Designing Our Next GPU Test Suite

240 Comments

View All Comments

• ottawanker - Sunday, March 14, 2010 - link

  Racing games are the only games I play on PC, and so they're the only ones that factor into my choice when buying a card. Often reviews don't have any racing games, or if they do its something from EA with extra motion blur and unrealistic effects.
  
  I understand that often racing games don't take up as many resources as some of the FPS, but often cards that are great at FPS seem to be crappy at racing games.
  
  At the moment Dirt 2 would be the obvious choice for a racing benchmark as its relatively recent and supports DirectX 11.

• Computer Bottleneck - Sunday, March 14, 2010 - link

  I would also like to see minimum frame rates measured.

• arpitnathany - Sunday, March 14, 2010 - link

  please try to include more RPGs and simulation along with FPSs.
  
  also a high end i7 does not represent real performance the majority are likely to get,i understand that this has to be done in fairness to the cards so as to bring out the max they can deliver,but no point if we the avg user do not realize that benefit, try using something upper midrange like an i5 750 or one of the low end i7s.

• Paulman - Sunday, March 14, 2010 - link

  #1. I agree. I would like to see especially MMORPG's tested, because some (like WoW) hav huge user bases. Even though I don't play WoW, I'm curious because I know player base = potential marketability of the card, so it's important. Someone's suggestion that you try something taxing (like a 25-person boss raid, or maybe a 3-screen Eyeinfinity-like setup) with WoW is a good one, since it's less GPU demanding.
  
  #2. Bad Company 2, because it's very popular and adequately GPU intensive.
  
  #3. PLEASE include benches with an Athlon II X4 or Core I3 class CPU, or at least a mid range CPU if possible. Theoretical GPU performance isn't always as practical as anticipating the actual performance you'd see when a mid-range GPU is paired up with a low/mid-range quad core CPU, which is a much more realistic rig. Especially when most games are GPU limited, especially with AA and at today's widescreen monitor resolutions. For example, you might see a 20% FPS difference at 1080p 4xAA in Crysis Warhead, let's say, between an Athlon II X4 and a Core I7 depending on the video card, but potentially a 100% difference if you changed the video card. So lower end CPU configs are valid To pair up with many GPU's, but t would be god to see if there is a signiicant performance hit, and in which games.
  
  #4. Lower resolutions and/or no AA would be very useful as well, especially for mid-range and below cards.
  
  P.S. And Mass Effect 2, just because it's so cool :P EVE-Online, for the same reason. Lol.

• slickr - Friday, March 19, 2010 - link

  I agree with this post. Having a mid range CPU to test the latest graphic cards would be great, as it would show just how much(or lack thereof)difference it makes.
  
  Maybe even a complete new medium system. Its different to see I7 980x ultra turbo extra 4ghz with 6gb of 2200mhz ram and GTX480, but its a whole different story when you put in a phenom II x4 955BE or core I3 CPU with 2GB of ram and then test from GTX480, down to 8800GT.
  
  Simply put, only 1% of people have such systems as the ones you test games on and thus a medium system would be much more real world like and give more realistic results people can identify with.
  
  I would also like benchmarks to be run only on bare operating system, meaning no software(antivirus, winamp, firewall, skype, msn, etc...), other that the required drivers, DX and game software.

• Paulman - Saturday, March 20, 2010 - link

  I agree with you guys agreeing with me, even as I had agreed with the OP.

• pepe2004x - Tuesday, March 16, 2010 - link

  Totally agree with number 3. I had spend ALOT updating a nice pc to make it the best I could for gaming (going from an EXCELLENT for general use Athlon le1640 to a athlon 2 x4 620, from 2 gigs of ram to 3, from a 4850 to a 5850, getting a xbox360 control, changing the mobo to one with a decent sound chip since I couldnt put a sound card near a gpu, etc, etc) and I just dont want to start thinking about getting a i5 or phenom and spend more. So, putting average cpus on the betchmatk would be cool. :S
  
  And yeah, I`m a little angry about having spent a lot on pc gaming :(

• Computer Bottleneck - Sunday, March 14, 2010 - link

  I would like to see triple monitor resolutions tested.
  
  Unfortunately this would necessitate Anandtech getting two Fermis from Nvidia.
  
  Oh well, maybe at a later date this could be accomplished?

• piesquared - Sunday, March 14, 2010 - link

  I agree, definitely need a multimonitor benchmarks, but even if 2 nvidia cards are not available. Eyefinity is on the market, and is defintitely not a gimmik, so it should be tested. Just like intel's turbo mode is constantly used because it is in hardware, the situation with Eyefinity is identical. And there's no doubt AT can afford 2 more monitors, so there's no excuse there. Benchmarking shouldn't just be about comparing competitor's hardware.

• dgthug - Monday, March 15, 2010 - link

  Eyefinity and Nvidia's multi-monitor solution would be beneficial for everyone (mainly enthusiasts though). Also in multi-gpu set-ups.
  
  Initially and periodically it would be beneficial to run multiple tests on multi-gpu configurations to test the affect different port utilization has on framerate.
  
  I believe two standard configurations should be tested. One consisting of 3 monitors at 1280x1024 (for people using older displays) and then another 3 at 1920x1080.