TWC Total War Attila benchmark thread - last update: 26. December 2016

[Praetorian_spqr]

@Praetorian_spqr
Actually, your settings are being downgraded automatically because your card only has 2GB of VRAM. At 0:37 in your video u can see that message popping up in the loading screen.
I think u have to check 'unlimited video ram' to avoid downgrading.

If u go to the %APPDATA%\The Creative Assembly\Attila\logs folder in Windows explorer u can check the gfx.log.txt file to see which settings were downgraded exactly.

OK, will do it... thx!

[alQamar]

Intel I7 4790K, @4.0GHz, 8 GB RAM 1600 MHz CL 10-10-10-30, MSI GTX 770, Windows 7 Ultimate 64bit, Resolution 1920x1080p

CPU-Z file: Attachment 321417

Results:
http://youtu.be/GoS-_Je8fpo?list=UU0...oE67atrFKZZM9g

Hey Praet, can you please run the benchmark with all presets where you don't get this downgrade message. Thanks again.

Please also try to post your results like this:
http://www.twcenter.net/forums/showt...1#post14376597

Sure, that's the most logical explanation. Or what else would prevent the game from running faster if you raise the CPU speed considerably? Only the RAM would be another (but unlikely) possibility. Faster CPU -> faster execution of program code -> higher framerate (if the graphics card can render the graphical part fast enough).


Why shouldn't 63 fps (performance) or 99 fps (high performance) be the limit of a GTX 670 in Attila? I don't think that these numbers are unreasonable.


Ok, I did a quick test with Maximum Quality Preset @ 1080:
1) 4770K@4.2Ghz + 1x7970: 33 fps avg, 100% GPU utilization
2) 4770K@4.2Ghz + 2x7970: 50 fps avg, 2 x 75% GPU utilization (=150% of 1 GPU)

That shows exactly that the 4770K@4.2Ghz was limited by the single 7970 (@1100Mhz) for the Maximum Quality preset.

Also, alQamar, you can just lower your screen resolution to reduce the GPU load. Then check whether the scaling still stops @ 3.8Ghz.

Ok Barbarian I will repeat this test with a lower resolution e.g. 1024x768, but this may take some time. Thanks for your help. Would be nice if have the time - to run all the preset with and without SLI. Your results and knowlegde was very helpful back in time.

Today I will update the results section with all results that have been posted according to the rules.

[alQamar]

Hey guys, despite the good activity and discussion in this thread only 3 people managed to post their results accordingly to the rules.
I have included Spartan IV, RPG32 and IamOdysseus ones.

Based on the CPU-Z log I was able to see that RPG32, Spartan IV and IamOdysseus have improper configurations though, so I contacted them via PM with instructions how to optimize their individual performance.

Please keep your benchmark results coming! Also low end and mid range computers are very welcome! So far we have seen only powerful builds. This thread welcomes anyone, regardless the specs, so its not mentioned for showing of but giving people an idea how good Attila performs at all on different hardware.

Very likely to the Rome 2 thread I am missing the participation of people with slow computers and laptops and also AMD users . You really do not need much experience to do the benchmark. The technical details listed in the results will be extracted by the CPU-Z log, so you don't have to bother with that stuff.

[Lord DGP]

CPU: AMD Phenom II x6 1100T BE @3,60 (OC, Turbo Core @ 4 Ghz) GPU: Gigabyte Radeon R9 290 WindForce 3X OC
RAM:8 GB RAM 1333 Mhz CL9 OS: Windows 7 64bit,Resolution: 1920x1080p

extreme quality (needs 3-4 GB VRAM):
min FPS: 11
avg FPS: 17,7
max FPS: 24

maximum quality:
min FPS: 14
avg FPS: 22,4
max FPS: 29

quality:
min FPS: 18
avg FPS: 26,7
max FPS: 36

performance:
min FPS: 26
avg FPS: 36,3
max FPS: 46

maximum performance:
min FPS: 29
avg FPS: 50,8
max FPS: 86

gfx.log says:

