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

[SPARTAN VI]

This might be interesting to u guys

https://www.reddit.com/r/pcmasterrac..._support_dx12/

TL;DR: Gefore 900 series doesn't support Async compute\shaders, nvidia cards will suck at dx12 games, especially if you wanna use VR. (also it does explain why did nvidia cards had worse FPS in dx12 than in dx11 in some cases in that RTS benchmark.)

Gotta call this FUD out. We're talking about Ashes of the Singularity, which is just one beta DX12 game. A couple weeks ago it got out that a GTX 980 Ti was being shown up by a R9 290X in that game. For those not keeping track, the GTX 980 Ti is about $650 USD and the R9 290X can be had for around $250 USD, so this was raising eyebrows in the PC gaming enthusiast communities. However, non-critical thinkers latched onto that as gospel that current Nvidia hardware wouldn't be able to keep up with AMD in ALL DX12 games.

Well, Guru3D just ran a review of the R9 Nano on 9/10/15 compared to a R9 390X and GTX 980 Ti and included the latest AotS beta build... and guess what? The GTX 980 Ti performed 20-30% better than both, while the R9 Nano ($650) and R9 390X ($400) performed about the same as each other. So as you can see, native async compute support is not the determining factor in DX12 performance after all.

Honestly, I'm a bit conflicted. I'd like to see AMD stake a more dominant claim in the PC gaming segment to at least upset the order of things and remain competitive. At the same time I'm glad that my GTX 970 will likely still be relevant with the coming DX12 releases, assuming I'll still have a GTX 970 when that day comes.

[RaymanTWC]

Been following the thread for ages, just thought I'd make an account to jump in the conversation though. I have some benches, and I'll post them along with some from the new patch to see if its improved anything, but I cant seem to open to game for whatever reason. Ill figure it out later.

Gotta call this FUD out. We're talking about Ashes of the Singularity, which is just one beta DX12 game. A couple weeks ago it got out that a GTX 980 Ti was being shown up by a R9 290X in that game. For those not keeping track, the GTX 980 Ti is about $650 USD and the R9 290X can be had for around $250 USD, so this was raising eyebrows in the PC gaming enthusiast communities. However, non-critical thinkers latched onto that as gospel that current Nvidia hardware wouldn't be able to keep up with AMD in ALL DX12 games.

Well, Guru3D just ran a review of the R9 Nano on 9/10/15 compared to a R9 390X and GTX 980 Ti and included the latest AotS beta build... and guess what? The GTX 980 Ti performed 20-30% better than both, while the R9 Nano ($650) and R9 390X ($400) performed about the same as each other. So as you can see, native async compute support is not the determining factor in DX12 performance after all.

The game being in beta or being just one example of DX12 features isn't really relevant. It showed maxwell simply cant do async compute natively, and it has to achieve it through heavier context switching. This should result in a either less of a performance increase, or a performance hit, which we saw. Since then async compute has been disabled on the benchmark for nvidia cards, so its no longer taking the hit and isn't being closed in on as much by cards half its price.
According to the devs, Oxide, the feature is only used moderately, and "pales with comparisons to some of the things which the console guys are starting to do". Once they come to PC its gonna be interesting at least.

Although I'll admit I haven't really seen the initial benches (by arstechnica I think?) been matched, and obviously async compute isn't the only important feature of DX12, but from what we know AMD do seem to have a big advantage going into dx12.

The nano is priced because of its niche form factor and short HBM supply, not performance. I don't really see the market for it honestly, but that's another discussion. Using its price there is a bit disingenuous.

(Also, please dont go around using stuff like "FUD" and "non-critical thinkers", just talk, man)

[SPARTAN VI]

The game being in beta or being just one example of DX12 features isn't really relevant. It showed maxwell simply cant do async compute natively, and it has to achieve it through heavier context switching. This should result in a either less of a performance increase, or a performance hit, which we saw. Since then async compute has been disabled on the benchmark for nvidia cards, so its no longer taking the hit and isn't being closed in on as much by cards half its price.
According to the devs, Oxide, the feature is only used moderately, and "pales with comparisons to some of the things which the console guys are starting to do". Once they come to PC its gonna be interesting at least.

Although I'll admit I haven't really seen the initial benches (by arstechnica I think?) been matched, and obviously async compute isn't the only important feature of DX12, but from what we know AMD do seem to have a big advantage going into dx12.

The nano is priced because of its niche form factor and short HBM supply, not performance. I don't really see the market for it honestly, but that's another discussion. Using its price there is a bit disingenuous.

(Also, please dont go around using stuff like "FUD" and "non-critical thinkers", just talk, man)

I should've pointed out specifically what I was responding to which was the "nvidia cards will suck at dx12 games" comment in the post I quoted. My point is that there's a lot of folks trying to spread fear uncertainty and doubt over what basically is an incomplete comparison. Of course you already know that.

I should have also mentioned that the R9 Nano is basically a smaller version of the R9 Fury, which is a $550 GPU but in the traditional dual-slot form factor.

[A Barbarian]

I should have also mentioned that the R9 Nano is basically a smaller version of the R9 Fury, which is a $550 GPU but in the traditional dual-slot form factor.

Actually, the Nano is a lower clocked Fury-X since it has the full version of the chip. The Fury is a cut down version.
Nano and FuryX have 4096 Stream processors and 256 texture units, the Fury 3584 and 224.
The Nano has a much lower power limit which in practice results in a lower clock than for the FuryX.

[A Barbarian]

Yeah, that kinda sucks. But until there are some DX12 games out which actually use that feature a lot there will be the next generation of GPUs out (Pascal / Arctic Islands). Until then, I enjoy the great performance and features (DSR, ShadowPlay, ..) of the 970s. Before the 970s, I had 2x7970s but primarily the lack of proper driver (no SLI support for many newer games) and feature support (no VSR for 7970s for a long time) made me to get the 970s. AMDs hardware is usually great, primarily their drivers/software is often lacking.

I hope that AMD will be able to offere a more competetive product in the future. That would be only good for competition and all users.

[karaislam]

amd fx 6300
r7 260x
8gb ram
500 watt psu
1366x768

can i play with this in ultra settings?

[A Barbarian]

@karaislam

Your CPU will be the main bottleneck. If u play with smaller armies and with lowered shadow settings, u could probably play with most other settings at 'ultra' (not extreme).

It also depends on how much FPS u expect. Even with a high-end PC I would set my target framerate somewhere between 25 and 40 fps and maximize the settings. Simply because even with the fastest computer and some decent (not extreme) settings, the framerate will dip quite often to below 30 fps. And some people can even live with 20 fps average.

My standard recommendation in such a case is to use the ultra (not extreme) preset as a starting point and then:

switch off (if not already off):
- SSAO
- Screen Space Reflections

Lower
- unit size to 'large' or 'medium'
- shadow quality to 'performance'
- particle effects to 'quality'

And possibly switching off 'blood effects'.

[karaislam]

@karaislam

Your CPU will be the main bottleneck. If u play with smaller armies and with lowered shadow settings, u could probably play with most other settings at 'ultra' (not extreme).

It also depends on how much FPS u expect. Even with a high-end PC I would set my target framerate somewhere between 25 and 40 fps and maximize the settings. Simply because even with the fastest computer and some decent (not extreme) settings, the framerate will dip quite often to below 30 fps. And some people can even live with 20 fps average.

My standard recommendation in such a case is to use the ultra (not extreme) preset as a starting point and then:

switch off (if not already off):
- SSAO
- Screen Space Reflections

Lower
- unit size to 'large' or 'medium'
- shadow quality to 'performance'
- particle effects to 'quality'

And possibly switching off 'blood effects'.

thanks for the info.

with that system u think i must buy rome 2 or attila?with rome 2 can i ve better performance?i must buy one of the game.what is ur opinion?

[A Barbarian]

Buy Rome 2.

In the base version it lacks some of the features present in Attila, like the family tree. But it will perform better, offer more variety in factions and units and have better mod support. And u should be able to pick it up relatively cheap on some sale. Personally, I am also not too much of a fan of the theme and 'mood' of Attila.

[Druout]

Buy Rome 2.

In the base version it lacks some of the features present in Attila, like the family tree. But it will perform better, offer more variety in factions and units and have better mod support. And u should be able to pick it up relatively cheap on some sale. Personally, I am also not too much of a fan of the theme and 'mood' of Attila.

I like Rome 2 better out of the two myself, but I use Attila for the multiplayer campaign.

[nosuchname]

@karaislam

Your CPU will be the main bottleneck. If u play with smaller armies and with lowered shadow settings, u could probably play with most other settings at 'ultra' (not extreme).

It also depends on how much FPS u expect. Even with a high-end PC I would set my target framerate somewhere between 25 and 40 fps and maximize the settings. Simply because even with the fastest computer and some decent (not extreme) settings, the framerate will dip quite often to below 30 fps. And some people can even live with 20 fps average.

