Steam for Linux > Limited Beta > 제목 정보
panosgeorgiadis 2012년 12월 7일 오후 4시 39분
Serious Sam 3 Performance review using 17 GPUs
I've spent the last three weeks benching with my PC running Ubuntu 12.04 32bit and a bunch of modern graphics cards, kindly provided by my overclockers friends.

http://ubuntuxtreme.com/serious-sam-3bfe-gpu-graphics-performance-in-ubuntu/

Any feedback, is appreciated :D
11개 중 1-11 표시중
< >
ThOR27 2012년 12월 7일 오후 4시 47분 
Awesome Article and benchmark! It's impressive as there is an ATI card on top of the chart, I'm a bit away about hardware nowadays, Do you think you got any impact in speed because of drivers?

Thanks!
drjoms 2012년 12월 7일 오후 4시 49분 
+1
dilweed 2012년 12월 7일 오후 4시 51분 
You, sir, are a gentleman and a scholar. Can you make the timedemo and / or methodology you used available? And if you need a GeForce 285 GTX, 9800 GX2, 9800 GTX, or a 460 SE to complete your set, just let me know ;).
echo 2012년 12월 7일 오후 5시 35분 
"Even the latest AMD Catalyst 12.11 beta driver doesn’t seem to work with my Radeon 4890, so I had to stay with Ubuntu 12.04 default proprietary driver. Also, there is no option for open source ATI drivers either — they don’t meet Steam’s minimum requirements. So, if you own a “legacy” AMD GPU, you have to replace it by another modern GPU as soon as possible…"

Yeh thanks to amd! I will never buy amd cards...
panosgeorgiadis 2012년 12월 8일 오전 6시 32분 
ThOR27님이 먼저 게시:
Awesome Article and benchmark! It's impressive as there is an ATI card on top of the chart, I'm a bit away about hardware nowadays, Do you think you got any impact in speed because of drivers?

Thanks!

Thank you :D In a matter of fact there is no an ATI card o top of the chart but a combination of using two Radeon HD 6970. The CFX means to CrossFireX, which is the hardware implementation for having two exact same Radeon GPUs working together on the same box.

Drivers-wise there is a difference about 10-15 FPS compared to Windows (but DirectX). This difference is not important but only to mediocore GPUs, especially when they score roughly 40 FPS.



dilweed님이 먼저 게시:
You, sir, are a gentleman and a scholar. Can you make the timedemo and / or methodology you used available? And if you need a GeForce 285 GTX, 9800 GX2, 9800 GTX, or a 460 SE to complete your set, just let me know ;).

Thank you very much for your kind words sir :) I would gladly accept your most generous offer but I think there will be way too expensive to ship all these GPUs in Greece. I used to work for Guru3D.com and VR-Zone as a volunteer editor, and during my time there I've met some nice people who are in charge of hardware reviews/sampling in Europe region. I put my faith that they will send me some samples to review hardware for Linux and thus make UbuntuXtreme a great place for Linux hardware enthusiasts.


zikembe님이 먼저 게시:
Yeh thanks to amd! I will never buy amd cards...

It's not their cards, but their crappy drivers :(


Thank you all for your comments, currently I am testing GeForce GTX 650 that my friend Antonis gave me to play with this week. The card is manufactured by MSI. Probably next week there will be a new article ;)

I will try some HD4000 benchmarks too, but I guess I have to find an Ivy Bridge processor first. Duh
z33ky 2012년 12월 8일 오전 8시 17분 
You write that SSAA "does the same as the MSAA does". This is not entirely correct.
MSAA will only render some buffers in a higher resolution, while SSAA will render all buffers in a higher resolution.
Also, is blur in Photoshop really the same as anti-aliasing? AA is not just a blur-filter, it renders the picture in a higher resolution and down-scales it.
The last frame of this animated image[i1.wp.com] is garbled for me.
AlenL 2012년 12월 8일 오전 8시 44분 
Hi panos! Thank you for taking your time for such an exhaustive research with all that hardware and options. Here's one small, hopefully constructive, feedback:

In the "How to setup CPU Governor indicator" section, screenshot shows highlighting "On demand", which is exactly what you don't want. If one reads the text carefully, it describes that powersaving should be off. But I guess most people will see your screenshot and select "on demand". I would recommend that that screenshot is remade with "performance" highlighted.

Thanks!
SUSEd 2012년 12월 8일 오전 8시 56분 
They doesnt tested intel HD graphics :'(
SUSEd님이 마지막으로 수정; 2012년 12월 8일 오전 8시 56분
kwahoo 2012년 12월 8일 오전 9시 33분 
dilweed님이 먼저 게시:
Can you make the timedemo and / or methodology you used available?
+1
I'm curious about the methodology since SS3 does not support timedemo (demo recording to be exact).
Kano 2012년 12월 8일 오전 10시 09분 
The whole benchmarks are FAKE. For at least 2 reasons
a) SS3 does not have got a reproduceable benchmark mode
b) i have got 2 hd 5670 and i am still searching for a game that runs faster with CF on Linux. That CF would work for SS3 with this scaling is completely unbelievable
panosgeorgiadis 2012년 12월 8일 오후 12시 43분 
z33ky님이 먼저 게시:
You write that SSAA "does the same as the MSAA does". This is not entirely correct.

The last frame of this animated image[i1.wp.com] is garbled for me.

Yes z33ky, you're right. I didn't want to delve into too much detail explaining what exactly does MSAA and SSAA, that's the reason why I said they are almost the same. Just because what Tom and Jerry would think looking both screenshots -- without knowing all the background details about buffer and rendering.

BTW the animated gif seems ok if you open it in another tab but for some unknown reason doesn't work properly in Wordpress blogspot. I am searching on it...


AlenL님이 먼저 게시:
I would recommend that that screenshot is remade with "performance" highlighted.

Thank you AlenL, I will replace the image to avoid misunderstandings. Glad to help :)


kwahoo님이 먼저 게시:
dilweed님이 먼저 게시:
Can you make the timedemo and / or methodology you used available?
+1
I'm curious about the methodology since SS3 does not support timedemo (demo recording to be exact).

That was the hard part because of lack of a timedemo I had to play the game again and again in order to be sure that the final average FPS is representing this GPU, in my system, with my settings, in Into the Spider's Nest level. That's why it took me about 3 week to complete all the measurements. Wayyy to much time consuming. Thankfully my next target is Team Fortress and yes sir, there is a timedemo utility in there ;)

Okay here's the procedure you asked for:

STEP 1: Fix the comma bug in keyboard layout
in order to run the benchmark you have to run from console (`) the command bmkStartBenchmarking(start,duration) where start and duration are integer numerical values. The problem is that there is a bug with comma in console mode, thus go to ~/Steam/SteamApps/common/Serious Sam 3/Temp/RCon/History and insert the command (with comma) in the end of file Local.txt. The go back in game and press Up Arrow

STEP 2: Load the game's level "Into the Spider's Nest"

STEP 3: Start benching while playing for 5 minutes.
bmkStartBenchmarking(3,300)

Write down the results.

STEP 4: Restart the game

STEP 5: Load the game's level "Into the Spider's Nest".

STEP 6: Start benching while playing for 5 minutes.
bmkStartBenchmarking(3,300)

trying to simulate the previous gameplay.

write down the results.

STEP 7: Compare these two benchmark runs.

If the results don't differ more than 5 FPS average, do the math: (First Bench + Second Bench) /2 = Final FPS score

STEP 8: Replace the GPU, and goto STEP 1.


shished님이 먼저 게시:
They doesnt tested intel HD graphics :'(

I use an 2500K overclocked at 4.5 GHz, which is a Sandy Bridge processor using Intel HD 3000 graphics -- some may say it is outdated due to the HD4000 -- now that new Intel Ivy Bridge CPUs come along with it. If I get an Ivy Bridge processor then I could get you some scores. But just for the reference, all these IGPU are considered to be for entry-gaming, not even mediocre settings.


Kano님이 먼저 게시:
The whole benchmarks are FAKE. For at least 2 reasons
a) SS3 does not have got a reproduceable benchmark mode
b) i have got 2 hd 5670 and i am still searching for a game that runs faster with CF on Linux. That CF would work for SS3 with this scaling is completely unbelievable

a) I used bmkStartBenchmarking(start,duration) function. That's why I had to run the same benchmark with the same GPU twice in order to get the average FPS. It may be not be zero point exact, but it's okay. Next time I will bench with Team Fortress where there is a timedemo function. Where's the FAKE on that ?

