Can this THG chart possibility be correct?

[topper743]

So.....I play games online mostly UT2K2 and the Clan I like the best is now hosting Call of Duty 4. I am also looking to replace my current video card. I am considering a HD 4850 or a 9800 gtx+ to play at 16x10 res. So I turn to my trusty THG performances charts and what do I find...this mess.

http://www.tomshardware.com/charts/gaming-graphics-charts-q3-2008/Call-of-Duty-4-v1-6,742.html

How is it that a single hd 4870 beats a hd 4870 cf (2 card setup) by 15 fps? How is it that single gtx 280 beats a gtx 280 sli and a gtx 260 sli by 5fps and 8fps respectively? How is it that the same gtx 280 and 260 sli'ed setups, the 280 only beats the 260 by not quite 4 fps? All that money and only 3FPS!!!Why is it that a hd 4850 beats the 9800x2 by 3fps? Maybe it's the sli or cf thing? Then why do the 8800 series sli'ed out perform all the rest of the cards? How is it that a single 4870 actually beat a gtx 280 sli setup? This chat has a lot of questions. It provokes more questions than it answers. Some will say that some games are optomized for Nvidia, not COD4 apparentlly, not well anyway. COD4 must not like SLI or CF either except that the 8800gts sli with 512 mb ram (per the chart) is the over all winner besting the 280's, the 260's the 4870's and the 4850s. All this might be true, could be true and must be true. Is it true? Could the chart be wrong? Might someone have misposted the findings? Maybe one of the forums proctors (Turpit) can ask this question of the authors? Ask them what is up with this wacky chart. Many more oddities in this chart. Now I have no idea which VC I want. Comments?

 

[zenmaster]

Personally, I don't trust the charts.

I suspect the data was collected over time and not in one fell swoop of testing. The result is that the tests may have been done on different hardware and different software revisions.

I've just seen too many things not seem right to trust them.
I would go to one of THG's reviwe of the 4850 or other card.
I suspect the numbers there would tend to be more reliable.

Now, it is possible for SLI or CF to have 0 benefeit in some games and a small drop in performance. Still, I would try to check other sources to verify.

 

[JAYDEEJOHN]

Since the focus of the european writers has been much higher than the north american writers, the scope, accuracy, timing and reliability has gone down hill IMO

 

[shortstuff_mt]

The charts are crap. I wouldn't base any decisions off of them. Check some other major site's charts.

Those charts were run with a CPU that didn't run fast enough to feed the cards, especially in the crossfire/SLI setups. Some of these new cards will continue seeing improvements past 4GHz.

Toms should be embarrassed these charts ever made it out the door.

This article makes an attempt at explaining things. They did overclock their CPU in it, but it still couldn't fully drive the crossfire/SLI configs on some games.

http://www.tomshardware.com/reviews/GeForce-Catalyst-overclocking,2037.html

I personally have lost a lot of respect for Toms because of the obviously bad charts they posted and have stood by. I agree with jaydeejohn, Toms is definitely in a downward spiral, especially the North American site.

 

[JDocs]

It is possible for dual card to loss to a single card in a senario where a single card can sap the CPU core used for graphics rendering. Then a dual card would still require overhead to manage the second card. This however should not account for more than 5 FPS at absolute maximum.

Something seems off about these results.

 

[knotknut]

Just like this site says " Toms Guide".
Picking out the right stuff drives me nuts. I want the best for my needs at the best price, and no shipping charges. I forgot, I want it no later that tomorrow.
You just have to decide what will be the best for your needs, abilities and budget and go for it. Make a decision and get on with it. You know what ever you buy will be out dated soon after you get it in your computer anyway.

 

[Dekasav]

From what I've seen, the HD 4850 is exactly what you want, though. Handles CoD4 at 16x10 very well.