Wednesday, August 4, 2010

Bonus Upgrade!

Left: ASUS 8800 GT, Right: EVGA GTX 460 SC

It's time for an upgrade!  In anticipation of Starcraft 2, and realizing that I don't exactly have much cash to blow on gaming right now, I decided to enter a few hardware giveaways online.  One of them was run by Puget Systems, a business that specializes in custom PCs.  All I had to do was use their website to design a custom gaming PC (gee, never tried that before) and advertise it on Facebook.  They gave away two GeForce GTX 460 768MB video cards, and I was lucky enough to be randomly selected for the second one.  (By the way, Puget Systems just introduced a new line of products that use an aquarium to cool a PC.  I will say this: the presentation looks amazing.)

Up until now, I have been using an ASUS GeForce 8800 GT, which I bought for $130 from Newegg in June 2008.  The new GTX 460 is of the EVGA Superclocked variety, which is currently going for $210 on Newegg.  NVIDIA released the 8800 GT in October 2007, and the GTX 460 in July 2010.  Both were very similar in that, upon release, they provided a uniquely low price/performance ratio as far as NVIDIA cards go.  Both are also quite efficient in the realms of power, heat, and noise.  The GTX 460 is a true successor to the 8800 GT.

I decided this would be a good time to do some benchmarking for fun.  I basically want to see if I can put some measurements on just how much of an upgrade this is.  How will it enhance the games that I currently play?  Is there enough extra oomph for a larger monitor in the future?  Is my computer going to turn into a space heater?

Testing Details

I don't have a whole lot of free time right now, with two small children and another baby due any day.  Therefore, I decided to keep things simple.  I didn't spend time making sure all the drivers were consistent, though I did record the driver numbers for all tests.  I didn't spend much time researching the different types of antialiasing, though I think it was consistent across the board.  I only used games with automated benchmarks.  If I was actually doing a serious hardware comparison, I would have paid more attention to these details.

Here are my system specs:

Operating System: Windows XP Pro SP3, 32-bit / Linux Mint 9, 64-bit
Processor: Intel Core 2 Duo E8400 (OC: 3.6 GHz, 1600 FSB)
Motherboard: Foxconn X38A (PCI-E 2.0 x16)
Memory: OCZ Platinum DDR2-800 2x1GB (5-5-5-15)
Hard Drive: Seagate Barracuda 250GB/16MB
Video Cards: ASUS 8800 GT 512MB / EVGA GTX 460 768MB SC
Power Supply: Apevia 650W
Case: NZXT Tempest

I know; I could use a RAM upgrade.  Thankfully, my motherboard also supports DDR3.  And yes, I have entered a few giveaways just for this purpose!  I also have a copy of Windows 7 Home Premium sitting on my shelf.  I'll install the 64-bit flavor some day - maybe after the first service pack or a RAM upgrade.

Drivers used:

8800 GT Windows: 254.08
8800 GT Linux: 195.36.15
GTX 460 Windows: 258.96
GTX 460 Linux: 256.44

All tests ran at 1600x1200 (I'm still using a CRT), 16x Quality AA, with all in-game details set as high as possible.  Vsync and triple buffering were disabled.  For Company of Heroes, AA was enabled via the game menu.  For Doom 3 and Nexuiz, AA was forced via the NVIDIA control panel.  I used such high levels of AA in order to keep the framerates down to reasonable numbers at this resolution.

Company of Heroes was run in Windows, Nexuiz in Linux, and Doom 3 in both.  Wine was not used in Linux, though I could probably get CoH to run with it if I had the time.  If Wine can run my heavily-modded version of Oblivion (which it does), CoH should be no problem.  I may try it in the future.  Nexuiz was run via the Phoronix Test Suite, an automated benchmarking tool that I had never used before.  I hear it runs in Windows too, so I may visit that in the future as well.  I am likely to re-run some tests when I upgrade to Windows 7.