device_description = "AMD Radeon R9 200 Series"
vendor_id = 0x1002
device_id = 0x67b1
in_database = TRUE
found_vendor_device_ids = FALSE
matched_device_description = "Radeon R9 200 Series"
final_quality = 3
m_gpu_memory_type_final = 1
vista_or_later = TruE
sixty_four_bit_OS = tRUE
integrated_gpu = fAlSe
uma_64_bit = FAlSE
available.unlimited_video_memory = TRUe
current.unlimited_video_memory = FalsE

[SPARTAN VI]

Here are my benchmark results with 2x GTX 770 2GB in SLI. I've made corrections to my configuration that alQamar found as well.

Intel i5 3570K, @4.4 GHz, 16 GB RAM 1600 MHz CL 11-11-11-30, SLI 2x GTX 770 2GB, Windows 7 Home 64bit, Resolution 1920x1080

extreme quality:
min FPS: 2
avg FPS: 20.2
max FPS: 39

maximum quality:
min FPS: 31
avg FPS: 40.6
max FPS: 54

quality:
min FPS: 40
avg FPS: 49.4
max FPS: 67

performance:
min FPS: 56
avg FPS: 66
max FPS: 80

maximum performance:
min FPS: 93
avg FPS: 116.5
max FPS: 162

Spoiler Alert, click show to read: 

JON-PC.txt

[alQamar]

great to see that Spartan. Your settings are now perfect, as you fixed all configuration issues.
Glad to see your SLI is working now, I assume that at least. If you actually can redo the benchmark for the non SLI config, I would be happy to update this in the original thread aswell.

I stick to my opinion that the performance in Attila is really absymal. A modern quadcore CPU with 4,4 GHz is far above the usual equipment of a usual gamer, aswell as having a SLI consisting of 2 770GTX is really ambitious.

To achieve 66 fps avg using performance settings only, a preset that is somehow comparable to Rome 2 "high" settings is laughable slow. We have to bear in mind that a lot of demanding stuff is disabled or downgraded already!
This speed is something you could have expected for Shogun 2 maxed out or Rome 2 ultra or extreme preset.

Attila needs performance patches (not again LoD adjustments), so Rome 2 needs them still.

[SPARTAN VI]

great to see that Spartan. Your settings are now perfect, as you fixed all configuration issues.
Glad to see your SLI is working now, I assume that at least. If you actually can redo the benchmark for the non SLI config, I would be happy to update this in the original thread aswell.

I stick to my opinion that the performance in Attila is really absymal. A modern quadcore CPU with 4,4 GHz is far above the usual equipment of a usual gamer, aswell as having a SLI consisting of 2 770GTX is really ambitious.

To achieve 66 fps avg using performance settings only, a preset that is somehow comparable to Rome 2 "high" settings is laughable slow. We have to bear in mind that a lot of demanding stuff is disabled or downgraded already!
This speed is something you could have expected for Shogun 2 maxed out or Rome 2 ultra or extreme preset.

Attila needs performance patches (not again LoD adjustments), so Rome 2 needs them still.

I'll rerun the single card benchies, but I doubt they'll change much, if at all, because I was running my RAM in single channel.

In the mean time, I figured I'd lay out the SLI performance scaling I've seen. I'll revise these figures if the single GTX 770 performance changes.

(edit) Shows what I know. Values have been modified:

+04% improvement avg fps @ Extreme preset
+43% improvement avg fps @ Max quality preset
+04% improvement avg fps @ Quality preset
-01% decrease avg fps @ Performance preset
+09% improvement avg fps @ Max performance preset

[SPARTAN VI]

Here are my benchmark results with SLI disabled (i.e. single GTX 770 2GB) after the corrections were made to my configuration that alQamar found. Basically, I was running my RAM in single channel, so I moved the module over 1 slot, and my UEFI/BIOS reset my RAM speeds to 1333MHz instead of 1600MHz. After correcting these, I see as high as a 10% improvement in some places, but little change at max quality/extreme presets.

