VGA Charts 2008: 101 Configurations Tested

[caamsa]

@randomizer I wish they had thrown in a tri-core for good measure. But that was an interesting review. Despite showing its age the K-8 imo still does rather well. Launched in 2003 and then the X2 came out in 2006 for support of DDR2 but was still basically the same chip. I upgraded to a 4850 and saw a nice boost in frames. My X2 6000+ will keep me happy for at least another year since my monitor is 1280x1024.

 

[randomizer]

@caamsa: I'm hoping my E6600 tides me over until Sandy Bridge

 

[blerb]

[citation][nom]Anti fanboy[/nom]ATI fanboys have only a reason to whine, if and only IF Tom's used a more recent Nvidia drivers than ATI's (something i don't know)Otherwise just shut up[/citation]
lol you just used a biconditional statement (Obviously school is brainwashing me!!!)

 

[blotch]

Why admit there is a CPU bottleneck and still waste that many hours testing. All the 200, 4800, sli and crossfire cards will be bottlenecked so why even test them. I hate when all these sites spend all this time testing cards only to use an stock dual core cpu. And using an old driver is a big problem just look at those stupid GPU charts, the Q3 08 chart has the 4870 in CF performing lower then the 3870 in CF....maybe cuz last time i checked they were still using CAT 7.1 for the charts. Until they actually update that stupid chart they should just burn the link!

 

[waffle911]

I may just be another biased fanboy, but still...
The Catalyst Drivers 8.8 blah blah blah... 4870X2 blah blah blah...

CPU Bottleneck. WTF. This is basic high school science procedure- its called "Level of Treatment." If you do not test and experiment within a reasonable environment, your results are incomplete and invalid because the range was either too broad or to narrow to provide sufficient data. In this case, if you create an artificial ceiling, then cards with overall higher performance will not perform to their highest capacities, and are thus not fairly compared to other cards. If this is supposed to be a comparison of graphics cards, eliminate the CPU as a restricted variable. This is like chemistry: If you are testing the scale of reactions of different quantities of baking soda to vinegar, then you should provide more vinegar than is stoichiometrically required to assure that as much of the baking soda reacts as possible.

If you prefer a different example, then perhaps engines:
You can have one engine that requires more fuel to make more horsepower than another engine, but when limited to the same amount as the second engine the first cannot achieve its optimum performance. It's not like a Lamborghini has the appetite for gasoline of a lawnmower.

These are VGA charts. We want absolute performance values. Just because "most people" are running outdated CPU's (myself among them) doesn't mean they won't upgrade at some point, nor does it discount the fact that there exists a crowd that has the latest and greatest overclocked and thermo-electrically-cooled CPU and want the graphics setup to match their completely murderous overkill thrill rig. They are out there. If you test them, they will come. Also, for platform fairness, let's maybe go for a pair of AMD boards (one AMD one Nvidia) with a Phenom 9950+BE. Then you can test both brands of cards on their hometown chipsets to find their absolute performance. I haven't found Intel chipsets to be as supportive of Crossfire as AMD chipsets; if you disagree, please explain why and I might change my mind.
At the very least, do this for the cards that are obviously capped.

I'm sorry for the rant, but I'm tired of Tom's whole downward spiral in media quality. I miss the way it was before the 10th anniversary.

 

[haplo602]

Is there an option to get the results in a usable way ? example:

I select a benchmark from the first page, I get then a nice graph for ALL the cards. The I select the cards I am interested in and get a freaking TABLE of all benchmarks for the selected cards. Where is my bar graph for the selected benchmark and selected cards so I have a quick an nice overview ?

 

[snarfies1]

[citation][nom]sandmanwn[/nom]yeah I would rather they sacrificed some older generations to include the newer models like the 4870X2. Its not likely that you would ever recommend a forth or fifth generation model or even be able to purchase them so why waste valuable time on something nobody really wants other than nostalgic merits. [/citation]

I would assume they are there for people who are thinking of upgrading FROM those cards. Not everyone buys a new video card every 6 months, and they need a baseline for comparison.

