My "Underpowered" GPU Works Just Fine, Thanks

[nanoflower]

JustAnEngineer,

Here's an example of what older hardware can do when combined with newer hardware.
http://www.3dmark.com/3dm/166001?
Ice Storm Score 34051 Graphics Score 62687 Physics Score 13103

Cloud Gate Score 4770 Graphics Score 17874 Physics Score 1338

Fire Strike Score 2290 Graphics Score 2999 Physics Score 1787 Combined Score 975


As you can see where my hardware lacks is in the physics area because my E5200 (even at 3GHz) is no comparison to your 3570K. But with the 650TI the overall performance isn't bad especially when you consider that this benchmark is really a 2K screen from what I've heard.

[MadManOriginal]

This article link was in the Friday Shortbread. Sadly they don't include any modern CPUs but it does show that some modern games definitely benefit from quad cores. It would have been nice if they'd included something like a 2c/4t i3 and a quad core i5 for comparison sake, but if you really want to do the work you can extrapolate from various reviews. It also shows that with a modern card, 'old' CPUs still give playable framerates - the Core 2 architecture really was a huge leap forward and is still usable.

Anyway, yes, older video cards that at least have unified shaders can still be decent. Running out of VRAM might be the biggest issue and cause the greatest performance drop-off. Plus if you only play older games then an older card won't suddenly become slower, power draw and noise then become the main upgrade reasons.

[CampinCarl]

Okay! So, with my main rig (which is in my sig), I tested four games. Chivalry, Planetside 2, Battlefield 3, and Warhammer 40K: DoW 2. These are four games I play frequently, so I felt they might be fun to benchmark. Here are my notes:

Okay, so, some test data from my main rig. First off, I should mention a few things:
Running 12.10 drivers
Programs still running in background:
Firefox 18.0.2 with 1 Tab (TR Home page)
Pidgin
Ventrilo
Sticky Notes
XFire
Steam
Notepad
MSSE
CoreTemp

Plus drivers that have systray icons:
Asus Xonar Driver Panel (Uni Xonar v1.54)
Logitech Gaming Software 8.3.86(G15)

Now, with that bit of bookkeeping out of the way, I tested a plethora of random games.
I tried to give ~5 minutes of playtime from each game, since I was only playing through once.

Chivalry -
50 Players on a 64 player server, playing Team Objective on Darkforest, first segment.
Installed to 1TB Samsung Spinpoint F3 HDD
Resolution: 1680x1050
Graphical Settings: All set to highest possible.

Planetside 2 -
Playing as NC on Connery server, from spawn @ Indar Warpgate, flying in a Reaver to Zurvan Amp Station,
and participating in the capture of Zurvan from the VS through infantry combat as Heavy Assault,
plus HE vanguard as an Engineer.
Actual play time of about 15 minutes between the two files.
Installed on 1TB Samsung Spinpoint F3 HDD
Resolution: 1680x1050
Graphical Settings: See attached screen shot.

Battlefield 3 -
Playing Multiplayer, 58 Player server on Conquest Large Operation Firestorm
PLUS same server on Kharg Island Conquest Large
Resolution 1680x1050
Graphical Settings: See attached screen shot.

Warhammer 40K DoW II
Built-in Test (avg. over 3 tests)
Resolution: 1680x1050
Graphical Settings: Ultra Preset
Results:
Min 54.56 52.89
Avg 87.98 88.54
Max 180.63 182.45
Plus!
First level of Calderis part of the original campaign!
Same settings, as FRAPS frametime data.

The FRAPS data (FRAMETIMES!) can be found from a google drive account I created:
Chivalry
Planetside 2 Part 1
Planetside 2 Part 2
BF3 Op. Firestorm
BF3 Kharg Island
Warhammer 40K DoW II Part 1
Warhammer 40K DoW II Part 2

I'll do some graphs and get them on imgur asap.
Edit: Config settings!
BF3
Planetside 2

[tanker27]

You guys kill me with your 3dmark numbers and exacting framerates. If the game is playable satisfactorily on the resolution and settings you like then what does it really matter? This is nothing more than flexing epeen. (notwithstanding the microstutter that ticks me off)

[JustAnEngineer]

You guys kill me with your 3dmark numbers and exacting framerates. This is nothing more than flexing epeen.

