ATI launch drivers vs Catalyst 10.6's, performance in detail.

[TurtleCrusher]

You heard it here first. Higher minimums don't mean anything.

Have you been drinking tonight?

On an unrelated note, single core processors FTW and Windoze sucks. Defragment your SSD's often. Don't forget to pick up a Killer NIC to improve your gaming experience.

[sarcasm]

The above comment is really the dumbest thing I've heard in a long time. Do you even game or do you just post in the video forums?

 

[Sylvanas]

I don't understand the indifference being put forth by some on these improvements. I read through the article, I didn't just skip to the conclusion and I thought the gains were quite substantial as far as driver releases go. Minimums went up in Splinter Cell and AVP by nearly 10fps and the around +5fps in other titles, averages were consistently higher from 3-6fps. I think earth-shattering performance can be better brought to bear by developers releasing patches for their game code rather than Nvidia and ATI having to optimise for flawed designs in most cases.

The nature of AMD's drivers and architecture is tied significantly to compiler optimisations (drivers) and extracting ILP. There will always be headroom here for improvements, expect to see more 1-5% increases in future drivers. The good news here is any improvements to ILP improving utilisation of their VEC5 layout of shaders will affect older generations of cards aswell- back to the R600.

 

[Ancalagon44]

happy medium, what is the point of this thread? Like the other one, you dig through pages of stuff to find a single bad thing to say about ATI.

Please.... go play somewhere else.

 

[NoQuarter]

You heard it here first. Higher minimums don't mean anything.

Have you been drinking tonight?

On an unrelated note, single core processors FTW and Windoze sucks. Defragment your SSD's often. Don't forget to pick up a Killer NIC to improve your gaming experience.

[sarcasm]

The above comment is really the dumbest thing I've heard in a long time. Do you even game or do you just post in the video forums?

Wait.. what? You put sarcasm in your post and didn't realize his post was sarcasm?
--


I want to point out that driver performance enhancements are generally per game enhancements. The driver team actually analyzes the game code to see what low level improvements can be made to speed things up for *that* game. That means a newer game like Splinter Cell: Conviction isn't going to be any faster on 10.6 over 9.10 because neither driver has game specific enhancements for it yet, and much older games (pre-9.10) will most likely already have been enhanced and see no further improvement either.

So if you give them, say, a 2 month lag on game release to driver improvements it's pretty much the games that came out after August and before April that are going to see changes between 9.10 and 10.6. And what games do we see improvements in? AvP (10&#37 Metro 2033 (10%) Dirt 2 (15%) BFBC2 (10% w/AA, no change with no AA). Games released before or after the period saw no change (Crysis Warhead, Splinter Cell: Conviction). This is pretty true over at WSGF's driver comparison also:
http://widescreengamingforum.com/wiki/WS_and_WTH_Benchmarking_-_Radeon_HD_5870_1GB_Driver_Comparison

So while drivers do 'mature' the nature of this process isn't really understood properly.. universal speed improvements are very rare and what we see as 'maturing' is just per game optimization shortly after a game's release. Since the games used for benches don't change very often it gives the appearance the card has just become faster overall over time, when that's not the case.

 

[netxzero64]

with the updated drivers... take a look

http://www.techspot.com/review/283-geforce-gtx-400-vs-radeon-hd-5800/

 

[nemesismk2]

I enjoyed reading this comparison by HardOCP. I usually only update my drivers if I am having a problem or there are new features which the drivers now have. I think that when a driver is released which seriously increases performance then the driver has probably disabled something which should be enabled!

 

[NoQuarter]

I enjoyed reading this comparison by HardOCP. I usually only update my drivers if I am having a problem or there are new features which the drivers now have. I think that when a driver is released which seriously increases performance then the driver has probably disabled something which should be enabled!

Nah, the game devs often end up with very generic 3d code that can be greatly enhanced by the driver team to take specific advantage of the rendering techniques their hardware uses. I don't think a lot of devs get as close to the metal as they used to and that leaves lots of room for improvements without taking 'shortcuts' or reducing image quality.


It's sort of how devs are able to squeeze more and more power out of consoles because they learn how to specialize their code for that hardware - for PC's ATI/Nvidia does this step for them because the dev can't really specialize their code for every GPU out there.

 

[Will Robinson]

with the updated drivers... take a look

http://www.techspot.com/review/283-geforce-gtx-400-vs-radeon-hd-5800/

Ouch the Green Team won't like that review.^_^
Kinda puts paid to the myth of GTX470 being faster than HD5850 once and for all.
Would have liked to seen them use the Cat 10.6 drivers though...

 

[Lonyo]

Ouch the Green Team won't like that review.^_^
Kinda puts paid to the myth of GTX470 being faster than HD5850 once and for all.
Would have liked to seen them use the Cat 10.6 drivers though...

That article was already posted a while ago and a long discussion ensued mainly consisting of people saying it was wrong (guess who) and other people pointing out it wasn't wrong.

 

[MrK6]

Your common sense is not welcome here.

My bad :awe:

That article was already posted a while ago and a long discussion ensued mainly consisting of people saying it was wrong (guess who) and other people pointing out it wasn't wrong.

People can talk until they're blue in the face (and I'm sure they will), but the simple fact that they're having trouble selling the GTX 470 at even $300 speaks volumes.

 

[ASTOCADDIDS]

The ATi drivers for Linux are poor at best ive heard.

 

[Voo]

The ATi drivers for Linux are poor at best ive heard.

2005 called and.. ah we all know.

My 4870 works completely fine under linux, it's always nice if members with only a handful posts, want to share their 3rd hand knowledge they just "heart somewhere" with all of us, though.

 

[netxzero64]

That article was already posted a while ago and a long discussion ensued mainly consisting of people saying it was wrong (guess who) and other people pointing out it wasn't wrong.

well, its either fanboyism or true performance on the benchmark results... but i think its true because of the fact that 5870 and GTX 480 are still 15% apart in performance...

 

[tincart]

That article was already posted a while ago and a long discussion ensued mainly consisting of people saying it was wrong (guess who) and other people pointing out it wasn't wrong.

Followed by happy_medium posting some of his own fudged benchmark numbers in an amusing attempt to prove that the TechSpot numbers were useless. Seemed almost a truism to claim that in-game benchmarking will give you a decent idea of in-game performance but some people didn't like the results.

 

[PingviN]

I don't see why it's bad drivers havn't improved performance much. I mean, it's good that a $350 card performs as it should when you buy it. I'd prefer that to waiting for 6 months before getting what I actually paid for.