 

[eklipz330]

guys c'mon... this article is huge, and if they actually tested all those cards, imagine how long it would have taken... i can TOTALLY understand if theres flaws in this article, like no 4870x2, or early drivers... Gj guys

 

[sandmanwn]

[citation][nom]snarfies1[/nom]I would assume they are there for people who are thinking of upgrading FROM those cards. Not everyone buys a new video card every 6 months, and they need a baseline for comparison.[/citation]
If you are upgrading from 5 generations ago it doesnt matter what it is your looking at today, its way faster. All that really matters is the price. Find your price range and then use the chart to find the best card for you.

If you have an x1300 do you really need a chart to tell you the new budget x4000 is way faster??? And they aren't just one, but two versions behind on drivers, AGAIN, so the numbers on the charts don't reflect what you will see if you purchased them even right this moment.

 

[jonxor]

I also agree with most of the complaints of others on the Drivers issue and CPU used. Obviously the top performing cards will not reach their peak, and so the entire top 1/8th of the chart is practically useless because the CPU has become the limiting factor. The drivers issue is more understandable, if you started the test when 8.6 were out, although still provides slanted results. I'd also really like to see what a pair of 4870x2 cards perform at. All of these things that are left out, just seems like you took so much time to build a well-engineered mansion, but just covered the roof in cardboard.

If you had just used a much faster CPU (OC'ed Q6600 even. come on, it's not hard) and the better drivers, then I would have heralded this article as the final, definitive, Indisputible Argument Ender for the Generations. It's not to late to make changes!

 

[Viper5030]

Why no test results for the 9800GT? I know it's pretty much just a 8800GT with some minor changes, but it would have been nice to see that proven through the benchmarks.

It's becoming a very popular card on the market now and it would have been nice to at least mention it.

 

[Guest]

I'm not a TOM fan, but to be honest, there are some unfair critics to this chart, i agree the right catalyst drivers should have been used, but that is fault of AMD too who released the cards without proper drivers.

You want absolute performance charts? thats ok, go to anandtech, they have a nice QX9770 @ 3.20GHz ready for you in multiple charts with the fastest cards, but then don't complain if you buy a much faster card (according to anandtech) only to realize it give you a minimal or zero improvement over a much cheaper card because you are using a stock E6600 (like me) or even a slower processor. Apart from showing the theoretical performance of the cards, those charts at anandtech give very little practical information if you are planning to buy a card and have a mainstream processor (and upgrading the processor could imply upgrading the motherboard and perhaps memory as well, so you have to think it twice).

Those charts showing fast cards limited aren't useless, they show you that if you plan to buy those cards first be sure to have a processor faster than a X6800 (at least). Ideally we should have multiple charts for many processors but thats time-impossible, so there have to be a choice and that turn out to be the X6800 which is more o less according with a mainstream gamer, so that's ok with me. I do not discard absolute performance charts but they alone are very dangerous, as you could end up wasting a lot of money.

Finally i recommend retesting the x4000 (after a break) with the correct drivers and update the charts.

 

[Guest]

Wonderfull. Good job guys.
Thanks

 

[Guest]

If they have time to use the most recent drivers for Nvidia then surely they have time to do it for ATI too. The new drivers gave me better performance on my HD4870 so I think they're really just avoiding it now. My 7900GS, in my other PC, performs about the same as the chart so for that I'm all good.

 

[kRogue]

Um, on the comment that "GeForce 6 only support OpenGL 1.5" .. completely wrong; GeForce 6 even support OpenGL 2.1 (infact even crappy old FX cards had limited GL2.0 support). The GeForce 6 has all what is needed for GL2.0 support: full branching in vertex and fragment shaders, vertex texture fetch (something that ATI still don't have for the OpenGL API), in addition, GeForce 6 also has: floating point buffers, multiple render targets... well you get the idea...

 

[genored]

F**k i hate when this happens. TOM's has been bought. I'v seen AMD vs Intel test that are totally retarded and now this, old drivers and "didn't have time to test the best card on the market". THIS IS F**KING retarded. I'm so f**king pissed.

To be so stupid that you'd think people wouldn't notice. WTF why don't you just change the name of the site to Tom's INTEL NVIDA FANBOY site...

 

[randomizer]

[citation][nom]genored[/nom]WTF why don't you just change the name of the site to Tom's INTEL NVIDA FANBOY site...[/citation]
Because that would actually provide a justified reason for calling the site biased. Who would do something that stupid?

