[hardocp] 2012 AMD Video Card Driver Performance Review

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

jimhsu

Senior member
Mar 22, 2009
705
0
76
Including FC3 was indeed a weird decision...

According to computerbase relative perf. of NV and AMD cards is virtually unchanged since GTX 670 release (May 2012).
Pretty shocking considering how often we hear about AMD drivers pulling ahead with each release.
With newer drivers GCN does pull ahead vs HD6000, but not so much vs Fermi, and not at all vs Kepler.

http://www.computerbase.de/artikel/grafikkarten/2012/test-nvidia-geforce-gtx-670/4/
http://www.computerbase.de/artikel/grafikkarten/2013/test-17-grafikkarten-im-vergleich/3/

On the contrary, I see a very noticeable increase in performance between those 2 articles on the high res tests (where you're more likely to see a GPU bottleneck). Look at especially the 5760 8xAA/16xAF chart, where 7950 manages up pull ahead of 680 (??). This trend of the AMD cards being ahead in these ultra-tough high res conditions (where you need every single FPS you can get) seems to repeat itself.
 
Last edited:

blackened23

Diamond Member
Jul 26, 2011
8,548
2
0
Who bought one of these cards and wasn't able to use it "in a sense"?

I bought 7970s at release and enjoyed them once a few things were straightened out - however that being said, the release situation wasn't a good one. For several days, a downloadable driver was not available - and then RC11 popped up on their website, but the performance was far from ideal. In fact, most of the release reviews were based on that RC11 driver and the performance was sub-par in comparison to later releases. Surround also crashed consistently with this RC11 driver, i'm sure you can imagine quite a few people being upset over that one. Also, one website that I remember had to benchmark with catalyst version 11 which was designed for the Cayman series of cards, i'm sure you can imagine how that went.

On top of this, there was no driver available for windows vista (I know I know) for a couple of weeks.

Like I said, I thought the 7970 was a great card and I enjoyed it during my brief ownership; however I'd be kidding myself if I stated the release driver situation was ideal. I really hope a lesson was learned from that, AMD doesn't need to repeat that nonsense again -- I don't think the H article is unfair. Fact of the matter is AMD is making large improvements with the driver situation for their GCN cards, and the article took note of that. Some of those games had a 20-35% increase from drivers alone.

I've come to the opinion that the best hardware means nothing without great software to accompany it. I think AMD realizes this and is making improvements in that respect.
 
Last edited:

Tuna-Fish

Golden Member
Mar 4, 2011
1,324
1,462
136
I can't tell if this says that the AMD driver team is extremely competent and managed to pull a miracle out of their cards, or if they were wholly incompetent and finally figured out how to write drivers for their cards after nvidia brought on competition.

60% improvement on the 7950 from the RC to the 12.11 release??

The big reason for this is that GCN is a huge change in the core. Most of the "driver" is the shader compiler, whose job is to turn the HLSL and GLSL shaders into machine code that suits the card. Writing optimizing compilers is really hard, in the absolute sense. Compilers need to be ~5 years old before they are considered mature.

Because GCN is more flexible than their previous architecture, you can always easily translate old code into new code that runs on GCN, however, that new code won't make any use of the awesome flexibility of the new platform. Building a better compiler that makes good use of GCN just takes time. I think there are still plenty of improvements to come.
 

f1sherman

Platinum Member
Apr 5, 2011
2,243
1
0
On the contrary, I see a very noticeable increase in performance between those 2 articles on the high res tests (where you're more likely to see a GPU bottleneck). Look at especially the 5760 8xAA/16xAF chart, where 7950 manages up pull ahead of 680 (??). This trend of the AMD cards being ahead in these ultra-tough high res conditions (where you need every single FPS you can get) seems to repeat itself.

Come now... 5760 8xAA. For real? 8xAA is garbage anyway, use SSAA if you have that much power.

How about something meaningful. Perhaps 5760 4xAA?

May 2012 -------------- February 2013

7950 - 100% ----------- 100%
670 - 102% ----------- 102%
680 - 110% ----------- 112%
7970 - 118% ----------- 116%

And it actually looks like NV did a hair better job with drivers ;)

EDIT: Actually measurably better it seems (at these settings anyway)
Because that 7950 in Feb. 2013 is actually Boost edition. Surprise surprise...
 
Last edited:

SolMiester

Diamond Member
Dec 19, 2004
5,331
17
76
Got a link to any review of the products that said they were crippled upon release?

I like the fact the engineers have continued to invest the time and effort to improve the experience I have with the hardware I have purchased. It actually increases the value of my purchase over time.
mGPU Eyefinity was upon release for 6mths +
 

Tweak155

Lifer
Sep 23, 2003
11,447
262
126
This thread isn't a review.

The OP has a list to a review of the drivers... try to use some real abstract thinking here... Also you didn't answer my question, so I don't see the point in replying anymore to your statements.
 

Phynaz

Lifer
Mar 13, 2006
10,140
819
126
You question was a snide remark that didn't deserve an answer. Therefore it didn't get one.

So far you haven't found a single review to backup your claim that these cards were unusable.
 

Tweak155

Lifer
Sep 23, 2003
11,447
262
126
You question was a snide remark that didn't deserve an answer. Therefore it didn't get one.

So far you haven't found a single review to backup your claim that these cards were unusable.

Are you unable to go back to the OP as I pointed out and click the link? Fine:

http://www.hardocp.com/article/2013/02/18/2012_amd_video_card_driver_performance_review#.USU_oKXhA6V

Sheesh. If you can't deduce that performance improved dramatically through software from that REVIEW, I can't help you. I'm half wondering if you even read the OP?
 

ICDP

Senior member
Nov 15, 2012
707
0
0
Including FC3 was indeed a weird decision...

According to computerbase relative perf. of NV and AMD cards is virtually unchanged since GTX 670 release (May 2012).
Pretty shocking considering how often we hear about AMD drivers pulling ahead with each release.
With newer drivers GCN does pull ahead vs HD6000, but not so much vs Fermi, and not at all vs Kepler.

http://www.computerbase.de/artikel/grafikkarten/2012/test-nvidia-geforce-gtx-670/4/
http://www.computerbase.de/artikel/grafikkarten/2013/test-17-grafikkarten-im-vergleich/3/

I agree totally with the fact that relative performance at stock speeds is largely unchanged. Both Nvidia and AMD have improved driver performance since release, so AMD do not deserve special mention in this regard.

When Kepler was released it beat AND on price perf based on the MSRP at the time. That swung in AMD's favour when they dropped prices to win price/perf. It is price/perf that AMD is winning as in general HD 7970 GE is only ~10 faster but it is generally ~$50-$60 cheaper. So when some say AMD have turned it around it is mainly meant in respect to price/perf. Right from 1st release my experience with both HD 7950, HD 7970 and GTX 680 has been that it literally comes down how good each card overclocks. I have tested a HD 7950 that OCs to 1250 core beat the crap out of GTX 680 that couldn't reach 1200 core and a 7970 that couldn't go past 1050 core. Even at an overclock of 1150 a HD 7950 is giving OC edition GTX 680 speeds.

It's not just about the performance for most people when it comes to proclaiming a winner. What helped AMD is that 7970 GE is (marginally) faster and a fair bit cheaper than GTX 680.
 
Last edited: