long as I remember, every time someone links pclab.pl game benches they are favouring NV by 40-50% or showing the most out of whack scores from anywhere else on the Internet. I don't know anyone who takes that site seriously on our forum.

 

[Silverforce11]

As was discussed during the beta, when DX11 had terrible stutters, it seems AMD does not even bother to optimize for DX11 in games where Mantle is available.

 

[xthetenth]

As was discussed during the beta, when DX11 had terrible stutters, it seems AMD does not even bother to optimize for DX11 in games where Mantle is available.

Makes sense, I wouldn't either if Mantle did a job like that.

 

[Head1985]

Let me show other websites which show R9 290X much faster than GTX 970 especially at 1440p.

http://gamegpu.ru/action-/-fps-/-tps/battlefield-hardline-test-gpu.html

pclab.pl is known to be extremely Nvidia biased. :thumbsdown:

I only 100% trust german sites.
Only Pcgameshardware and computerbase make some good test with GTX970 3.5GB issue.
Other EN sites only copy PR NV bullshits and make just bad tests.

And i dont trust techspot and gamegpu.ru either.
Their results are wierd.

Pclabs i dont know.They test games in max details with 16xAF forced in drivers just like pcgameshardware.And their results atleast for NV cards are on spot-
GTX780Ghz +-same performance as 780TI
OC GTX970 match stock 980
http://pclab.pl/art62452-8.html
Just like it is in real world.I trust them more than most EN sites.
Maybe its because forced 16xHQAF in drivers is why radeons sucks in pclabs.They have more performance hit when forced 16xHQAF in drivers than NV cards.

 

[raghu78]

I only 100% trust german sites.
Only Pcgameshardware and computerbase make some good test with GTX970 3.5GB issue. Other EN sites only copy PR NV bullshits and make just bad tests.

And i dont trust techspot and gamegpu.ru either. Their results are wierd.

Pclabs i dont know.They test games in max details with 16xAF forced in drivers just like pcgameshardware.And their results atleast for NV cards are on spot-
GTX780Ghz +-same performance as 780TI
OC GTX970 match stock 980
http://pclab.pl/art62452-8.html
Just like it is in real world.I trust them more than most EN sites.

You would trust any site which shows results favourable to Nvidia. not that its any surprise

 

[Head1985]

Nope..
You know i have GTX970 and i make here first GTX970 3.5GB issue topic
http://forums.anandtech.com/showthread.php?t=2416150&highlight=
I trust german sites because they are not NV PR puppets like most EN sites.

 

[AtenRa]

As was discussed during the beta, when DX11 had terrible stutters, it seems AMD does not even bother to optimize for DX11 in games where Mantle is available.

Why would they ??? if the hardware supports Mantle, running the game in DX-11 mode is stupid. It is like running the game in DX-9 when both the game and hardware supports DX-11. Would anyone complain if drivers were not optimized for DX-9 in that situation ???

But i would like to see frametimes in DX-11 with older AMD hardware, HD6xxx series and older.

 

[DiogoDX]

Guru3D

http://www.guru3d.com/articles_pages/battlefield_hardline_vga_graphics_performance_review,6.html

 

[Head1985]

Yeah and again GTX980 absurd faster than 780TI and GTX970
30% faster than both..Yeah sure..
GTX780 only 3 Fps slower than GTX780TI witch have 25% more SP
Thats why i dont trust these sites.
Atleast results on pclabs they make some sense.

 

[Silverforce11]

Yeah and again GTX980 absurd faster than 780TI and GTX970
30% faster than both..Yeah sure..
GTX780 only 3 Fps slower than GTX780TI witch have 25% more SP
Thats why i dont trust these sites.
Atleast results on pclabs they make some sense.

It seems you were not aware that since Maxwell's launch, Kepler (780/ti) has fallen behind by a much larger gap than the original 5-10%. Compare launch review where the 980 is barely faster than 780ti to more recent reviews..

Some say NV is maximizing/optimizing Maxwell because its a new architecture so there's more performance to extract so the gap grows over time. Some say NV is not bothering to optimize for Kepler and focus on Maxwell only. Believe what you will.

980 should be 20-25% faster than 970 if drivers are optimized to take advantage of the increased cores, rops and bandwidth (which isn't 224GB/s at all, its only 192GB/s up to 3.5GB and for the remaining 0.5GB, well you know..).

 

[Silverforce11]

Why would they ??? if the hardware supports Mantle, running the game in DX-11 mode is stupid. It is like running the game in DX-9 when both the game and hardware supports DX-11. Would anyone complain if drivers were not optimized for DX-9 in that situation ???

But i would like to see frametimes in DX-11 with older AMD hardware, HD6xxx series and older.

I'm just saying, some people were extremely quick to diss "AMD drivers, DX11 LOLOL".. even after people mentioned "it's BETA" or "not Mantle".

At least its clear now AMD will focus on Mantle if its available. Which is good but also bad, because those on VLIW4/5 cards of old will suffer.

 

[Head1985]

It seems you were not aware that since Maxwell's launch, Kepler (780/ti) has fallen behind by a much larger gap than the original 5-10%. Compare launch review where the 980 is barely faster than 780ti to more recent reviews..

Some say NV is maximizing/optimizing Maxwell because its a new architecture so there's more performance to extract so the gap grows over time. Some say NV is not bothering to optimize for Kepler and focus on Maxwell only. Believe what you will.

980 should be 20-25% faster than 970 if drivers are optimized to take advantage of the increased cores, rops and bandwidth (which isn't 224GB/s at all, its only 192GB/s up to 3.5GB and for the remaining 0.5GB, well you know..).

But its not 15-20Max vs GTX970 witch have 99.99% reviews.
Its 30% thats just wrong and impossible and you ignoring fact that GTX780 with 2300SP and 6Ghz memory have 3fps less than 780TI with 2880SP and 7Ghz memory
Just look at this and tell me if these results dont looks 100% more accurate

980 is 16% faster than 970 and 14% vs 780TI just like it should be and not absurd 30%!!!!!!!

 

[Silverforce11]

980 is 16% faster than 970 and 14% vs 780TI just like it should be and not absurd 30%!!!!!!!

Latest Titan X review:

You can do maths to see how much faster the 980 is above 780ti.

 

[Head1985]

You again ignoring fact that 780 have 3 less Fps than 780TI.And you know 980 vs 970 is 13% and top of that its not even same game.
Ok we are finish here i presume if you ignoring my arguments.

 

[casiofx]

Latest Titan X review:

You can do maths to see how much faster the 980 is above 780ti.

Man it's really punch in the face for those peeps who bought 780 Ti at full MSRP just right before 970/980 were out.

Fastest performance deterioration ever.

 

[Spjut]

I'm just saying, some people were extremely quick to diss "AMD drivers, DX11 LOLOL".. even after people mentioned "it's BETA" or "not Mantle".

At least its clear now AMD will focus on Mantle if its available. Which is good but also bad, because those on VLIW4/5 cards of old will suffer.

The VLIW4/5 cards presumably need different DX11 optimizations than AMD's GCN cards.