Final Thoughts
So there you have it: you’ve seen the results from each of the games we tested comparing release drivers to current drivers and you’ve seen the performance gains. To summarise they are as follows:
- In 3DMark the HD 7970 gained 30.8% in FireStrike and 28.9% in FireStrike Extreme while the R9 290X gained 2.5% in FireStrike and there was no change in FireStrike Extreme.
- In Battlefield 4 the HD 7970 gained 19.9% while the R9 290X gained 1.3%.
- In Bioshock Infinite the HD 7970 gained 15.4% while the R9 290X saw no change.
- In Metro Last Light the HD 7970 gained 6.4% while the R9 290X gained 1.8%.
- In Sleeping Dogs the HD 7970 gained 9.2% while the R9 290X saw no change.
- In Tomb Raider the HD 7970 gained 6.6% while the R9 290X saw no change.
The broad trend that we can see was not unexpected in the slightest. The HD 7970 has made some significant performance gains across most games and benchmarks because the HD 7970 was the first graphics card of a new architecture, the 28nm GCN design. The gains that have been seen with the HD 7970 are mainly architectural ones and that is why they are so high. For three years worth of driver progress there are some impressive gains in performance, such as in Battlefield 4, Bioshock Infinite and Sleeping Dogs. Of course, the R9 290X is totally the opposite. Because it uses the same architecture as the HD 7970 all the optimisations had already been made by the time it was released in October 2013: nearly 2 years on from the HD 7970’s release. As a result we see very little change from driver updates with most games and benchmarks showing no change, or a 2.5% or less boost in performance.
On the whole I think it is fair to say driver updates can make a huge difference: anyone comparing reviews of the HD 7970 to the R9 280X may be misled by the results: the R9 280X is faster because reviews are based on newer drivers will all the architectural optimisations. The same applies to anyone comparing HD 7000 series cards with their R9 2xx series equivalents, such as the HD 7950 and the R9 280 or the HD 7870 and the R9 270X. AMD has done a great job with its driver updates in terms of performance gains. From now on forward I doubt we’ll see any more performance increases for GCN 1.0 and 1.1 based cards, most newly released driver updates will just be bug fixes for things like game issues with new games or problems with CrossFireX profiles and frame latency issues.
I hope you have found this article interesting, please stay tuned for our Nvidia driver analysis article where will do the same to Nvidia’s two flagship cards: the GTX 680 from the GTX 600 series and the GTX 780 Ti from the GTX 700 series. We will also be comparing and contrasting the AMD and Nvidia results to see who really does the best job with driver optimisations!
Thank you to all our partners who provided the hardware and software that made this driver analysis possible.
send me the 290 my hd 7970 is failing me
stop mining BTC/LTC then
None of your tested games was released when 7970 launch drivers became available. At the same time, all of your tested games were released by the time 290X launch drivers were available (even BF4 was specifically targeted by those betas). In that sense your findings are flawed. 14.7 doesn’t bring much to the table for 290X not because it is GCN and a known architecture but mainly because the launch driver was already optimized for those titles that you tested. Try the same approach with Watch Dogs or something as fresh and the picture will be different.
That;s true – most games were released pre- 11.12 RC11 but that doesn’t make the results invalid IMO, they are still built from pre-existing established APIs. I think the point about the GCN architecture that you’ve mentioned is more relevant, I also mentioned that in the conclusion: there is minimal scope for major gains in GCN performance as the architecture is already well optimised with drivers. Thanks for your input, will consider.
This is definitly interesting. Too bad it’s only for high-end gpu’s. Try testing this on lower-end and it would be worse. Also, the problem with AMD drivers is consistency: for mobility there are long known bugs with black screens and no video aceleration. There is still some bugged mouse cursor around. They mostly fix something and break something.
(i’m not an AMD hater btw)
Agreed. AMD drivers have bug issues (but so do many Nvidia driver releases). In that regard this article was not exploring bugs and driver issues but pure in-game performance, but yes that’s also something to consider that we may explore.
I was sent here from AMD FaceBook http://www.facebook.com/AMDUK?brand_redir=1
Am willing to bet the low end have more performance than the high end. And its the exact opposite for Nvidia.
There is no GCN 2.0 yet. GCN 1.0 and GCN 1.1.
There is GCN 2, its the R9 290X (AMD have never mentioned GCN 1.1)
no there isn’t, GCN 2.0 has not even been announced yet. the 290X is GCN 1.1, it is very similar to HD7970 (GCN1.0).
http://www.anandtech.com/show/7457/the-radeon-r9-290x-review/2
How much longer until we can see Nvidia’s progress?
Still working on it, we’ve done the GTX 780 Ti testing and are currently sourcing a GTX 680 (we have one at HQ but trying to get it here for testing) for the equivalent testing.
http://www.eteknix.com/examining-nvidias-driver-progress-since-launch-drivers-gtx-780-ti-gtx-680/
I find this very interesting because i have a Gigabyte HD 7970 ghz edition gpu, which has slightly faster speeds than the XFX card. I have been thinking about upgrading my card to a 290x but i think now i will not. My card plays just about anything at max settings and i have had no problems since i bought it. Seeing the improvements from 7970 to 290x is nice but not enough for me to upgrade, yet.