That's right. How did yours measure up?

The point of the benchmarks is that they provide an objective measurement that we can compare without everyone having to get together with their PCs in the same room so that we can compare the game experiences on each one.

There are times when I'd like more performance while gaming. It hasn't yet reached the point that I'm willing to spend money to make it better. Knowing what sort of relative performance other folks get with their systems provides a reference.

[tanker27]

But thats just it, it's not objective. Drivers for AMD and Nvidia have been cheating for years! And 3dmark is far from objective.

I dont really care what my 'number' is. It comes down to if I can play a game or use a program. If I have problems I will troubleshoot and sometimes that may lead me to new hardware. But is hasnt yet, like I said before I am rocking an i7 920 (1366) with a 560 Ti and I have no problems playing anything new at max.

I dont mean to thread bash (even though I recognize I have), I'll just let you all carry on.

[Chrispy_]

Without detracting from the obvious awesomeness that is TR's GPU reviews, I find Anand's BENCH tool incredibly useful for this exact subject.

[axeman]

Rocking a 6770 on an i3 system. Max resolution for my monitor is 1680x1050, and if it can't do medium type detail at that res, I drop to whatever other 16:10 option there is, like 1440x900 or 1280x800. I'm a pretty casual gamer now, and giant, hot, noisy, power sucking GPUs are not for me. I'm not interested in anything that requires my desktop to have > 350w PSU nowadays.

You could build an i5 3570k and fast mid-range GPU like a GTX 660 system that wouldn't draw more than 250W (not overclocked) max...I'm talking unrealistic loads like Prime95+Furmark. In gaming such a system would probably pull 150W-200W.

i3-2120 + 7870 here; I do not break 200 W while gaming.

You probably still need a > 300w PSU so you have the requisite current on the 12v rails, though. I'm more looking at it from the standpoint of larger GPUs putting out more heat than the CPU and having somewhat noisy coolers the majority of the time. BTW, I think my rig draws something like 120w. Including the monitor (I wouldn't have believed it unless I seen the kill-a-watt myself)

[CampinCarl]

You guys kill me with your 3dmark numbers and exacting framerates. If the game is playable satisfactorily on the resolution and settings you like then what does it really matter? This is nothing more than flexing epeen. (notwithstanding the microstutter that ticks me off)

That's kind of the point of the thread, though. It's an attempt to build up some data on configs that may nor may not be tested by sites like TR and Anandtech. This would allow another person to come in, look at the data, and say "Hmm...this person here has a config similar to what I'm looking at, and play at settings I want to play at, but in game [xyz] the performance isn't quite good enough!" and make a decision from there. It doesn't even necessarily need to be objective, I think that frametimes (not 99th percentile, but looking at a whole frametime plot) would definitely allow people to get a feel for how it plays.

There's a new Top Gear on tonight so I probably wont' be able to do much more testing, but I did manage to download the same games I tested on my main rig yesterday, onto my old, last night. Hopefully I can at least do those tests tonight. I don't have a huge resolution monitor to test various screen sizes, but I will try and get some different settings configs tested on the monitors I have available (22" 1680x1050, 21.5" 1600x900). Hopefully that'll help people. Maybe someday I'll get one of those 27" Korean IPS displays.

[DPete27]

I find Anand's BENCH tool incredibly useful for this exact subject.

I use Anand-Bench occasionally as well, but they suffer the same shortcomings as TR. They can only test so many games at so many video settings. What if none of the games you play/will play are on their list? Or if need be, how much do you have to lower the video detail in order to achieve "playable" framerates? (anandtech has to retain some level of consistency in their settings to allow various cards to be compared to each other) You can still enjoy a game at medium-low video settings if your budget is tight.