Core/Memory clocks:

ASUS 8800 GT Stock: 600/900
ASUS 8800 GT Overclocked: 650/1000
Ref. GTX 460 Stock: 675/900 (not tested)
EVGA GTX 460 Superclocked: 763/950
EVGA GTX 460 Overclocked: 825/1000
EVGA GTX 460 Overvolted: 900/1050

The GTX 460 used a voltage of 0.975 by default, and I set it to 1.025 when overvolted.  I didn't try to push the cards in order to get the maximum overclock possible.  I just used what I figured would be safe based on my own research.  I could have probably gotten more out of the new card if I really cared to do so and had the time for testing.  I was unable to overclock the GTX 460 in Linux because the driver didn't seem to recognize the Coolbits option.  It worked fine for my 8800 GT; perhaps the 460 is still too new.  Fortunately, this is a factory "superclocked" edition, I still have speeds in Linux that are higher than reference.


The Company of Heroes results are fairly straightforward, and impressive.  Obviously, I have more frames than I know what to do with.  All indicators suggest Starcraft 2 will be silky smooth as well.

Something must be abnormal with my Windows Doom 3 setup.  Yes, framerates increased, but not nearly as much as Linux; the Linux number is closer to what we should expect.  The Windows performance is inconsistent with the other benchmarks too.  I didn't notice any differences in detail between Windows and Linux, though fine details can be hard to see at these settings.  I know that AA was definitely working in Linux.  I'll probably revisit this in Windows 7, but for right now I am assuming that it's a fluke.  If anyone has any suggestions, I'm all ears.

Nexuiz is an open source FPS built on a heavily modified Quake 3 engine.  It uses several advanced graphical effects, such as HDR, so it's a favorite for Linux benchmarks.  The results here are nothing special, but it is interesting to note that the difference is not as large as the other two games.  This was a great opportunity to try out the Phoronix Test Suite ( for the first time.

This chart is quite impressive.  Even with a hefty overclock, it's still cooler than my old card at stock.  It should be noted that this card actually clocks itself way down (core, memory, and voltage) when idle, which explains the chilly idle temps.  Also, it was interesting to note that overclocking and overvolting only affects the speeds and voltage under load.  When idle, all the numbers are the same across the board.

(It should also be noted that there was a bit of dust on my 8800 GT when I finally uninstalled it.  It wasn't too bad because I clean regularly, but it was enough to possibly make a small difference.)

I really like the physical cooling system on the new hardware as well.  Rather than use a fan in the front that pushes air across the heatsink and out a rear exhaust, it simply blows air directly onto the top of the heatsink.  The ASUS 8800 GT has a similar system.  This has one disadvantage - the hot air is not all vented out the back of the case.  However, it has advantages that I consider to be more important:

1) The GPU itself runs cooler.  If the GPU is the hottest component in my system, I want it to have the best cooling.  In addition, my NXZT Tempest has amazing ventilation, so the hot air gets sucked right out anyway.
2) The fan runs quieter.  It spins consistently around 40%, and I can't hear it over the other components.  That's saying something, because my entire PC is pretty quiet.
3) It's easier to clean.  In this department, though, the ASUS 8800 GT is even better because it doesn't have a big plastic shield over the top.  The plastic shield is probably the only negative thing I can find with the GTX 460.  It looks cool, but I don't buy win internal components for looks.  Some manufacturers, such as MSI, have removed them.


My machine will certainly be able to handle more advanced games with ease, and at higher resolutions.  In fact, with results like this, why would I want a GTX 480?  It's not like I'd get higher detail, but I would get more noise, heat, and power consumption.  This card is a perfect fit for my goals.  I think I'll keep it set to 825/1000 with no overvolting, as the card can handle this without breaking a sweat.

Perhaps a more demanding game and a 1920*something monitor will present more of a challenge, but I wouldn't count on it.  It will be interesting to see if Windows 7, and maybe 4GB of DDR3, will provide any measurable differences.  I'd like to see if I can find a simple way to benchmark The Elder Scrolls IV: Oblivion and Empire: Total War, as both are more demanding than the games above.  Oblivion needs no introduction, and I use a lot of mods that make it more of a resource hog.  Empire isn't nearly as popular, but it does some crazy things with memory, drawing massive armies across battlefields.  I'd like to see it in more benchmarks because it's rather unique in this regard.  As always, I'm open to suggestions.

If you haven't figured it out by now, I'm a happy camper.  Thanks, Puget Systems!

No comments: