cool article, though I was a little surprised after the value angle title to find no "value" orientated cards like the 3650 series (no point including the 8600GT though, everyone knows how crap that was). I'd find it pretty interesting to see how playable, and with what detail, lower end cards are at the resolutions the majority of people play at.

I understand the arguments behind testing at crazy resolutions, but all that left me with was the vague understanding that any of these cards will be fine at up to 1650x1050, rather than which one is the best value for me. I also don't see much merit to the future proofing argument, especially for graphics cards. Surely it'd make more sense to by a mid-range card every year for 2-3 yrs than to blow a similar amount on a high end card and hope that you've somehow guessed right about future game requirements, so that the high-end card will last longer than a year and a half?

Still that's just me :)

I wish techsites developed some standard to more clearly differentiate multi SLI/CF card set-ups in listings, it's so annoying how they just intersperse them and don't even give them another color in graphs and you painfully have to stare and try to mentally get a picture of it all.
Oh well, maybe in another 100000 years eh.

Where are the 9600GT X2 cards? Will they ever become available?

Such a card will change the entire market and I want one!

My attempt at the cost of power for the cards
http://techreport.com/forums/viewtopic.php?f=3&t=59023

No 256MB 3850? Would have loved to see this. These cards can be had dirt cheap (like 110$), way cheaper than the cheapest 256MB 8800GT or 384MB 8800GS, and the ATI HD series suffers not nearly as bad as the current GeForce cards under memory pressure (which usually rules out the 256MB 8800GT card as a serious contender). Should offer excellent value.

Interesting stuff, but I wonder about how the statistics play out a bit. Frames per second per ten dollars is a decent metric, but it doesn't take some critical stuff into account.

For instance, while I understand that testing at high resolutions is important to discern the difference between graphics cards, it's still not as useful from a value standpoint. Somebody who's shelled out for a 30" monitor to game on probably doesn't care as much about value as someone with a more pedestrian 20" or 22" model. It'd be interesting to see how the performance-per-dollar stuff changes when lower resolutions are put in play.

Also, pure framerate per ten dollars can be somewhat misleading. Imagine a $20 video card that offers an average of 7 fps across the games tested (not that it exists, of course - this is a thought experiment). It would absolutely dominate the performance-per-dollar ranking, but for people who wanted to actually play games, a $150 card capable of sustaining playable framerates would be superior. Alternately, an $2000 setup capable of 500 fps in everything might look pretty good, but for the majority of gamers, a less expensive card that still offered playable framerates would probably be a better buy. It'd be interesting to see some weighting, so 0-30 fps is vital, 30-60 fps contributes a significant amount, and over 60 falls off rapidly. Defining the weighting function would of course be another source of argument, and coming up with a good one would be a lot of work - but it's interesting to think about.

I'd love to see the final averaged graphs to be updated with the real time prices of the video cards, by taking the average of the cheapest stock clocked card available at several retailers. I know this would be non-trivial to implement and would probably require some regular maintenance, but it would add tremendously to the value of this article for the coming few months.

As it stands now, it's a snapshot. A damn informative one, but a snapshot still.

Stuff like this makes me love my 8800GT even more than before. I'm gonna marry that card someday.

Quick question... Since readers have replied to a hardware survey back in Feb.2008 and the majority responded with a monitor size between 17 and 22 inches... Why are we still seeing test resolutions of 2560x1600 when a closer resolution to you average readership would be 1680x1050? Just asking...

Where's the Voodoo 3000? I think it presents the best value on the market now. Could you please include it next time?

Great article guys! IMO, these are the types of articles that interest people the most, along with system recommendation guides.

very nice. i'd just like to suggest one more thing to add to the graphs- a trendline. having a line of best fit for all these models would certainly help people see the "average" amount of additional performance they should expect per extra dollar, as well as how far above the trend the best value cards are.

The conclusion mimics a statement I made in Fox's "SLI debate" thread - the 9600GT and the HD3850 are the two configs that make multi-gpu worth it. It was pretty obvious looking at the most recent benches along side some online prices. I'm sensing that it's a pretty consistent trend with multi-gpu that there are really only one or two products on offer that you can team together and get some real value. Teaming two (or more) high end products together simply costs way to much, and teaming two low end products together is just a doubling of disappointing performance.

Edit: and BTW, these value articles are just about the best things you (TR) guys can do in terms of what's in the best interests of your readers. I hope they become a staple item.

Over the past few pages, we've seen three things. The first is that scatter plots are really cool. Seriously, just look at 'em. Organizing this type of data could hardly get any better.

Quoted for truth.

Is it just me or are these sentances a little difficult to understand?

"Last time we attempted to quantify the value propositions of a large cross-section of competing products, we concentrated on microprocessors. "

Shouldn't it read THE last time we attempted...

And this needs a re-write. I had to read it three times!

"This is an iffy exercise, this attempt to quantify a value proposition that often involves considerations one can't easily boil down to numbers. We have no illusions that this little thought experiment will yield the only or best information you need in order to decide which graphics card to buy".

I just read it again and still don't get what you guys are trying to say???

I wish Techreport stuck a 8800gs into the mix. This card is by far the cheapest of the bunch and perform faster than 3850 and equal terms with 9600gt minus AA performance. It would easily be neck and neck with 9600gt price performance per $$$.

Any reason why the 8800GTS 512MB wasn't included in the SLI results, it looks to perform the same as the 9800GTX, but costs less, surely it would be better value than the 8800GT in this case?

Very interesting. Well done. I'm fairly pleased how my video card plays out, considering the state the market was in at the time I upgraded last.
Anyway you can take power consumption into consideration in future versions of this without borking the system you have? (maybe track the card's individual power consumption x number of cards etc)
Edit: I should read the posts THEN post myself. *roll*

It's been a long time since I've used graphs for anything constructive.
But the Average performance graph shows a gap between the two "hero" SLI setups and the 9800 GX2 without a large jump along the X axis.
How could that be read?
In addittion, as stated in the article, the platforms these cards were run on differ. I would like to see the delta between the nvidia chipsets and various Intel chipset mb's running the GX2. I don't know if that is what this data shows or not.

I know people hate this stuff, but some cheapskates may like this.

$99 after mir.

http://www.newegg.com/Product/Product.aspx?Item=N82E16814131096

Nice article.

Since your basing this on the value angle you could have included running costs. Use the idle and load power ratings and apply a formula similar to the following. PC is on 8 hours a day 80% of the time GPU idle, the other 20% GPU load (for a heavy gamer).

Then add the cost of the power consumed by the cards in 1 year at whatever you local rates are.

Nice compilation of data for us readers although some readers would have liked to see other types of data added to the mix as well.

Please hose!

http://slashdot.org/firehose.pl?op=view&id=660180