All-inclusive GPU Charts

I know benchmarking can be quite the time consuming pain, but comparing many cards in one test can be useful for people looking to upgrade.

You assemble a PC with a fairly high end CPU (since you want GPU to be the limitting factor as much as possible) and run a bunch of gaming benchmarks with a bunch of graphics cards from the past couple years.  You can even add new cards when they come out to the tests if you only use the PC for benchmarking, so that Windows doesn't "age" and hurt performance.

I wouldn't mind seeing an older CPU used for some tests as well (but only as a bonus edition, a high end CPU comes first).  It wouldn't have to be for as many cards and benchmarks.  I'm thinking this could be useful for people with older CPUs who are wondering if a new GPU is worth it.  You could compare a 3Ghz Intel Core 2 against an older AMD Athlon 64 5000+ with each system testing several GPUs in newer games like Crysis and Call of Duty 4.  How much worse does a Radeon 4850 perform at Crysis on an aging AMD dual core vs a new Intel?  Is it worth upgrading from a older card to a newer one on such an "old" CPU?

i second this. this is the only reason i go to toms.... (toms is a bit biased btw!) 

Good Idea! I third it.

I would be happy to offer up my X2-4400 as a workbench rig.  You just send me those super highend cards; and I will put em to the test.  Heck I even have a Nforce 4 SLI board, so send em 2 at a time.  I will include not just bench scores but over all gaming impressions.

Same Ice... thats the only thing I use them for as well. But they are very one sided some times

It would be hard to keep an apples-to-apples comparison given other changing platform components, but I agree that it's a worthy endeavor.  I would like to see how HH tackles it.

At the very least, keep older benchmarks in some easily searchable place.  Maybe have a database of cards and tests...  we can select a vid card and see the results for the available test, or see all the cards and their results in a single test - with notations about the other system components.

3vi1:

It would be hard to keep an apples-to-apples comparison given other changing platform components, but I agree that it's a worthy endeavor.  I would like to see how HH tackles it.

At the very least, keep older benchmarks in some easily searchable place.  Maybe have a database of cards and tests...  we can select a vid card and see the results for the available test, or see all the cards and their results in a single test - with notations about the other system components.

 

Guys, We find these charts are grossly out of date and out of rev with other sites and it's because GPU revs and driver revisions change like the wind.  You can't possibly keep up with them and have accurate results with the dozens of GPUs on the market at any given time.  That said, on our major launches, if you look at the charts, we include all relevant GPUs in a given segment, from competitive solutions to the previous gen solutions. This does give you a solid view of where they stand.  I agree, I'd love to have a big database of numbers here but I'm not convinced it's practical or accurate even after a month or two of a recent update.

I'm all ears for ideas on this though.

yeah that would be a full time job for someone just strictly trying to keep everything up to date.

Older cards already have pretty decent drivers. Don't do updated benchmarks on their drivers, just list what driver version was used in the test. On current Gen cards run new benchmarks for newer drivers, but once the next gen comes out, then file the numbers and start the process on the new current gen cards.

To keep things the same for comparison, you could have 1 system that the specs never change on. And run maybe 2 synthetic benchamarks and 1 or 2 real world game bench marks. Have a second system with current up to date equipment to test with new cards that come out so people can see how it will perform with the best of today and the best of yesterday. The Best equipment from 1 or 2 years ago performs about on par with mainstream stuff now. This will allow people who went all out and bought a sweet system a year or 2 ago see how a vid card upgrade will do, and current mainstreamers can also get an idea of where they would be. After 2 years of have a system around... retire it and start with a new system.

Or you could just do some math and figure percentages of performance gains and issue estimated numbers.

example... an 8800 GTX get's XX fps in a system with X equipment. An upgrade to a Faster processor resulted in a gain in XX% fps. So it can be assumed a 9800 GTX will also gain XX% fps.

Or an 8800 GTX got XX fps in System A which is year old top of the line equipment. In System B with current top of the line equipment the card saw an XX% increase in FPS. It can be assumed all cards tested using System A or a similar configuration will gain XX% fps.

I would think at least an estimate would be better than nothing... however a disclaimer would need to be in place :-P

yes we really need to kill off toms, they've gone downhill quite a bit over the years =\

Agreed, I used to go to them for every benchmark back in the day. Now it's a little frustrating to find a site that's as comprehensive as they used to be.