 

[Spider D]

First I'd like to say I'm happy that the charts are finally being updated again. The guys at Tom's put in a lot of hours, and I truly appreciate all the effort. However, I'd also like to say that I think this chart is simply foundation for the next update.
Honestly, the results of the newer high end cards are far too often inconsistent. I don't want to get into the ATI vs Nvidia debate at this time. Yes, some results are most definitely flawed (like the overall score where a single GTX 280 beats dual HD4870s, even though a single 4870 usually comes close, or can sometimes even beat a GTX280). However, even GPUs of the same company display results that make no sense.
Note: I'm focusing on 1920x1200 since that can show full HD resolution (1920x1080).
Apparently according to the overall test results dual 8800GTs or two 9600GTs are better than a 9800GX2? The 9800GX2 is two slightly underclocked 9800GTXs, a card which will noticeably out perform the 8800GT and 9600GT in every benchmark that can be run.
In other tests the 8800GTS 512 in SLI beats the 8800 Ultra in SLI, and the 9800GX2. Did you know that a GTX 260 in SLI can beat a GTX 280 SLI? How about a 8800 GTS 512 being superior to a 8800GTX? Even more amazingly, two 3870s in Cross Fire will beat three or four 3870s in CF, and can actually best a 4870 in CF! Not 4850s, which are cards that stomp mud holes in the 3870, but 4870s. 70!
I just skimmed over these charts, and within seconds I found results that contradicted so many other website's tests. As well as results which are flatly opposite to Tom's own previous articles!
The abysmal results of the HD4000 series demands some kind of retest with newer drivers. I've said it before, and I'll say it again, no one wants to buy hardware based on faulty information. Spending your hard earned money buying the wrong hardware is unbelievably annoying.
Also, I agree with eliminating the CPU bottleneck. All the performance inconsistencies between GPU's of the same generation are almost certainly the result of CPU limitations. The whole point of these charts is compare the GRAPHICS power!
As it stands right now, all these tests are only truly relevant to people with the same CPU as the test rig. What if someone has a Phenom X3, or X4? Or maybe just an Athlon X2? What about the Celeron L, Pentium Dual Core, or those with lower quality Conroe cores? What if they have a Core 2 Quad, or an extreme Quad? Only those with an X6800 can find any real value in such results. My guess would be less than 1% of the readers here have that CPU!
A few people have posed the idea of an article devoted to finding the CPU bottle neck, at various performance points. I STRONGLY second such a topic! You don't need to test every single graphics card, like in this chart. Or even every CPU. Just a few well selected CPUs and GPUs. If your CPU is choking a 8800GT, then you can be pretty sure similar CPUs will choke the 8800GTS/GTX/Ultra 9800GT/GTX/GTX+ etc. I think readers (myself included) have been desperately longing for just such an article for a long time.
While on the subject of CPUs, I'd like to thank randomizer especially, for the quad vs dual link. I had no idea a stock speed Q6600 (2.4GHz) can easily match an e8400 overclocked to 3.3GHz, when running SLI or Cross Fire, in gaming performance.
Lastly, please don't update once a year. Please go back to doing regular updates like you use to. Then you can simply just add cards like the 4870X2, 9800GTX+, 8800GSO/9600GSO etc.

 

[Guest]

Very very good, but I missed 9500gt, 9800gt and hd4670 (maybe it wasn't out then)

 

[Lozil]

Nice Review...
Visit http://free-and-useful.blogspot.com

 

[ailgatrat]

Is it just me, or did this article get reposted???

 

[awilgus]

Its nice to see some testing all these video cards and cut through some the hype with so many to choose from in various configurations of a single and multiple cards. Running intense graphics such as games have always expensive, slow and jerky with pix-elation requiring a fast computers and good video cards.
This certainly helps with the selection thanks.

 

[Hellbound]

Well, looks like I made the right choice. I purchased two 8800 GT's a month ago for next to nothing.. Great performance in SLI.

 

[monsta]

101 configurations! Would of been great to see the 9800GX2 Quad SLI in there.

 

[aevm]

Nice work, despite the drivers/4870x2/dual-core CPU issues. Next time could you please use a Q9550.