4870 1GB and GTX280 - Pretty much equals?

[SlowSpyder]

I'm looking over a few numbers and have to say I'd consider the 4870 1GB equal in performance to the GTX280. Why is the GTX280 still more money?

http://www.nvnews.net/articles...es_tested/page_4.shtml

The GTX280 wins more benches then it loses, but the benches it wins are pretty well over 60FPS, in the slower benches where the extra FPS would count more (ex: 33FPS > 27FPS is more important then 119FPS > 110FPS) the Radeon seems to be ahead more often.

*edit - For the above article I thought that they were testing a GTX280, it is a GTX260 216... don't know why I kept thinking it was a 280 when it's clearly labeled 260 216. Still the AT article below show the 4870 1GB and GTX280 to be pretty close, but given chizow's reply I guess you'll have to wonder if it's the testing or if the 4870 1GB and GTX280 are actually that close.

http://www.anandtech.com/video/showdoc.aspx?i=3415&p=3

In the Anandtech review they ran 7 tests. 4 go to the GTX280, 3 to the 4870.

http://www.anandtech.com/video/showdoc.aspx?i=3463&p=2

Anandtech tested FC2 at a lot of different resolutions and settings. The cards pretty well trade blows with one another depending on the resolution and quality settings. This is a TWIMTBP title but it doesn't seem that the Nvidia hardware has a clear edge to me.

Am I missing something, or should the 4870 be considered more of a competitor with the GTX280? The reason I ask is when looking around at video card prices it seems like Nvidia certainly prices the GTX280 as a clear cut superior card then the 4870 1GB.

 

[chizow]

AT's results are dated and are clearly the outlier compared to other sites and the reason is quite simple:

AT uses archived benchmarks for previously reviewed parts

I've mentioned it in comments before, but Derek's notorious for ignoring criticisms. So what this means is those GTX 280 benches are old, probably going back as far as launch drivers.

4870X2 Review dated Aug. 12th 2008

4870 1GB Review dated Sep. 25th 2008

Look at the results for the 4870, GTX 280 and 4870X2, they're all the same. The drivers listed are also launch drivers, so 8.7 for ATI and 177.34 for Nvidia. While this is poor testing methodology in general, its not such a problem by itself except for the fact many new parts simply will not run on older drivers due to outdated .infs. Basically this suggests they are using archived results for old parts but benching new parts with new drivers.

While I personally don't put much faith in "miracle drivers", I do recognize small gains from release to release which can over time result in significant performance differences. The recent Big Bang II drivers show more of the miracle driver phenomena but clearly shows comparing new parts with new drivers to old parts with archived results with old drivers is a flawed testing methodology. I think most would agree with me.

Catalyst Performance Improvements from 8.7 to 8.11

GeForce Performance Improvements from 178.24 to 180.48

On a more serious note, the quality, frequency, consistency and reliability of AT video reviews has dropped off catastrophically. Besides the problems listed above, the latest "reviews" of the 180.48 drivers and Far Cry 2 are so poorly executed its pathetic. I've already covered archiving results, but beyond that, Derek admits to manipulating benchmark results to cover up ATI's poor drivers:

ATI Driver rant

In our tests, we initially wanted to take the 3 run average for each test. This was not something we could do with AMD hardware as even our benchmark sessions were marred with ridiculous stuttering and slow downs. We would have performance range from 25 to 55 frames per second on any given test. Rather than take the average, we decided to take the highest performance run for NVIDIA and AMD. It is worth noting that most of the performance results for NVIDIA were within less than a frame per second difference, so average versus max performance run isn't that different.

Wow, just wow. Combined with the lack of any disclosure of any info on testbed and drivers for either the Far Cry 2 and failing to review any of the relevant titles in the 180.48 review, along with some of the issues I've brought up above, AT reviews simply aren't worth reading right now. Honestly, these are pretty serious mistakes and if they continue, Derek should seriously consider resigning or letting others handle video reviews.

 

[SickBeast]

When the two cards just came out, they were close, but the GTX280 was generally faster overall.

Now, with improved drivers, it's no contest and the GTX280 wins handily across the board (except perhaps at 8XAA).

I personally like the GTX260 instead of the 4870. It seems to be a good value at this point compared to what is on the market.

 

[Zillatech]

IMO ~

The GTX 260 (216) is now the best "Overall" card but the HD4870 is right there too. I think once the Nvidia 55nm versions hit the streets it will be the tipping point in Nvidia's favor. They will use a little less power & have a little bit more features & overclock better too.

All things being equal, Nvidia usually wins because that's what people know and trust. ATI has done a fantastic job with the 48xx series and lets hope they deliver again with the up coming 58xx series to keep Nvidia honest.

My only wish now is that Nvidia could shrink their card sizes a bit. They are still just too big to fit comfortably in a Mid-Tower ATX case without worrying about the hard drive cage placement.

BTW, Newegg is selling the GTX 260 (216) cards for much cheaper now. I'm tempted to buy one but I'm waiting until the 55nm cards hit retail because Power & Heat are important to me.

GTX 260 (216) vs. HD4870 1GB

 

[error8]

Originally posted by: chizow
AT's results are dated and are clearly the outlier compared to other sites and the reason is quite simple:

AT uses archived benchmarks for previously reviewed parts

I've mentioned it in comments before, but Derek's notorious for ignoring criticisms. So what this means is those GTX 280 benches are old, probably going back as far as launch drivers.

4870X2 Review dated Aug. 12th 2008

4870 1GB Review dated Sep. 25th 2008

Look at the results for the 4870, GTX 280 and 4870X2, they're all the same. The drivers listed are also launch drivers, so 8.7 for ATI and 177.34 for Nvidia. While this is poor testing methodology in general, its not such a problem by itself except for the fact many new parts simply will not run on older drivers due to outdated .infs. Basically this suggests they are using archived results for old parts but benching new parts with new drivers.

While I personally don't put much faith in "miracle drivers", I do recognize small gains from release to release which can over time result in significant performance differences. The recent Big Bang II drivers show more of the miracle driver phenomena but clearly shows comparing new parts with new drivers to old parts with archived results with old drivers is a flawed testing methodology. I think most would agree with me.

Catalyst Performance Improvements from 8.7 to 8.11

GeForce Performance Improvements from 178.24 to 180.48

On a more serious note, the quality, frequency, consistency and reliability of AT video reviews has dropped off catastrophically. Besides the problems listed above, the latest "reviews" of the 180.48 drivers and Far Cry 2 are so poorly executed its pathetic. I've already covered archiving results, but beyond that, Derek admits to manipulating benchmark results to cover up ATI's poor drivers:

ATI Driver rant

In our tests, we initially wanted to take the 3 run average for each test. This was not something we could do with AMD hardware as even our benchmark sessions were marred with ridiculous stuttering and slow downs. We would have performance range from 25 to 55 frames per second on any given test. Rather than take the average, we decided to take the highest performance run for NVIDIA and AMD. It is worth noting that most of the performance results for NVIDIA were within less than a frame per second difference, so average versus max performance run isn't that different.

Wow, just wow. Combined with the lack of any disclosure of any info on testbed and drivers for either the Far Cry 2 and failing to review any of the relevant titles in the 180.48 review, along with some of the issues I've brought up above, AT reviews simply aren't worth reading right now. Honestly, these are pretty serious mistakes and if they continue, Derek should seriously consider resigning or letting others handle video reviews.

One thing is clear: ATi totally fails in FarCry 2. After all this time they haven't been able to create a driver that can make the game work how it should.