Intel i5 3570K, @4.4 GHz, 16 GB RAM 1600 MHz CL 11-11-11-30, GTX 770 2GB, Windows 7 Home 64bit, Resolution 1920x1080

extreme quality (unlimited memory checked):
min FPS: 14
avg FPS: 19.9
max FPS: 24

maximum quality:
min FPS: 23
avg FPS: 28.4
max FPS: 36

quality:
min FPS: 38
avg FPS: 47.6
max FPS: 71

performance:
min FPS: 46
avg FPS: 67.6
max FPS: 79

maximum performance:
min FPS: 73
avg FPS: 106.1
max FPS: 129

Spoiler Alert, click show to read: 

JON-PC.txt

[alQamar]

I'll rerun the single card benchies, but I doubt they'll change much, if at all, because I was running my RAM in single channel.

In the mean time, I figured I'd lay out the SLI performance scaling I've seen. I'll revise these figures if the single GTX 770 performance changes.

+77% improvement @ Extreme preset
+44% improvement @ Max quality preset
+16% improvement @ Quality preset
+13% improvement @ Performance preset
+20% improvement @ Max performance preset

Fair enough, the scalation of SLI is great in higher settings but why they seem so unemployed in lower ones - so no improvement happens there.

Thanks for the new results without SLI I will take them in!

How can you run the extreme preset with just 2 GB VRAM Spartan? I cannot without getting downgraded.

[SPARTAN VI]

Fair enough, the scalation of SLI is great in higher settings but why they seem so unemployed in lower ones - so no improvement happens there.

The results you quoted were before I fixed my setup. I've since ninja edited that post; the SLI scaling is much less pronounced, but just measuring averages doesn't paint the entire picture. I'd still take the -1% average fps because when looking at the graph, you see the SLI fps was constantly in a higher range whereas the single GPU was all over the place. I'd predict the higher variance would translate into stuttering.

How can you run the extreme preset with just 2 GB VRAM Spartan? I cannot without getting downgraded.

Every time I run the Extreme preset, I checked the box for unlimited video memory. I'm assuming the only difference between the Max Quality preset and the Extreme preset is Texture Filtering is bumped up from 2xAF to 4xAF, and Anti-Aliasing is bumped from MLAA to 4xAA.

[alQamar]

do you think this would be better readable? Thanks for comments!
>>

The results you quoted were before I fixed my setup. I've since ninja edited that post; the SLI scaling is much less pronounced, but just measuring averages doesn't paint the entire picture. I'd still take the -1% average fps because when looking at the graph, you see the SLI fps was constantly in a higher range whereas the single GPU was all over the place. I'd predict the higher variance would translate into stuttering.



Every time I run the Extreme preset, I checked the box for unlimited video memory. I'm assuming the only difference between the Max Quality preset and the Extreme preset is Texture Filtering is bumped up from 2xAF to 4xAF, and Anti-Aliasing is bumped from MLAA to 4xAA.

Ok Spartan, so basically I would need to remove your computer from the extreme preset, because it should be run natively with VRAM, by ticking unlimited VRAM he will use the RAM of your computer, interestingly enough it doesn't slow you and more over not even makes a difference whether you have dualchannel or not... this game is a miracle. Sad we will never see a developer comment why this actually works out, while technically it seems impossible to me.

[SPARTAN VI]

do you think this would be better readable? Thanks for comments!
>>




Ok Spartan, so basically I would need to remove your computer from the extreme preset, because it should be run natively with VRAM, by ticking unlimited VRAM he will use the RAM of your computer, interestingly enough it doesn't slow you and more over not even makes a difference whether you have dualchannel or not... this game is a miracle. Sad we will never see a developer comment why this actually works out, while technically it seems impossible to me.

That's fine. I can agree with that. I made a mistake to the SLI extreme benchmark anyway that I just sent a PM to you about.

I like the format we currently have in the OP. If you post it as excel screenshots, we wont be able to use our browsers to search for details.

[alQamar]

That's fine. I can agree with that. I made a mistake to the SLI extreme benchmark anyway that I just sent a PM to you about.

I like the format we currently have in the OP. If you post it as excel screenshots, we wont be able to use our browsers to search for details.

a good point against it...

how about that? the problem except the point you made is this is also blurry and I dont know how it looks like when you don't have a 1080p resolution... many still don't have that.

[IamOdysseus]

a good point against it...

how about that? the problem except the point you made is this is also blurry and I dont know how it looks like when you don't have a 1080p resolution... many still don't have that.

For searching details, you could upload the excel and then we can look for details in the file

[SPARTAN VI]

Couple of corrections alQamar.

I have two PNY GTX 770 2GB, not MSI. My username has a roman number 6, (VI), not 4 (IV).

[alQamar]

Couple of corrections alQamar.

I have two PNY GTX 770 2GB, not MSI. My username has a roman number 6, (VI), not 4 (IV).

ok I will edit that, thanks for pointing that out.

Hope we get some more pro / cons how to present the results in future, keep up the discussion and keep more results coming.
Perhaps it would be helpful to include this thread in your signatures too, as it is now moved to the tech subforum I don't think many will see it, mostly because the tech subforum was never really helpful for Rome 2 players as CA don't reply there on requests.

While I still want to keep this thread clean and objective you can (hopefully) enjoy some related memes in the Attila Humour thread.

[SPARTAN VI]

Hoping to see some R9 290X results. I'm still on the fence for my $300~ GPU recommendations, so I'm really curious to see how the GTX 970 stacks up against the R9 290X here.

[A Barbarian]

Hoping to see some R9 290X results. I'm still on the fence for my $300~ GPU recommendations, so I'm really curious to see how the GTX 970 stacks up against the R9 290X here.

At the moment, the GTX970 and the R9 290X are about the same speed until u use 4xMSAA. Then the GTX970 is faster. MSAA optimization is really lacking in the AMD drivers. That was always the case when compared to NVIDIA and even more so in Attila. According to these benchmarks activating 4xMSAA in Attila costs the GTX980 9% performance, thereas for the 290x the framerate drops by 32% .

AMDs hardware is potentially more capable (e.g. much more memory bandwidth) but NVIDIA makes better use of their hardware (e.g. texture compression, much better driver optimizations for slower CPUs compared to AMD, ...).

AMD also just posted a job opening for GPU driver optimization . So maybe in 6 months we can see the beginning of better driver optimizations from AMDs side

[alQamar]

At the moment, the GTX970 and the R9 290X are about the same speed until u use 4xMSAA. Then the GTX970 is faster. MSAA optimization is really lacking in the AMD drivers. That was always the case when compared to NVIDIA and even more so in Attila. According to these benchmarks activating 4xMSAA in Attila costs the GTX980 9% performance, thereas for the 290x the framerate drops by 32% .

AMDs hardware is potentially more capable (e.g. much more memory bandwidth) but NVIDIA makes better use of their hardware (e.g. texture compression, much better driver optimizations for slower CPUs compared to AMD, ...).

AMD also just posted a job opening for GPU driver optimization . So maybe in 6 months we can see the beginning of better driver optimizations from AMDs side

The job offer is a really interesting find Barb, thanks for your comments, when it comes to GPU tech - not limited though - I am not able to hold a candle to you.

So what's your choice about the result posting ? Excel or text based as it is now?

[booker]

I will post my benchmark later today. But one thing in noticing is that max quality are giving 29fps. Same as some folks in here. But quality only gives me a 36fps compared to 45+ of the same people that had 29/30fps on max quality.

It doesn't make sense.