The unfortunate truth is that, unless you only play the popular AAA titles, it's difficult to know how graphically demanding a given game really is. (I realize that it's safe to assume that the most popular games used for benchmarking represent the most demanding loads and that lesser known and lower budget titles would presumably be easier to run, but again, that's a qualitative assumption)
The impossible solution to all this is to benchmark every game on a variety of GPUs with a variety of video settings. Notebookcheck.net has a pretty extensive list of benchmarked games for variety of mobile GPUs. I haven't been able to find a similarly extensive benchmark list for desktop GPUs, hence the purpose of this thread.

[MadManOriginal]

axeman - true, but I always assume a high quality PSU. Maybe that's a mistake in the broader sense but for people reading TR or using the SBA forum who are recommended high quality PSUs it's not a bad assumption. Now there aren't many high quality 350W-range PSUs, there are many more in the 450W range, but more companies are starting to make good, low-power PSUs.

I hadn't looked up numbers, but now that I do <=200W DC is probably about right in gaming loads for the system I stated. Here's a TR review showing at the wall power (so PSU efficiency is a factor) with a higher power consumption platform and CPU (X79 and 4c/8t CPU) of ~230W. At a good 90% efficiency that works out to DC draw of ~207W. (The lower the PSU efficiency, the lower the DC draw for a given AC input.)

[axeman]

But thats just it, it's not objective. Drivers for AMD and Nvidia have been cheating for years! And 3dmark is far from objective.

I dont really care what my 'number' is. It comes down to if I can play a game or use a program. If I have problems I will troubleshoot and sometimes that may lead me to new hardware. But is hasnt yet, like I said before I am rocking an i7 920 (1366) with a 560 Ti and I have no problems playing anything new at max.

I dont mean to thread bash (even though I recognize I have), I'll just let you all carry on.

I'm not a developer, and I haven't paid attention to this in recent years either, but I also recall hearing that 3DMark doesn't even really use an actual game engine; so it's entirely meaningless. Also, there was something about it using uncompiled shader code which no game developer would ever do either. TR Experts, feel free to weigh in here...

[PainIs4ThaWeak1]

3930k @ 1.30v/44x-plier (4.44Ghz) & GTX 680 SLI ... Guess I really don't have to elaborate. :| (But don't think I'm attempting to throw it in anyone's face here either...)

But BL2 @ 1920x1080, full eye candy (e.g. all quality options set to their maximum states), capped @ 120fps, is more "fluid" than I ever realized a game could be. I don't have FPS numbers, but I'm sure they're a plenty on TR/around the web. But I think what is important to me, is that the experience playing games with my current setup is more... "responsive" (or "fluid" as I mentioned earlier) - To me its the kind of "noticable" that comes from comparing a video @ 30FPS vs. 60FPS.

[Firestarter]

You guys kill me with your 3dmark numbers and exacting framerates. If the game is playable satisfactorily on the resolution and settings you like then what does it really matter? This is nothing more than flexing epeen. (notwithstanding the microstutter that ticks me off)

It matters because what you may consider satisfactory may be too slow for my tastes. It's a subjective measure, one that has value in addition to objective benchmarks, but you cannot replace one with the other. Objective benchmarks alone cannot show the complete picture, and neither can subjective impressions.

[Firestarter]

Edit: Config settings!
Planetside 2

FYI: PS2 has hidden 'ultra' settings. You can enable them by editing your UserOptions.ini file in the game directory. To give you an idea, I have a 2500K and an overclocked HD7950, and I use these options:

Code: Select all

[Rendering]
GraphicsQuality=4
TextureQuality=0
ShadowQuality=3
RenderDistance=4000.000000
Gamma=0.000000
MaximumFPS=250
UseLod0a=0
VSync=0
OverallQuality=-1
LightingQuality=4
FogShadowsEnable=1
EffectsQuality=4
TerrainQuality=4
FloraQuality=4
ModelQuality=4
ParticleLOD=3
MotionBlur=0
AO=1
VerticalFOV=74


That's pretty much 'ultra', but with shadows at 'high' (ultra shadows kill framerate). The most important one is TextureQuality=0, which gives you a lot more texture detail for almost 0 performance cost.