My standard recommendation in such a case is to use the ultra (not extreme) preset as a starting point and then:

switch off (if not already off):
- SSAO
- Screen Space Reflections

Lower
- unit size to 'large' or 'medium'
- shadow quality to 'performance'
- particle effects to 'quality'

And possibly switching off 'blood effects'.

worst are those fanboys on the steam and offical total war forum, pretending like they can run this game maxed out without dropping below 40-30 fps. there was even a guy who was insisting that he could run the game on ultra settings with 100+ fps and they were telling me that my pc simply was broken. I even post video footage of other people with high end system who had performance issues as well, but no they keep on lying, damn I hate fanboys, they just want to ruin everything for everybody.

the game runs like on my gtx 980 ti and intel i5 4690k

[alQamar]

worst are those fanboys on the steam and offical total war forum, pretending like they can run this game maxed out without dropping below 40-30 fps. there was even a guy who was insisting that he could run the game on ultra settings with 100+ fps and they were telling me that my pc simply was broken. I even post video footage of other people with high end system who had performance issues as well, but no they keep on lying, damn I hate fanboys, they just want to ruin everything for everybody.

the game runs like on my gtx 980 ti and intel i5 4690k

time for mega laugh but don't feel too much concerned about those myths... some even claimed the same for Rome 2 and their pretty laptops I fully agree with you about the existence of such fanboys they deny to admit they are not satisfied with the performance of either Rome 2 and Attila.

[nosuchname]

amd fx 6300
r7 260x
8gb ram
500 watt psu
1366x768

can i play with this in ultra settings?

you serious, hell no. thuis game is poorly optimized

[A Barbarian]

Hi alQamar, do u have some time to update the benchmarks with the latest submissions?

[alQamar]

oh my... I have read the link and linked it to my approriate tech thread.
http://www.twcenter.net/forums/showt...-Rome-2/page18

Still I take the reddit news with a grain of salt as long we do not have other DX12 performance references like those from Futuremark etc.

Hi alQamar, do u have some time to update the benchmarks with the latest submissions?

I will work on it tomorrow, I am ill for this week so I can spend time on TWC.

[Fanest]

Beside what RymanTWC wrote problem is that nvidia also falsely advertised their cards for being able to use this feature (much like CA with rtw2) and thats also causing quite a bit of dissatisfaction in this story.

[Totalheadache]

Beside what RymanTWC wrote problem is that nvidia also falsely advertised their cards for being able to use this feature (much like CA with rtw2) and thats also causing quite a bit of dissatisfaction in this story.

the whole Dx12 async thing was a load of bollocks. there will be more pre dx12 launch....as there is with gpu launches.

Read guru3d.com/forum for real tech news not from anywhere else as it's most likely complete tosh.

And yes it was FUD.

p.s Nano is somewhat of a fail card...better off buying a 980ti, and the "its for small builds reason" thats crud as well.

[RaymanTWC]

the whole Dx12 async thing was a load of bollocks. there will be more pre dx12 launch....as there is with gpu launches.

Sorry, I dont understand you here. What do you mean?

Read guru3d.com/forum for real tech news not from anywhere else as it's most likely complete tosh.

Ive always used Tomshardware and Anandtech myself. Although I'm pretty sure theyre all owned under the same guys now, those two have much better reviews than guru3d imo. And man I hate guru3d's charts.
There are plenty of other good, reputable review sites too, including ars technica, who did the initial benchmarks. Limiting yourself to one is just exposing yourself to whatever biases that site may have, same with all media sources.

p.s Nano is somewhat of a fail card...better off buying a 980ti, and the "its for small builds reason" thats crud as well.

I mean, there are cases that cant fit a full size card in, so its supposed to fill that niche and show off the space savings HMB brings. Though I agree in the vast majority of even mini-itx builds you can fit one, and you're better off with a 980ti.
There's probably a better place to discuss this though.

[Fanest]

there will be more pre dx12 launch....as there is with gpu launches.

elaborate please

Read guru3d.com/forum for real tech news not from anywhere else as it's most likely complete tosh.

links to relevant thread please

p.s Nano is somewhat of a fail card...better off buying a 980ti, and the "its for small builds reason" thats crud as well.

its the most powerful mini itx graphic card designed for this reason only so u see why?

https://youtu.be/rsOkaDp2mlQ?t=4m30s

https://www.youtube.com/watch?v=OCqTw92tC9Q

[nebucodonosor]

I've been following the tread for a while , yet never posted. Thank you Al'Qamar for taking the time and provide us with benchmarks on how Attila runs on different systems .
In my case , with the following specs , I can't complain about performance but indeed despite the advertising the engine is poorly optimize to accommodate battles with more than 10000.00 units in battle.
The game benchmark with the following system :
MOBO : Gygabite P67 B3 UD3 , CPU : i5-3.3 OC at 4.2 GHZ , GTX 980 TI MSI , 8G or RAM 1666 , Windows 7 64
Game settings : All on MAX settings except , Shadows - HIGH and Reflections - HIGH , with all extra features on SAAQ , VSINC etc , I get MIn 24 FPS and MAX 40 FPS with an average of 32.9 FPS .
The game is quite enjoyable and can't complain about performance , except in cases when I have 15000.00 - 20000.00 soldiers in the battle.