Don't so much accurate, you may say. Yes, true.
Fake results, you say. No, sorry but my results are real.

b) Well I am not using 2x5670, but I am using 2x6970 which is A COMPLETELY DIFFERENT architecture thanks to CAYMAN XT. In fact Cayman GPUs are proven to be so good in working together that AMD developed another dual-core GPU called HD 6990 which is in fact 2x6970 into a single PCB. If you like, take look at my last year's review for CrossFireX performance depending on DirectX9, DirectX10, DireftX11, Tessellation, MSAA and different resolutions. http://www.insomnia.gr/reviews/_/hardware/videocards/παρουσίαση-radeon-6970-και-radeon-6950-r20?pg=30 (it's in greek language so please use google translate)

I use a pair of HD 6970 using CaymanXT architecture. These cards work so good together that AMD decided to make them a sandwich and release AMD Radeon HD6990.

So depending on:
GPU architecture
Drivers
OpenGL, DirectX9, DirectX10, DirectX11
Game engine itself
Resolution
Anti Aliasing
Tessellation


the results could vary from one to another.

For example:
In Call of Duty: Modern Warfare 2, crossfireX doesn't work
- Radeon 6970: 140 FPS
- 2xRadeon 6970: 139 FPS
link: http://www.guru3d.com/articles_pages/radeon_hd_6990_review,14.html

or there is very low scaling
- Radeon 6950: 137 FPS
- 2xRadeon 6950: 150 FPS
link: http://www.guru3d.com/articles_pages/radeon_hd_6950_crossfirex_review,5.html


However, if you pick another game, like Far Cry 2 there is double performance (huge scaling)
- Radeon 6970: 83 FPS
- 2x Radeon 6970: 154 FPS
link: http://www.guru3d.com/articles_pages/radeon_hd_6990_review,15.html

one more
- Radeon 6950: 76 FPS
- 2x Radeon 6950: 147
link: http://www.guru3d.com/articles_pages/radeon_hd_6950_crossfirex_review,6.html

In a matter of fact I am using a 6950 flash into 6970 using TechPowerUp's mod (http://www.techpowerup.com/articles/overclocking/vidcard/159). Here's a video of mine, last year, showing how to do the firmware mod in Windows (http://www.youtube.com/watch?v=0aQ5gQ7EUSo&list=UUdbK4xiJwxGLlRKTSPbq27w&index=2) -- but it's in greek :( cause that video was made for OutOfSpecs overclocking Team.

Just for your information I used to make reviews for: Intel, OCZ, AMD, nVIDIA, G.Skill, ASUS, Gigabyte, MSI, Corsair, Coolermaster, BitFenix, Xigmatek and other hardware manufacturers. I don't want to show off, but I am forced to in order to defend my integrity here. If you still don't believe me, please take a look at my linkedin profile: http://www.linkedin.com/in/drpaneas

I love linux, I use it for ten years now, and today (thank God) Valve is trying to change things here for the better. Better drivers, better support, better hardware, better gaming etc. That's why I've made Ubuntu Xtreme, in order to help and support this movement. I am not earning any money from it, but paying the hosting instead. So if you think that I am posting false and fake stuff then don't visit UbuntuXtreme.

Phoronix tests with obsolete gams, like nexuiz, openarena, Doom3, Unreal Tourname etc. How do you expect to see scaling in there? You have to use Unigine Heaven 3.0 and play with settings in order to find out what's happeing with AMD, Linx, and Crossfirex. But just using games from 1998-2004 era and benching with 2012 modern hardware, no verdicts for me.

Things are in BETA PHASE, so strange things might happen. Don't judge too easily, but try to figure out what's wrong. That has been told, *please* be more careful next time you are about accuse someone.


Panos out.
panosgeorgiadis님이 마지막으로 수정; 2012년 12월 8일 오후 2시 52분
11개 중 1-11 표시중
< >
페이지당: 15 30 50
게시된 날짜: 2012년 12월 7일 오후 4시 39분
게시글: 11